@prefix vivo: . @prefix edm: . @prefix ns0: . @prefix dcterms: . @prefix skos: . vivo:departmentOrSchool "Medicine, Faculty of"@en, "Population and Public Health (SPPH), School of"@en ; edm:dataProvider "DSpace"@en ; ns0:degreeCampus "UBCV"@en ; dcterms:creator "Slogrove, Amy L."@en ; dcterms:issued "2015-12-01T03:19:05"@en, "2015"@en ; vivo:relatedDegree "Doctor of Philosophy - PhD"@en ; ns0:degreeGrantor "University of British Columbia"@en ; dcterms:description "Background: Universal infant morbidity risk factors (poor birth outcomes, suboptimal breastfeeding, poverty) occur more frequently in HIV exposed uninfected (HEU) than HIV unexposed uninfected (HUU) infants. HEU infants’ unique exposures, including in utero exposure to HIV products and maternal immune compromise, may potentiate HEU infants’ infectious morbidity risk. The primary objective was to determine whether HEU infants experience greater infectious morbidity than HUU infants through HIV exposure-specific pathways beyond universal infant morbidity risk factors. Methods: This prospective cohort study identified low risk HIV-infected and HIV-uninfected mothers and their term newborns from a single community midwife unit in Kraaifontein, South Africa. The primary outcome, at least one infectious cause hospitalization or death before six months of age, was classified according to modified WHO case-definitions and compared between HEU and HUU infants. Complete outcome determination on all infants was possible through linkage with the electronic provincial hospital administration system and mortality registry. Adjusted odds ratios (aOR) were calculated by multivariable logistic regression including stratified analyses conditioned on breastfeeding. Results: One hundred and seventy six (94 HEU, 82 HUU) mother-infant pairs were included. HIV-infected mothers were older (median 27.8 vs. 24.7 years, p<0.01) and HEU infants less often breastfed (35/94 (37%) vs. 81/82 (99%), p<0.001). The groups were similar on maternal education, antenatal course, household characteristics, birth weight, gestational age and immunizations. Incidence rate ratio of all-cause sick clinic visits in HEU compared to HUU infants was 0.82 (95% CI 0.58,1.16). The primary outcome occurred in 17 (18%) HEU and 10 (12%) HUU infants (p=0.38), giving an aOR of 1.45 (95% CI 0.44,4.55). In stratified analysis comparing only infants with any breastfeeding, HEU infants had an aOR for a very severe infectious cause hospitalization or death of 4.2 (95% CI 1.00,19.2, p=0.05). Seven of 17(41%) HEU and 1/10 (10%) HUU primary outcome events occurred after 90 days of age (p=0.07). Conclusion: Amongst term infants with similar social circumstances, a higher probability of very severe infectious morbidity was observed in breastfed HEU compared to breastfed HUU infants. HEU infant risk may be driven through HIV exposure-specific pathways unrelated to universal infant morbidity risk factors."@en ; edm:aggregatedCHO "https://circle.library.ubc.ca/rest/handle/2429/55603?expand=metadata"@en ; skos:note "THE PATTERN AND PATHWAYS OF INFECTIOUS MORBIDITY IN SOUTH AFRICAN HIV EXPOSED UNINFECTED INFANTS by Amy L. Slogrove Master of Medicine (Paediatrics), Stellenbosch University, 2009 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF Doctor of Philosophy in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Population and Public Health) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) November 2015 © Amy L. Slogrove, 2015 ii Abstract Background: Universal infant morbidity risk factors (poor birth outcomes, suboptimal breastfeeding, poverty) occur more frequently in HIV exposed uninfected (HEU) than HIV unexposed uninfected (HUU) infants. HEU infants’ unique exposures, including in utero exposure to HIV products and maternal immune compromise, may potentiate HEU infants’ infectious morbidity risk. The primary objective was to determine whether HEU infants experience greater infectious morbidity than HUU infants through HIV exposure-specific pathways beyond universal infant morbidity risk factors. Methods: This prospective cohort study identified low risk HIV-infected and HIV-uninfected mothers and their term newborns from a single community midwife unit in Kraaifontein, South Africa. The primary outcome, at least one infectious cause hospitalization or death before six months of age, was classified according to modified WHO case-definitions and compared between HEU and HUU infants. Complete outcome determination on all infants was possible through linkage with the electronic provincial hospital administration system and mortality registry. Adjusted odds ratios (aOR) were calculated by multivariable logistic regression including stratified analyses conditioned on breastfeeding. Results: One hundred and seventy six (94 HEU, 82 HUU) mother-infant pairs were included. HIV-infected mothers were older (median 27.8 vs. 24.7 years, p<0.01) and HEU infants less often breastfed (35/94 (37%) vs. 81/82 (99%), p<0.001). The groups were similar on maternal education, antenatal course, household characteristics, birth weight, gestational age and immunizations. Incidence rate ratio of all-cause sick clinic visits in HEU compared to HUU infants was 0.82 (95% CI 0.58,1.16). The primary outcome occurred in 17 (18%) HEU and 10 (12%) HUU infants (p=0.38), giving an aOR of 1.45 (95% CI 0.44,4.55). In stratified analysis comparing only infants with any breastfeeding, HEU infants had an aOR for a very severe infectious cause hospitalization or death of 4.2 (95% CI 1.00,19.2, p=0.05). Seven of 17(41%) HEU and 1/10 (10%) HUU primary outcome events occurred after 90 days of age (p=0.07). Conclusion: Amongst term infants with similar social circumstances, a higher probability of very severe infectious morbidity was observed in breastfed HEU compared to breastfed HUU infants. HEU infant risk may be driven through HIV exposure-specific pathways unrelated to universal infant morbidity risk factors. iii Preface The work presented in Chapters 2 and 3 of this dissertation was conducted solely by Dr. Slogrove under the supervision of Drs. Julie A. Bettinger and Joel Singer. The work presented in Chapters 4 to 7 was conducted by Dr. Slogrove in collaboration with the Mother Infant Health Study principal investigators (Drs. Monika M. Esser, David P. Speert, Mark F. Cotton and Tobias R. Kollmann) and PhD students (Dr. Gareth Mercer and Ms. Moleen Zunza). Dr. Slogrove took primary responsibility for writing the study proposal with guidance from Dr. Bettinger and input from Dr. Gareth Mercer. Dr. Slogrove designed the maternal and infant health questionnaires. The household and infant feeding questionnaires were designed by Dr. Gareth Mercer and Ms. Moleen Zunza respectively. Dr. Slogrove supervised but was not directly involved with data collection. She assisted with the study database design and performed database quality control and data cleaning. Dr. Slogrove conducted all statistical analysis independently under the supervision of Drs. Bettinger, Singer and MacNab. As of August 2015, none of the work from this dissertation had been published. Findings from Chapters 6 and 7 were presented respectively at the International HIV Pediatrics Workshop and International AIDS Society Conference in Vancouver, Canada, July 2015. This study was approved by the University of British Columbia Children and Women’s Health Centre of British Columbia Research Ethics Board (H12-01181 and H13-03518), the Stellenbosch University Faculty of Medicine & Health Sciences Health Research Ethics Committee (S12/01/009 and N13/10/139) and the Western Cape Provincial Health Impact Assessment Committee (2012RP22 and 2013RP191). iv Table of Contents Abstract ....................................................................................................................................................... ii!Preface ........................................................................................................................................................ iii!Table of Contents ...................................................................................................................................... iv!List of Tables .............................................................................................................................................. xi!List of Figures .......................................................................................................................................... xiv!List of Abbreviations ................................................................................................................................ xv!Acknowledgements ................................................................................................................................ xvii!Dedication .............................................................................................................................................. xviii!Chapter 1 Introduction ............................................................................................................................... 1!1.1! General HIV exposed uninfected infant morbidity & mortality ........................................................ 2!1.2! South African context ..................................................................................................................... 8!1.3! Universal infant morbidity and mortality risk factors ....................................................................... 9!1.3.1! Preterm and small for gestational age newborns ................................................................... 9!1.3.2! Infant feeding ........................................................................................................................ 10!1.3.3! Growth and nutrition ............................................................................................................ 11!1.3.4! Infectious pathogen exposure ............................................................................................... 11!1.3.5! Maternal wellbeing ................................................................................................................ 12!1.3.6! Poverty and social circumstances ........................................................................................ 12!1.4! Infant risk factors unique to HEU infants ...................................................................................... 13!1.4.1! In-utero environment altered by maternal HIV disease ........................................................ 13!1.4.1.1! HIV disease dynamics ................................................................................................... 14!1.4.1.2! Antiretroviral therapy versus antiretroviral prophylaxis .................................................. 15!1.4.1.3! Maternal HIV disease in the pre-antiretroviral therapy era ............................................ 15!1.4.1.4! Maternal HIV disease in the antiretroviral therapy era .................................................. 16!1.4.2! Immunological differences in HEU infants ............................................................................ 17!1.4.3! Antiretroviral exposure in HEU infants .................................................................................. 18!v 1.4.4! Cotrimoxazole preventive therapy ........................................................................................ 20!1.5! HEU infant infectious morbidity .................................................................................................... 21!1.5.1! Respiratory tract infections ................................................................................................... 21!1.5.2! Diarrhoea .............................................................................................................................. 21!1.5.3! Bacterial sepsis ..................................................................................................................... 22!1.5.3.1! Bacterial sepsis in neonates .......................................................................................... 22!1.5.3.2! Bacterial sepsis in infancy ............................................................................................. 23!1.5.4! Congenital viral infections ..................................................................................................... 23!1.6! Cape Town HEU pilot study ......................................................................................................... 24!1.7! Community context ...................................................................................................................... 25!1.8! Study aims, hypotheses and objectives ....................................................................................... 27!1.8.1! Study aim .............................................................................................................................. 27!1.8.2! Primary hypothesis and objective ......................................................................................... 27!1.8.3! Secondary hypothesis 1 and secondary objective 1 ............................................................. 28!1.8.4! Secondary hypothesis 2 and secondary objective 2 ............................................................. 28!Chapter 2 The Paediatric Infectious Event Tool for Research (PIET-R) .............................................. 29!2.1! Rationale for the development of the PIET-R .............................................................................. 29!2.1.1! Definitions from clinical trial research ................................................................................... 29!2.1.2! Definitions from infectious disease surveillance ................................................................... 30!2.1.3! Definitions from child health management ............................................................................ 31!2.2! PIET-R case-definitions ............................................................................................................... 31!2.3! Designing, pilot testing and evaluating the PIET-R ...................................................................... 35!2.4! PIET-R evaluation hypotheses and objectives ............................................................................. 36!Chapter 3 PIET-R reliability and validity evaluation: methods, results and discussion .................... 37!3.1! Methods ....................................................................................................................................... 37!3.1.1! Study design ......................................................................................................................... 37!3.1.2! Study population and setting ................................................................................................ 37!3.1.3! Sample selection .................................................................................................................. 37!vi 3.1.4! Eligibility criteria .................................................................................................................... 38!3.1.5! Study procedures and data collection ................................................................................... 38!3.1.5.1! Determination of gold standard diagnosis ..................................................................... 38!3.1.5.2! Testing of reliability and validity ..................................................................................... 38!3.1.5.3! Study documentation ..................................................................................................... 39!3.1.6! Data management ................................................................................................................ 39!3.1.7! Ethical considerations ........................................................................................................... 40!3.1.8! Sample size calculation ........................................................................................................ 40!3.1.9! Analytic strategy ................................................................................................................... 40!3.1.9.1! Primary objective: reliability ........................................................................................... 40!3.1.9.2! Secondary objective: validity ......................................................................................... 42!3.2! Results ......................................................................................................................................... 43!3.3! Discussion .................................................................................................................................... 45!Chapter 4 Mother Infant Health Study methods .................................................................................... 48!4.1! Study design, population and setting ........................................................................................... 48!4.2! Study methodology ...................................................................................................................... 48!4.2.1! Eligibility criteria .................................................................................................................... 48!4.2.2! Study procedures .................................................................................................................. 49!4.2.2.1! Study site preparation .................................................................................................... 49!4.2.2.2! Enrolment and informed consent process ..................................................................... 49!4.2.2.3! Co-ordination of study visits .......................................................................................... 51!4.2.2.4! Health record reviews .................................................................................................... 51!4.2.2.5! Interviews and infant physical examination ................................................................... 51!4.2.2.6! Maternal investigations .................................................................................................. 52!4.2.2.7! Infant investigations ....................................................................................................... 52!4.2.2.8! Laboratory procedures .................................................................................................. 52!4.2.2.9! Outcome determination – infectious cause hospitalization or death ............................. 53!4.2.2.10! Additional sub-studies ................................................................................................. 54!vii 4.2.3! Data Management ................................................................................................................ 54!4.3! Study variable definitions ............................................................................................................. 54!4.3.1! Maternal variables ................................................................................................................ 54!4.3.2! Household variables ............................................................................................................. 56!4.3.3! Infant variables ..................................................................................................................... 56!4.3.4! Hospitalization variables ....................................................................................................... 58!4.4! Ethical Considerations ................................................................................................................. 58!Chapter 5 Analytic methods .................................................................................................................... 60!5.1! Study objectives ........................................................................................................................... 60!5.2! Definition of major determinants and outcomes ........................................................................... 60!5.2.1! Primary determinant: HIV exposure status ........................................................................... 60!5.2.2! Definition of breastfeeding status ......................................................................................... 60!5.2.3! Definition of maternal ARV regimen during pregnancy ......................................................... 61!5.2.4! Definition of primary outcome – infectious cause hospitalization or death ........................... 61!5.2.5! Definition of secondary outcomes – severe and very severe infectious cause hospitalization or death ............................................................................................................................................. 61!5.3! Sample size calculation ................................................................................................................ 62!5.4! Analytic strategy ........................................................................................................................... 63!5.4.1! Description of the analytic cohort .......................................................................................... 63!5.4.2! Analysis of primary objective ................................................................................................ 64!5.4.2.1! Control of confounding .................................................................................................. 64!5.4.3! Analysis of secondary objective 1 ......................................................................................... 66!5.4.4! Analysis of secondary objective 2 ......................................................................................... 67!Chapter 6 Results - Do HEU infants have a greater probability of infectious morbidity than peer HUU infants? ............................................................................................................................................. 68!6.1! Cohort background ....................................................................................................................... 68!6.2! Comparison of HEU and HUU infants .......................................................................................... 72!6.2.1! Maternal characteristics ........................................................................................................ 72!viii 6.2.2! Household characteristics ..................................................................................................... 74!6.2.3! Infant characteristics ............................................................................................................. 76!6.3! Comparison of infants with and without the primary outcome by HIV exposure group ................ 85!6.3.1! Maternal characteristics ........................................................................................................ 85!6.3.2! Household characteristics ..................................................................................................... 86!6.3.3! Infant characteristics ............................................................................................................. 89!6.4! Characterization of hospitalization events .................................................................................... 93!6.4.1! Infectious cause hospitalizations .......................................................................................... 93!6.4.2! Non-infectious cause hospitalizations ................................................................................... 96!6.5! Determining the effect of HIV exposure on infectious morbidity .................................................. 97!6.5.1! Multivariable analysis ............................................................................................................ 97!6.5.2! Stratified analysis ................................................................................................................ 102!6.6! Summary of main findings .......................................................................................................... 103!Chapter 7 Results - In the antiretroviral therapy era, is HEU infant infectious morbidity associated with the extent of maternal HIV disease? ............................................................................................. 104!7.1! Description of all HIV-infected mothers ...................................................................................... 104!7.2! Sub-group: HEU infant comparison by maternally indicated cART or VTP prophylaxis ............ 105!7.2.1! Maternal characteristics ...................................................................................................... 105!7.2.2! Household characteristics ................................................................................................... 109!7.2.3! Infant characteristics ........................................................................................................... 111!7.3! Sub-group: HEU infants - comparison of predisposing characteristics by primary outcome ..... 118!7.3.1! Maternal characteristics ...................................................................................................... 118!7.3.2! Household characteristics ................................................................................................... 118!7.3.3! Infant characteristics ........................................................................................................... 122!7.4! Determining the effect of maternal ARV regimen during pregnancy on HEU infant risk for infectious morbidity .............................................................................................................................. 124!Chapter 8 Discussion ............................................................................................................................. 126!8.1! The mothers, households and infants in this study .................................................................... 127!ix 8.2! Infectious morbidity in HEU compared to HUU infants .............................................................. 128!8.2.1! All cause infectious morbidity ............................................................................................. 128!8.2.2! Lower respiratory tract infectious morbidity ........................................................................ 129!8.2.3! Diarrhoeal morbidity ............................................................................................................ 130!8.2.4! Pattern of infectious morbidity: the timing and severity of infectious events ....................... 130!8.2.4.1! Timing of infectious morbidity ...................................................................................... 131!8.2.4.2! Severity of infectious morbidity .................................................................................... 132!8.2.5! The relationship between maternal CD4 count and infant infectious morbidity .................. 132!8.2.6! Infectious morbidity risk beyond suboptimal breastfeeding ................................................ 134!8.3! Potential immunologic explanations for HEU infant infectious morbidity ................................... 135!8.3.1! Deficient transplacental antibody transfer ........................................................................... 136!8.3.2! Delayed functional immune development ........................................................................... 136!8.3.3! Breast milk immunologic quality ......................................................................................... 137!8.4! The extent of maternal HIV disease and HEU infant infectious morbidity .................................. 138!8.4.1! Maternal disease stabilisation versus disease progression ................................................ 139!8.4.2! Maternal immune suppression versus immune activation .................................................. 140!8.4.3! Improved maternal physical well-being ............................................................................... 141!8.5! Potential effects of ARV exposure observed in HEU infants ...................................................... 142!8.5.1! Anaemia in zidovudine-exposed HEU infants ..................................................................... 142!8.5.2! Shorter length in tenofovir-exposed HEU infants ................................................................ 143!8.5.3! Central nervous system defects in HEU infants with first trimester efavirenz exposure ..... 144!8.6! Challenges and limitations ......................................................................................................... 145!8.6.1! Implementation challenges ................................................................................................. 145!8.6.2! Analytic challenges ............................................................................................................. 146!8.6.3! Additional study limitations ................................................................................................. 148!8.7! Recommendations and implications .......................................................................................... 149!8.7.1! The importance of an appropriate control group ................................................................. 149!8.7.2! Looking beyond incidence of morbidity ............................................................................... 150!x 8.7.3! Implications ......................................................................................................................... 150!Chapter 9 Conclusion ............................................................................................................................. 151!References .............................................................................................................................................. 153!Appendices ............................................................................................................................................. 177!Appendix A PIET-R case-definitions ................................................................................................... 177!Appendix B PIET-R Evaluation ............................................................................................................ 180!B.1! PIET-R Hospitalization Event Gold Standard Source Document .......................................... 180!B.2! PIET-R: Abstraction Source Document ................................................................................. 181!B.3! PIET-R: Final Classification and Grade Source Document ................................................... 184!B.4! PABAK descriptive classification according to Byrt ............................................................... 185!Appendix C Study variables ................................................................................................................ 186!Appendix D Subgroup of all HEU infants – additional tables ............................................................... 191!Appendix E Population effect of HIV exposure .................................................................................... 199! xi List of Tables Table 1.1 Summary of studies comparing infectious morbidity and all-cause mortality in African HEU and HUU infants ................................................................................................................................................. 5!Table 2.1 PIET-R case definitions for lower respiratory tract infections and diarrhoeal disease ............... 33!Table 3.1 Contingency table for inter-observer agreement of diarrhoea case-definition ........................... 41!Table 3.2 Contingency table for sensitivity and specificity of diarrhoea case-definition ............................ 42!Table 3.3 Inter-observer agreement of PIET-R diagnostic classifications ................................................. 44!Table 3.4 Validity of PIET-R diagnostic classifications .............................................................................. 45!Table 4.1 Summary of study procedures ................................................................................................... 50!Table 5.1 Sample size calculation scenarios ............................................................................................. 62!Table 6.1 Baseline maternal and infant characteristics compared by mother-infant pairs retained in the cohort at two weeks and those lost before two weeks of infant age .......................................................... 69!Table 6.2 Infant reasons for non-completion of 6-month follow-up ............................................................ 71!Table 6.3 Demographic and health characteristics of HIV-infected and HIV-uninfected mothers ............. 73!Table 6.4 Household characteristics compared by infant HIV exposure group ......................................... 75!Table 6.5 Infant birth and healthcare characteristics compared by infant HIV exposure group ................ 77!Table 6.6 Infant feeding characteristics compared by infant HIV exposure group .................................... 79!Table 6.7 Infant haemoglobin and anaemia compared by infant HIV exposure group .............................. 81!Table 6.8 Infant anthropometric measurements compared by infant HIV exposure group ....................... 82!Table 6.9 Maternal demographic and health characteristics compared by infant outcome group ............. 87!Table 6.10 Household characteristics compared by infant outcome group ............................................... 88!Table 6.11 Infant birth and healthcare characteristics compared by infant outcome group ...................... 89!Table 6.12 Infant feeding characteristics compared by infant outcome group .......................................... 90!Table 6.13 Infant haemoglobin and anaemia compared by infant outcome group .................................... 91!Table 6.14 Infant anthropometric measurements compared by infant outcome group ............................. 92!Table 6.15 Table of cumulative proportions of infants with infectious cause hospitalizations (including death) over time compared by HIV exposure group .................................................................................. 94!xii Table 6.16 Comparison of HEU and HUU infants by highest grade of severity and type of major infectious events ........................................................................................................................................................ 96!Table 6.17 Odds ratios for the associations between maternal and infant characteristics and the primary outcome, adjusted for HIV exposure ......................................................................................................... 98!Table 6.18 Odds ratios for the association between infant feeding mode and the primary outcome, adjusted for HIV exposure ......................................................................................................................... 99!Table 6.19 Logistic regression models for the odds of the primary and secondary outcomes in HEU infants relative to HUU infants ................................................................................................................. 101!Table 6.20 Stratified analysis of the effect of HIV exposure on the primary and secondary outcomes, conditioned on breastfeeding and adjusted for maternal age .................................................................. 103!Table 7.1 Maternal demographic & obstetric characteristics compared by maternal pregnancy ARV regimen .................................................................................................................................................... 107!Table 7.2 Maternal HIV characteristics compared by maternal pregnancy ARV regimen ....................... 108!Table 7.3 Household characteristics compared by maternal pregnancy ARV regimen ........................... 110!Table 7.4 Infant birth and healthcare characteristics compared by maternal pregnancy ARV regimen .. 112!Table 7.5 Infant feeding characteristics compared by maternal pregnancy ARV regimen ...................... 113!Table 7.6 Infant haemoglobin and anaemia compared by maternal pregnancy ARV regimen ............... 114!Table 7.7 Infant anthropometric measurements compared by maternal pregnancy ARV regimen ......... 115!Table 7.8 Maternal demographic and obstetric characteristics compared by HEU infant outcome group ................................................................................................................................................................. 119!Table 7.9 Maternal HIV characteristics compared by HEU infant outcome group ................................... 120!Table 7.10 Household characteristics compared by HEU infant outcome group .................................... 121!Table 7.11 Infant birth and healthcare characteristics compared by infant outcome group .................... 122!Table 7.12 Infant feeding characteristics compared by HEU infant outcome group ............................... 123!Table 7.13 Odds ratios for the associations between maternal and infant characteristics and the primary outcome, adjusted for maternal ARV regimen ......................................................................................... 125!Table A.1 Table of PIET-R case-definitions ............................................................................................. 177 Table B.1 PABAK descriptive classification ............................................................................................. 185!xiii Table C.1 Table of variables .................................................................................................................... 186!Table D.1 Maternal characteristics compared by HEU infant outcome group ......................................... 191!Table D.2 Maternal CD4 and HIV viral load compared by HEU infant outcome group ........................... 192!Table D.3 Infant characteristics compared by HEU infant outcome group .............................................. 193!Table D.4 Infant feeding characteristics compared by HEU infant outcome group ................................. 194!Table D.5 Infant haemoglobin and anaemia compared by HEU infant outcome group ........................... 195!Table D.6 Infant anthropometric measurements compared by HEU infant outcome group .................... 196!Table D.7 Infant haemoglobin and anaemia compared by in-utero ZDV-exposure ................................. 197!Table D.8 Infant anthropometric measurements compared by in-utero TDF-exposure ......................... 198!Table E.1 Attributable fraction and population attributable fraction for estimates of the risk ratio for infectious cause hospitalization or death in HEU relative to HUU infants in South Africa ....................... 199! xiv List of Figures Figure 1.1 Schematic representation of HEU infant pathways to infectious morbidity ................................ 7!Figure 5.1 Conceptual framework of the relationship between HIV exposure and infant infectious morbidity .................................................................................................................................................... 65!Figure 5.2 Schematic representation of stratified analysis conditioned on the presence of any breastfeeding ............................................................................................................................................. 66!Figure 6.1 Disposition of participants over the duration of study ............................................................... 70!Figure 6.2 WHO weight-for-age Z-score compared by infant HIV exposure group ................................... 83!Figure 6.3 WHO length-for-age Z-score compared by infant HIV exposure group .................................... 83!Figure 6.4 WHO weight-for-length Z-score compared by infant HIV exposure group ............................... 84!Figure 6.5 Kaplan-Meier curves of time to first infectious cause hospitalization or death in HEU and HUU infants ........................................................................................................................................................ 94!Figure 6.6 Infant age at time of primary outcome compared by HIV exposure group ............................... 95!Figure 7.1 Infant WHO weight-for-age Z-score compared by maternal pregnancy ARV regimen ........... 116!Figure 7.2 Infant WHO length-for-age Z-score compared by maternal pregnancy ARV regimen ........... 116!Figure 7.3 Infant WHO weight-for-length Z-score compared by maternal pregnancy ARV regimen ....... 117! xv List of Abbreviations 3TC Lamivudine AE Adverse event AEFI Adverse event following immunization AIDS Acquired immune deficiency syndrome APR Antenatal Pregnancy Registry ARV Antiretroviral CAD Canadian Dollar cART Combination antiretroviral therapy CDC Centers for Disease Control CI Confidence interval CMV Cytomegalovirus CPT Cotrimoxazole preventive therapy CRF Case report form DAIDS Division of AIDS DCHS Drakenstein Child Health Study EFV Efavirenz FTC Emtricitabine GBS Group B Streptococcus Hb Haemoglobin HCZ Head-circumference Z-score HEI HIV exposed infant HEU HIV exposed uninfected HIV Human Immunodeficiency Virus HR Hazard ratio HUU HIV unexposed uninfected Ig Immunoglobulin IMCI Integrated Management of Childhood Illnesses IPT Isoniazid preventive therapy KID-CRU Children’s Infectious Diseases Clinical Research Unit LAZ Length-for-age Z-score LBW Low birth weight LMIC Low-middle income country LPV/r Lopinavir/ritonavir LRTI lower respiratory tract infection xvi M.tb Mycobacterium tuberculosis MOU Midwife Obstetric Unit MUACZ Mid-upper-arm-circmference Z-score NHLS National Health Laboratory Service NVP Nevirapine OR Odds ratio PABAK Prevalence adjusted bias adjusted kappa PCP Pneumocystis pneumonia Pe Proportion expected agreement PIET-R Paediatric Infectious Event Tool for Research PMTCT Prevention of mother to child transmission PNA Proportion negative agreement Po Proportion observed agreement PPA Proportion positive agreement RaR Rate ratio RCT Randomized control trial RiR Risk ratio RNA Ribose nucleic acid SASPID South African Society for Paediatric Infectious Diseases SATS South Africa Thoracic Society SGA Small for gestational age TB Tuberculosis TDF Tenofovir disoproxil fumarate UBC University of British Columbia VEC Vaccine Evaluation Center VTP Vertical transmission prevention WAZ Weight-for-age Z-score WHO World Health Organisation WLZ Weight-for-length Z-score ZAR South African Rand ZDV Zidovudine ZEBS Zambia Exclusive Breastfeeding Study ZVITAMBO Zimbabwe Vitamin A for Mothers and Babies Project xvii Acknowledgements The study families who gave so generously. Monika Esser and Mark Cotton for igniting my “HEU” spark eight years ago and for fanning the flame in the years since with your unwavering belief in the value of this work. David Speert for your vision and confidence in my potential to succeed (and for sharing the beauty of Whistler). My PhD supervisors Julie Bettinger and Joel Singer, for taking a chance when you didn’t know me from a bar of soap, for your dedication to my task, unfailing support and exquisite guidance in navigating this journey. My supervisory committee members Drs. Ying MacNab, Mark Cotton and David Speert, for always being available and giving freely of your time. The Mother Infant Health Study team, including the project steering committee (Monika Esser, David Speert, Tobi Kollmann, Mark Cotton and Julie Bettinger), the Vaccine Evaluation Center (VEC) team in Vancouver and the Children’s Infectious Diseases Clinical Research Unit (KID-CRU) team in Cape Town for your commitment and expert contributions to this project. Arlene and Marcha, for your skilled study supervision as well as your moral support and friendship. Kim and Wenli, for your incredible patience, belief and friendship. Maureen for your organisation and friendly administrative support. The UBC Peter Wall Institute for Advanced Studies for funding the Mother Infant Health Study and my personal funders the Canadian Institutes of Health Research through the Canada Hope Fellowship and Canadian HIV Trials Network International Fellowship as well as the South African National Health Scholars Programme. The School of Population & Public Health at the University of British Columbia for the outstanding teaching and Beth Hensler for the thankless administrative support. The Department of Paediatrics & Child Health at Stellenbosch University under the leadership of Prof. Marianna Kruger, for keeping me in the family while supporting my pursuit of a PhD at UBC. The Stellenbosch University Ukwanda Rural Health Centre in Worcester, for giving me a quiet space to write. Gareth, Moleen, Cilla, Aneesa, Adrie, Helena, Regan, Rachel, Lyndsay, Alden and Amy for sharing in the highs and lows of this PhD journey together. Arlene & Les and Jenny & Richard for giving me homes-away-from-home in Vancouver. Nicole, Lynda, Rosalyn and Aneesa for your treasured friendship that has spanned decades and continents. My mom (Sandy) dad (Ashley), brother (Doug) and extended family (Slogrove, Gaskell, Swart, Cooper and Fulmer) for your unconditional love, support and encouragement. Oosie and Ben who have made the biggest sacrifices of all and continually rejuvenate my spirit. xviii Dedication Oostewalt and Benjamin, my guiding stars 1 Chapter 1 Introduction Human Immunodeficiency Virus (HIV) is thought to have originated in south-eastern Cameroon in West Africa during the early 20th century, subsequently making its way along the trade routes to East and then Southern Africa (1). HIV in Sub-Saharan Africa has spread predominantly through heterosexual transmission. In Africa, over 50% of all HIV infections are in women, thus large numbers of HIV exposed children are born to HIV-infected mothers (2). The developed world has seen HIV transmission concentrated largely in identifiable high-risk groups, initially blood transfusion recipients, intravenous drug users and men-who-have-sex-with-men, as opposed to being generalized in the population as in Africa. With increasing migration from Africa to North America and Europe, these continents are starting to see an increased heterosexual contribution and greater numbers of children born to HIV-infected mothers (3,4). Globally, in 2012, of 35 million people living with HIV, about 1 million were in North America, and 25 million in Sub-Saharan Africa. Sub-Saharan Africa represents 70% of all HIV-infected people and 85% of HIV-infected children on a continent with only 12% of the world’s population (5). HIV prevalence in Africa varies markedly from 1.2% in Ghana and Ethiopia to almost 25% in Swaziland (5). In South Africa HIV prevalence amongst adults aged 15 to 49 years is 18%. One of the many tragedies of the HIV epidemic is the ability of the virus to transmit vertically from mothers to their children during pregnancy (in-utero), labour (intra-partum) and via breast milk (6). There is undoubtedly still much to be done in order to achieve the elimination of paediatric HIV-infection, however prevention of this vertical transmission of HIV from mothers to children is one of the public health successes of the 21st century. Globally it is estimated that annually 1.4 million HIV-infected women become pregnant resulting in more than a million HIV exposed infants born each year (5). Expanding vertical transmission prevention (VTP) programmes have resulted in marked reductions in the numbers of HIV-infected children, but those lucky enough to avoid HIV-infection still bear consequences of HIV exposure (5). Success of these VTP programmes across the globe is not uniform however. North America and Europe have seen virtual elimination of paediatric HIV-infection, compared with variable success in Africa where some countries have instituted interventions that have reduced vertical transmission to under 5%, while others still struggle with poor VTP programme coverage in fewer than 50% of pregnant HIV-infected women (5,7,8). In South Africa 30% of pregnant women are HIV-infected but as a result of a large effective public VTP programme less than 3% of the 300 000 HIV-exposed South African infants born annually are HIV-infected by six weeks of age (8,9). This large and growing population of HIV-exposed but uninfected (HEU) infants however may not achieve equivalent child health outcomes to their HIV unexposed uninfected (HUU) peers. 2 This dissertation seeks to better understand this high-risk group of HEU infants within the context of general infant vulnerability in South Africa, specifically the pathways to and pattern of infectious morbidity in HEU compared to HUU infants. See section 1.8 for the specific aim, hypotheses and objectives. Throughout this dissertation the term HEU will be used to refer to HIV exposed infants confirmed to be HIV uninfected. The terms HIV-exposed infants and HIV-exposure refer to both HIV-infected and HIV-uninfected infants born to HIV-infected mothers. 1.1 General HIV exposed uninfected infant morbidity & mortality Awareness has only been generated during the last decade that being born to an HIV-infected mother, that is being HIV exposed, carries consequences even for the HIV-uninfected HEU infant and child (3,10–12). When one considers the course an HEU child takes starting from conception, there are numerous potential pathways, particularly during infancy, that could mediate greater risk for mortality as well as poor health and development outcomes. The HIV-exposed foetus is exposed not only to an in-utero environment altered by maternal HIV but also to antiretroviral drugs in order to prevent vertical HIV infection (13,14). In addition the foetus may be exposed to infectious agents besides HIV, including Mycobacterium tuberculosis (M.tb) and cytomegalovirus (CMV), more often than HIV-unexposed foetuses (15–18). These in-utero exposures contribute in varying degrees to poorer birth outcomes, including higher rates of prematurity and low birth weight in HIV-exposed compared to HUU newborns (19,20). After birth, close contact with HIV-infected parents during infancy may result in greater exposure to infectious pathogens including Pneumocystis jirovecii and M.tb (21,22). These infants, even if HIV-uninfected, can be exposed to long courses of antiretroviral (ARV) drugs to ensure they remain HIV-infection free. Other prophylactic antimicrobial agents include trimethoprim-sulphamethoxazole for Pneumocystis pneumonia (PCP) prevention and isoniazid preventive therapy for primary tuberculosis (TB) prophylaxis (14,23). Although unavoidable, these drug exposures during infancy are not without concern and their full effects on HEU infants’ growth and development are not clearly defined (12). Due to the risk of HIV transmission via breast milk, infant feeding choices are complex in the setting of HIV, with many HIV-exposed infants having been deprived of the immunological and nutritional benefits of breast milk (24). HIV-affected families may experience greater compromise in their social circumstances with enhanced poverty resulting in greater infant morbidity and mortality (25,26). These infants and children are also more likely to lose one or both parents with particularly maternal mortality holding dire consequences for any child (27). Most of what is known about HEU infant infectious morbidity comes from observations made early in the epidemic describing the natural history of HIV infected children compared to HIV-uninfected children, both 3 HIV exposed and unexposed, or from secondary analyses of VTP clinical trials often without an HUU control group (10,28–33). Of the 16 published African studies comparing adequately defined HEU to HUU infants for all cause mortality or infectious morbidity, only two were designed to primarily compare these two groups (Table 1.1). These two studies were limited to the outcomes of invasive pneumococcal disease and pneumonia treatment failure at 48 hours (Table 1.1 study 13 & 15) (34,35). Eleven of the 16 studies were conducted prior to 2009 in the context of limited antiretroviral interventions for either maternal health or VTP (28–30,34–47). Early, prospective cohort studies, designed to describe the natural history of HIV infection in Rwandan and Malawian children, concluded that HEU infants were no different from HUU infants for morbidity and mortality (Table 1.1 study 1 & 3) (29,30). The Rwandan cohort study observed an inconclusive elevated hazard for hospitalizations and pneumonia (hazard ratio (HR) 1.4 (95% confidence interval (95%CI) 0.9, 2.0) and 1.3 (95% CI 0.9,1.4) respectively) (29). The Malawian study found no difference in a host of morbidities and, for mortality, analysed HEU infants and HUU infants as a single group of HIV-uninfected infants, potentially masking differences between these two groups (30). The first study to definitively demonstrate a higher mortality and morbidity in HEU compared to HUU African infants was the Zimbabwe Vitamin A for Mothers and Babies Project (ZVITAMBO) (Table 1.1 study 5) (37,38). Conducted from 1997 to 2000, ZVITAMBO was a randomized placebo controlled trial of Vitamin A for preventing vertical transmission of HIV. This large study was conducted prior to the availability of ARVs for either treatment or prevention of HIV transmission with 94% of infants breastfed until 12 months of age. The study demonstrated an almost four times greater risk of mortality (rate ratio (RaR) 3.9, 95% CI 3.2, 4.8) at 12 months and two times greater risk of mortality (RaR 2.0, 95% CI 1.2, 3.5) at 24 months in HEU compared to HUU infants (37). The relative risk for mortality reached 5.5 (95% CI 3.9, 7.9) between eight weeks and 6 months of age, although the absolute risk for HEU mortality, as expected for all infants, was highest during the neonatal period. Eighty percent of HEU mortality occurred during the first 6 months of life. Morbidity experienced by HEU infants was also greater than in HUU infants, but not to the same extent as the elevation in mortality (38). An increase in all-cause hospitalizations was significantly higher during the neonatal period (RaR 1.5, 95% CI 1.2, 2.0) driven by an elevation in LRTI in HEU neonates (RaR 2.7, 95% CI 1.6, 4.7). All-cause hospitalizations between one and six months also occurred more frequently in HEU infants but the 95% confidence interval crossed the null (RaR 1.2, 95% CI 0.9, 1.6). Smaller Southern and Eastern African studies confirmed the elevated morbidity and mortality in HEU infants in this region. The Rakai Community Study conducted in Rakai, Uganda observed a slightly greater risk for mortality up to 18 months (RaR 1.16, p<0.005) (Table 1.1 study 4) (36). The Mashi study (Table 1.1 study 6) similar to ZVITAMBO, observed an at least four times greater risk for mortality and double the risk for hospitalization at six and 24 months (39). A South African VTP randomized control trial (RCT) observed an insignificant increased infant mortality hazard in HEU compared to HUU infants (adjusted hazard ratio (aHR) 1.3, 95% CI 0.7, 2.6) (Table 1.1 study 8) (41). A 4 pooled analysis of population cohort studies conducted in Uganda, Tanzania and Malawi prior to 2000 showed an increased hazard for mortality in children born to HIV-infected mothers of 2.9 (95% CI 2.3, 3.6) times that of children born to HIV-uninfected mothers after adjusting for the association between maternal mortality and child mortality (48). The analysis did not distinguish between HIV-infected and HEU children. Cohorts of HEU infants and children in Europe and the Americas reported rates of hospitalization and other morbidities in HEU infants without comparison to HUU control groups. The European Collaborative Study followed more than 1600 HEU infants from 1985 to 2002 and observed a hospitalization rate of 264/1000 child-years not associated with sociodemographic factors (3). Recently a retrospective cohort of Belgian HEU infants described a substantially higher rate of hospitalization for Group B Streptococcus and Streptococcus pneumoniae infections in HEU infants compared to the general infant population (49). The Women and Infants Transmission Study followed HIV-infected and HEU children in Puerto Rico and the United States between 1989 and 2001 and described morbidity and mortality to two years of age (50). Hospitalization occurred in 24% of HEU infants in the first two years of life with respiratory tract infections occurring most frequently (16%), followed by gastroenteritis (11%). The National Institute of Child Health and Human Development International Site Development Initiative Perinatal Study, a prospective study in Latin America and the Caribbean of HIV-infected mothers and their infants with ongoing enrolment since 2002, described infectious disease morbidity in 462 HEU infants up to six months of age between 2002 and 2004 (51). Hospitalizations occurred at least once in 17.5% of infants, the majority (53%) of admissions due to lower respiratory tract infections. 5 Table 1.1 Summary of studies comparing infectious morbidity and all-cause mortality in African HEU and HUU infants Country Year Study Type Sample size (HEU:HUU) Outcome Follow-up Findings (always HEU vs. HUU; point estimate (95% CI)) 1 (29) Rwanda 1989-94 Prospective cohort (138:209) Natural history of HIV in children 0-60 months Severe pneumonia aHR 1.4 (0.9,2.0); hospitalization aHR 1.3 (0.9,1.4); unknown feeding 2 (28) DRC (Zaire) 1989-92 Prospective cohort (139:191) Diarrhoea incidence 0->12 months Persistent diarrhoea incidence 4.9 vs. 2.7/100py (p=0.23) (unadjusted); most some breastfeeding. 3 (30) Malawi 1994-97 VTP RCT (499:119) Morbidity and mortality 6-36 months No difference in numerous morbidities. Mortality 0-12 months compared HIV-infected to HIV-uninfected; mortality 13-36 months 46.3/1000py vs. 35.7/1000py (no CI); all some breastfeeding 4 (36) Uganda 1994-98 Community surveillance (269:3183) Mortality 0-24 months Mortality 18 months RaR 1.16 (p<0.05); 98% some breastfeeding 5 (37,38) Zimbabwe 1997-00 VTP RCT (3135:9210) Morbidity and mortality 0-24 months Mortality 8 weeks to 6 months RaR 5.5 (3.9,7.9); mortality RaR > 2 (1.2,7.9) throughout 0-24 months; neonatal hospitalization RaR1.5 (1.2,2.0); post-neonatal hospitalization RaR1.2 (0.9,1.6); all had some breastfeeding 6 (39) Botswana 2000-05 VTP RCT (534:137) Morbidity and mortality 0-24 months Mortality RiR 4.2 (p<0.01); hospitalizations RiR 2.17 (p<0.001); all had some breastfeeding 7 (40) South Africa 2001-02 Hospital cohort (41:75) Severe pneumonia 48 hour treatment failure Infants aOR 6.02 (1.55,23.38); feeding unknown 8 (41) South Africa 2001-05 Breastfeeding peer support intervention cohort (936:115) Diarrhoeal morbidity and all-cause mortality 0-18 months Diarrhoea at 6 months aHR 1.45 (0.75,2.79); mortality 12 months aHR 0.77 (0.49,1.21); any breastfeeding 73% vs. 80% 6 Country Year Study Type Sample size (HEU:HUU) Outcome Follow-up Findings (always HEU vs. HUU; point estimate (95% CI)) 9 (42) South Africa 2004-07 RCT (1255:5804) Neonatal sepsis incidence 0-1 month Lower rate of early onset sepsis (20.6 vs. 33.7/1000 live births, p<0.05), similar late onset sepsis (5.8 vs. 4.1 per 1000 live births, p=0.6); feeding unknown 10 (43) Malawi 2008-09 Retrospective cohort PMTCT programme evaluation (128:200) Morbidity and mortality 0-20 months No difference in hospitalization (univariable analysis); mortality compared HEI to HUU infants; majority HEI and HUU mixed fed by 3 months 11 (44) Mozambique 2008-2009 Prospective cohort clinical, haematologic, immunologic profiles (Total 153) Incidence of out-patient visits and admissions 0-12 months RaR for out patient visits 0.79 (0.63,0.99); RaR for hospitalization 1.51 (0.71,3.18); all breastfed at 6 months 12 (45) South Africa 2009-11 Prospective cohort exploring innate immmune differences (27:28) Infectious cause hospitalization incidence 0-12 months RaR for hospitalization 2.74 (0.75,8.78); all but 1 HEU infants formula fed, all HUU some breastfeeding 13 (34) South Africa 2009-13 IPD Laboratory surveillance (253:377) IPD incidence and mortality 0-12 months IPD RaR 0-6 months 3.6 (2.8,4.6); mortality due to IPD aOR 1.76 (1.09,2.85); feeding unknown 14 (46) Uganda 2010-2013 Malaria prevention RCT (186:389) Morbidity and mortality 6-24 months No differences breastfed HEU and breastfed HUU; 84% and 99% breastfed at 6 months 15 (35) Botswana 2012-2013 Hospital cohort (64:153) Pneumonia 48 hour treatment failure 1-23 months Treatment failure: RiR1.83 (1.27,2.64); mortality RiR 4.31 (1.44,12.87); no adjustment for breastfeeding 16 (47) South Africa 2012-2014 Birth cohort study (130:567) Pneumonia incidence, severity, risk factors 12 months All pneumonia: aRaR 1.62 (1.01,2.61); severe pneumonia: RaR 4.04 (1.51,10.8); adjusted for any formula feeding before 6 months aHR – adjusted hazard ratio; aOR - adjusted odds ratio; CI – confidence interval; DRC – Democratic Republic of Congo; HEI – HIV exposed infant (HIV-infection not excluded); HEU – HIV exposed uninfected; HUU – HIV unexposed uninfected; IPD – invasive pneumococcal disease; RaR – rate ratio; RiR – risk ratio; RCT – randomized control trial; VTP – vertical transmission prevention7 Despite some inconsistencies, overall the existing evidence, largely from the pre-antiretroviral therapy era, suggests that HEU infants experience greater morbidity and mortality than HUU infants. In considering differing risks between these two groups of infants one can consider two groups of risk factors for infant morbidity and mortality (Figure 1.1). Firstly, there are circumstances that will increase the risk of infant infectious morbidity and mortality universally in infants regardless of HIV exposure. However, some of these circumstances may occur more frequently in HIV-exposed compared to HUU infants. These include being a preterm or small for gestational age newborn, being fed with replacement rather than breast milk in the first years of life, maternal mortality, TB exposure, and poverty. Secondly, there are risk factors unique to HIV-exposed infants that may increase risk for infectious morbidity and mortality. These include exposure to the in-utero environment altered by the presence of HIV products and maternal immune compromise as well as in-utero and postnatal exposure to ARV and other drugs. Figure 1.1 Schematic representation of HEU infant pathways to infectious morbidity Maternal(HIV( HEU(infant( Infec1ous(morbidity(Subop1mal(infant(feeding(Preterm(/(Small(for(gesta1on(Maternal(mortality(Poverty(Pathogen(exposure(Universal*infant*factors*HIV(product(exposure(Maternal(immune(compromise(HEU1unique*factors**An1retroviral(drug(exposure(8 The context of South African infant vulnerability is set next in section 1.2 before considering universal infant risk factors in section 1.3 and HEU unique risk factors in section 1.4. The evidence for HEU infant infectious morbidity is discussed in section 1.5. The pilot study that preceded the current study and the community in which the current study was set are described in sections 1.6. and 1.7 respectively. Finally, the specific aim, hypotheses and objectives employed are detailed in section 1.8. 1.2 South African context South Africa, the second largest economy in Africa with a population of 53 million, experiences high levels of poverty and inequality (52). Adult unemployment has been stagnant at around 25% for the last 10 years with unemployment and poverty concentrated in households with children (53). Forty two percent of South African children live in households where both parents are unemployed and 60% live in households falling below the lower bound poverty line, a monthly household income below 575 South African Rand (ZAR), approximately 50 Canadian Dollars (CAD) (53,54). Despite universal health care provided by the South African government to all pregnant women and children below 6 years of age, South Africa experienced an increase in the under-5 mortality rate (the number of child deaths before the fifth birthday per 1000 live births), peaking in 2005 at somewhere between 60 and 80 per 1000, depending on the modelling estimates used (55). This increasing mortality rate was largely due to escalating HIV-infection in pregnancy and inadequate prevention of vertical transmission of HIV (56). A watershed event occurred in 2002 when the South African Constitutional Court ordered the South African government and the National Department of Health to supply the simple VTP strategy of single dose nevirapine to all pregnant HIV-infected women. In the subsequent decade the increasing sophistication of VTP strategies and their expanding implementation in South Africa has seen HIV vertical transmission plummet from approximately 20% in the absence of interventions in 1999 to 3.5% at 6 weeks of age by 2010 (8). Alongside this, further child health gains have been realized through the addition of pneumococcal conjugate vaccine and rotavirus vaccine to the national immunization programme in 2009 (57). These vaccines target pneumonia and diarrhoeal disease respectively, two of the top five causes of infant mortality (58). Evaluations of vaccine effectiveness in South Africa show a 50% reduction in invasive pneumococcal disease in infants and 54% effectiveness for rotavirus vaccine in reducing hospitalization due to rotavirus diarrhoea (59,60). The South African under-5 mortality rate, determined by the Global Burden of Disease Study 2013, has now almost halved at 37/1000 (61). Even with these child health gains, the South African under-5 mortality rate is currently at least double that of countries with a similar gross national income per capita, e.g. Peru and Thailand, or of countries with similar health expenditure per capita, e.g. Brazil (52). Irrespective of HIV exposure, the vast majority of South African infants are still vulnerable to high rates of morbidity and mortality (58). 9 Approximately 1.1 million children are born in South Africa every year with over 300 000 being HIV exposed but uninfected (8,9,62). In a pilot prospective cohort study in Cape Town, South Africa, comparing HEU and HUU infants we saw an increased need for hospitalization due to infections in HEU infants (see section 1.6 for further detail) (45). Although an increased risk for morbidity and mortality in HEU infants was observed in neighbouring Zimbabwe and Botswana, these countries differ in the size of their populations, their economies and their access to healthcare (37–39,52). The observations in Zimbabwe and Botswana may not be directly generalizable to South African HEU infants. Considering the size of the South African HEU infant population, even a moderate additional increase in morbidity adds to the burden on the public healthcare sector. Direct comparison of HEU and HUU infants in South Africa to determine the contribution of HIV exposure to infant morbidity and mortality and to discern the pathways driving morbidity and mortality in HEU infants will facilitate determining appropriate interventions to reduce morbidity and mortality in HEU infants without compromising also vulnerable HUU infants. 1.3 Universal infant morbidity and mortality risk factors Regardless of HIV exposure, chances of dying during infancy are markedly increased by the following: 1) adverse birth outcomes, i.e. preterm birth, low birth weight and small for gestational age; 2) not being afforded the immunological and nutritional benefits of breastfeeding; 3) compromised growth and nutrition; 4) exposure to TB and other infectious diseases; 5) maternal ill health or death and 6) poverty. These universal risk factors for infant mortality and the evidence for whether they occur more frequently in HEU infants will be explored. 1.3.1 Preterm and small for gestational age newborns Almost 40% of all deaths in children under five years of age occur in the neonatal period (63,64). Preterm birth (born before 37 weeks completed gestation) is the most important single cause globally and in South Africa of neonatal mortality. In the Southern African region an infant’s risk of dying in the neonatal period is doubled if born preterm or small for gestational age (birth weight below the 10th centile for gestational age) and the mortality risk associated with preterm birth endures through the first year of life (65). Infants born to HIV-infected mothers compared to those born to HIV-uninfected mothers are more often preterm and small for gestational age (SGA) (19). Despite the controversy in the literature as to why, this observation is consistently seen across the world (19,20,66). A surveillance study of birth outcomes in 30 000 Botswanan women estimated that HIV-infected women have a 1.3 (95% CI 1.3,1.4) times greater odds of a preterm delivery and a 1.8 (95% CI 1.7,1.9) times greater odds of an SGA baby; however the study did not specifically differentiate the HIV-infected from HEU newborns (19). The study also observed a positive association between neonatal death and preterm birth or SGA compared to term or appropriate 10 for gestational age newborns. A smaller study in Cameroon showed HEU newborns were almost twice as likely as HUU newborns to be SGA (67). These poor birth outcomes do hold long term consequences for HEU infants. In Zambian and Tanzanian HEU breastfed infants, low birth weight (below 2500g at birth) was independently associated with infant mortality (10,68). In Kenyan HEU infants preterm birth, low birth weight (LBW) and SGA were associated with a six fold increase in neonatal mortality and a more than two fold increase in infant mortality (69). 1.3.2 Infant feeding Breastfeeding provides important protection against infectious disease mortality in all children (70). Lack of breastfeeding, at a minimum, doubles the risk of infant mortality in the first six months of life (64). Breastfeeding provides profound protection against diarrhoeal disease mortality but also against respiratory tract infections (70). Due to the risk of HIV transmission via breast milk many infants in less developed countries at high risk of infectious disease morbidity and mortality were denied the benefits of breastfeeding, either through complete replacement with infant formula milk or through an attenuated duration of breast milk exposure in attempts to lessen the risk of postnatal HIV transmission (24). The only randomized trial primarily comparing breastfeeding and replacement formula feeding in African HIV exposed infants was conducted in Nairobi, Kenya from 1992 to 1998 and included 212 infants randomized to breastfeeding and 213 infants randomized to formula feeding (71). After adjustment for infant HIV infection, the two-year mortality hazard was similar in breast and formula fed HIV exposed infants (aHR 1.1, 95% CI 0.7,1.7). A substantial amount of crossover occurred from formula to breastfeeding, although results by intent-to-treat and true feeding mode were similar. Nutritional status in the first six months tended to be better in the breastfed infants, yet incidence of diarrhoea and pneumonia did not differ by feeding mode (72). Despite these similar outcomes in a clinical trial setting, the benefits of breastfeeding and the extent to which it ameliorates infectious disease and nutritional risks in all infants cannot be denied. The Mashi study conducted in Botswana between 2001 and 2003 was a randomized trial comparing breastfeeding with infant ARV prophylaxis to formula feeding for the outcome of HIV-free survival (73). Despite a lower HIV-infection rate in the formula fed infants, mortality at 7 months was significantly higher in formula fed than breast fed HIV exposed infants with no survival benefit derived by 18 months of age. The Zambia Exclusive Breastfeeding Study (ZEBS), a randomized trial comparing early cessation of breastfeeding at 4 months to prolonged breastfeeding, included 749 HIV exposed infants followed to two years of age (74). Breastfeeding cessation at any age prior to 2 years was independently associated with a substantially elevated hazard for mortality and the absence of breastfeeding was associated with a shorter time to first hospitalization (10,75). Two comparisons of gastroenteritis related mortality in Uganda and Malawi before and after the World Health Organisation (WHO) guidance on early cessation of 11 breastfeeding, showed increased mortality in the later studies with shortened breastfeeding duration (76,77). This further enhanced the growing understanding of the detrimental effects of early weaning in African HIV exposed infants. Studies evaluating the safety of attenuated durations of breastfeeding in HEU infants have observed that the risk for infectious morbidity and mortality in HEU infants is highest in early infancy but increases again substantially in later infancy following weaning from breastfeeding (32,75). As these studies did not compare to HUU infants, it is possible that the weaning period is a vulnerable stage for all infants irrespective of HIV exposure. 1.3.3 Growth and nutrition Acute and chronic malnutrition, as evidenced by wasting (weight-for-length Z-score below -2 standard deviations from the mean WHO Child Growth Standards) and stunting (length-for-age Z-score below -2 standard deviations from the mean WHO Child Growth Standards) respectively, universally increase a child’s odds of overall mortality and specifically mortality due to pneumonia, diarrhoea and measles even after adjusting for non-nutritional determinants of infectious disease mortality (78,79). Acute and chronic malnutrition are prevalent across Africa with 2.7% (95% CI 1.0, 6.8%) of Southern African children under 5 years severely wasted and a remarkable 30.2% (95%CI 25.4, 35.6%) of children stunted. The prevalence of stunting reaches 50% (95% CI 42.3, 57.9) in Eastern Africa (79). Southern and Eastern African HEU infants, when compared to WHO or other growth standards, appear to experience marked growth failure following weaning in later infancy (80–82). However, there is little evidence for a difference in growth and nutritional outcomes when directly comparing African HEU infants to appropriate HUU control infants experiencing similar community and socioeconomic constraints to infant growth (83–88). Despite poorer birth growth outcomes in HIV exposed infants, in general, in-utero antiretroviral exposure for VTP has not yet been linked with long-term growth consequences for HEU infants (89,90). 1.3.4 Infectious pathogen exposure The nature of infectious diseases is that exposure to a pathogen is essential for infection, that in turn is essential for disease to develop. All exposed individuals do not necessarily become infected and all infected individuals will not necessarily develop disease. Using TB as an example, if never exposed to the M.tb organism one cannot become infected or develop symptomatic disease; on exposure some will become infected with M.tb , and of those infected, some will develop asymptomatic latent TB infection without disease and others will progress to symptomatic TB disease (91). It makes sense then that HIV-exposed children living with HIV-infected adults, who are themselves prone to higher rates of infectious disease, will experience greater exposure to infectious pathogens than HIV-unexposed children. There is evidence that HEU infants are more often exposed to M.tb and Streptococcus pneumoniae than HUU infants (22,92). 12 In a prospective cohort study of M.tb-exposed newborns in South Africa, 63% were also HIV exposed compared to the local antenatal HIV seroprevalence of approximately 19% (93). During screening for an Isoniazid Preventive Therapy (IPT) trial in South Africa, 10.1% (95% CI 8.0, 12.4) of HIV exposed infants were already exposed to M.tb in the home by 3 to 4 months of age (22). There are no data, to our knowledge, of M.tb exposure rates in South African HUU infants. A Kenyan study observed a prevalence of a positive TSPOT-TB test, indicating M.tb infection, in 10.9% (95% CI 6.1, 17.7%) of HEU infants at six months of age (94). Despite this high M.tb exposure prevalence, in HEU infants without a known TB contact IPT was not beneficial as primary prophylaxis for preventing TB disease or death (95). HEU infants have been observed to have higher rates of invasive pneumococcal disease (IPD) compared to HUU infants and part of their risk for IPD may be due to greater exposure to Streptococcus pneumoniae (34). Streptococcus pneumoniae is more often transmitted from children to their mothers than from mothers to children, this is so for both HIV exposed and HIV unexposed infants. However, HIV exposed infants in comparison to HIV unexposed infants experience increased exposure to Streptococcus pneumoniae from their HIV-infected mothers than HIV unexposed infants do from their HIV-uninfected mothers (92). Vaccinating HIV-infected women during pregnancy with pneumococcal conjugate vaccine was not successful though in reducing nasopharyngeal carriage in their unvaccinated HEU infants younger than 6 months of age (96). 1.3.5 Maternal wellbeing The connection between maternal wellbeing and infant health is well described. Southern African children of mothers recently deceased or terminally ill, irrespective of maternal HIV-status, have an almost 4 times greater hazard of death (aHR 3.9, 95%CI 2.8, 5.5) (48). HEU infants specifically have a 3.65 (95% CI 1.92, 6.95) times greater odds of infant mortality following maternal death than those with mothers still alive (27). ZEBS found an independent association between maternal mortality and infant and child mortality (10,75). In 2007, when there was little access to antiretroviral therapy, HIV-infected women in South Africa were 6 times (maternal mortality ratio 6.2, 95% CI 3.6, 11.1) more likely to experience pregnancy related mortality (both antenatal and post-partum) than HIV-uninfected women (97). 1.3.6 Poverty and social circumstances Globally, poverty is associated with poor child health outcomes. In South Africa, where free healthcare is provided to children under-6 years of age and to pregnant and lactating women, children born in the poorest quintile have an infant mortality rate four times that of those born in the wealthiest quintile (54). Families living in poverty are more likely to be affected by HIV. Further, HIV affected families are more likely to experience additional vulnerabilities, including loss of earnings, greater expenditure on the direct and hidden costs of healthcare, food insecurity, the responsibility of caring for children in the extended family or community orphaned by HIV and stigma in the community, which may alter health seeking 13 behaviour (26). This can lead to economic, physical and emotional instability for HIV exposed children, even if HIV-uninfected (25,98). In Zambian HIV exposed infants, food insecurity was independently associated with child mortality up to 2 years of age (75). Quantifying the extent to which poverty is associated with morbidity and mortality in HEU infants and children is challenging but this determinant of risk for infant mortality must be considered. There is good evidence that a number of universal risk factors for infant morbidity and mortality occur more often in HEU compared to HUU infants. HIV exposed infants are more often born preterm and SGA. HEU infants have less often benefited from optimal breastfeeding than HUU infants with evidence for the dire impact this has had on their health and survival. HEU infants more frequently experience infectious pathogen exposure, specifically M.tb and Streptococcus pneumoniae. Their mothers experience greater mortality and their families shoulder additional social and economic challenges. 1.4 Infant risk factors unique to HEU infants HEU infants are subjected to a number of additional unique exposures not experienced by HIV unexposed infants. This altered course commences in-utero, where maternal HIV disease as well as ARV drugs, received either for maternal therapy or for prophylaxis, result in a perturbed in-utero environment. Postnatally HEU infants themselves may receive ARVs as well as other prophylactic agents, such as cotrimoxazole, that have benefits but may also negatively influence the HEU infant’s course in the short and the long term. 1.4.1 In-utero environment altered by maternal HIV disease By definition HIV-exposed infants, born to HIV-infected mothers, experience a maternal in-utero environment that is different to that experienced by HIV-unexposed infants. Through the demonstration of HIV-specific immune responses in HEU newborns and infants it can be inferred that infants born to HIV-infected mothers are physically exposed to HIV in-utero through viral products able to cross the placental barrier but not competent to establish HIV infection (13,99–101). In addition to this physical viral product exposure, the in utero environment may differ immunologically due to the ramifications of HIV infection on the maternal immune system. HIV has its impact on the immune system through two broad mechanisms: firstly the well recognized immune suppression represented grossly by declining CD4 count; additionally through contrasting immune activation driven by the antigenic pressure of HIV viral particles (102–104). HIV is a disease with a long spectrum, from mild asymptomatic disease with initially subtle immune aberrations, progressing without treatment to a chronic debilitating severely immune compromised state in which the HIV-infected person is at risk for opportunistic infections as well as non-infectious complications (102,105). Treatment with combination antiretroviral therapy (cART), depending on when it 14 is commenced, will alter the progression along this spectrum to varying degrees (106). A pregnant HIV-infected woman may fall anywhere along the HIV disease spectrum, resulting in marked heterogeneity in the in-utero environment experienced by HIV exposed foetuses. Basic HIV disease and treatment dynamics are explained below before elaborating on changes to the maternal in-utero environment brought about by cART. 1.4.1.1 HIV disease dynamics HIV infection results in immune compromise by causing both immune suppression, primarily through infection of CD4 T-lymphocytes, and dysfunctional immune activation from continuous antigenic pressure due to the presence of circulating HIV (102,103,107,108). Three parameters, two biological markers and one clinical staging, that can be used to crudely quantify the extent of HIV disease include the CD4 count, the HIV viral load and the WHO Clinical Staging. CD4 T-lymphocytes are a component of the human immune system essential to control microbial infections; they become depleted during the progression of HIV disease (103). The CD4 lymphocyte count, although only one of a host of immunological parameters that are altered by HIV, is widely used as an indicator of the degree of immune suppression in an HIV-infected person. The normal CD4 T-lymphocyte count in HIV-uninfected African adults ranges from approximately 500 to 2000 cells/µl (109). Severe immune suppression is considered to be present once the CD4 count drops below 200 cells/µl and the risk for opportunistic infections increases (110). The HIV viral load measures HIV RNA copies in one milliliter of blood. High viral loads reflect more HIV replication and rapidity of disease progression. When HIV replication is below detectable limits on routine laboratory tests, this is referred to as viral suppression and indicates minimal HIV replication. The severity of HIV disease can also be staged according to the WHO Clinical Stage, where stage 1 represents asymptomatic HIV, stage 2 represents HIV disease with mild symptoms, stage 3 represents disease with advanced symptoms and stage 4 disease with severe symptoms (110). During acute HIV infection the viral load rises dramatically and the CD4 count drops slightly. With engagement of the still functioning adult immune system, HIV replication is brought under control, the viral load drops and the CD4 lymphocytes recover. The HIV-infected adult can remain in this state of equilibrium for an extended period with a low viral load and adequate but slowly declining CD4 count. Without antiretroviral therapy, ultimately HIV will escape control by the immune system and the virus will replicate with an increasing viral load and progressive immune suppression as evidenced by a rapidly declining CD4 count. Early in the epidemic it was recognized that not only T-lymphocytes, but also B-lymphocyte and humoral immune function is altered. This results in hyperactivity of B-lymphocytes and a non-specific increase in immunoglobulin’s (Ig) or hypergammaglobulinaemia, representing dysfunctional immune activation (107). Combination antiretroviral therapy alters this process by suppressing viral replication sufficiently to achieve an undetectable viral load, rapidly reducing the dysfunctional immune 15 system activation and with time allowing for CD4 lymphocyte and other altered immune functions to reconstitute. 1.4.1.2 Antiretroviral therapy versus antiretroviral prophylaxis It is important to distinguish the use of ARV drugs for therapy and for prevention of vertical HIV transmission. cART is defined as the use of at least three ARV drugs for treatment to suppress viral replication in established HIV infection. ARVs used for VTP prophylaxis in pregnant women reduce the risk of HIV infection in the infant but are not intended to treat maternal HIV disease. Mono or dual agent prophylactic regimens given to the pregnant mother, although effective in reducing vertical transmission of HIV, will not halt HIV replication and thus these women are likely to have ongoing viral replication and some dysfunctional immune activation. In contrast, mothers on cART with good adherence to treatment are likely to have suppressed viral replication and immune reconstitution provided there is no viral resistance. 1.4.1.3 Maternal HIV disease in the pre-antiretroviral therapy era Studies conducted prior to the availability of cART in Southern and Eastern Africa recognized that HEU infants born to mothers with advanced HIV disease, most often represented by a low CD4 count, had greater mortality than those born to mothers with less advanced disease (10,31,36,37). In Zambia and Tanzania, HEU infants of mothers with an antenatal CD4 count of below 350 cells/µl were twice as likely to die than infants of mothers with CD4 counts above 350 cells/µl (10,31). In ZEBS there was a significant dose-dependent reduction in infant mortality at four months of age associated with increasing maternal CD4 count as well as an association of the time to first hospitalization with antenatal maternal CD4 count (10). And although a CD4 count of below 200 cells/µl was associated with earlier cessation of breastfeeding, early breastfeeding cessation mediated only a small proportion of the effect of advanced maternal HIV on infant mortality (111). Therefore, there are pathways other than breastfeeding cessation contributing to infant mortality in severely immune suppressed mothers. In Tanzanian HEU infants, maternal HIV viral load of above 50 000 copies/ml at delivery was associated with a seven fold greater risk of death in HEU infants up to two years of age (31). In the ZVITAMBO study HEU infant all-cause and thrush-specific sick-clinic visit rates were inversely associated with maternal antenatal CD4 count and were increased for any maternal CD4 count below 800 cells/µl indicating a detrimental effect in infants of HIV-infected mothers with both advanced and minimal immune suppression compared to HUU infants (38). In contrast, a pooled analysis of African VTP studies that observed an association between maternal mortality and infant mortality did not find an association between infant mortality and maternal antenatal CD4 count of below 500 cells/µl (27). These studies preceded cART availability. Pregnant women received only prophylaxis for VTP irrespective of how severely immune suppressed they were. Thus, pregnant women in these studies with advanced HIV disease in the absence of cART would have had severe immune suppression as well as marked dysfunctional immune activation. While mothers with good 16 CD4 counts would not yet have had severe immune suppression and would have had less immune activation than mothers with more advanced disease. 1.4.1.4 Maternal HIV disease in the antiretroviral therapy era During the period 2009 to 2013 in South Africa, access to cART expanded particularly for pregnant women. Pregnant South African women were already receiving VTP prophylaxis, but between 2009 and 2013 they became eligible for cART for their own health with CD4 counts below 350 cells/µl or with WHO stage 3 or 4 disease (112). This is referred to as maternally-indicated cART and once commenced is continued lifelong. All pregnant women with a CD4 count of 350 cells/µl or greater with WHO stage 1 or 2 disease received a time-limited mono or dual agent prophylactic ARV regimen referred to as VTP prophylaxis. Currently, as of 2015, all HIV-infected pregnant and breastfeeding South African women, irrespective of CD4 count or WHO stage receive lifelong cART (113). The scenario considered throughout this dissertation is for 2009 to 2013 and is referred to as the maternally-indicated cART era. The maternally-indicated cART era, compared to the pre-cART era, changed the relationship between CD4 count and viral load during pregnancy, and thus also the relationship between immune suppression and immune activation. During the pre-antiretroviral therapy era an inverse relationship existed between maternal CD4 count and HIV viral load, the lower the CD4 count the higher the HIV viral load in the setting of advanced HIV disease. Thus mothers with advanced HIV, would have had both substantial immune suppression and dysfunctional immune activation. In the maternally-indicated cART era, pregnant women with the lowest CD4 counts and severe immune suppression received cART with subsequent reduced immune activation and resolving immune suppression. Pregnant women with CD4 counts above 350 cells per µl, not receiving cART, would have ongoing viral replication with dysfunctional immune activation in the absence of severe immune suppression. Although still an extremely heterogeneous group, this largely resulted in two groups of pregnant HIV-infected women: 1) on maternally indicated cART who either during or prior to pregnancy had advanced HIV qualifying for cART, but during pregnancy may have any CD4 count and likely an undetectable or low viral load; 2) on VTP prophylaxis with a CD4 count of 350 cells/µl or more, who had not yet reached the threshold for cART, with a detectable viral load during pregnancy. Although pregnant women on maternally indicated cART may no longer have a low CD4 count during pregnancy, they have all reached a stage of severe immune suppression, with a CD4 count of lower than 350 cells/µl, that rarely completely recovers despite an improved CD4 count (114,115). Thus pregnant women on cART whose CD4 count has risen to above 350 cells/µl are still immunologically different to pregnant women on VTP prophylaxis with equivalent CD4 counts above 350 cells/µl. Furthermore, pregnant women on cART with high CD4 counts differ from pregnant women on VTP because their HIV viral load is likely to be undetectable or 17 low. These immunological and virological maternal differences may result in differing in-utero environments for their infants. From the pre-antiretroviral therapy era studies it is difficult to distinguish between the effect of maternal immune suppression and maternal immune activation, and whether one or both of these maternal immunological aberrations were driving infant morbidity and mortality in mothers with advanced maternal HIV disease. The question in the maternally-indicated cART era is whether the relationship observed during the pre-antiretroviral therapy era of increased infant mortality with advanced maternal HIV, represented by both immune suppression and immune activation, remains. Alternatively, has cART given to pregnant women for their own HIV disease also improved the outcome of their infants through reduced HIV product exposure to the foetus, less immune activation and improved maternal immune function? 1.4.2 Immunological differences in HEU infants Differences in the innate, cell-mediated and humoral immune responses of HEU infants compared to HUU infants have all been observed. The immune system functions to protect the body against invasion from external pathogens and does this by responding in two broad ways. The first is through the innate immune response, that is germline encoded and present from birth, and acts immediately but non-specifically to kill offending organisms either by the release of inflammatory modulating cytokines primarily from monocytes, or through ingestion by professional phagocytes. The second is the adaptive or acquired response that acts less rapidly on first exposure than the innate response but more specifically by acquiring the capacity to recognize antigens. The adaptive immune response is able to establish immune memory that can be recalled at a later stage for an anamnestic, efficient and specific response to a diverse array of pathogens. The adaptive response further works through two mechanisms referred to as cell-mediated and humoral immunity. The cell-mediated response utilizes immune cells, most prominently T- and B-lymphocytes, and is primarily responsible for defense against intracellular pathogens such as HIV. The humoral response utilizes antibodies and complement for direct lysis of pathogens or as opsonins to enhance phagocytosis. Quantitative and functional differences in cells of the HEU innate immune system have been observed. HEU infants exhibit a pro-inflammatory response in early life compared to HUU infants that is characterized by a skewing of cytokine response to antigenic challenge (116). These differences appear to completely resolve by 12 months of age (116–118). Changes seen in cell-mediated immunity provide evidence for direct exposure of HEU infants to HIV. HIV-specific cell-mediated responses suggest that HEU infants, although HIV-uninfected, have interacted while in-utero with HIV viral products or even live replicating virus in some cases (101,119,120). These HIV-specific responses are not observed in HUU infants. Quantitative and functional changes observed in 18 the major lymphocyte populations may occur as a result of either direct HIV exposure or ARV drug exposure (121,122). The degree of HIV exposure as represented by the maternal HIV viral load may play a role in determining the magnitude of these changes (121,123). Unlike the changes observed in the innate immune system, cell-mediated immune changes following in-utero exposure to HIV may be longer lasting, possibly through to adolescence (101,124). Changes in HEU infant humoral immunity related to both specific and non-specific antibody responses have been described. Newborns and young infants do not have autologous antibody at birth and initially rely on passive humoral immunity provided by the transplacental transfer of maternal antibodies to the foetus. It is well established that this transfer is less efficient in HIV-infected women, particularly in those with high HIV viral loads, resulting in deficient passively acquired immunity in their infants during the first months of life (125–129). Even following maternal vaccination against influenza during pregnancy, HEU infants receive less transplacental maternal antibody to influenza than HUU infants and experienced a higher attack rate of confirmed influenza during the first six months of life (130). This opens a potential window of vulnerability for infectious morbidity in the early months of HEU infant life. HEU infants, however, seem to respond appropriately, even robustly, to vaccination, with equivalent levels of vaccine specific antibodies at four to six months of age when compared to HUU infants (128,131,132). Chronic immune activation in HIV-infected adults results in elevated levels of non-specific immunoglobulin (Ig) G antibodies that do not afford protection. Total IgG antibody levels in HEU infants are higher than in HUU infants up to 2 years of age and are positively correlated with maternal total IgG levels, suggesting that in-utero HIV exposure results in non-specific alterations to the HEU infant’s humoral immune response (133). Exposure to maternal antiretroviral therapy in-utero is associated with normalization of total IgG in the HEU infant, suggesting that suppression of maternal HIV viral load through effective antiretroviral therapy removes the antigenic stimulus for dysfunctional humoral immune activation in the HIV-infected mother and its consequences for the infant (133). Although it seems that the occurrence of these innate, cell-mediated and humoral immune differences are temporally related to the period of highest risk for infectious morbidity in HEU infants, such an association has not yet been directly tested and little is understood about the clinical significance of this altered immunological trajectory of HEU infants. 1.4.3 Antiretroviral exposure in HEU infants ARV drug exposure is an inevitable consequence of HIV exposure in regions of the world where access to ARV based vertical transmission preventive interventions are expanding. This ARV exposure may begin at conception or is introduced during pregnancy as soon as maternal HIV-infection has been confirmed (14). ARV exposure then continues postnatally by administration of prophylactic ARV drugs to the infants themselves or in breastfeeding infants exposed via breast milk to maternally administered 19 drugs (134). Although there are documented consequences of exposure to some ARV drugs for HIV exposed infants, generally their use is considered reasonably safe (135,136). In terms of in-utero and postnatal ARV exposure and whether this may be related to an increased risk for infectious morbidity during infancy, the important drugs to consider in the South African programme are the nucleoside/nucleotide reverse transcriptase inhibitors zidovudine (ZDV), tenofovir (TDF), lamivudine (3TC) and emtricitabine (FTC), the non-nucleoside reverse transcriptase inhibitors efavirenz (EFV) and nevirapine (NVP) and the protease inhibitor lopinavir/ritonavir (LPV/r). ZDV causes anaemia in neonates after in-utero or postpartum exposure (137,138). This is, however, transient and corrects after withdrawal of ZDV. Other haematological aberrations associated with ZDV exposure, include altered neutrophil and platelet counts (137). In the general population, chronic anaemia in childhood is associated with increased infectious morbidity and mortality (139). However, the clinical significance of these haematological alterations in HEU infants and whether they may be associated with an increased infectious morbidity risk has not been determined. TDF has nephrotoxic effects that, through altered bone metabolism, can result in reduced bone mineral density when used for treatment in HIV-infected children (140–143). The association between stunting and increased infectious morbidity risk makes potential TDF associated bone toxicity an important consideration in African HEU infants (79). Although there is concern about in-utero exposure to TDF and its effects on the developing foetal skeletal structure, in two large US based cohorts there was no evidence of an association between in-utero TDF exposure and foetal growth (144,145). These studies did however identify a lower length-for-age Z-score in TDF-exposed infants at one year of age and a lower weight-for-age Z-score at six months of age. The importance of these observations was uncertain but highlighted the need for further surveillance of TDF-exposed infants. In Ugandan and Zimbabwean infants exposed to TDF in-utero there was no evidence of poor bone health or inferior growth at 2 years of age compared to HIV exposed infants not exposed to TDF (146). Pharmacovigilance programmes, including the Antiretroviral Pregnancy Registry in North America, have not identified short-term safety concerns with 3TC or FTC in-utero exposure (136,147). Unrelated to infectious morbidity is the possibility of increased HEU infant morbidity related to mitochondrial toxicity secondary to NRTI exposure (148–152). The clinical significance of these mitochondrial changes is unclear but has potential for poorer neurodevelopmental outcomes, myocardial aberrations and oncogenic effects, warranting further pharmacovigilance (153–155). There has been concern regarding EFV use during the first trimester of pregnancy and risk for neural tube defects following animal studies and case-reports of such an association (147,156). A comprehensive systematic review and meta-analysis found no evidence of birth defects associated with first trimester EFV use, although the numbers were too small to comment specifically on risk related to neural-tube defects (157). There is no reason to believe that EFV would be associated with an increased infectious morbidity risk. Prolonged postnatal NVP exposure up to 6 months of age in breastfed infants is 20 safe when used at prophylactic doses with low rates of NVP associated rash or hepatotoxicity, known side-effects of NVP when used at treatment doses in children and adults (158). There are also no specific concerns related to mechanisms for increased infectious morbidity. There is controversy in the literature as to whether in-utero ARV exposure, specifically protease inhibitors (PI) and initiation of cART during pregnancy, are additional determinants of adverse birth outcomes, including PTB, SGA and LBW, in HEU infants. Increased risk of PTB with PI-based cART in pregnancy has been observed in Africa and Europe, but not in the United States (20,66,159). ZDV-related anaemia and TDF-related long bone growth deficits may potentially be pathways to infectious morbidity in HEU infants, but there is no definitive evidence in this regard. ARV exposure does not seem to be directly associated with infectious morbidity risk in HEU infants. 1.4.4 Cotrimoxazole preventive therapy Cotrimoxazole preventive therapy (CPT) is recommended by the World Health Organisation (WHO) for all HIV exposed infants from six weeks of age until HIV infection is excluded and all potential HIV exposure through breastfeeding has ceased (14). Cotrimoxazole is a broad-spectrum antibiotic that effectively reduces the risk of death due to Pneumocystis jirovecii pneumonia (PCP) in HIV-infected infants (160). Pneumocystis jirovecii is an opportunistic fungal infection causing PCP in immune compromised hosts and is responsible for substantial mortality in HIV-infected infants (161). Early infant diagnosis of HIV can be challenging in much of Africa and the rationale behind the WHO recommendation is to prevent infant mortality due to PCP in undiagnosed HIV-infected infants. Once HIV-exposure has ceased (i.e. via breastfeeding) and HIV-infection has been excluded, CPT should be discontinued (14). Although cases of PCP have been described in HEU infants, these are few considering the growing number of HEU infants that are exposed to prolonged CPT and the dwindling number of HIV infected infants at risk for PCP (40,162,163). The benefit of continuing CPT in HEU infants is unclear and there are potential harms to consider. In the Breastfeeding Antiretroviral Nutrition study conducted in Malawi, CPT reduced the risk of malaria in the short-term in HEU infants, but this effect seemed to wane with time (164). CPT had no beneficial effect in reducing severe illness or death, low weight for age or anaemia. Two South African studies describe a small potential benefit in reduction of lower respiratory tract infections in HEU infants receiving CPT, however this was associated with an increased risk of diarrhoea in breastfed HEU infants in one study and after adjusting for ever or never breastfeeding in the second (165,166). A third South African study observed that CPT was dosed excessively by caregivers almost 50% of the time and not administered on week-ends in almost 50% of HIV exposed infants, posing the risk of either potential toxicity or suboptimal dosing respectively (167). Whether prolonged CPT in HEU infants is of any benefit to their health and survival is currently under investigation in at least one randomized clinical trial (clinicaltrials.gov identifier NCT01229761). 21 1.5 HEU infant infectious morbidity The causes of morbidity and mortality in South African HEU infants are no different to those in HUU infants, but may occur more often or with greater severity. In South Africa, as for most of Sub-Saharan Africa, substantial infant morbidity and mortality results from neonatal disorders including complications of preterm birth and neonatal sepsis, that can account for up to 34% of mortality (63,168). Pneumonia and diarrhoeal disease each account for up to 15% of infant mortality (168). Malaria, although not prevalent in South Africa, is an important additional cause often in the top 5 causes of infant mortality in many African countries and malnutrition is all too often an underlying risk factor for these infectious causes of mortality (64,79). 1.5.1 Respiratory tract infections HEU infants experience substantial morbidity from lower respiratory tract infections (LRTI), often the most common reason for hospitalization (50,75,169). But whether this respiratory morbidity is greater than that of HUU infants in the same circumstances is not clearly understood due to a lack of studies with direct comparisons between HEU and HUU infants. A large perinatal cohort study in South America and the Caribbean observed that 9% of HEU infants (49/547) were hospitalized at least once for an LRTI in the first six months of life, 81% being due to viral bronchiolitis (169). In the ZEBS study, 31 of 39 HEU infant hospitalizations up to 4 months of age, were due to pneumonia/sepsis (10). In South African and Botswanan studies HEU infants more often than HUU infants failed empiric treatment for pneumonia 48 hours after initiation (aOR (adjusted odds ratio) 6.0, 95% CI 1.5, 23.4 and risk ratio (RiR) 1.83, 95% CI 1.27, 2.64, respectively) (35,40). As expected in all children, risk of acute respiratory infections was elevated in HEU infants with acute or chronic malnutrition and breastfeeding was protective against acute respiratory infections in the first 12 months of life (170). 1.5.2 Diarrhoea One of the earliest observations of increased infectious morbidity in HEU infants was from a prospective cohort study in Kinshasa, Democratic Republic of Congo between 1989 and 1992, in which 139 HEU and 199 HUU infants were observed to 24 months of age (28). HEU infants had almost double the incidence of persistent diarrhoea (4.9 vs. 2.7 cases / 100 child-years) and the hazard of persistent diarrhoea was increased further in HEU infants of mothers who died (HR 10.4, p<0.005). The ZEBS study documented a hospitalization rate for diarrhoea in the first 24 months of HEU infant life of 8 per 100-child years and a diarrhoea-related mortality of 4 per 100-child years (75). An increased risk of diarrhoea was associated with early weaning between 4 and 6 months of age compared to prolonged breastfeeding (RR 1.8, 95% CI 1.3, 2.4). The risk of diarrhoeal disease is intricately tied to feeding practices (see 1.3.2). In a non-22 randomized intervention cohort designed to determine the effect of exclusive breastfeeding on HIV transmission and to determine effectiveness of peer support on rates of exclusive breastfeeding, there was no difference between HEU and HUU infants in rates of acute, persistent or any diarrhoea (41). In all infants exclusive breastfeeding in the first 6 months of life was associated with fewer diarrhoeal days than mixed or no breastfeeding. 1.5.3 Bacterial sepsis Bacterial sepsis refers to a systemic response to a bacterial infection. Irrespective of HIV exposure, the risk for bacterial sepsis is highest during the neonatal period (the first 28 days of life) (171). The immature neonatal immune system is often unable to contain bacterial infections (172). As the immune system undergoes maturation during infancy and childhood it gains the ability to respond appropriately to bacterial infections and the risk of bacterial sepsis lessens. Not all bacteria require the same host immune response and so defects in specific arms of the immune system may increase susceptibility to some bacterial infections but not to others. There is evidence that HEU neonates and infants are at increased risk for some specific bacterial infections, namely Group B Streptococcus during the neonatal period and Streptococcus pneumoniae (pneumococcus) during infancy (34,173,174). 1.5.3.1 Bacterial sepsis in neonates Although risk for all-cause early (0 to 7 days of life) or late (8 to 28 days of life) onset neonatal sepsis is not increased in HEU compared to HUU neonates, HEU infants are at increased risk specifically for Group B Streptococcus (GBS) sepsis in the neonatal and early infant periods (42,173,174). In a large cohort of South African neonates, including 1255 HEU and 5804 HUU neonates enrolled in an RCT of chlorhexidine vaginal washes to reduce early onset neonatal sepsis, there was no difference between HEU and HUU neonates in early or late onset neonatal sepsis (42). This persisted in a second analysis using propensity score matching to deal with differences in important variables between HIV-infected and HIV-uninfected mothers. This study looked at all causes of neonatal sepsis and included presumed (based on clinical presentation) and confirmed (based on isolation of bacteria from blood or other sterile site) neonatal bacterial sepsis. GBS infection though, is significantly increased in Belgian and South African neonates. A Belgian retrospective cohort study looking specifically at confirmed GBS neonatal sepsis observed an almost 20 times (RiR 19.7, 95% CI 7.5, 51.7) greater risk of Group B Streptococcus infection in HEU compared to HUU neonates and infants between 2001 and 2008 (173). This increased risk was also associated with more severe manifestations of disease including septic shock, meningitis and late-onset disease. In Soweto, South Africa, HIV-exposed infants had a 70% greater risk of early-onset GBS disease (RaR 1.69, 95% CI 1.28, 2.24) and three times the risk of late-onset GBS disease (RaR 3.18, 95% CI 2.34, 4.36) (174). 23 HEU neonates may have elevated risk for necrotizing enterocolitis (NEC). NEC is an acute intestinal necrosis syndrome of unknown aetiology, likely resulting from mucosal injury due to bacterial infection or other pro-inflammatory insults in the setting of poor host protective mechanisms after injury, occurring most frequently, but not exclusively, in preterm neonates (175). NEC occurs more often in HIV exposed than HIV unexposed neonates and may occur with more severe manifestations in HIV exposed neonates. A French case-control study observed a six times greater odds (aOR 6.6, 95% CI 1.26, 34.8) of NEC in HIV exposed compared to HIV unexposed gestational age matched neonates (176). One South African retrospective cohort study observed a greater odds of mortality in HIV exposed neonates with NEC requiring surgery compared to HIV unexposed neonates requiring surgery (OR 4.8, 95% CI 1.7,14.2) (177). A second South African retrospective cohort however observed no difference in outcomes from Stage 3 (advanced) NEC in HIV exposed and HIV unexposed neonates (178). These findings do not seem confounded by the increased risk of preterm birth in HIV exposed neonates as gestational age was similar in the two groups in both South African studies. 1.5.3.2 Bacterial sepsis in infancy During the first six months of life HEU South African infants are at greater risk for invasive pneumococcal disease that includes pneumonia, bacterial septicaemia and meningitis. An incidence rate ratio for invasive pneumococcal disease of 3.6 (95% CI 2.8, 4.6) in HEU relative to HUU infants was recently observed in a nationwide invasive pneumococcal disease laboratory surveillance study (34). The same study found HEU infants were at greater risk for mortality due to invasive pneumococcal disease (aOR 1.8, 95% CI 1.1, 2.9) independent of infant pneumococcal vaccination. The risk for all-cause bacterial sepsis has not been evaluated in South African HEU infants outside of the neonatal period. However evidence from the French Perinatal Cohort, that observed 7638 HEU infants between 2002 and 2010, showed an association between serious bacterial infections in the first year of life and maternal CD4 count near delivery (179). The aHR for a serious bacterial infection were 1.7 (95% CI 1.2, 2.6) and 1.2 (95% CI 0.8, 1.9) for infants of mothers with CD4 count below 350 cells/µl and 350-500 cells/µl respectively compared to above 500 cells/µl. Maternal antenatal CD4 count was not associated with serious viral infections. 1.5.4 Congenital viral infections HIV exposed infants are more often congenitally infected with viral infections than HIV unexposed infants, most notably CMV, but also Human Herpes Virus 6 (180,181). The odds of congenital CMV infection in HIV exposed infants born to mothers with CD4 counts below 200 cells/µl is double that of HUU infants (180,181). A retrospective cohort in the United States from 1997 to 2005 documented a 3% prevalence of congenital CMV in HEU infants associated with a lower mean birth weight, lower gestational age and higher maternal HIV viral load (17). A South African cross-sectional study has recently documented a birth prevalence of congenital CMV in HEU infants of 2.9% (95% CI 1.9, 4.4%), that was also associated 24 with a maternal CD4 count of below 200 cells/µl (aOR 2.9, 95% CI 1.2,7.3) (18). A higher breast milk CMV viral load was associated with a lower weight-for-age and length-for-age Z-score in 6 month old HEU infants (182). Valacyclovir given to pregnant HIV-infected women from 34 weeks gestation to 12 months post-partum was not effective in reducing the acquisition of CMV by 12 months of age or the time to CMV acquisition in the HEU infants, compared to placebo (183). The French Perinatal Cohort though observed a reduction in prevalence of congenital CMV over time as cART was introduced, decreasing from 3.5% in 1997-98 to 1.2% in 2003-04 (181). Although HEU infants experience a substantial burden of respiratory tract infections, there is no evidence yet that the incidence is different to that of HUU infants in the same context. Risk for diarrhoea in HEU infants is associated with weaning from breast milk or no breastfeeding. There is no evidence that HEU infants experience greater all-cause neonatal sepsis, but may be at greater risk than HUU infants specifically for Group B Streptococcal sepsis, NEC and congenital CMV during the neonatal period. There is good evidence that HEU infants are at greater risk for invasive pneumococcal disease compared to HUU infants, but risk for all-cause post-neonatal sepsis has not been determined in comparison to HUU infants. 1.6 Cape Town HEU pilot study Between 2005 and 2008 we documented a case series of severe infections in 8 HEU infants in Cape Town (163). Cases included PCP, CMV colitis, haemorrhagic varicella, community acquired Pseudomonas septicaemia and Group A Streptococcal meningitis and endocarditis. Some children had subtle immunological defects such as isolated B cell or CD4+ T cell lymphopenia. Following this observation, a hypothesis-generating pilot study focusing on exploration of innate and adaptive immune system differences between HEU and HUU infants was conducted. This study enrolled mothers and their newborns from the postnatal ward of Tygerberg Academic Hospital during March and April 2009 with follow-up for 24 months. At this time the South African VTP programme consisted of maternal cART for mothers with a CD4 count below 200 cells/µl, maternal short-course ZDV with single-dose NVP at delivery for mothers with a CD4 count of 200 cells/µl or greater and short-course ZDV for all infants. Infant formula was provided for 6 months by the government VTP programme for mothers choosing not to breastfeed. cART, although provided by the government health system, was not yet widely available and monitoring of maternal CD4 counts during pregnancy was not yet well implemented. Infants in this pilot study did not benefit from the addition of rotavirus and pneumococcal conjugate vaccines to the routine infant immunization schedule, that occurred later in 2009. 25 Twenty-seven HEU and 28 HUU infants were enrolled. The two groups differed in terms of race composition and maternal social habits such as smoking and alcohol use during pregnancy. All but one HEU infant was formula fed from birth and all HUU infants received some breastfeeding, the median duration being 12 weeks. Observations related to the immune function of HEU infants confirmed previous observations of lower levels of vaccine-associated maternally derived antibodies compared to HUU infants but subsequent excellent quantitative response to vaccination (131). A novel observation was made related to the innate immune system that HEU infants compared to HUU infants display a heightened pro-inflammatory response in the first months of life that normalizes by 12 months (116). It was postulated that this early, exaggerated immune response might partially explain some of the observed HEU vulnerability to infections. Although the pilot study was not designed to evaluate the rate of infectious morbidity we observed that HEU infants had a 2.7 (95% CI 0.8, 8.8) times greater risk for hospitalization due to an infectious event in the first 12 months of life despite appropriate immunization, good nutritional status and mothers generally without advanced HIV disease (45). This supplementary observation of a possible elevation in infectious morbidity in HEU infants, particularly an increased risk for severe infections requiring hospitalization, and the burgeoning of this population of infants in South Africa provided the impetus for the current study. The substantial differences between the pilot study HEU and HUU infants in race, maternal social habits and infant feeding were taken into consideration in the design of the current study to reduce imbalance in these potential confounders. 1.7 Community context Kraaifontein, the community in which the current study was set, is situated within the Northern sub-district of the City of Cape Town, approximately 30 kilometers east of the downtown center of Cape Town, the largest city in the Western Cape Province of South Africa. The City of Cape Town has a population of approximately 3.7 million people with extreme income inequality and a Gini co-efficient of 0.6. The unemployment rate in the City is 23.9%, disproportionately concentrated amongst young black residents (184). The unemployment rate around Kraaifontein is greater than 30% with less than a 50% high school graduation rate and some of the highest rates of poverty in the city. Seventy-eight percent of the residents of the City of Cape Town live in formal (brick built) housing, 87% have access to water piped into their dwelling or yard as well as electricity for cooking and 88% use a flush toilet connected to sewerage (184). Equivalent figures for the Northern sub-district or specifically the community of Kraaifontein are unpublished, but likely differ substantially from the City as a whole. A large proportion of the Kraaifontein community reside in areas that transitioned from informal settlements, only recently receiving basic municipal services and infrastructure. Informal houses built of hardboard, corrugated iron 26 and plastic sheets are the norm, and three to four households generally share a single flush toilet and outdoor tap with piped water situated in a courtyard between the households. Although households may have access to electricity via a “pay-as-you-go” system, financial resources for purchasing of electricity may not be reliably available. During this study 89.5% of children under 1 year of age in the Northern sub-district were fully immunized, similar to 91.6% in the City of Cape Town. However, the Northern sub-district experienced a substantially higher rate of severe acute malnutrition compared to the average for Cape Town, 5.0 / 1000 compared to 2.8 / 1000 children under 5 years of age respectively (184). The antenatal HIV prevalence of the Northern sub-district is equivalent to the City at 18% and substantially lower than the South African national antenatal HIV prevalence of 29% (9,185). Pregnant women in Kraaifontein receive free care, according to South African National policy, at the Kraaifontein Midwife Obstetric Unit (MOU), a public primary level community based obstetric service. The MOU is managed and staffed by registered nurse midwives and deals with low-risk deliveries of uncomplicated pregnancies. There are no medical doctors on site, and specialist obstetricians at the secondary level hospital, approximately 10 kilometers away, provide support to the MOU staff. Complicated cases, including all patients in preterm labour, antepartum haemorrhage, severe hypertensive diseases of pregnancy and complicated labour, are referred to the secondary hospital through reliable ambulance transport. Free healthcare is provided for all children younger than six years of age as well as all HIV-infected people. Referral from a primary healthcare practitioner is required to be seen at the secondary level hospital emergency department. During this study the South African National Prevention of Mother to Child Transmission (PMTCT) programme provided free vertical transmission preventive care at all public clinics in the country, including Kraaifontein MOU. In the Western Cape Province in 2010, 98% of pregnant women receiving antenatal care were tested for HIV and 94% of those testing HIV-positive received cART or ARV prophylaxis (8). During the study the national PMTCT programme provided the following maternal and infant interventions (112): A. Maternal interventions a. cART (FTC or 3TC plus TDF plus EFV) to all pregnant women with a CD4 count below 350 cells/µl or WHO stage 3/4 disease b. Antiretroviral prophylaxis to all pregnant women not already on cART with a CD4 count of 350 cells/µl or more, including ZDV from 14 weeks gestation plus single dose NVP, FTC and TDF in labour and 3 hourly ZDV in labour c. Repeat CD4 count at 6 weeks post-partum and 6-monthly thereafter B. Infant interventions 27 a. NVP commenced within 72 hours of birth and continued until 7 days after cessation of all breastfeeding or until 6 weeks of age if not breastfed b. Infant HIV-PCR test for diagnosis or exclusion of HIV-infection at 6 weeks of age and again 6 weeks after cessation of all breastfeeding c. CPT commenced at 6 weeks of age and continued until no longer breastfeeding and confirmed to be HIV-uninfected d. Maternal infant feeding counselling and support in the mother’s own choice to exclusively breastfeed or replace breast milk with infant formula milk. The public programme provided free infant formula for 6 months to mothers choosing to formula feed their infants. 1.8 Study aims, hypotheses and objectives HEU infants experience the same universal risk factors for infectious morbidity as HUU infants, but there is evidence that at least some of these risk factors occur more often in HEU than HUU infants, specifically poor birth outcomes, avoidance of breastfeeding, maternal mortality and poverty. In addition HEU infants experience unique exposures to HIV particles and antiretroviral drugs that may set them on an altered immunological trajectory than HUU infants. The infectious morbidity in HEU infants is largely due to common childhood infections, pneumonia and diarrhoea, but may also include TB, bacterial sepsis and congenital CMV. HEU infants comprise 30% of the South African infant population and deserve dedicated study to inform appropriate child health strategies in South Africa. 1.8.1 Study aim This study aimed to establish whether there was a difference in the pattern and pathways of infectious morbidity between HEU and HUU infants from a single community in Cape Town in similar socioeconomic circumstances and after accounting for differences in breastfeeding exposure. It also aimed to determine whether, within HEU infants, the association observed in the pre-cART era between advanced maternal HIV disease and increased infant infectious morbidity and mortality still existed in the maternally-indicated cART era. 1.8.2 Primary hypothesis and objective Primary hypothesis: HEU infants have a higher probability of all-severity infectious morbidity in the first six months of life compared to HUU infants. 28 Primary objective: to determine whether HEU infants experienced a higher probability of all-severity infectious cause hospitalizations or death compared to HUU infants in the first six months of life in infants from a single community after adjusting for differences in breastfeeding exposure. 1.8.3 Secondary hypothesis 1 and secondary objective 1 Secondary hypothesis 1: HEU infants experience a greater severity of infectious morbidity in the first six months of life compared to HUU infants. Secondary objective 1: to determine whether HEU infants have a higher probability of severe or very severe infectious cause hospitalizations or death compared to HUU infants in the first six months of life in infants from a single community after adjusting for differences in breastfeeding exposure. 1.8.4 Secondary hypothesis 2 and secondary objective 2 Secondary hypothesis 2: HEU infants born to mothers on maternally indicated cART have a higher probability of infectious morbidity compared to HEU infants born to mothers on VTP prophylaxis. Secondary objective 2: to determine whether HEU infants born to mothers on maternally indicated cART have a higher probability of infectious cause hospitalization or death compared to HEU infants born to mothers on VTP prophylaxis. See Chapter 5.2 for definition of the major determinants and outcomes as given in the hypotheses and objectives. Before describing the main study methods in Chapter 4, Chapter 2 follows with a discussion of the rationale for a study specific tool for outcome classification and a description of the design process and pilot testing of the outcome classification tool, the Paediatric Infectious Event Tool for Research (PIET-R). Chapter 3 details the methods and results of the formal reliability and validity evaluation of the PIET-R conducted on a separate sample of infant hospitalization events. 29 Chapter 2 The Paediatric Infectious Event Tool for Research (PIET-R) HEU infants and children may have a more complicated course and be considered high-risk for various morbidities and mortality. Therefore clinicians may have a lower threshold for hospitalization of HEU infants than HIV unexposed infants (186). This potential hospitalization bias in conjunction with the desire to evaluate a difference in severity of infectious morbidity prompted the design of a study specific infectious event classification and grading system, the Paediatric Infectious Event Tool for Research (PIET-R). The rationale for the PIET-R is detailed in section 2.1, its design and pilot testing are described in section 2.3 and the hypotheses and objectives of the validity and reliability evaluation are given in section 2.4. The methods, results and discussion of the PIET-R evaluation study are presented in Chapter 3. 2.1 Rationale for the development of the PIET-R Currently available outcome classification schemes used in infectious disease research rely heavily on special investigations and specialist opinion (187,188). The definitive diagnosis of an infectious disease relies on isolating and identifying the causative organism through a variety of laboratory based investigations not readily available outside of research settings in low-middle income countries. The ability to classify and grade clinical infectious events in a standardized manner using only signs and symptoms determined by history and physical examination conducted by non-specialist health care workers may facilitate outcome determination in accordance with good epidemiologic principles while efficiently using available research resources. In considering appropriate case-definitions and severity grading schemes to use, existing definitions used in clinical trial research, infectious disease surveillance, and international and local child health management were reviewed for applicability to address our research objectives, as discussed below. Definitions and grading systems were sought for the infectious disease syndromes causing the greatest morbidity in South African infants, these being LRTIs (including pneumonia, bronchiolitis and TB) and infectious diarrhoeal disease (58). 2.1.1 Definitions from clinical trial research The terminology “adverse event” (AE), although designed for clinical trials of pharmaceutical products is widely used in observational studies unrelated to the administration of pharmaceutical products and therefore we looked at case-definitions from clinical trial research (189). Two international efforts at standardizing adverse event reporting relevant to paediatric infectious disease clinical trials are the National Institutes of Health Division of AIDS Table for Grading the Severity of Adult and Paediatric 30 Adverse Events (DAIDS AE Grading Table) and the Brighton Collaboration case definitions for adverse events following immunization (AEFIs) (187,188). The DAIDS AE Grading Table is a descriptive terminology used in HIV research and grades AEs on a scale of one (mild) to five (death) (187). The Table is also extensively used to grade infectious and other morbidity in secondary observational studies from which much of the evidence related to infectious morbidity in African HEU infants has been generated (32,38,76,77,190). The DAIDS AE Grading has a single “infection” category for all infections other than HIV-infection and grades events according to need for local or systemic antibiotic therapy and/or impaired ability to perform usual social and functional activities. Most infections resulting in hospitalization would be associated with use of systemic antibiotics or impair ability to perform usual social and functional activities graded as ‘3’. A life-threatening event or death are graded as ‘4’ or ‘5’ respectively. The DAIDS AE Grading Table does not account for differences in severity of non-life threatening grade ‘3’ events that require hospitalization. The non-specific single category of “Infection”, the subjectivity of determining the presence or absence of usual social and functional activities in infants, and the poor discriminatory power of the grading for events resulting in hospitalization made the DAIDS AE Grading Table inappropriate for measuring this study’s primary outcome. The Brighton Collaboration developed standardized case definitions for reporting of AEFIs to enhance comparability across studies and facilitate vaccine safety determination (188). These definitions are based on expert working group consensus following extensive review of the literature and are being formally tested and evaluated. The Brighton Collaboration has taken cognizance that definitions are required for differing geographic settings with varying resources and study questions, and have included in the classifications three different “levels of diagnostic certainty” depending on the complexity of information available to make the diagnosis. Due to the standardized and rigorous process conducted in developing the Brighton Collaboration case definitions, and their relevance to low resourced settings they were reviewed for appropriateness for the main study’s outcome classification. Of the 30 currently published Brighton Collaboration AEFI definitions, only diarrhoea overlapped with the conditions causing hospitalization in our cohort (191). The Brighton case-definition of diarrhoea is based on the World Health Organization (WHO) definition and classification of diarrhoea in children that was ultimately used as the PIET-R case-definition and is discussed further in section 2.1.3 below (192). 2.1.2 Definitions from infectious disease surveillance Infectious disease surveillance, utilizing both clinical as well as laboratory case-definitions, most often focuses on surveillance of organism-specific infectious diseases and seldom on infectious disease syndromes. The South African National Department of Health and the United States Centers for Disease Control and Prevention (CDC) provide case-definitions for notifiable communicable diseases (193). Of 31 potential use for our study were the case-definitions for surveillance of TB that have both a clinical and laboratory based case-definition. The locally used definition devised for clinical management, as opposed to surveillance, of childhood TB in the Southern African region as discussed in 2.1.3 was considered more appropriate (23). The clinical management definitions provide clearer and more specific clinical criteria on which to make a clinical diagnosis of TB than the South African National Department of Health surveillance or CDC clinical case-definition. 2.1.3 Definitions from child health management Algorithms to identify common childhood diseases have been developed to assist a spectrum of health care workers in diagnosing and treating common childhood diseases. The WHO Integrated Management of Childhood Illnesses (IMCI) uses simple clinical case definitions for diarrhoea and pneumonia, designed with high sensitivity to assist in identification and management of these conditions by primary health care workers (194). IMCI assigns degrees of severity to diarrhoea and pneumonia with “severe” disease warranting immediate referral to hospital. The use of IMCI became a national strategy in South Africa and is widely used across the country (195). Because HEU infants have been observed in some settings to experience a high burden of bronchiolitis, a LRTI of viral aetiology, but in other settings been observed to be at greater risk for severe bacterial but not viral infections, we wanted to be able to distinguish risk for probably viral as opposed to probably bacterial LRTIs (34,169,179). Unlike the WHO IMCI definition, the South African Thoracic Society (SATS) guidelines differentiate between community acquired pneumonia and bronchiolitis (196,197). Thus the WHO IMCI definition for pneumonia was augmented with a separate definition of bronchiolitis based on the SATS guidelines. The WHO IMCI, The International Union of Tuberculosis and Lung Disease and the South African Society for Paediatric Infectious Diseases (SASPID) all use a common definition to clinically diagnose TB (23,194,198). For our primary outcome determination where organism-specific infectious aetiology was impractical and considered unnecessary for determining a difference in overall infectious morbidity, IMCI and the South African guidelines were considered most relevant and adapted for our outcome determination. 2.2 PIET-R case-definitions The case-definitions for LRTI and diarrhoea were prioritized, focussing on syndromic definitions rather than laboratory or expert specialist diagnosis. LRTI encompassed pneumonia, bronchiolitis and TB as defined in the SATS and SASPID guidelines based on WHO definitions (23,196–198). Diarrhoea included acute and persistent diarrhoea as defined by WHO IMCI (194). The PIET-R case-definitions for lower respiratory tract infections and diarrhoea are described in Table 2.1 and completed based on information from a history and physical examination only. 32 The case-definitions for “severe” events were based on criteria for hospital admission or referral in local guidelines and WHO IMCI respectively. An additional grading of “very severe” was defined as an event meeting criteria for a “severe” event, with persistence of the severe criteria for at least 48 hours following admission. When available simple investigations such as pulse oximetry for measurement of oxygen saturation, chest radiography and M.tb microbiology can also be used to meet the case-definitions, however these investigations are not required. Although less common, important additional conditions included in the PIET-R case-definitions were dysentery, viral exanthemas including measles and varicella zoster, bacterial skin infections, meningitis, invasive bacterial infections and congenital infections. Where available WHO case-definitions were used for these additional case-definitions. However some of the less common conditions, specifically invasive bacterial infections such as septic arthritis or congenital infections do rely on laboratory etiologic or specialist diagnosis. The comprehensive list of PIET-R case-definitions are in Table A.1, Appendix A. 33 Table 2.1 PIET-R case definitions for lower respiratory tract infections and diarrhoeal disease Classification Definition of case Definition of severe case A. Lower respiratory tract infections 1. Pneumonia History of cough or difficulty breathing PLUS tachypnea for age or oxygen required WITHOUT wheeze At least 1 of • Lower chest wall indrawing or nasal flaring • Not able to drink or breastfeed • Vomiting everything • Convulsions during this illness • Lethargic or unconscious • Oxygen saturation < 92% on room air • Any oxygen required 2. Bronchiolitis History of cough or difficulty breathing PLUS wheeze or evidence of hyperinflation on physical exam At least 1 of • Tachypnea for age • Lower chest wall indrawing or nasal flaring • Not able to drink or breastfeed • Vomiting everything • Convulsions during this illness • Apnea on history or witnessed • Lethargic or unconscious • Oxygen saturation < 92% on room air • Any oxygen required 3. Tuberculosis Started on TB treatment in hospital OR A close TB contact PLUS chest X-ray suggestive of pulmonary TB OR Bacteriological confirmation of TB on sputum, gastric aspirate or lymph node biopsy OR A close TB contact PLUS 2 of • Persistent non-remitting cough > 14/7 • Persistent fever > 14/7 • Unsatisfactory weight gain • Fatigue or reduced playfulness Hospital diagnosed extrapulmonary or miliary TB OR At least 1 of • Tachypnea for age PLUS lower chest wall indrawing or oxygen required • Headache, neck stiffness, drowsiness, irritability, convulsions • Hepatosplenomegaly • Peripheral oedema • Distended abdomen with or without ascites • Angulation of the spine/gibbus 34 Classification Definition of case Definition of severe case B. Diarrhoeal disease 1. Acute diarrhoea Liquid stools (more unformed than usual) with increased number of stools for < 14 days At least 2 of • Lethargic or unconscious • Sunken eyes • Not able to drink or drinking poorly • Skin pinch takes > 2 seconds to return 2. Persistent diarrhoea Liquid stools (more unformed than usual) with increased number of stools for > 14 days At least 2 of • Lethargic or unconscious • Restless or irritable • Sunken eyes • Not able to drink, drinking poorly, drinking eagerly, thirsty • Skin pinch takes > 1 second to return 35 2.3 Designing, pilot testing and evaluating the PIET-R The PIET-R was designed first to provide case-definitions for specific disease syndromes based on clinical signs and symptoms; secondly to incorporate a severity scale that could discriminate degrees of severity amongst hospitalized cases; and thirdly to provide a standardized form for abstracting the necessary attributes from medical records to apply to the case definitions. The PIET-R comprises two separate documents: 1) the PIET-R case-definitions (Appendix A) for classification of the infectious events and grading of severity, either severe (hospitalization definitely indicated) or mild-moderate (hospitalization not definitely indicated) and 2) the PIET-R source document (Appendix B.2) in which clinical information from hospital charts is abstracted in a standardized format according to the presence or absence of 54 attributes comprising the infectious event case-definitions. During the design phase, three general paediatricians were consulted for input on appropriateness of the PIET-R case-definitions. Revisions were made before Dr. Slogrove and 2 others (a medical student and a registered nurse) piloted the PIET-R source document on chart abstractions. Following further minor revisions, the PIET-R source document was used for all hospital chart abstractions in the main study. To better understand the performance of the PIET-R case definitions and the standardized abstraction source document when used in a research setting, and for awareness of the potential limitations of the tool, the reliability (precision) and validity (accuracy) of the LRTI and diarrhoea case-definitions were formally evaluated in a separate study on a set of infant hospitalization events different to those of the main study. The predominant concerns were how reliably the necessary attributes to meet the case-definitions would be abstracted by non-specialist members of the study team and how valid the diagnostic classification would be in comparison with the diagnosis from a specialist paediatrician. Also taken into consideration was the potential for wider use in low-middle income countries (LMICs) by research staff without any formal healthcare training. Lay counsellors trained in HIV adherence counselling, usually without any post-secondary education qualifications, have become valuable members of research teams in LMICs. Therefore, even though lay counsellors were not used for the main study hospital chart abstractions, a lay counsellor was initially included in the evaluation. 36 2.4 PIET-R evaluation hypotheses and objectives The two conditions most commonly associated with hospitalization in South African children, LRTI and diarrhoea, were prioritized for evaluation. Primary hypothesis: the reliability (inter-observer agreement) of the final PIET-R classification and grade as measured by the lower bound of the 95% confidence interval for the prevalence-adjusted-bias-adjusted Kappa (PABAK) is > 0.61. Primary objective: to determine the reliability, as measured by the PABAK, of the PIET-R classification and severity grade for diarrhoea and LRTI following abstraction of hospital information by a lay counsellor, registered nurse, general practitioner and paediatrician. Secondary hypothesis: the sensitivity and specificity of the infectious event classification and grade obtained from the PIET-R case-definitions is > 80% and 70% respectively. Secondary objective: to determine the sensitivity and specificity of the PIET-R classification and severity grade for the diarrhoea and LRTI case-definitions compared to the gold standard (paediatrician determined) diagnosis. The Chapter 3 will present the results from the formal evaluation of the PIET-R. 37 Chapter 3 PIET-R reliability and validity evaluation: methods, results and discussion 3.1 Methods 3.1.1 Study design The PIET-R evaluation study was a cross-sectional study of LRTI and diarrhoea hospitalization events identified from the general paediatric ward of Worcester Provincial Hospital (WPH). Diarrhoea and respiratory tract infections peak in this region during the summer and winter months respectively. Due to this seasonality the study was conducted during December 2013 and January 2014 for diarrhoea and during July to September 2014 for LRTI. 3.1.2 Study population and setting WPH was unlikely to receive admissions from the main study cohort and was used to ensure that hospitalization events in the main study and in the PIET-R evaluation did not overlap. WPH is located in Worcester, a rural town of 90 000 people, situated 120 kilometers outside of the city of Cape Town in the Western Cape Province. Similar to the hospital in the main study, WPH is a secondary level publicly funded hospital. It has 40 paediatric beds and three specialist general paediatricians that provide paediatric care to the immediate area and 7 surrounding primary level hospitals. The general paediatric ward averages 210 admissions per month, 60% of admissions are infants aged one to 12 months and the most common reasons for admission are infectious diseases including infectious diarrhoea (34% of general paediatric admissions), LRTI (23%), and undefined sepsis (12%). The hospital thus had an adequate profile and number of admissions to identify sufficient diarrhoea and LRTI cases for the evaluation. 3.1.3 Sample selection The sample included the first 25 eligible diarrhoeal disease hospitalization events identified during December 2013 and January 2014 and the first 25 eligible LRTI events hospitalized during July 2014 to September 2014. Prior to the routine week-day morning ward round in the general paediatric ward, the research assistant (a junior paediatric resident) used the ward admission register to identify hospitalization events meeting study eligibility criteria, according to infant age and diagnosis, and brought them to the attention of the ward paediatrician for inclusion in the study and determination of the gold standard diagnosis. 38 3.1.4 Eligibility criteria Infants between one and twelve months of age at the time of hospitalization in the general paediatric ward of WPH were eligible for inclusion in the study if they had a paediatrician diagnosis of at least one of the following conditions: LRTI (including pneumonia, TB or bronchiolitis) or diarrhoea (including acute diarrhoea or persistent diarrhoea). HIV-infected infants as well as neonates less than 28 days of age were excluded as both these groups may present with atypical clinical signs and symptoms of infections. HIV-infection was excluded with appropriate testing on all infants according to the WPH paediatric ward protocol. Eligible infants were only included for one hospital admission each, but each hospital admission could be associated with a diarrhoea event, an LRTI event or both. 3.1.5 Study procedures and data collection The study was conducted in two separate stages, firstly determining of the gold standard diagnosis, as described in section 3.1.5.1 and secondly testing of the reliability and validity of the PIET-R source document and case-definitions, as described in section 3.1.5.2. The case report forms for collection and recording of all study data are described in section 3.1.5.3. 3.1.5.1 Determination of gold standard diagnosis A single paediatrician, involved with neither this study nor the main cohort study, and blinded to the PIET-R source document and classification case-definitions, determined the gold standard diagnosis and severity grade for each case. This paediatrician was asked to assign a grade of “severe” to events that definitely warranted hospitalization and a grade of “mild-moderate” to events that may not have required hospitalization. The “very severe” grading was not evaluated as it was not possible to give the paediatrician an instruction on what should objectively be graded as “very severe” while keeping him blinded to the case-definitions. The paediatrician assigned a gold standard diagnosis and severity grade based on his clinical examination and results of available investigations, including chest X-rays, blood investigations and other diagnostic tests. The gold standard diagnosis and severity grading was recorded on a standardized gold standard case report form (CRF) (Appendix B.1) 3.1.5.2 Testing of reliability and validity A general paediatrician, general practitioner, registered nurse and a lay-counsellor, conducted this testing phase and are referred to as the “observers”. These four different types of healthcare professionals were chosen to evaluate the reliability of the PIET-R across differing levels of paediatric expertise. During the study, none of the observers had been employed at WPH in the previous 5 years and all worked in Cape Town. Prior to commencing the chart abstractions, the four observers simultaneously received a short training session from Dr. Slogrove on how to complete the PIET-R source document using the hospital charts and how to determine the diagnostic classification and grade according to the PIET-R case definitions. The observers, blinded to the gold standard diagnosis, 39 independently completed chart abstractions recording the abstracted information onto the PIET-R source document (Appendix B.2). They then used the abstracted information on the PIET-R source document to classify and grade the hospitalization events according to the PIET-R case definitions and recorded their final grade and severity classifications on a third standardized CRF (Appendix B.3). As all observers came from outside of Worcester and the hospital charts could not leave WPH, this phase of the study was conducted by all observers simultaneously over a two-day period at WPH. The lay counsellor, without any formal healthcare training and without prior experience working with hospital charts could only complete half of the event abstractions and classify less than a quarter of the events with the information abstracted. Therefore, the study strategy was revised to exclude the lay counsellor’s information in the final analysis. 3.1.5.3 Study documentation A study log containing the personal healthcare number and the date of admission and discharge for the event was kept separately from all other study documentation and CRFs. All CRFs were identified using an anonymous hospitalization identifier assigned uniquely to each hospitalization event. Three CRFs were used to collect study data. The first CRF was the Gold Standard CRF used by the paediatrician to indicate for each of the possible diagnostic classifications (LRTI including pneumonia, bronchiolitis, TB and diarrhoea including acute and persistent diarrhoea) whether this diagnosis was present and graded as mild-moderate, present and graded as severe or not present. Also collected on this CRF was the infant’s age in months, dates of admission and discharge and an indication that the infant met the study eligibility criteria. The second CRF was equivalent to the PIET-R source document onto which the four observers abstracted the required information from the hospital charts. The observers used the third CRF to record the final diagnostic classification and grade they had assigned to each event according to the PIET-R case-definitions. The observers indicated on this CRF for each of the diagnostic classifications whether this diagnosis was present and mild-moderate, present and severe or not present. They were also able to indicate if an event was unclassified according to the PIET-R case-definitions. 3.1.6 Data management Each hospitalization event was allocated an anonymous alphanumeric identifier for identification on all study CRFs. All data from the CRFs was entered into an electronic OpenClinica database housed on a secure server at the University of British Columbia (UBC). Dr. Slogrove developed the study database with the support of the data team at the Vaccine Evaluation Centre, UBC. The database was routinely backed up to the UBC server, following standard operating procedures. Following single data-entry, quality checks were performed using logic checks written in R and run on the extracted data. Logic check discrepancies were verified with the source documents and edited as appropriate by Dr. Slogrove. 40 3.1.7 Ethical considerations The study was conducted according to the principles of ICH Good Clinical Practice for clinical research and taking into account the Declaration of Helsinki. The Stellenbosch University’s Health Research Ethics Committee of the Faculty of Health Sciences (N13/10/139), Worcester Hospital, the Provincial Government of the Western Cape Health Impact Assessment Committee (2013RP191) and the institutional review board of the University of British Columbia (H13-03518) granted approval of the study. As this study was conducted by way of review of patient records and did not impact on clinical care decision-making, a waiver of individual informed consent was approved. 3.1.8 Sample size calculation Fifty hospitalization events were sought, 25 LRTI and 25 diarrhoea events. It was estimated that it would be possible to conduct no more than 25 hospital chart abstractions per day over the two-day period available. Each hospitalization event could be associated with either a lower respiratory tract infection event, a diarrhoea event or both. It was decided a priori that an estimated inter-observer agreement as measured by the prevalence-adjusted-bias-adjusted-kappa (described in section 3.1.9) with a lower 95% confidence interval bound of greater than or equal to 0.61, indicating at a minimum “good” agreement , would indicate adequate inter-observer agreement was observed (199). On a sample size of 50 this would require an estimated prevalence-adjusted-bias-adjusted-kappa of 0.73. 3.1.9 Analytic strategy Evaluating the reliability of the diagnostic classifications between various observers was prioritized as the primary objective, as without adequate inter-observer agreement appropriate evaluation of the validity of the case-definitions against the gold standard diagnoses could not be assessed. 3.1.9.1 Primary objective: reliability The primary objective was to determine the reliability of the PIET-R diagnostic classifications for diarrhoea and LRTI and also the reliability of identifying severity of diarrhoea and LRTI events by 3 levels of healthcare professionals; a specialist paediatrician, general practitioner and a registered nurse (it was not possible to do this for the lay counsellor as noted above). Various measures of inter-observer agreement for dichotomous variables exist, each with its own limitations. The crude percent positive agreement is intuitive and easy to calculate but does not take agreement by chance into account. The kappa statistic improves on this by giving a measure of inter-observer agreement after accounting for the proportion of agreement that could be expected by chance alone (200). Kappa can take on values between -1 and 1, values greater than zero representing agreement greater than chance and values less than zero representing agreement less than chance. 41 However kappa is dependent upon the prevalence of true “positivity” in the population and tends towards zero at extremes of prevalence of positivity. Kappa also rewards differential assessment of positivity between observers (201). Thus when the marginal totals of the contingency table (f1, f2, g1, g2 in Table 3.1 below) are unbalanced, kappa can be paradoxically higher than when marginal totals are balanced. In response to the limitations in kappa, Byrt et.al described the prevalence-adjusted-bias-adjusted-kappa (PABAK) (202). The PABAK adjusts for a bias index defined as the difference between two observers in the proportion of positive responses, and a prevalence index defined as the difference between the probability of a positive response and the probability of a negative response in the population. PABAK, like kappa, has values between -1 and 1. Table 3.1 Contingency table for inter-observer agreement of diarrhoea case-definition Observer 2 Diarrhoea No diarrhoea Observer 1 Diarrhoea a b g1 No diarrhoea c d g2 f1 f2 N In reference to Table 3.1 above, the following measures of inter-observer agreement were calculated between each pair of observers for each of the classifications a) all diarrhoea, b) severe diarrhoea, c) all LRTI, d) severe LRTI: • Proportion positive agreement (PPA) = 2a/(2a + b + c) • Proportion negative agreement (PNA) = 2d/(2d + b +c) • Proportion observed agreement (Po) = (a + d)/N • Proportion expected chance agreement (Pe) = (f1 g1 + f2g2)/N2 • Kappa = (Po – Pe)/(1-Pe) • Bias index (the difference in proportion of yes for each of two observers) = (b-c)/N • Prevalence index (the difference between the probability of a positive response and the probability of a negative response in this sample) = (a-d)/N • Prevalence-adjusted-bias-adjusted-Kappa (PABAK) = 2Po – 1. An a priori cut-off for adequate reliability was set as the lower bound of the 95% confidence interval for the estimate of PABAK being greater than or equal to 0.61, representing at a minimum good agreement beyond chance according to the scale suggested by Byrt (see Appendix B.4). For the inter-observer agreement of “all diarrhoea” the LRTI events without concurrent diarrhoea were considered as “no diarrhoea” and for “all LRTI”, diarrhoea events without concurrent LRTI considered as 42 “no LRTI”. Thus for the “all diarrhoea” and “all LRTI” classifications all 50 events were included. For the “severe diarrhoea” classification, only the events first classified by the observers as “all diarrhoea” were included for the evaluation of “severe diarrhoea” as opposed to “mild-moderate diarrhoea” and likewise for the “severe LRTI” events. 3.1.9.2 Secondary objective: validity The secondary objective was to determine the validity of the PIET-R diagnostic classifications “all diarrhoea” and “all LRTI” and grades “severe diarrhoea” and “severe LRTI” as determined by the three observers and tested against the gold standard paediatrician diagnosis. The sensitivity, specificity, positive predictive value and negative predictive value were determined for each classification individually for each of the observers and for all observers combined. When calculating the validity indices for all observers combined, a true positive was considered when all three observers correctly classified the condition as present, a true negative when all three observers correctly classified the condition as absent, a false positive if at least one observer falsely classified the condition as present and a false negative if at least one observer falsely classified the condition as absent. With reference to Table 3.2 below, the validity indices were calculated as follows: Sensitivity = true positive / (true positive + false negative) Specificity = true negative / (true negative + false positive) Positive predictive value = true positive / (true positive + false positive) Negative predictive value = true negative / (true negative + false negative) Table 3.2 Contingency table for sensitivity and specificity of diarrhoea case-definition Gold Standard Diagnosis Diarrhoea No diarrhoea Observer Classification Diarrhoea True positive False positive No diarrhoea False negative True negative To calculate confidence intervals for PABAK its variance was calculated as 4Po (1-Po)/N (203). Confidence intervals for the validity indices were calculated according to the adjusted Wald method using the on-line calculator at http://www.measuringu.com/wald.htm (204). Statistical analysis was conducted by Dr. Slogrove using R version 3.1.0 (2014 R Foundation for Statistical Computing, Vienna, Austria) and Microsoft Excel for Mac 2011 version 14.4.7 (2010 Microsoft Corporation). 43 3.2 Results Fifty hospitalizations of eligible infants one to 12 months of age were identified with a gold standard diagnosis and grading determined. Twenty-eight events had a gold standard diagnosis of diarrhoea, 26 considered by the gold standard paediatrician as “severe diarrhoea”. Twenty-six events had a gold standard diagnosis of an LRTI, 19 considered as “severe LRTI”. All gold standard events were classifiable according to a PIET-R case-definition except for a single gold standard diarrhoea event that was unclassified by the registered nurse and a single LRTI event that was unclassified by all three observers. The primary objective was to determine the reliability of the final PIET-R classification and grade achieved through measures of inter-observer agreement. For “all diarrhoea” events there was high positive, negative and overall observed agreement with a low bias index resulting in identical kappa and PABAK values between 0.88 (95% CI 0.74, 1.00) and 0.96 (95% CI 0.88, 1.00) for the three pairs of observers (Table 3.3). For “severe diarrhoea”, observed agreement ranged from 0.74 to 0.81, but agreement by kappa and PABAK were substantially lower ranging from 0.42 (95% CI 0.09, 0.75) to 0.62 (95% CI 0.32, 0.92) and 0.48 (95% CI 0.14, 0.82) to 0.62 (95% CI 0.32, 0.92) respectively. For “all LRTI” events, overall observed agreement was between 0.82 and 0.9 with a low bias index and identical kappa and PABAK values ranging from 0.64 (95% CI 0.43, 0.85) to 0.80 (95% CI 0.64, 0.97). For “severe LRTI” events, overall agreement was between 0.9 and 0.91. With a high prevalence index kappa and PABAK differed considerably, kappa ranging from 0.45 (95% CI -0.19, 1.00) to 0.62 (95% CI 0.15, 1.00) and PABAK ranging from 0.80 (95% CI 0.54, 1.00) to 0.82 (95% CI 0.58, 1.00). By PABAK inter-observer agreement for “all diarrhoea” was very good according to Byrt's descriptive classification and met the a priori hypothesis of at least good agreement (the lower bound of the 95% CI > 0.61) for all 3 observer pairs. The only other classification to meet the a priori criteria for good agreement was “all LRTI” for the general practitioner-paediatrician pair. For “severe diarrhoea” the lower confidence interval bounds were in the range of only slight agreement and “severe LRTI” in the range of fair agreement and thus did not meet the a priori criteria of at least good agreement. The point-estimates for “severe diarrhoea” indicate that agreement could be fair and for “severe” LRTI could be good. 44 Table 3.3 Inter-observer agreement of PIET-R diagnostic classifications PPA PNA Po BI PI Kappa (95%CI) PABAK (95%CI) All Diarrhoea Nurse-GP 0.98 0.98 0.98 -0.02 0.14 0.96 (0.88, 1.00) 0.96 (0.88, 1.00) Nurse-Paediatrician 0.96 0.95 0.96 0.00 0.12 0.92 (0.81, 1.00) 0.92 (0.82, 1.00) GP-Paediatrician 0.95 0.93 0.94 0.02 0.14 0.88 (0.74, 1.00) 0.88 (0.74, 1.00) Severe Diarrhoea Nurse- GP 0.83 0.70 0.79 -0.21 0.29 0.56 (0.27, 0.84) 0.58(0.28, 0.88) Nurse-Paediatrician 0.84 0.78 0.81 -0.04 0.15 0.62 (0.32, 0.92) 0.62(0.32, 0.92) GP -Paediatrician 0.81 0.59 0.74 0.19 0.37 0.42 (0.09, 0.75) 0.48(0.14, 0.82) All LRTI Nurse- GP 0.82 0.82 0.82 -0.06 0.02 0.64 (0.43, 0.85) 0.64(0.42, 0.86) Nurse-Paediatrician 0.82 0.85 0.84 0.00 -0.04 0.68 (0.48, 0.88) 0.68(0.48, 0.88) GP -Paediatrician 0.90 0.90 0.90 0.06 0.02 0.80 (0.64, 0.97) 0.80(0.64, 0.96) Severe LRTI Nurse- GP 0.94 0.67 0.90 -0.10 0.71 0.62 (0.15, 1.00) 0.80(0.54, 1.00) Nurse-Paediatrician 0.94 0.67 0.90 -0.10 0.70 0.62 (0.15, 1.00) 0.80(0.54, 1.00) GP -Paediatrician 0.95 0.50 0.91 0.00 0.83 0.45(-0.19, 1.00) 0.82(0.58, 1.00) BI – bias index; GP – general practitioner; PABAK – prevalence-adjusted-bias-adjusted kappa; PI – prevalence index; PPA – proportion positive agreement; PNA – proportion negative agreement; Po – proportion observed agreement The secondary objective was to determine the validity of the PIET-R diagnostic classifications, through measures of sensitivity, specificity, positive and negative predictive value. The “all diarrhoea” classification had the highest sensitivity and specificity ranging from 0.96 (95% CI 0.81, 1.00) and 0.90 (95% CI 0.71, 1.00) respectively for all observers combined to 1.00 (95% CI 0.90, 1.00) and 1.00 (95% CI 0.87, 1.00) for the nurse (Table 3.4). The “severe diarrhoea” grading performed less well with a sensitivity of 0.50 (95% CI 0.32, 0.68) for all observers combined and slightly better for individual observers ranging from 0.54 (95% CI 0.35, 0.71) for the nurse to 0.77 (95% CI 0.58, 0.89) for the general practitioner. Positive predictive value for “severe diarrhoea” was good ranging from 0.93 (95% CI 0.66, 1.00) for all observers to 1.00 (95% CI 0.83, 1.00) for the paediatrician. With only two “mild-moderate” or “negative severe diarrhoea” events it was not possible to interpret the specificity and negative predictive value of “severe diarrhoea”. The “all LRTI” classification performed consistently across all validity indices with an overall sensitivity and negative predictive value of 0.73 (95% CI 0.52, 0.87) for all observers and overall specificity and positive predictive value of 0.79 (95% CI 0.59, 0.91). The “severe LRTI” grading performed adequately in terms of sensitivity and positive predictive value, but with only 7 “negative severe LRTI events”, the specificity and negative predictive values varied widely across observers. 45 Table 3.4 Validity of PIET-R diagnostic classifications Sensitivity Specificity Positive Predictive Value Negative Predictive Value All Diarrhoea All observers 0.96 (0.81, 1.00) 0.90 (0.71, 1.00) 0.93 (0.77, 0.99) 0.95 (0.76, 1.00) Nurse 1.00 (0.90, 1.00) 1.00 (0.87, 1.00) 1.00 (0.90, 1.00) 1.00 (0.87, 1.00) GP 1.00 (0.90, 1.00) 0.95 (0.76, 1.00) 0.96 (0.81, 1.00) 1.00 (0.86, 1.00) Paediatrician 0.96 (0.81, 1.00) 0.95 (0.76, 1.00) 0.96 (0.81, 1.00) 0.95 (0.76, 1.00) Severe Diarrhoea All observers 0.50 (0.32, 0.68) 0.00 (0.00, 0.78) 0.93 (0.66, 1.00) 0.00 (0.00, 0.20) Nurse 0.54 (0.35, 0.71) 0.50 (0.10, 0.90) 0.93 (0.68, 1.00) 0.08 (0.00, 0.38) GP 0.77 (0.58, 0.89) 0.50 (0.10, 0.90) 0.95 (0.76, 1.00) 0.14 (0.10, 0.53) Paediatrician 0.62 (0.42, 0.78) 1.00 (0.22, 1.00) 1.00 (0.83, 1.00) 0.09 (0.00, 0.40) All LRTI All observers 0.73 (0.52, 0.87) 0.79 (0.59, 0.91) 0.79 (0.59, 0.91) 0.73 (0.52, 0.87) Nurse 0.96 (0.70, 0.97) 0.96 (0.78, 1.00) 0.96 (0.78, 1.00) 0.96 (0.70, 0.97) GP 0.85 (0.66, 0.94) 0.79 (0.59, 0.91) 0.81 (0.63, 0.92) 0.82 (0.63, 0.94) Paediatrician 0.85 (0.66, 0.94) 0.92 (0.73, 0.99) 0.92 (0.73, 0.99) 0.85 (0.66, 0.94) Severe LRTI All observers 0.82 (0.58, 0.95) 0.16 (0.01, 0.58) 0.74 (0.51, 0.89) 0.25 (0.00, 0.71) Nurse 0.82 (0.58, 0.95) 0.33 (0.09, 0.70) 0.78 (0.54, 0.92) 0.40 (0.12, 0.77) GP 1.00 (0.84, 1.00) 0.60 (0.23, 0.88) 0.89 (0.67, 0.98) 1.00 (0.47, 1.00) Paediatrician 0.94 (0.72, 1.00) 0.25 (0.03, 0.71) 0.85 (0.63, 0.96) 0.50 (0.09, 0.91) GP – general practitioner The sensitivity of all classifications except “severe diarrhoea” was greater than 80% for individual observers against the gold standard and met the a priori set limit for adequate sensitivity of 80%. The lower bound of the 95% confidence intervals indicate though that the sensitivity could be as low as 66% for “all LRTI” and 58% for “severe LRTI”. With an insufficient number of negative severe events for both diarrhoea and LRTI, specificity could not be evaluated. 3.3 Discussion The PIET-R was designed to provide standardized case-definitions with a grading scheme for the most frequent infectious disease syndromes in South Africa. The tool used WHO and local clinical definitions and the definitions and grading scheme were applied using signs and symptoms recorded in hospital 46 notes. The precision and accuracy of the PIET-R case-definitions were formally evaluated in this small study in order to guide interpretation of the main study results. The attributes to be applied to the case-definitions can be considered as relevant as they were adequately recorded in the admission notes for the observers to classify all but one diarrhoea and one LRTI event. The overall proportion of inter-observer agreement was between 74% and 98%, and all classifications for all observer pairs had agreement better than chance (PABAK >0), but only “all diarrhoea” had very good agreement, with “severe diarrhoea” being the least precise. To attain a classification of any LRTI or any diarrhoea required only a few signs or symptoms be well reported in hospital notes and hence relatively good agreement between observers was reached. To classify an event as severe required one additional attribute for “severe LRTI” and two additional attributes for “severe diarrhoea”, possibly explaining why the severity classifications and severe diarrhoea specifically performed poorly. For validity of the case definitions “severe diarrhoea” had low sensitivity for all observers against the gold standard. Despite the wide implementation in primary healthcare clinics in South Africa of IMCI, on which the definition of severe diarrhoea was based, hospital clinicians still often use older definitions to quantify dehydration, quantifying the extent as 5% or 10% dehydration. The PIET-R definition for severe diarrhoea may warrant including assessment of “10% dehydrated” to meet criteria for a severe event. With few events considered as “mild-moderate” or negative severe events, it was not possible to interpret the specificity and negative predictive value of the severe diarrhoea and severe LRTI classifications and is a short-coming of the study. Inclusion of additional “mild-moderate” cases seen in the paediatric emergency or out-patient department, that did not require admission to the paediatric ward, would have been advantageous to ensure adequate numbers of “mild-moderate” cases for appropriate evaluation of specificity and negative predictive value of the severity grading. Overall, the PIET-R classifications had reasonable to good positive predictive value. It is reassuring that the PIET-R does not over classify event occurrence or severity. Using this tool would underestimate rather than overestimate the number of all diarrhoea and all LRTI events and underestimate severity. Due to logistic and resource considerations, this initial PIET-R evaluation was limited to assessing the most common infectious event classifications, LRTI and diarrhoea. The complete PIET-R tool was not evaluated. Moreover the evaluation was confined to HIV-uninfected infants one to 12 months of age. In this initial evaluation only one hospital was used, with the same team of paediatric doctors during the evaluation period. The paediatric department of WPH sees patients with a similar disease profile and has doctors of a similar level of experience to the hospital receiving the majority of hospitalizations for the main study. Therefore, although these results may not be widely generalizable, they adequately represent the tools performance in the main study. 47 It is unfortunate that the lay-counsellor, without any medical training, could not complete all of the abstractions in the time available. The observers all received a short introductory session on how to use the PIET-R tool, but Dr. Slogrove gave very little input to the interpretation of what was documented in the hospital notes or navigating the hospital admission charts so as not to bias the abstraction and classification by observers. With additional training and guided practice a lay-counsellor may well be able to adequately abstract information from hospital notes to the PIET-R source document and follow the algorithms to reach a case-definition. Without this additional guidance though, a lay counsellor should not be given the responsibility of chart abstractions and diagnostic classification. The PIET-R did not perform as well as hypothesized. This may partly be due to the limited nature of the evaluation and short-comings in the study design. However, there are no published validity or reliability evaluations of the DAIDS adverse event grading, Brighton case-definitions or other infectious disease case-definitions or classification schemes for comparison. In conclusion, sufficient information according to the case-definition attributes was available in the clinical hospital admission notes. All classifications had agreement better than chance, but only “all diarrhoea” met the a priori expectation for minimum “good” inter-observer agreement. Validity according to sensitivity met a priori expectations for all classifications except “severe diarrhoea” and the overall high positive predictive values for the severe classifications indicate that severity was not overestimated. Specificity and negative predictive value could not be adequately assessed. 48 Chapter 4 Mother Infant Health Study methods To test the primary study hypothesis that South African HEU infants have a greater probability of infectious morbidity than HUU infants, the Mother Infant Health Study recruited a prospective cohort of these two groups of infants. Central to testing this hypothesis was the assumption that universal infant risk factors, specifically socioeconomic position, poor birth outcome (preterm birth, small for gestational age) and infant feeding differences, do not account for the entirety of the difference between HEU and HUU infants, and that HEU unique exposures also play a role. As such, this study was designed to reduce potential socioeconomic differences between infants and to consider differences in infectious disease risk between term HEU and HUU infants beyond what can be accounted for by breastfeeding differences. 4.1 Study design, population and setting Mothers and their infants were recruited at birth from the Kraaifontein midwife obstetric unit (MOU), Cape Town, Western Cape Province, South Africa. A single MOU providing care for pregnant women from low socio-economic communities was chosen to ensure roughly comparable household socio-economic position among study participants. Participants were followed at the Children’s Infectious Diseases Clinical Research Unit (KID-CRU), Tygerberg Hospital, Cape Town until infants were six months of age. Study enrolment took place Monday through Friday from 15 July 2012 to 30 June 2013 and 6 month follow-up was completed in December 2013. HIV-infected and HIV-uninfected mothers were frequency matched on race/ethnicity to limit variation in social habits including smoking, alcohol and drug use (205,206). To control for seasonal patterns in infectious morbidity HIV-unexposed infants were chosen to match HIV-exposed infants with the closest date of birth, but no more than 30 days different in age. 4.2 Study methodology 4.2.1 Eligibility criteria To further ensure comparable socio-economic status, HIV-infected and HIV-uninfected mothers from four well-defined neighbourhoods with similar low socio-economic position were eligible for inclusion. Mothers with low-risk obstetric histories and HIV-infected mothers with uncomplicated HIV disease were included to minimize maternal comorbid factors that are independently associated with universal infant morbidity and mortality. Mothers 18 years and older from whom written informed consent could be obtained and who delivered a live-born baby at Kraaifontein MOU were eligible for inclusion. Mothers and their infants 49 were ineligible if the mother was unable to communicate in English, Afrikaans or Xhosa (the three main languages of the Western Cape Province), if the mother had no means of telephonic communication or intended to move away from Cape Town before the infant’s first birthday. Mothers who had received no antenatal care prior to admission in labour or had delivered before arrival at Kraaifontein MOU were excluded. High-risk mothers including HIV-infected mothers on third line cART, mothers with pre-eclampsia, eclampsia or a multiple pregnancy were also excluded. Infants born before 34 weeks gestation, birth weight below 2000g or with any severe terminal congenital birth defects or genetic abnormalities (e.g. Trisomy 13 or 18) were excluded. HIV-infected infants and infants with a diagnosis of a primary immune deficiency or oncological disorder were withdrawn from the study. Eligibility was not restricted according to infant feeding choice as the local infant feeding policy was in transition during the study. 4.2.2 Study procedures The study consisted of an enrolment visit within 72 hours of delivery, and follow-up visits at two weeks of age, two months, four months and six months of age. See Table 4.1 for a summary of the study procedures. 4.2.2.1 Study site preparation Prior to commencement of the study the health management team of Kraaifontein MOU was engaged in a discussion on the aims of the study and what measures the study team could take to minimize interference with clinic activities. A short orientation session about the study was held with the labour ward and postnatal ward staff before commencement of enrolment. There was no additional load placed on the nursing staff or counsellors at the facility. The study coordinator maintained open communication with the MOU manager to ensure that the study was not impinging on their service delivery. The Kraaifontein monthly aggregate statistics of the total number of deliveries and proportion delivered by HIV-infected and HIV-uninfected mothers for the period of enrolment were collected in order to determine the representativeness of the study sample compared to the population. 4.2.2.2 Enrolment and informed consent process Prior to enrolment, the study team provided information about the study to all pregnant women attending the Kraaifontein antenatal clinic for prenatal visits. After receiving information about the study, mothers were approached following delivery and given up to 3 days to consider participation in the study before written informed consent was requested. The informed consent procedure and all study interviews were offered in the mother/caregivers preference of Afrikaans, English or Xhosa. At the start of each study visit the study team requested ongoing informed consent affirmation. 50 Table 4.1 Summary of study procedures Enrolment Visit 1 Visit 2 Visit 3 Visit 4 Hospitalized Window (in reference to infant age) 0-72 hours 14 +/-4 days 8 +/-2 weeks 16 +/-2 weeks 26 +/-2 weeks 0-194 days Retrieval of PMTCT Programme results: Maternal antenatal CD4 count X Infant 6-week HIV-PCR (HEU) X Health record reviews: Maternal Obstetric Care Record X Infant Road To Health Book X X X X X X Interviews with mother/caregiver: Maternal health history X Socioeconomic interview X Infant health & feeding history X X X X Physical examinations: Maternal anthropometry X Infant physical examination & anthropometry X X X X Maternal investigations: HIV serology (HIV-uninfected) X CD4 count (All) X HIV viral load (HIV-infected) X Infant investigations: Full blood count (All) X X X X HIV-PCR (HEU) X X HIV serology (HUU) X X DBS card stored for HIV-PCR (All) X X X Hospitalization record abstraction X DBS – dried blood spot; HEU – HIV exposed uninfected; HUU – HIV unexposed uninfected; HIV-PCR – HIV polymerase chain reaction; PMTCT – prevention of mother to child transmission 51 4.2.2.3 Co-ordination of study visits In order to reduce and manage losses to follow-up, at enrolment detailed contact information, including address with landmarks, relatives contact details (with consent) and intended clinic to be used for routine infant visits, was collected. At each study visit, a plan was made with the mother for the most suitable date for the next study visit and how she would travel to the study site for the visit. Mothers were offered two transport options, either to be transported by the study bus or to be reimbursed for cost of public transportation to the study site. Mothers received a text-message reminder and phone call (with consent) before the scheduled study visit. Mothers contributed approximately five hours of their time for each study visit, from the time they were picked up by the study bus in the morning to the time they were dropped off at home again. For any mother-infant pairs that did not attend a study visit, attempts to trace the pair were made by the study team using available contact information. If this was unsuccessful, the local health clinic was contacted for information about the possible whereabouts of the mother and infant. Mother-infant pairs were withdrawn from study follow-up if contact was not re-established after two months of active tracing. Even if lost-to-contact, all infants were maintained in the study for determination of the primary outcome (see section 4.2.2.9 below). 4.2.2.4 Health record reviews Maternal obstetric records were reviewed at enrolment and necessary data abstracted onto a standardized study enrolment source document. The maternal obstetric record, used in all public healthcare facilities in the province, is completed and updated by registered nurse midwives at the first and all subsequent antenatal clinic visits. It also contains the comprehensive record of the delivery, the outcome of the delivery and all care given to the mother and newborn immediately post delivery. See Table C.1 in Appendix C for maternal variables abstracted. At every follow-up visit the infant Road To Health Book was reviewed. This is a caregiver held record containing comprehensive information about the infant’s birth, growth and immunizations received at the local health clinic. Documentation of immunizations and the dates they were received was abstracted from the Road to Health Book onto the study source document at each visit. 4.2.2.5 Interviews and infant physical examination All study interviews were conducted according to standardized questionnaires. At enrolment a research nurse conducted an interview for maternal obstetric, demographic and contact details. At the two-week follow-up visit the research medical officer conducted a maternal health history, a research nurse performed maternal anthropometric measurements and a research counsellor conducted a detailed socioeconomic and household questionnaire with the mother. At every follow-up visit the research medical officer conducted an infant health history and physical examination, a research nurse performed infant anthropometric measurements and a research counsellor conducted a detailed infant feeding history with the mother or caregiver. The research medical officer conducting the infant health history and 52 physical examination was not blinded to infant HIV-exposure status so as to be able to provide appropriate medical care and referrals as needed. Where there was a clinical indication to suspect HIV-infection at any study visit on any infant, the HIV-PCR was conducted immediately at that study visit following appropriate counselling. These clinical indications for suspicion of HIV-infection included any infectious cause hospitalization, moderate or severe acute malnutrition, failure to thrive, diagnosis of TB or at the discretion of the research medical officer. If the infant was confirmed to be HIV-infected, the infant was referred for appropriate HIV care and treatment and the mother and infant withdrawn from the study. 4.2.2.6 Maternal investigations At enrolment, within 72 hours of delivery, maternal venous blood samples were collected for CD4 count on all mothers and HIV viral load on HIV-infected mothers. At the two-week visit HIV-uninfected mothers had their HIV status confirmed with a rapid 4th generation HIV ELISA antibody test using standard South African HIV testing algorithms (112). Mothers that were found to have seroconverted postnatally were appropriately counselled and referred for HIV care. Their infants immediately had an HIV PCR test performed (as described in 4.2.2.7 below) to exclude infant HIV infection and were referred for appropriate prophylactic measures if HIV-uninfected or for HIV antiretroviral treatment if found to be HIV-infected. 4.2.2.7 Infant investigations At each visit from two weeks of age infant blood was sampled for a haemoglobin measurement. Prior to the two month visit, results of the infant HIV-PCR test performed by the primary healthcare clinics on all HIV exposed infants at six weeks of age, according to the South African National PMTCT programme guidelines, were accessed from the National Health Laboratory Service (NHLS) database according to the infant’s province-wide unique healthcare number. In order to exclude HIV infection, at 6 months of age infant blood was sampled for an HIV-PCR on HIV exposed infants and an HIV ELISA on HIV unexposed infants. In addition blood was stored on a dried blood spot card at each visit for retrospective exclusion of HIV-infection in infants lost to follow-up prior to six months of age. Prior to all blood sampling a local anaesthetic ointment was applied to the infants’ skin. The maximum safe blood volume that could be drawn at each visit was never exceeded (207). 4.2.2.8 Laboratory procedures The NHLS processed all laboratory investigations. The maternal CD4 lymphocyte count was performed by standard NHLS Immunology Laboratory procedures using the Beckman Single Platform PanLeucoGating method. HIV viral load was performed by the NHLS Virology Laboratory and measured as a quantitative HIV RNA level using the Abbott RealTime HIV-1 Assay, with a limit of detection of 40 copies/ml. The Abbott 4th Generation ELISA Axsym was used to screen for HIV-infection in all HUU 53 infants and the Roche COBAS AmpliPrep/ COBAS Taqman HIV-1 Test version 2.0 HIV-PCR was used to exclude HIV-infection in all HEU infants. 4.2.2.9 Outcome determination – infectious cause hospitalization or death Near-real time hospitalization of infants was tracked using telephone caregiver contact in conjunction with regular telephone contact with the ward clerks of the paediatric wards at the two hospitals most likely to receive admissions of study infants. An unblinded trained member of the research team (one registered nurse and one medical student) reviewed the admission notes as soon as the hospitalization was identified and completed the de-identified standardized source document of the Paediatric Infectious Event Tool for Research (PIET-R). At roughly six-month intervals during the study and on completion of all six-month infant follow-up, the Western Cape Province electronic hospital administrative system was searched for every infant enrolled in the study for a record of hospital admission. The hospital records of any additional hospitalizations identified through these searches were reviewed and a PIET-R source document completed as described above. This allowed for identification and inclusion of hospitalizations for all infants including those lost to face-to-face follow-up prior to completing six months of follow-up. Two paediatricians, (Dr. Slogrove and another paediatrician not otherwise involved with any aspects of the study) both blinded to HIV exposure status, used the PIET-R source document for each hospitalization to independently classify and grade the infectious events according to the PIET-R case-definitions (see Appendix A). A grading according to the DAIDS Grading of Adult and Paediatric Adverse Events was also given to each event by both paediatricians independently, so as to be able to compare to previous studies that have used DAIDS (187). Where the classification and grading differed between the two paediatricians, consensus was reached through discussion and these differences and decisions documented. Hospital emergency department visits that did not result in an overnight admission to the paediatric ward were not included as hospitalization events and were counted instead as sick-clinic visits. At completion of the last six-month follow-up a number of mechanisms were used to identify infant deaths to be included in the primary and secondary outcomes. Firstly, attempts were made to recontact all families of infants that were lost to face-to-face follow-up before six months of age. Secondly, the Western Cape Province mortality registry was searched based on date of birth for all infants not completing six months of follow-up. It is possible that some infant deaths are not officially registered and undergo informal burial services in the community and would not be identified through these mechanisms. To understand whether we were likely to have misidentified a substantial number of infant deaths and underestimated the primary outcome we used local infant and child mortality rates to estimate the number of infant deaths we could reasonably expect in our cohort. Published Western Cape Province and City of Cape Town infant and under-5 mortality estimates for 2010 were used, indicating an infant mortality rate of 22.3/1000 and under 5 mortality rate of 27/1000 (168,184). Our study excluded HIV-infected infants and high-risk neonates, these two factors contributing up to 64% of under-5 mortality in this area (58). 54 Based on this, we altered the estimate of infant mortality for what could be expected in our cohort to 7.9/1000 and compared this to the number of infant deaths ultimately identified. 4.2.2.10 Additional sub-studies In addition to the study presented here for Dr. Slogrove’s dissertation, a number of other sub-studies conducted by collaborators explored specific aspects of HEU infant health in comparison to HUU infants. These included i) an evaluation of maternal placental pathology, ii) comparison of infant feeding practices and growth outcomes between breast and formula fed HEU infants up to 12 months of age iii) neurodevelopmental outcomes of HEU compared to HUU infants at 12 months of age iv) CMV and toxoplasmosis prevalence and timing of acquisition, v) description of haematological differences between HEU and HUU infants and vi) a fathering efficacy study. The methods and results of these sub-studies are not presented in this dissertation. 4.2.3 Data Management All mothers and infants in the study were assigned a unique anonymous study identifier and all data was collected on coded de-identified source documents at enrolment and each study visit. A patient demographic form containing the personal identification of enrolled mother-infant pairs was stored separately from the source documents. Data recorded on the source documents was entered into an OpenClinica database on an on-going basis by a single study data-capturer at KID-CRU, Tygerberg Hospital. Quality control logic checks were run on the data entered in the database and logic check discrepancies resolved by review of source documents. The electronic database was password-protected and access restricted to the study investigators. The database was routinely backed up to the University of British Columbia, Vaccine Evaluation Center server, following a standard operating procedure. 4.3 Study variable definitions The definitions of the most important study variables are given below. These variables are derived from data collected through abstraction from health records, maternal or caregiver interviews and infant physical examinations. The comprehensive list of study variables including their source, type and response options can be found in Appendix C. 4.3.1 Maternal variables • HIV status – binary variable, either HIV-infected or HIV-uninfected; according to the maternal obstetric record and confirmed with rapid HIV antibody test at two week study visit 55 • Age – numeric variable, maternal age at delivery in years and days; calculated from maternal date of birth according to maternal interview and infant date of birth according to maternal obstetric record • Primiparous – binary variable, either yes (this is the first time the mother has given birth) or no (the mother has given birth before); according to the maternal obstetric record • Feeding intention at delivery – nominal variable, the mothers intention at delivery to either exclusively breastfeed, exclusively formula feed or mixed feed (both breast and formula) her baby; according to maternal baseline interview • Timing of HIV diagnosis – binary variable, diagnosis of HIV made either pre-pregnancy (> 280 days before infant birth) or during pregnancy (280 days or less before infant birth); derived from the date of HIV diagnosis as given by the mother during the maternal baseline interview, and infant date of birth; • ARV regimen during pregnancy – nominal variable, none, VTP prophylaxis or cART; according to documentation on maternal obstetric record and report from maternal baseline interview • Timing of initiation of cART – nominal variable, pre-pregnancy (> 280 days before infant birth), 1st trimester (197-280 days before infant birth), 2nd trimester (99-196 days before infant birth), 3rd trimester (1-98 days before infant birth); derived from the date of cART initiation, as given by the mother during the maternal baseline interview and infant date of birth • Timing of initiation of ZDV – nominal variable, 1st trimester (197-280 days before infant birth), 2nd trimester (99-196 days before infant birth), 3rd trimester (1-98 days before infant birth); derived from the date of ZDV initiation, as documented on the maternal obstetric record, and infant date of birth • Antenatal absolute CD4 count – numeric variable measured in cells/µl; as recorded on maternal obstetric record with date not more than 280 days prior to delivery; also categorized as an ordinal variable with categories <350 cells/µl, 350-499 cells/µl and 500 cells/µl or more • Delivery absolute CD4 count – numeric variable; measured in cells/µl on maternal blood taken at delivery; also categorized as an ordinal variable with categories <350 cells/µl, 350-499 cells/µl and 500 cells/µl or more • Delivery CD4 % – numeric variable; CD4 percentage of total lymphocyte count on maternal blood collected at delivery • Absolute change in CD4 count – numeric variable; calculated as the difference between the antenatal and the delivery absolute CD4 count • Percentage change in CD4 count – numeric variable; calculated as the difference between the antenatal and delivery absolute count divided by the antenatal absolute CD4 count • HIV viral load – numeric variable; number of copies of HIV per ml of blood measured on maternal blood collected at delivery (the limit of detection of the laboratory test was 40 copies/ml, 56 measurements of <40 copies/ml were imputed with values of 39 copies/ml to calculate summary measures of log10 HIV viral load); also categorized as an ordinal variable with categories <40 copies/ml, 40-999 copies/ml, 1000-9999 copies/ml and 10 000 copies/ml or greater 4.3.2 Household variables All household information was collected by maternal interview according to a standardized questionnaire at the two-week study visit. Information was collected in reference to the household in which the infant resided should this have differed to the household in which the mother was resident. • Type of house – nominal variable; stand alone house, apartment in apartment block, house/flat/room in backyard, shack in back yard, shack not in back yard, other • Water supply – nominal variable; water piped into dwelling, water piped into yard, public tap • Sanitation – nominal variable; flush toilet (connected to sewerage), flush toilet (with septic tank), no facility/bush/field • Fuel for cooking – nominal variable; electricity, gas, paraffin, firewood, other • Number of rooms in the house – numeric variable; total number of rooms in the house including living areas, kitchen and bedrooms • Number of people in the household – numeric variable; total number of people resident in household including mother and infant • Distance to clinic – numeric variable; distance to nearest primary health care clinic measured in minutes taken to walk from home to the clinic according to maternal report • Household assets – binary variables; possession of the following assets in the household at time of interview (yes/no) – radio, television, computer, fridge, home phone, cell phone, bicycle, motorcycle, car or truck, donkey or horse, sheep, goats or cattle 4.3.3 Infant variables • HIV exposure status – binary variable; HIV exposed uninfected (HEU) or HIV unexposed uninfected (HUU), determined from maternal HIV-infection status and infant HIV-infection status; HEU infants were infants confirmed to be HIV-uninfected at 6 weeks and 6 months of age by infant HIV-PCR testing and born to HIV-infected mothers, and HUU infants were infants confirmed to be HIV-uninfected at 6 months of age by HIV ELISA testing and born to mothers confirmed to be HIV-uninfected at the 2 week study visit by HIV ELISA testing • Gestational age at birth – numeric variable; number of weeks post-last normal menstrual period at time of delivery, as recorded in the maternal obstetric record, determined by nurse midwifes. • Birth weight - numeric variable; infant weight at birth measured in grams; as recorded on the maternal obstetric record 57 • Birth length – numeric variable; infant length at birth measured in centimetres; as recorded on the maternal obstetric record • Birth head circumference – numeric variable; infant head circumference at birth measured in centimetres; as recorded on the maternal obstetric record • Infant weight – numeric variable; infant weight measured in kilograms at each study visit • Infant length – numeric variable; infant supine length measured in centimetres at each study visit • Infant head circumference – numeric variable; infant head circumference measured in centimetres at each study visit • Immunizations up to date – binary variable; yes immunizations up to date at time of study visit according to the South African National Expanded Programme for Immunization, or no immunizations not up to date at time of study visit; determined from documentation of immunizations received in the infant Road To Health Book • Current TB contact – binary variable; yes (there is an adolescent or adult currently on TB treatment that is resident in the same household as the infant), no (there is no adolescent or adult currently on TB treatment resident in the household); as reported by caregiver interview at each study visit • Current TB prophylaxis – binary variable; yes (the infant is currently receiving IPT) or no (the infant is not currently receiving IPT); as reported by caregiver interview at each study visit • Number of all-cause clinic visits – numeric variable; number of times the infant has been to the primary health care clinic since the last study visit; as reported by caregiver interview at each study visit • Number of sick clinic visits – numeric variable; number of times the infant has been to the primary health care clinic for visits other than for immunizations or routine growth monitoring since the last study visit; as reported by caregiver interview at each study visit • Symptoms at time of sick clinic visit - nominal variable; cough, fever, diarrhoea, poor feeding, poor growth or other symptoms associated with sick clinic visit, more than one symptom could be reported; as reported by caregiver interview at each study visit • Infant breastfeeding status – nominal variable; exclusively breastfed, partially breastfed, not breastfed; derived from detailed maternal 48 hour and seven day recall of what the infant received orally for nutritional and non-nutritional purposes, asked at each study visit o Exclusively breastfed – infant received breast milk only in the 48 hours and seven days prior to the study visit (drops or syrups consisting of vitamins, mineral supplements or prescribed medicines allowed); definition for exclusive breastfeeding must also have been met at all prior study visits otherwise the infant was considered as partially breastfed 58 o Partially breastfed – infant received some breast milk in the 48 hours and seven days prior to the study visit but also received artificial feeds including infant formula, other liquids or solids o Not breastfed – infant received no breast milk in the 48 hours and seven days prior to the study visit • Infant haemoglobin – numeric variable; haemoglobin measured in g/dl on infant blood collected at each study visit; also categorized for grade of anaemia as normal or grade 1 to 4, by the Division of AIDS Table for Grading the Severity of Adult and Paediatric Adverse Events (Version 1.0, December 2004; Clarification August 2009) 4.3.4 Hospitalization variables • Age at hospitalization – numeric variable; age in days calculated from date of admission and infant date of birth • Length of stay – numeric variables; number of days that the infant was hospitalized for each individual hospitalization; calculated from date of admission and date of discharge • Infectious event type – nominal variable; categories of infectious events according to PIET-R definitions • Severe infectious event – nominal variable; an infectious cause hospitalization that met criteria for a grading of “severe” according to the PIET-R definitions • Very severe infectious event – nominal variable; an infectious cause hospitalization that met criteria for a grading of “severe” for 48 hours or more according to the PIET-R definitions 4.4 Ethical Considerations This prospective cohort study was conducted according to the principles of ICH Good Clinical Practice for clinical research and taking into account the Declaration of Helsinki. The Health Research Ethics Committee of the Faculty of Medicine & Health Sciences of Stellenbosch University (S12/01/009), the Western Cape Provincial Health Impact Assessment Committee (2012RP22) and the institutional review board of the University of British Columbia (H12-01181) approved the study. Written informed consent was obtained from all mothers of enrolled mother-infant pairs. The mother/primary caregiver was able to withdraw herself and her infant from the study at any stage. Every attempt was made to avoid inadvertent disclosure of HIV-status by participation in the study and attendance for study visits. Contact by telephone or home visits if necessary to track mothers and infants lost to follow-up were only conducted with prior permission from the mother. For confirmation of HIV-59 uninfected mothers HIV status, the two-week visit was selected as opposed to at delivery, to ensure that pre and post-test counselling could be conducted in an appropriate confidential environment at the study site (KID-CRU, Tygerberg Hospital) rather than the busy Kraaifontein MOU. If an HIV-infected mother or primary caregiver of an HEU infant foresaw that an alternative caregiver would be attending the next study visit with the infant, but particularly the two month study visit, the mother/primary caregiver was asked whether she consented to the infant’s HIV-PCR result being disclosed to the specified alternative caregiver. If the mother/primary caregiver did not attend the study visit and did not consent to results being discussed with a specified alternative caregiver, arrangements were made for follow-up by telephone with the mother/primary caregiver with regards to the infant’s HIV and other necessary results. Enrolled mothers were compensated for their transportation costs and time to attend the study visits. No further financial incentives were provided for participation in the study. 60 Chapter 5 Analytic methods This chapter reviews the study objectives and defines the major determinants and outcomes before describing how sample size calculations were performed and the statistical analysis conducted to achieve the study results presented in Chapters 6 and 7. 5.1 Study objectives The primary objective of this study was to determine whether HEU infants experience an increased probability of infectious cause hospitalizations or death compared to HUU infants in the first six months of life after adjusting for differences in breastfeeding exposure. A secondary study objective was to determine whether HEU infants experience increased probability of severe or very severe infectious cause hospitalizations or death compared to HUU infants. And the final objective was to determine whether HEU infants born to mothers on maternally indicated cART during pregnancy have an increased probability of infectious cause hospitalization or death, compared to HEU infants born to mothers on VTP prophylaxis. 5.2 Definition of major determinants and outcomes The most important determinant variables for analysis of the study objectives included infant HIV exposure status, infant breastfeeding status and for the sub-group comparison of HEU infants, maternal ARV regimen during pregnancy. 5.2.1 Primary determinant: HIV exposure status The primary determinant for the primary and first secondary objective, infant HIV exposure status, was a binary variable, HIV exposed uninfected (HEU) or HIV unexposed uninfected (HUU), defined according to infant HIV exposure and infection status. HEU infants were confirmed HIV-uninfected infants born to an HIV-infected mother (see section 4.3.3). 5.2.2 Definition of breastfeeding status Infant breastfeeding status was a nominal variable with three categories: exclusively breastfed (infant received breast milk only in the 48 hours and 7 days prior to the study visit), partially breastfed (infant received some breast milk in the 48 hours and 7 days prior to the study visit but also received infant formula, other liquids or solid food) and not breastfed (infant received no breast milk in the 48 hours and 7 61 days prior to the study visit). For infants to be considered as exclusively breastfed at a visit, they must have been considered to be exclusively breastfed at all prior visits. If a visit was missed but the infant was not lost to follow-up and the feeding mode at the visits on either side of the missed visit was consistent, then the consistent feeding mode of these visits was assigned to the missed visit. If the feeding mode on either side of the missed visit was inconsistent then the feeding mode of the visit prior to the missed visit was assigned to the missed visit. For infants lost to follow-up, feeding mode was assigned for missed visits according to the last reported feeding mode carried forward. For infants that died, their feeding mode was censored at the time of death. Due to very few observations of partially breastfed HEU infants and partially breastfed infants with the primary outcome, this category was collapsed for the multivariable analysis and feeding mode was categorized as breastfed (including both exclusively breastfed and partially breastfed) or not breastfed at each visit. 5.2.3 Definition of maternal ARV regimen during pregnancy The determinant for secondary objective 2 of maternal ARV regimen during pregnancy was a binary variable and was defined as follows: i) maternally indicated cART - mothers receiving a triple ARV drug regimen during pregnancy that was commenced either before or during pregnancy for maternal indication of a CD4 count below 350 cells/µl or WHO stage 3/4 disease; ii) VTP prophylaxis - mothers receiving VTP prophylaxis during pregnancy, with a CD4 count of 350 cells/µl or greater during pregnancy and WHO stage 1/2 disease. 5.2.4 Definition of primary outcome – infectious cause hospitalization or death The primary outcome was defined as infant hospitalization for a presumed or confirmed infection or death in the first six months of life. The upper limit of six months of life was set as 194 days old (180 + 14 days), equivalent to the upper limit of the window for the final six-month follow-up visit. All deaths were included in the primary outcome irrespective of cause. 5.2.5 Definition of secondary outcomes – severe and very severe infectious cause hospitalization or death For the secondary outcome infectious cause hospitalizations were graded as mild-moderate or severe according to the PIET-R definitions. Severe infectious cause hospitalizations were defined as infectious cause hospitalizations graded as severe, with criteria for a severe event present for any duration during hospitalization. Very severe infectious cause hospitalizations were defined as infectious cause hospitalizations for which criteria for a severe event were present for at least 48 hours following admission. See Chapter 2 for a description of the PIET-R and Chapter 3 for its evaluation. See Appendix A for the infectious cause hospitalization diagnostic classification and severity grading definitions. See Appendix C for a complete list of study variables and definitions. 62 5.3 Sample size calculation Sample size calculation for the primary objective was based on a test of proportions for independent groups and performed with the OpenEpi Version 3.03 sample size calculator using the Fleiss method with continuity correction. To calculate the appropriate sample size for determining whether HEU infants experience an increased probability of infectious cause hospitalizations or death compared to HUU infants, data from the Cape Town HEU Pilot Study, described in Chapter 1.6, were used (45). In the pilot study a risk difference of 23% (35% vs. 12%) was seen between HEU and HUU infants for at least one infectious cause hospitalization before six months of age. To detect this same size difference, 61 infants were required in each group at six months of age to achieve 80% power. Towards the end of the pilot study and following acquisition of funding for the current study, the South African infant immunization schedule was expanded to include the addition of rotavirus and pneumococcal conjugate vaccines, and a policy change towards promoting exclusive breastfeeding with ARV prophylaxis for HIV exposed infants was introduced. It was therefore not expected that an event rate as high as that in the pilot study would again be observed. Various scenarios of reduced event rates were considered: scenario 1 in that both HEU and HUU infants experienced a one third reduction in events; scenario 2 in that HEU infants experienced a greater reduction in events than HUU infants but still had double the event rate of HUU infants; scenario 3 in that HEU infants experienced a greater reduction in events than HUU infants but still had 50% more events than HUU infants. Table 5.1 indicates the sample sizes required with alpha at 5% and beta at 20% considering the three different scenarios and their estimated event rates. Table 5.1 Sample size calculation scenarios HEU with the outcome (%) HUU with the outcome (%) Sample Size (per group) Pilot study 35 12 61 Scenario 1 23 8 103 Scenario 2 16 8 283 Scenario 3 12 8 927 HEU – HIV exposed uninfected; HUU – HIV unexposed uninfected Taking into consideration the possibility of up to 10% of HIV exposed infants (HEI) becoming HIV-infected, and based upon an attrition rate of 33% in both groups at 6 months in the pilot study, considering scenario 1, a minimum of 174 HEI and 156 HUU mother-infant pairs needed to be enrolled to retain 103 infants in each group in follow-up at 6 months. 63 Sample size calculations for the secondary objectives were performed in a similar manner and are not shown. 5.4 Analytic strategy All statistical analysis was performed by Dr. Slogrove using R version 3.1.0 (2014 R Foundation for Statistical Computing, Vienna, Austria) and the HIV exposure groups were kept blinded until finalization of the primary outcome analysis. 5.4.1 Description of the analytic cohort Of all the enrolled infants, those that returned for the first study visit at two weeks of age and who were confirmed to be HIV-uninfected at the end of follow-up were included in the analytic cohort. Available aggregate data on all deliveries at the Kraaifontein MOU were collected in order to compare the MOU population with the study sample in terms of maternal age, the proportion of all HIV-infected mothers, the proportion of HIV-infected mothers on cART during pregnancy and the proportion of HIV-infected mothers intending to breastfeed. For description of the analytic cohort continuous numeric variables were explored for normality by inspection of histograms and box-and-whisker plots. Measures of central tendency and spread were described as mean and standard deviation for normally distributed numeric variables and as median and interquartile range if not normally distributed. Categorical variables were described as absolute frequencies and percentages. For evaluation of the anthropometric indices, measurements of weight, length, head circumference and mid-upper-arm-circumference were transformed to corresponding WHO child growth standards for weight-for-age, length-for-age, weight-for-length, head-circumference-for-age and mid-upper-arm-circumference-for-age Z-scores. These WHO child growth standards are gender specific and were calculated according to the exact day of infant age at the time of anthropometric measurements. WHO child growth standard Z-scores were calculated in R using the WHO igrowup_standard function (07/10/2013) written for R (208). Individual associations with variables of interest and the outcome groups (those with at least one infectious cause hospitalization or death compared to those without an infectious cause hospitalization or death) as well as individual associations with variables of interest and the exposure groups (HEU and HUU) were evaluated using the Student’s 2-sample t-test for normally distributed numeric variables, Wilcoxon rank sum test for non-normally distributed numeric variables, Pearson’s Chi-squared test for categorical variables that met the appropriate assumptions and Fisher’s exact test if assumptions for the Chi-squared test were violated. A two-sided alpha level of 0.05 was set as the limit of statistical significance and a p-value of less than 0.05 interpreted as statistically significant without correction for multiple comparisons. 64 The incidence rate of sick-clinic visits was calculated including all sick-clinic visits recorded in the numerator over the number of days of infant follow-up directly observed until the last study visit in the denominator. The risk ratio for the primary outcome of at least one infectious cause hospitalization or death was calculated with each infant that had at least one primary outcome event included only once in the numerator over all infants in the analytic cohort including those lost to face-to-face follow-up in the denominator. The age at the time of primary outcome was explored graphically with Kaplan-Meier curves. 5.4.2 Analysis of primary objective Primary comparison: Odds ratio adjusted for breastfeeding exposure of at least one infectious cause hospitalization or death before six months of age in all HEU infants compared to all HUU infants. For this comparison logistic regression was chosen to determine adjusted odds ratios (OR) and their 95% confidence intervals. All infants with at least one infectious cause hospitalization of any grade (mild-moderate, severe and very severe) were included in the primary outcome. Although there were infants lost to face-to-face follow-up during the six months of observation, it was possible to determine the outcome of all infants including those lost to follow-up through the province wide electronic hospital administration system and mortality registry. As such, logistic regression was deemed an appropriate method for calculating adjusted estimates of association for the primary comparison. 5.4.2.1 Control of confounding Consideration of variables as confounders was based on existing evidence in the literature, the presence of an association between the potential confounding variable with the exposure of interest (HIV exposure group) and the outcome, and the alteration of the direction or magnitude of the OR adjusted for the potential confounder relative to the unadjusted OR. The most important confounders of the relationship between in-utero HIV exposure and infectious morbidity considered a priori were adverse birth outcomes (preterm birth, low birth weight), breastfeeding exposure and socio-economic circumstances (Figure 5.1). Breastfeeding exposure can also be considered to be on the causal pathway between HIV exposure and infectious morbidity. However as the primary objective of this study was to determine whether an association exists between in-utero HIV exposure and infectious morbidity through pathways other than those related to breastfeeding exposure, it was deemed appropriate to control for breastfeeding exposure for this specific hypothesis. Control for socio-economic differences, preterm birth or low birth weight and maternal comorbidities were taken into consideration in the study design, through restriction of eligibility criteria. All maternal demographic, health and obstetric factors, as well as household and infant birth factors were evaluated for confounding. However, as this was a small study, priority was given 65 to controlling for variables established in the literature to be important determinants of infectious morbidity. Figure 5.1 Conceptual framework of the relationship between HIV exposure and infant infectious morbidity Variables considered a priori to be on the causal pathway between HIV exposure and infectious morbidity and thus not considered for adjustment as confounders included the maternal CD4 count, infant anaemia and infant anthropometric indices (weight-for-age, length-for age and weight-for length). Maternal CD4 count was considered to be on the causal pathway as in-utero exposure to a mother with an immune system that is suppressed to varying degrees, roughly inversely proportional to the CD4 count, may result in infant immune system aberrations leading to greater risk for infectious morbidity. Anaemia was considered to be on the causal pathway as HIV exposed infants exposed to ARVs in-utero, particularly ZDV, have been observed to have higher rates of anaemia during infancy and anaemia is a risk factor for infectious morbidity. And impaired infant growth, as measured by infant anthropometric indices, was considered to be on the causal pathway as HIV and ARV exposed infants may have compromised growth in-utero or postnatal, and infants with sub-optimal weight- or length-for-age are at greater risk for infectious morbidity. Various strategies for the most appropriate manner in which to control for breastfeeding exposure were considered. Fixed effects logistic regression was performed adjusting for infant breastfeeding status as breastfed or not breastfed at two weeks (representing ever breastfed or never breastfed respectively) and six months (representing breastfeeding for the entire duration of risk up to six months or breastfeeding for less than the entire duration of risk or not at all) individually in separate models. Nested logistic regression models including and excluding infant breastfeeding status were compared using the likelihood ratio test, Maternal((HIV(HEU((infant(Infec1ous(morbidity(Subop1mal(infant(feeding(Preterm(/(low(birth(weight(Socioeconomic(circumstances(Confounders*HIV(product(exposure(Maternal(immune(compromise(HEU*unique*pathways*66 and a two sided p-value of <0.05 was considered to indicate a significant difference in models. It was decided a priori that adjustment for breastfeeding would be retained on conceptual grounds, irrespective of a significant difference in nested models. Stratification was also used as a strategy to control for breastfeeding exposure. HEU and HUU infants were compared for the primary outcome (infectious cause hospitalization or death in the first six months of life), conditioned on the presence or absence of any breastfeeding at two weeks, and separately conditioned on the presence or absence of any breastfeeding at six months (Figure 5.2). A generalized linear mixed effects logistic regression model was considered. The glmer function in the lme4 package of R was used to evaluate the utility of a mixed effects model. Figure 5.2 Schematic representation of stratified analysis conditioned on the presence of any breastfeeding HEU – HIV exposed uninfected; HUU – HIV unexposed uninfected; OR – odds ratio 5.4.3 Analysis of secondary objective 1 Secondary comparison 1a: Odds ratio for at least one severe infectious cause hospitalization or death in HEU infants compared to HUU infants adjusted for breastfeeding exposure Secondary comparison 1b: Odds ratio for at least one very severe infectious cause hospitalization or death in HEU infants compared to HUU infants adjusted for breastfeeding exposure The secondary comparisons between HEU and HUU infants for at least one severe infectious cause hospitalization or death or at least one very severe infectious cause hospitalization or death in the first six months of life were evaluated in the same manner as the primary comparison described above. The outcome “severe infectious cause hospitalization or death” is inclusive of events classified as “very severe”. The outcome “very severe infectious cause hospitalization or death” includes only those events classified as “very severe”. Infant breastfeeding status at two weeks and six months was controlled for in Any breastfeeding HEU infants All Infants OR for infectious cause hospitalization or death breastfed HEU infants vs. breastfed HUU infants OR for infectious cause hospitalization or death not breastfed HEU infants vs. not breastfed HUU infants No breastfeeding HUU infants HEU infants HUU infants 67 separate models and a stratified analysis conditioned on breastfeeding status at two weeks and six months was performed. 5.4.4 Analysis of secondary objective 2 Secondary comparison 2: Odds ratio for at least one infectious cause hospitalization or death before six months of age comparing HEU infants born to mothers on maternally indicated cART during pregnancy and HEU infants born to mothers on VTP prophylaxis. The objective of this subgroup analysis was not to understand the effects of the cART or VTP ARV drugs on HEU infants. Rather, it was to understand the role of the extent of maternal HIV disease in terms of current or prior immune suppression and whether the association between advanced maternal HIV and increased infant morbidity and mortality observed prior to the availability of antiretroviral therapy still exists in the era of maternally indicated cART. During the time of the study, South African adults, including pregnant women, were only eligible for initiation of cART if they had a CD4 count of below 350 cells/µl or WHO stage 3/4 disease. All pregnant women not already on cART and with a CD4 count of greater than 350 cells/µl were given VTP prophylaxis for the duration of pregnancy. This meant that mothers on cART had at some stage experienced severe immune suppression as opposed to mothers on VTP prophylaxis who had not yet been severely immune suppressed. The maternal antenatal CD4 count was thus directly associated with receipt of either cART or VTP prophylaxis and thus also directly associated with the potential of having a detectable or undetectable HIV viral load. In this context the effect of current or prior severe immune suppression and the effect of cART cannot be disentangled and, as such, infant groups for this sub-group analysis were defined according to maternally indicated cART or maternal VTP prophylaxis during pregnancy. Pregnant women that received no ARVs at all during pregnancy and pregnant women with antenatal CD4 counts below 350 cells/µl, indicating maternal immune suppression, but only receiving VTP prophylaxis were excluded from this subgroup analysis. The association between maternal ARV regimen during pregnancy and the primary outcome, at least one infectious cause hospitalization or death in the first six months of life, was evaluated with logistic regression multivariable models to determine adjusted odds ratios. This was performed in the same way as described for the primary comparison of HEU and HUU infants. In addition to the a priori confounders considered for the primary comparison of socioeconomic status and breastfeeding exposure, timing of HIV diagnosis was also considered as a potential confounder in this sub-group analysis. 68 Chapter 6 Results - Do HEU infants have a greater probability of infectious morbidity than peer HUU infants? In this chapter the results of the primary and first secondary objective are presented. Firstly the study cohort is described in comparison to the study population and the disposition of study participants over the study duration is presented in section 6.1. The Bivariable analysis of maternal, household and infant characteristics is presented compared by the exposure groups (HEU and HUU infants) in section 6.2 and subsequently compared by infants with and without the primary outcome of at least one infectious cause hospitalization or death in section 6.3. The hospitalizations are characterized in further detail in section 6.4 and finally the multivariable analysis to determine the effect of HIV exposure on the odds of infectious cause hospitalization or death is presented in section 6.5. Maternal HIV specific detail and the results of the HEU infant subgroup analysis for the secondary objective 2 are presented in Chapter 7. 6.1 Cohort background During the study enrolment period from 15 June 2012 to 30 June 2013, 1384 deliveries occurred at Kraaifontein Midwife Obstetric Unit (MOU); 363 (26.2%) deliveries in HIV-infected women. Two hundred and sixty four (19.1%) mothers gave informed consent and were enrolled at delivery, 136 HIV-infected mothers and 128 HIV-uninfected mothers and their newborns. The age of mothers in the study sample and the MOU population did not differ significantly. The proportion of HIV-infected mothers was higher in the study sample (51.5%, 136/264) compared to the MOU population (26.2%, 363/1384), as was expected by the study design. Of the HIV-infected mothers, 36.4% (132/363) of the MOU population and 54.4% (74/136) of the study sample were receiving maternally indicated cART during pregnancy. There was no difference in the MOU population and study sample in the proportion of HIV-infected mothers intending to exclusively breastfeed (45.2% and 42.1% respectively). Feeding intention is not routinely collected at the MOU on HIV-uninfected mothers. One hundred and seventy eight (67.4%) mothers and infants returned at two weeks of age. Two of these infants were found to be HIV-infected, one at the two week visit and the second at the two month visit. These infants were withdrawn from the study and are excluded from all analysis, resulting in 176 mothers and their confirmed HIV-uninfected infants in the analytic cohort. There were no significant differences in available baseline characteristics between mothers and infants who returned at two weeks of age and those who did not (Table 6.1). Of 176 mother-infant pairs eligible for this analysis (94 HEU and 82 HUU), 134 (76.1%) were followed up until six months of age, 75 (80.0%) HEU infants and 59 (72.0%) HUU 69 infants. The disposition of study participants over the duration of the study is shown in Figure 6.1 and the reasons for non-completion of six-month follow-up are shown in Table 6.2. Table 6.1 Baseline maternal and infant characteristics compared by mother-infant pairs retained in the cohort at two weeks and those lost before two weeks of infant age (numeric variables as median (IQR), categorical variables as number (%), unless otherwise indicated) Total N=264 Retained at 2 weeks N=178 Lost before 2 weeks N=86 P-value Maternal Characteristics Age 26.8(23.3,30.4) 26.8(23.3,30.5) 26.8(23.6,29.8) 0.84 Black African race 241(91.3) 162(91.0) 79(91.9) 1.00 Xhosa or Zulu speaking 205(77.7) 141(79.2) 64(74.4) 0.23 Never married 177(67.0) 119(66.9) 58(67.4) 0.28 Completed secondary education 86(32.6) 63(35.4) 23(26.7) 0.21 Primiparous 62(23.5) 41(23.0) 21(24.4) 0.93 Gestation at 1st antenatal visit (weeks) 21(17,27) 20(16,27) 22(19,28) 0.29 Number of antenatal visits 5(4,6) 5(4,6) 4(3,5) 0.07 Intention to breastfeed 181(68.6) 118(66.3) 63(73.3) 0.32 Delivery CD4 count (cells/µl) 409(308,588) 409(303,592) 419(330,555) 0.81 HIV-infected mothers 136(51.5) 96(53.9) 40(46.5) 0.29 Antenatal CD4 count (cells/µl) 421(294,537) 422(285,538) 409(299,516) 0.93 Pregnancy ARV regimen: (N = 136) 0.11 cART 74(54.4) 47(49.0) 27(67.5) VTP 58(42.6) 45(46.9) 13(32.5) None 4(2.9) 4(4.2) 0(0.0) Infant Birth Characteristics Male 130(49.2) 85(47.8) 45(52.3) 0.57 Gestational age – mean (SD) 38.9(1.6) 38.9(1.6) 39.0(1.6) 0.75 Weight in grams – mean (SD) 3170(411) 3165(413) 3181(411) 0.78 Low birth weight (<2500g) 12(4.5) 10(5.6) 2(2.3) 0.35 Length in cm – mean (SD) 49.0(3.5) 49.0(3.6) 49.0(3.2) 0.98 Infant outcomes HIV infection 4(2.9) 2(2.1) 2(5.0) 0.60 All-cause hospitalization 35(13.3) 28(15.7) 7(8.1) 0.12 Death 1(0.4) 1(0.6) 0(0.0) 1.00 cART – combination antiretroviral therapy; SD – standard deviation; VTP – vertical transmission prevention prophylaxis 70 Figure 6.1 Disposition of participants over the duration of study Delivery: 264 mother-infant pairs (136 HIV exposed, 128 HIV unexposed) 2 weeks: 176 mother-infant pairs (94 HEU infants, 82 HUU infants) 2 months: 169 mother-infant pairs (92 HEU infants, 77 HUU infants) 4 months: 156 mother-infant pairs (86 HEU infants, 70 HUU infants) 6 months: 134 mother-infant pairs (75 HEU infants, 59 HUU infants) 2 HEU infants withdrawn (2 lost to follow-up) 5 HUU infants withdrawn (1 voluntary withdrawal, 1 moved away, 3 lost to follow-up) 1 HUU infant death 6 HEU infants withdrawn (1 voluntary withdrawal, 3 moved away, 2 lost to follow-up) 6 HUU infants withdrawn (4 voluntary withdrawal, 2 lost to follow-up) 11 HEU infants withdrawn (2 voluntary withdrawal, 3 moved away, 6 lost to follow-up) 11 HUU infants withdrawn (3 moved away, 8 lost to follow-up) 4 HIV-infected infants withdrawn 38 HEU infants withdrawn (12 voluntary withdrawal, 5 moved away, 21 lost) 46 HUU infants withdrawn (14 voluntary withdrawal, 6 moved away, 26 lost) 71 Table 6.2 Infant reasons for non-completion of 6-month follow-up (all variables are number (%)) Total N=176 HEU N=94 HUU N=82 Total lost to follow-up by 6 months of age* 42(23.9) 19(20.2) 23(28.0) Reasons: Death 1(0.6) 0(0.0) 1(1.3) Voluntary participant withdrawal 8(4.5) 3(3.2) 5(6.1) Family relocation 10(5.7) 6(6.4) 4(4.9) Unable to contact 23(13.1) 10(10.6) 13(15.9) HEU – HIV exposed uninfected; HUU – HIV unexposed uninfected; * 2 additional HEU infants missed the window for data collection for the 6 month visit but were not lost to follow-up 72 6.2 Comparison of HEU and HUU infants Ninety-four HEU infants and 82 HUU infants who returned at two weeks of age and were confirmed HIV-uninfected were followed up to six months of age. These infants are compared in terms of maternal demographic, socioeconomic and obstetric factors, household factors and the following infant factors: birth outcomes, healthcare characteristics, infant feeding, anaemia and anthropometric measurements. 6.2.1 Maternal characteristics Maternal demographic, obstetric and health characteristics are shown in Table 6.3. HIV-infected mothers were significantly older than HIV-uninfected mothers, with median age at delivery of 27.8 versus 24.7 years respectively (p = 0.008). Of the total cohort 91% of mothers were of black African race, 80% were isiXhosa or isiZulu speaking, and two thirds were never married with no difference in these characteristics between HIV-infected and uninfected mothers. The majority (94.9%) of mothers had completed school beyond primary education, but only 35.8% had completed secondary education. The median monthly maternal income prior to delivery was 1060 South African Rand (ZAR) (approximately 106 Canadian dollars). Education and income did not differ significantly between HIV-infected and uninfected mothers. HIV-infected mothers were less often primiparous than HIV-uninfected mothers (17.0% vs. 30.5%, p = 0.05) and had on average one more antenatal clinic visit than HIV-uninfected mothers (5 vs. 4, p = 0.03). The mothers in the cohort had a median gestational age at first antenatal clinic visit of 20 (interquartile range (IQR) 16,27) weeks that did not differ by HIV infection status. There were no mothers with gestational hypertension, gestational diabetes or antepartum or post-partum haemorrhage. Only two mothers tested positive for syphilis during pregnancy, one HIV-infected and one HIV-uninfected mother. Six mothers were treated for TB during pregnancy, four HIV-infected and two HIV-uninfected mothers, and two mothers, one in each group were treated for TB postnatally. There tended to be a greater proportion of HIV-infected mothers who smoked at any stage during pregnancy than HIV-uninfected mothers (11.7% (11/94) vs. 3.7% (3/82), p = 0.06), but there was no statistical difference in the proportion ever using alcohol during pregnancy (17.0% (16/94) vs. 11.0% (9/82), p = 0.32). Two HIV-infected mothers and no HIV-uninfected mothers reported use of illegal drugs while pregnant. There was no difference in previous stillbirths (6.4% (5/78) in multiparous HIV-infected mothers vs. 5.3% (3/57) in multiparous HIV-uninfected mothers, p = 1.0), or deaths in previous children under 5 years of age (7.7% (6/78) in multiparous HIV-infected mothers vs. 3.5% (2/57) in multiparous HIV-uninfected mothers, p = 0.47). The maternal CD4 absolute count and CD4 percent measured at delivery were both significantly lower in HIV-infected than HIV-uninfected mothers, with the majority (52.1% (49/94)) of HIV-infected mothers having CD4 counts below 350 cells/µl compared to 22.0% (18/82) of HIV-uninfected mothers (p<0.001). Body mass index measured at two weeks postnatal was equivalent in the two groups of mothers and no mothers required hospital admission or died during the six month follow-up period. 73 Table 6.3 Demographic and health characteristics of HIV-infected and HIV-uninfected mothers (numeric variables as median (IQR), categorical variables as number (%)) Total N=176 HIV infected N=94 HIV uninfected N=82 P-value Demographic characteristics Age 26.8(23.3,30.4) 27.8(23.8,31.1) 24.7(21.8,29.7) 0.008 Race: 0.78 Black African 161(91.5) 87(92.6) 74(90.2) Coloured 15(8.5) 7(7.4) 8(9.8) Language: 0.51 Afrikaans 18(10.2) 10(10.6) 8(9.8) English 6(3.4) 4(4.3) 2(2.4) isiXhosa or isiZulu 140(79.5) 76(80.9) 64(78.0) Other 12(6.8) 4(4.3) 8(9.8) Marital Status: 0.38 Never married 117(66.5) 64(68.1) 53(64.6) Married 53(30.1) 25(26.6) 28(34.1) Widow/Separated/Divorced 6(3.4) 5(5.3) 1(1.2) Education: 0.10 No or any primary 9(5.1) 6(6.4) 3(3.7) Some secondary 104(59.1) 61(64.9) 43(52.4) Completed secondary 63(35.8) 27(28.7) 36(43.9) Monthly income in ZAR 1060(285,2265) 1060(280,2290) 1040(385,2180) 0.79 Health characteristics Primiparous 41(23.3) 16(17.0) 25(30.5) 0.05 Gestation at 1st antenatal visit (weeks) 20(16,27) 20(16,26) 20(17,28) 0.19 Number of antenatal visits 5(4,6) 5(4,6) 4(3,5) 0.03 Postnatal BMI (kg/m2) 26.6(23.2,29.2) 26.6(23.1,28.9) 26.5(23.6,29.5) 0.42 Maternal CD4 at delivery Absolute count (cells/µl) 409(303,592) 343(235,501) 467(363,675) <0.001 Percent 33.4(24.5,40.6) 26.1(21.1,32.4) 39.3(36.2,45.9) <0.001 Categorized (cells/µl) <0.001 <350 67(38.1) 49(52.1) 18(22.0) 350-499 51(29.0) 21(22.3) 30(36.6) > 500 58(33.0) 24(25.5) 34(41.5) BMI – body mass index; ZAR – South African Rand 74 6.2.2 Household characteristics Characteristics of households in which the infants were resident are shown in Table 6.4. More than half (52.8% (93/176)) of infants’ were living in informal housing and 99.4% (175/176) had access to a flush toilet connected to sewerage, but only 47.2% (83/176) had water piped into the dwelling. Although 96.6% (170/176) had access to electricity, other sources of fuel including gas, wood and paraffin were used for cooking and heating. Houses had a median of 2 (IQR1,3) rooms and 4(IQR 3,5) occupants and were a median of 20 (IQR 15,30) minutes walk from the nearest primary healthcare clinic. HEU and HUU infant households were similar on all these characteristics as well as in possession of various household assets (Table 6.4). Thirty percent (53/176) of children were cared for during the day by somebody other than their mother during the first six months of life, but only 7.4% (13/176) attended a daycare center outside of the home. Although all were eligible, only 28.4% (50/176) of infants were receiving the government child support grant by six months of age, with no difference between HEU and HUU infants. 75 Table 6.4 Household characteristics compared by infant HIV exposure group (categorical variables as number (%)) Total N=176 HEU N=94 HUU N=82 P-value Type of house: 0.21 Stand alone house 70(39.8) 32(34.0) 38(46.3) Apartment in apartment block 7(4.0) 5(5.3) 2(2.4) House/flat/room in backyard 8(4.5) 4(4.3) 4(4.9) Shack in backyard 56(31.8) 37(39.4) 19(23.2) Shack not in backyard 29(16.5) 14(14.9) 15(18.3) Other 6(3.4) 2(2.1) 4(4.9) Water supply: 0.02 Water piped into dwelling 83(47.2) 39(41.5) 44(53.7) Water piped into yard 80(45.5) 51(54.3) 29(35.4) Public tap 13(7.4) 4(4.3) 9(11.0) Sanitation: 0.35 Flush toilet (with sewerage) 175(99.4) 94(100) 81(98.8) No toilet facilities 1(0.6) 0(0.0) 1(1.2) Fuel (cook): 0.78 Electricity 166(94.3) 88(93.6) 78(95.1) Gas 6(3.4) 3(3.2) 3(3.7) Paraffin 4(2.3) 3(3.2) 1(1.2) Fuel (heat): 0.57 Electricity 61(34.7) 37(39.4) 24(29.3) Gas 3(1.7) 1(1.1) 2(2.4) Paraffin 99(56.3) 49(52.1) 50(61.0) Firewood/Other 13(7.4) 7(7.4) 6(7.3) Fuel (light): 0.22 Electricity 170(96.6) 89(94.7) 81(98.8) Paraffin 6(3.4) 5(5.3) 1(1.2) Assets: Radio 103(58.5) 51(54.3) 52(63.4) 0.28 TV 153(86.9) 78(83.0) 75(91.5) 0.16 Computer 23(13.1) 10(10.6) 13(15.9) 0.42 Refrigerator 133(75.6) 69(73.4) 64(78.0) 0.59 Home phone 5(2.8) 4(4.3) 1(1.2) 0.37 Cell phone 170(96.6) 91(96.8) 79(96.3) 1.00 Bicycle 20(11.4) 11(11.7) 9(11.0) 1.00 Motorcycle or scooter 2(1.1) 2(2.1) 0(0.0) 0.50 Car or truck 15(8.5) 10(10.6) 5(6.1) 0.42 Horse or donkey 1(0.6) 1(1.1) 0(0.0) 1.00 Sheep, goat or cattle 1(0.6) 1(1.1) 0(0.0) 1.00 HEU – HIV exposed uninfected; HUU – HIV unexposed uninfected 76 6.2.3 Infant characteristics Infant characteristics are shown in Table 6.5. In the total cohort 47.7% (84/176) of infants were male with a mean gestational age of 38.9 (standard deviation (SD) 1.5) weeks and birth weight of 3171(SD 409) grams that did not differ significantly between HEU and HUU infants. Five percent (9/176) of infants were of low birth weight (<2500g) and 7.6% (10/132) of infants did not have complete immunizations by six months of age. There was no significant difference in TB exposure or treatment between HEU and HUU infants, seven HEU compared to two HUU infants reported a TB contact at some stage in the first six months of life and four HEU but no HUU infants had received TB treatment by six months of age. There was no difference in the number of all cause or sick clinic visits between HIV exposure groups with a median of 6 (IQR 5,7) all cause visits and 1 (IQR 0,1) sick-clinic visit per infant in the first six months of life (Table 6.5). Eighty-three infants had no sick clinic visits and 93 infants (47 HEU and 46 HUU) had a total of 140 sick-clinic visits with 70 visits in each HIV exposure group. The incidence rate of all cause sick clinic visits did not differ between HEU and HUU infants, 0.47 (95% CI 0.37, 0.59) per 100 infant days in HEU infants and 0.57 (95% CI 0.45, 0.72) per 100 infant days in HUU infants, for an incidence rate ratio of 0.82 (95% CI 0.58,1.16) in HEU relative to HUU infants. The incidence rate for respiratory symptom associated sick clinic visits was 0.27 (95% CI 0.20, 0.36) and 0.36 (95% CI 0.26 ,0.48) per 100 infant days in HEU and HUU infants respectively for a rate ratio of 0.75 (95% CI 0.48, 1.18). The incidence rate for diarrhoea associated sick clinic visits was also no different between HEU and HUU infants, 0.07 (95% CI 0.04, 0.13) and 0.06 (95% CI 0.03, 0.12) respectively for a rate ratio of 1.13 (95% CI 0.42, 3.24). The majority of visits were for respiratory tract symptoms, 57.1% (40/70) of HEU infant visits and 62.9% (44/70) of HUU infant visits. Sick-clinic visits for respiratory symptoms were more common than those for diarrhoeal symptoms. Thirty one percent (55/176) of all infants had at least one sick-clinic visit associated with respiratory tract symptoms compared to only 9.7% (17/176) of infants who had least one diarrhoeal associated sick-clinic visit. 77 Table 6.5 Infant birth and healthcare characteristics compared by infant HIV exposure group (numeric variables as median (IQR), categorical variables as number (%), unless otherwise stated) Total N=176 HEU N=94 HUU N=82 P-value Male 84(47.7) 46(48.9) 38(46.3) 0.85 Gestational age – mean (SD) 38.91(1.5) 38.7(1.5) 39.1(1.5) 0.06 Birth weight in grams – mean (SD) 3171(409) 3118(375) 3231(440) 0.07 Low birth weight 9(5.1) 6(6.4) 3(3.7) 0.51 Immunizations not up to date at 6 months (N=132) 10(7.6) 5(6.9) 5(8.5) 0.91 All-cause clinic visits per infant 6(5,7) 6(5,7) 6(5,7) 0.78 Sick clinic visits per infant 1(0,1) 1(0,1) 1(0,2) 0.12 Infants with at least one sick clinic visit: All cause 93(52.8) 47(50.0) 46(56.1) 0.43 Diarrhoea 17(9.7) 11(11.7) 6(7.3) 0.53 Respiratory 55(31.3) 26(27.7) 29(35.4) 0.88 Fever 8(4.5) 4(4.3) 4(4.9) 1.00 Total number of sick clinic visits: All cause 140 70 70 Diarrhoea 19(13.6) 11(15.7) 8(11.4) 0.24 Respiratory 84(60.0) 40(57.1) 44(62.9) 0.27 Fever 8(5.7) 4(5.7) 4(5.7) 1.00 HEU – HIV exposed uninfected; HUU – HIV unexposed uninfected; SD – standard deviation 78 Infant feeding patterns, shown in Table 6.6 were considerably different between the two groups. Almost all (97.6% (80/82)) HIV-uninfected mothers intended at delivery to exclusively breastfeed their babies compared to 39.4% (37/94) of HIV-infected mothers. Only one HUU infant never breastfed compared to 62.8% (59/94) of HEU infants who were never breastfed. The median duration of breastfeeding in infants that were breastfed was no different between HEU and HUU breastfed infants at 112 days (IQR 56,194). Feeding modes, either exclusive breastfeeding, partial breastfeeding or no breast feeding, were significantly different between HEU and HUU infants at all time points from two weeks to six months of age. By six months of age, although only 5% of infants in both groups were still being exclusively breastfed, only 32.2% (19/59) of HUU infants had ceased all breastfeeding compared to 84.9% (62/73) of HEU infants. 79 Table 6.6 Infant feeding characteristics compared by infant HIV exposure group (numeric variables as median (IQR), categorical variables as number (%)) Total N=176 HEU N=94 HUU N=82 P-value Mothers intention to exclusively breastfeed 117(66.5) 37(39.4) 80(97.6) <0.001 Days of any breastmilk in breastfed infants (N=116) 112(56,194) 112(56,194) 112(56,194) 0.18 Last visit any breastfeeding reported: <0.001 Never breastfed 60(34.1) 59(62.8) 1(1.2) 2 weeks 23(13.1) 8(8.5) 15(18.3) 2 months 27(15.3) 9(9.6) 18(22.0) 4 months 15(8.5) 7(7.4) 8(9.8) 6 months 51(29.0) 11(11.7) 40(48.8) Feeding mode at 2 weeks: (N=176) <0.001 Exclusive breastfeeding 99(56.3) 32(34.0) 67(81.7) Partial breastfeeding 17(9.7) 3(3.2) 14(17.1) No breastfeeding 60(34.1) 59(62.8) 1(1.2) Feeding mode at 2 months: (N=169) <0.001 Exclusive breastfeeding 71(42.0) 23(25.0) 48(62.3) Partial breastfeeding 22(13.0) 4(4.3) 18(23.4) No breastfeeding 76(45.0) 65(70.7) 11(14.3) Feeding mode at 4 months: (N=153) <0.001 Exclusive breastfeeding 33(21.6) 11(13.1) 22(31.9) Partial breastfeeding 33(21.6) 7(8.3) 26(37.7) No breastfeeding 87(56.9) 66(78.6) 21(30.4) Feeding mode at 6 months: (N=132) <0.001 Exclusive breastfeeding 7(5.3) 4(5.5) 3(5.1) Partial breastfeeding 44(33.3) 7(9.6) 37(62.7) No breastfeeding 81(61.4) 62(84.9) 19(32.2) HEU – HIV exposed uninfected; HUU – HIV unexposed uninfected, 80 HEU infants had significantly lower mean haemoglobin (Hb) than HUU infants at two months of age (10.5 (SD 0.9) g/dl vs. 10.8 (SD1.1) g/dl, p = 0.02), but this had resolved by six months of age. Although the proportion of infants with a grade 1 or more anaemia in the cohort was high at approximately 40% from two months to six months of age, this did not differ between HEU and HUU infants (Table 6.7). Considering infant growth parameters, shown in Table 6.8, HEU infants had significantly lower WHO mean weight-for-age Z-scores at two weeks and two months (Figure 6.2) and significantly lower length-for-age Z-scores (Figure 6.3) at birth, two weeks, two months and six months than HUU infants. There was no difference between HIV exposure groups though in mean weight-for-length Z-score (Figure 6.4), head circumference Z-score, mid-upper-arm-circumference Z-score or the proportion of infants that were stunted (length-for-age Z-score < -2) or wasted (weight-for-length Z-score < -2). 81 Table 6.7 Infant haemoglobin and anaemia compared by infant HIV exposure group (numeric variables as mean (SD), categorical variables as number (%)) Total N=176 HEU N=94 HUU N=82 P-value Mean haemoglobin (SD) 2 weeks (N=167) 14.3(1.9) 14.2(2.0) 14.4(1.8) 0.61 2 months (N=150) 10.6(1.0) 10.5(0.9) 10.8(1.1) 0.02 4 months (N=141) 11.0(0.9) 11.1(1.0) 10.8(0.9) 0.17 6 months (N=127) 11.2(1.0) 11.2(0.9) 11.1(1.0) 0.58 DAIDS anaemia grade 2 weeks: (N=167) 0.43 Normal 137(82.0) 71(81.6) 66(82.5) Grade 1 17(10.2) 8(9.2) 9(11.3) Grade 2 13(7.8) 8(9.2) 5(6.3) 2 months: (N=150) 0.83 Normal 91(60.7) 45(56.3) 46(65.7) Grade 1 41(27.3) 24(30.0) 17(24.3) Grade 2 16(10.7) 10(12.5) 6(8.6) Grade 3 2(1.3) 1(1.3) 1(1.4) 4 months: (N=141) 0.22 Normal 74(52.5) 47(59.5) 27(43.5) Grade 1 49(34.8) 23(29.1) 26(41.9) Grade 2 14(9.9) 7(8.9) 7(11.3) Grade 3 4(2.8) 2(2.5) 2(3.2) 6 months: (N=127) 0.58 Normal 76(59.8) 44(62.0) 32(57.1) Grade 1 38(29.9) 22(31.0) 16(28.6) Grade 2 10(7.9) 4(5.6) 6(10.7) Grade 3 3(2.4) 1(1.4) 2(3.6) DAIDS – Division of AIDS; HEU – HIV exposed uninfected; HUU – HIV unexposed uninfected; SD – standard deviation 82 Table 6.8 Infant anthropometric measurements compared by infant HIV exposure group (numeric variables as mean (SD), categorical variables as number (%)) Anthropometric measurements Total N=176 HEU N=94 HUU N=82 P-value Birth: (N=176) WAZ -0.28(0.88) -0.40(0.83) -0.17(0.92) 0.08 LAZ -0.22(1.88) -0.57(1.87) 0.18(1.88) 0.008 WLZ -0.39(1.79) -0.24(1.72) -0.56(1.87) 0.25 HCZ -0.10(1.17) -0.16(1.24) -0.03(1.09) 0.47 Stunted 28(15.9) 17(18.1) 11(13.4) 0.52 Wasted 26(14.8) 10(10.6) 16(19.5) 0.17 2 weeks: (N=176) WAZ -0.18(0.92) -0.40(0.81) 0.06(0.97) <0.001 LAZ -0.98(1.03) -1.19(0.93) -0.74(1.10) 0.004 WLZ 0.71(1.03) 0.65(0.98) 0.78(1.09) 0.41 HCZ 0.21(1.09) 0.10(1.08) 0.33(1.09) 0.15 Stunted 28(15.9) 15(16.0) 13(15.9) 1.00 Wasted 1(0.6) 0(0.0) 1(1.2) 0.47 2 months: (N=162) WAZ -0.10(1.10) -0.34(1.03) 0.17(1.12) 0.003 LAZ -1.03(1.06) -1.19(1.04) -0.84(1.06) 0.04 WLZ 1.23(1.12) 1.10(1.12) 1.38(1.12) 0.11 HCZ 0.42(1.11) 0.28(1.15) 0.59(1.04) 0.07 Stunted 29(17.9) 18(20.7) 11(14.7) 0.43 Wasted 0(0.0) 0(0.0) 0(0.0) NA 4 months: (N=146) WAZ 0.00(1.12) -0.09(1.14) 0.12(1.10) 0.25 LAZ -0.79(1.14) -0.91(1.40) - 0.65(1.13) 0.18 WLZ 0.86(1.14) 0.85(1.26) 0.88(0.98) 0.87 HCZ 0.63(1.03) 0.57(1.06) 0.70(1.00) 0.48 MUACZ 0.37(1.02) 0.28(1.08) 0.50(0.93) 0.18 Stunted 23(15.8) 14(17.3) 9(13.9) 0.74 Wasted 0(0.0) 0(0.0) 0(0.0) NA 6 months: (N=132) WAZ 0.18(1.19) 0.09(1.17) 0.29(1.23) 0.35 LAZ -0.53(1.03) -0.71(0.93) -0.31(1.10) 0.03 WLZ 0.75(1.29) 0.76(1.30) 0.72(1.28) 0.89 HCZ 0.76(1.12) 0.78(1.18) 0.74(1.06) 0.84 MUACZ 0.48(1.07) 0.38(1.09) 0.59(1.04) 0.27 Stunted 12(9.1) 8(11.0) 4(6.8) 0.60 Wasted 1(0.8) 1(1.37) 0(0.0) 1.00 HEU – HIV exposed uninfected; HUU – HIV unexposed uninfected; HCZ – head-circumference Z-score; LAZ – length-for-age Z-score; MUACZ – mid-upper-arm-circumference Z-score; WAZ – weight-for-age Z-score; WLZ – weight-for-length Z-score 83 Figure 6.2 WHO weight-for-age Z-score compared by infant HIV exposure group Figure 6.3 WHO length-for-age Z-score compared by infant HIV exposure group 0 1 2 3 4 5 6−2−1012Age in monthsZ−scoreHEUHUU* ** = P<0.050 1 2 3 4 5 6−2−1012Age in monthsZ−scoreHEUHUU** *** = P<0.0584 Figure 6.4 WHO weight-for-length Z-score compared by infant HIV exposure group 0 1 2 3 4 5 6−2−1012Age in monthsZ−scoreHEUHUU85 6.3 Comparison of infants with and without the primary outcome by HIV exposure group Twenty-seven (15.3%) infants experienced the primary outcome of at least one infectious cause hospitalization or death in the first 6 months of life, 17 (18.1%) HEU infants and 10 (12.2%) HUU infants (p = 0.38). Twenty-six of the outcomes were infectious cause hospitalizations and there was one death (without hospitalization) in an HUU infant. A total of 33 hospitalizations were observed in 28 infants up to six months of age, 21 hospitalizations in 18 HEU infants and 12 hospitalizations in 10 HUU infants. Of the 33 hospitalizations 27 were for infectious causes, 17 in 17 (18.1%) HEU infants and 10 in 9 (11.0%) HUU infants. One HUU infant had a second infectious cause hospitalization at 169 days of age. It was possible to re-establish telephonic contact with 24 (9 HEU and 15 HUU infants) of the 41 infants not maintained in follow-up to six months, all of whom were alive and healthy at six months of age. Of the 17 infants with whom telephonic contact was not re-established, three HEU infants were identified through the provincial electronic health administrative system as being hospitalized. The remaining 7 HEU and 7 HUU infants were not linked to any hospitalizations or deaths in the Western Cape Province hospital administrative system or the mortality registry. In our cohort of 176 infants we expected 1.4 infant deaths up to 12 months of age according to our estimate of infant mortality of 7.9/1000; that is in keeping with the single identified death up to six months of age. 6.3.1 Maternal characteristics Comparing maternal factors between infants with and without infectious cause hospitalization or death (Table 6.9) there tended to be a difference in median maternal age with mothers of infants with the outcome on average being younger than mothers of infants without the outcome (24.7 (IQR 21.0, 29.3) years vs. 27.0 (IQR 23.4, 30.6) years, p = 0.19). Maternal monthly income prior to delivery was lower in the group with an infectious cause hospitalization or death but not statistically significantly different to the group without an infectious cause hospitalization or death (ZAR 650 (IQR 265, 1900) vs. ZAR 1110 (IQR 373, 2292), p = 0.36). The highest level of maternal education did not differ between outcome groups, nor did gestational age at first antenatal clinic visit or the number of antenatal clinic visits attended. There was one mother in each group of infants with and without infectious cause hospitalization or death that tested positive and was treated for syphilis during antenatal screening. No mothers of infants in the outcome group had a diagnosis of TB during pregnancy or postnatal, compared to 8 (5.4%) mothers of infants without the outcome, 6 that received TB treatment during pregnancy and 2 that received treatment postnatal. There was no difference in reported smoking (11.1% (3/27) vs. 7.4% (11/149), p = 0.6), alcohol (14.8% (4/27) vs. 14.1% (21/149), p = 0.12) or drug use (0.0% (0/27) vs. 1.3% (2/149), p = 1.0) during pregnancy by mothers of infants with and without infectious cause hospitalization or death respectively. Neither was there a difference in previous stillbirths (4.3% (1/23) vs. 7.1% (8/112), p = 1.0) or previous 86 child deaths under 5 years of age (4.3% (1/23) vs. 6.3% (7/112), p = 1.0) in multiparous mothers of infants with and without the outcome respectively. The proportion of infants born to mothers with a CD4 count of <350 cells/µl was equivalent in infants with and without the primary outcome (37.0% (10/27) vs. 38.3% (57/149), p = 1.0). Although 22.0% (18/82) of HIV-uninfected mothers had CD4 counts of below 350 cells/µl, no primary outcome events occurred in infants of these mothers. 6.3.2 Household characteristics The circumstances of the households in which the infants resided were similar in both groups of infants with and without an infectious cause hospitalization or death, specifically in terms of the type of house, its size and number of occupants, water and sanitation infrastructure as well as fuel sources available (Table 6.10). The only household asset that differed significantly between the two groups was possession of a cell phone in the household, 11.1% (3/27) of infants with the outcome were living in households without a cell phone compared to 2.0% (3/149) of infants without the outcome (p = 0.05). 87 Table 6.9 Maternal demographic and health characteristics compared by infant outcome group (Primary outcome is at least one infectious cause hospitalization or death before 195 days old; numeric variables as median (IQR), categorical variables as number (%)) Total N=176 Infants with the outcome N=27 Infants without the outcome N=149 P-value Demographic characteristics Age 26.8(23.8,30.4) 24.7(21.0,29.3) 27.0(23.4,30.6) 0.19 Race: 0.71 Black African 161(91.5) 24(88.9) 137(91.9) Coloured 15(8.5) 3(11.1) 12(8.1) Education: 0.82 No or any primary 9(5.1) 1(3.7) 8(5.4) Some secondary 104(59.1) 18(66.7) 86(57.7) Completed secondary 63(35.8) 8(29.6) 55(36.9) Monthly income (ZAR) 1060(285,2265) 650(265,1900) 1110(373,2292) 0.36 Health characteristics Primiparous 41(23.3) 4(14.8) 37(24.8) 0.38 Gestation 1st antenatal visit (weeks) 20(16,27) 20(16,27) 20(16,27) 0.72 Number of antenatal visits 5(4,6) 5(4,6) 5(4,6) 0.74 BMI postnatal (N=173) 26.6(23.2,29.2) 26.1(23.3,29.6) 26.6(23.2,29.0) 0.93 Maternal delivery CD4 count Absolute count (cells/µl) 409(303,592) 502(265,843) 408(304,562) 0.23 Percent 33.4(24.5,40.6) 34.7(27.0,39.3) 32.7(24.0,40.7) 0.85 CD4 categorized: 0.03 <350 67(38.1) 10(37.0) 57(38.3) 350-499 51(29.0) 3(11.1) 48(32.2) > 500 58(33.0) 14(51.9) 44(29.5) ZAR – South African Rand 88 Table 6.10 Household characteristics compared by infant outcome group (Primary outcome is at least one infectious cause hospitalization or death before 195 days old; numeric variables as median (IQR), categorical variables as number (%)) Total N=176 Infants with the outcome N=27 Infants without the outcome N=149 P-value Type of house: 0.75 Stand alone house 70(39.8) 9(33.3) 61(40.9) Apartment in apartment block 7(4.0) 1(3.7) 6(4.0) House/flat/room in backyard 8(4.5) 0(0.0) 8(5.4) Shack in backyard 56(31.8) 11(40.7) 45(30.2) Shack not in backyard 29(16.5) 5(18.5) 24(16.1) Other 6(3.4) 1(3.7) 3(2.0) Water supply: 1.00 Water piped into dwelling 83(47.2) 13(48.1) 70(47.0) Water piped into yard 80(45.5) 12(44.4) 68(45.6) Public tap 13(7.4) 2(7.4) 11(7.4) Sanitation: 1.00 Flush toilet (with sewerage) 175(99.4) 27(100) 148(99.3) No toilet facilities 1(0.6) 0(0.0) 1(0.7) Fuel (cook): 0.79 Electricity 166(94.3) 27(100) 139(93.3) Gas 6(3.4) 0(0.0) 6(4.0) Paraffin 4(2.3) 0(0.0) 4(2.7) Fuel (heat): (3 unknown) 0.95 Electricity 61(35.3) 9(33.3) 52(35.6) Gas 3(1.7) 0(0.0) 3(2.1) Paraffin 99(57.2) 16(59.3) 83(56.8) Firewood/Other 10(5.8) 2(7.4) 8(5.5) Fuel (light): 0.59 Electricity 170(96.6) 27(100) 143(96.0) Paraffin 6(3.4) 0(0.0) 6(4.0) Number of rooms in house 2(1,3) 2(1,3) 2(1,3) 0.43 Number of people in household 4(3,5) 4(3,5) 4(3,5) 0.71 Distance to clinic (minutes walked) 20(15-30) 25(12.5-30) 20(15-30) 0.95 Assets: Radio 103(58.5) 14(51.9) 89(59.7) 0.58 TV 153(86.9) 24(88.9) 129(86.6) 1.00 Computer 23(13.1) 3(11.1) 20(13.4) 1.00 Refrigerator 133(75.6) 20(74.1) 113(75.8) 1.00 Home phone 5(2.8) 1(3.7) 4(2.7) 0.57 Cell phone 170(96.6) 24(88.9) 146(98.0) 0.05 Bicycle 20(11.4) 3(11.1) 17(11.4) 1.00 Motorcycle or motor scooter 2(1.1) 0(0.0) 2(1.3) 1.00 Car or truck 15(8.5) 4(14.8) 11(7.4) 0.25 89 6.3.3 Infant characteristics Infants with and without infectious cause hospitalization or death were of similar mean gestational age (38.6 (SD 1.4) vs. 39.0 (SD 1.5) weeks, p = 0.21) and birth weight (3191 (SD 416) vs. 3167 (SD 0.79) grams, p = 0.79). There was no significant difference in having incomplete immunizations at six months of age (4.2% (1/24) vs. 8.3% (9/108), p = 0.67), nor in TB exposure, TB prophylaxis or TB treatment received (Table 6.11). Forty six percent (68/149) of infants without the outcome of hospitalization or death did have at least one sick clinic visit, the majority of these for respiratory associated symptoms (55.9% (38/68)). Table 6.11 Infant birth and healthcare characteristics compared by infant outcome group (Primary outcome is at least 1 infectious cause hospitalization or death before 195 days old; numeric variables as median (IQR), categorical variables as number (%), unless otherwise stated) Total N=176 Infants with the outcome N=27 Infants without the outcome N=149 P-value HEU Infants 94(53.4) 17(63.0) 77(51.7) 0.38 Male 84(47.7) 16(59.3) 68(45.6) 0.27 Gestational age – mean (SD) 38.9(1.5) 38.6(1.4) 39.0(1.5) 0.21 Birth weight in grams – mean (SD) 3171(409) 3191(416) 3167(409) 0.79 Low birth weight (<2500g) 9(5.1) 1(3.7) 8(5.4) 1.00 Incomplete immunizations at 6 months (N=132) 10(7.6) 1(4.2) 9(8.3) 0.67 TB contact ever 9(5.1) 3(11.1) 6(4.0) 0.14 TB prophylaxis ever 7(4.0) 2(7.4) 5(3.4) 0.29 TB treatment ever 4(2.3) 2(7.4) 2(1.3) 0.11 All cause clinic visits per infant 6(5,7) 7(6,8) 6(5,7) 0.07 Sick clinic visits per infant 1(0,1) 2(1,2.5) 1(0,1) <0.001 Infants with at least one sick clinic visit: All cause 93(52.8) 25(92.6) 68(45.6) <0.001 Diarrhoea 17(9.7) 8(29.6) 9(6.0) 0.001 Respiratory 55(31.3) 17(63.0) 38(25.5) <0.001 Fever 8(4.5) 0(0.0) 8(5.4) 0.61 Total number of sick clinic visits: All cause 140 42 98 Diarrhoea 19(13.6) 8(19.0) 11(11.2) 0.33 Respiratory 84(60.0) 29(69.0) 55(56.1) 0.21 Fever 8(5.7) 0(0.0) 8(8.2) 0.11 HEU – HIV exposed uninfected; SD – standard deviation 90 Two-thirds of mothers intended at birth to exclusively breastfeed their infant’s and this did not differ by outcome group. Although differences in infant feeding did not differ significantly between the outcome groups, it is noticeable that at all time points there were proportionally more infants not breastfed in the group with the outcome compared to the group without the outcome (Table 6.12). Table 6.12 Infant feeding characteristics compared by infant outcome group (Primary outcome is at least 1 infectious cause hospitalization or death before 195 days old; numeric variables as median(IQR), categorical variables as number (%)) Total N=176 Infants with the outcome N=27 Infants without the outcome N=149 P-value Mothers intention to exclusively breastfeed 117(66.5) 17(63.0) 100(67.1) 0.84 Days of any breast milk in breastfed infants (N = 116) 112(56,194) 112(56,194) 112(56,194) 0.17 Last visit any breastfeeding reported: 0.46 Never breastfed 60(34.1) 12(44.4) 48(32.2) 2 weeks 23(13.1) 5(18.5) 18(12.1) 2 months 27(15.3) 4(14.8) 23(15.4) 4 months 15(8.5) 1(3.7) 14(9.4) 6 months 51(29.0) 5(18.5) 46(30.9) Feeding mode at 2 weeks: (N=176) 0.38 Exclusive breastfeeding 99(56.3) 13(48.1) 86(57.7) Partial breastfeeding 17(9.7) 2(7.4) 15(10.1) No breastfeeding 60(34.1) 12(44.4) 48(32.2) Feeding mode at 2 months: (N=169) 0.28 Exclusive breastfeeding 71(42.0) 9(34.6) 62(43.4) Partial breastfeeding 22(13.0) 1(3.8) 21(14.7) No breastfeeding 76(45.0) 16(61.5) 60(42.0) Feeding mode at 4 months: (N=153) 0.31 Exclusive breastfeeding 33(21.6) 4(16.7) 29(22.5) Partial breastfeeding 33(21.6) 2(8.3) 31(24.0) No breastfeeding 87(56.9) 18(75.0) 69(53.5) Feeding mode at 6 months: (N=132) 0.92 Exclusive breastfeeding 7(5.3) 1(5.0) 6(5.4) Partial breastfeeding 44(33.3) 4(20.0) 40(35.7) No breastfeeding 81(61.4) 15(75.0) 66(58.9) 91 Mean haemoglobin was significantly lower in infants with the outcome at four months (10.6 (SD 0.7) vs. 11.0 (SD 1.0) g/dl, p = 0.03), with a non-significantly greater proportion of infants with grade 1 or higher anaemia at four months (65.2% (15/23) vs. 44.1% (52/118), p = 0.08) (Table 6.13). There was no specific pattern to anaemia preceding or following hospitalization events, with 7 of 21 hospitalized infants having a grade 1 or 2 anaemia at the study visit directly preceding hospitalization and 7 of 21 hospitalized infants having a grade 1 anaemia at the study visit following but not preceding hospitalization. Anthropometric indices were not markedly different between infants with and without an infectious cause hospitalization or death (Table 6.14). Table 6.13 Infant haemoglobin and anaemia compared by infant outcome group (Primary outcome is at least 1 infectious cause hospitalization or death before 195 days old; numeric variables as mean (SD), categorical variables as number (%)) Total N=176 Infants with the outcome N=27 Infants without the outcome N=149 P-value Mean haemoglobin (SD) 2 weeks (N=167) 14.3(1.9) 14.2(1.9) 14.3(1.9) 0.86 2 months (N=150) 10.6(1.0) 10.6(0.9) 10.6(1.0) 0.74 4 months (N=141) 11.0(0.9) 10.6(0.7) 11.0(1.0) 0.03 6 months (N=127) 11.2(1.0) 11.0(0.9) 11.2(1.0) 0.24 DAIDS anaemia grade 2 weeks: (N=167) 0.51 Normal 137(82.0) 18(75.0) 119(83.2) Grade 1 17(10.2) 3(12.5) 14(9.8) Grade 2 13(7.8) 3(12.5) 10(7.0) 2 months: (N=150) 0.71 Normal 91(60.7) 15(71.4) 76(58.9) Grade 1 41(27.3) 5(23.8) 36(27.9) Grade 2 16(10.7) 1(4.8) 15(11.6) Grade 3 2(1.3) 0(0.0) 2(1.6) 4 months: (N=141) 0.08 Normal 74(52.5) 8(34.8) 66(55.9) Grade 1 49(34.8) 13(56.5) 36(30.5) Grade 2 14(9.9) 1(4.3) 13(11.0) Grade 3 4(2.8) 1(4.3) 3(2.5) 6 months: (N=127) 0.61 Normal 76(59.8) 9(47.4) 67(62.0) Grade 1 38(29.9) 8(42.1) 30(27.8) Grade 2 10(7.9) 1(5.3) 9(8.3) Grade 3 3(2.4) 1(5.3) 2(1.9) DAIDS – Division of AIDS; SD – standard deviation 92 Table 6.14 Infant anthropometric measurements compared by infant outcome group (Primary outcome is at least one infectious cause hospitalization or death before 195 days old; numeric variables as mean (SD), categorical variables as number (%)) HCZ – head-circumference Z-score; LAZ – length-for-age Z-score; MUACZ – mid-upper-arm-circumference Z-score; WAZ – weight-for-age Z-score; WLZ – weight-for-length Z-score Anthropometric measurements Total N=176 Infants with the outcome N=27 Infants without the outcome N=149 P-value Birth: [N=176] WAZ -0.28(0.88) -0.27(0.88) -0.29(0.88) 0.91 LAZ -0.22(1.88) -0.75(2.00) -0.13(1.85) 0.14 WLZ 0.39(1.79) 0.18(1.80) -0.49(1.78) 0.09 HCZ -0.10(1.17) -0.07(1.06) -0.11(1.19) 0.86 Stunted 28(15.9) 5(18.5) 23(15.4) 0.77 Wasted 26(14.8) 2(7.4) 24(16.1) 0.38 2 weeks: (N=176) WAZ -0.18(0.92) -0.20(0.98) -0.18(0.91) 0.90 LAZ -0.98(1.03) -0.96(0.98) -0.99(1.05) 0.89 WLZ 0.71(1.03) 0.64(1.31) 0.72(0.98) 0.76 HCZ 0.21(1.09) 0.30(0.75) 0.19(1.14) 0.52 Stunted 28(15.9) 3(11.1) 25(16.8) 0.58 Wasted 1(0.6) 1(3.7) 0(0.0) 0.15 2 months: (N=162) WAZ -0.10(1.10) -0.10(1.40) -0.10(1.05) 0.99 LAZ -1.03(1.06) -0.97(1.19) -1.03(1.04) 0.81 WLZ 1.23(1.12) 1.14(1.15) 1.24(1.12) 0.70 HCZ 0.42(1.11) 0.54(1.20) 0.40(1.10) 0.61 Stunted 29(17.9) 2(9.1) 27(19.3) 0.38 Wasted 0(0.0) 0(0.0) 0(0.0) NA 4 months: (N=146) WAZ 0.00(1.12) 0.17(1.44) -0.03(1.06) 0.53 LAZ -0.79(1.14) -0.76(1.26) -0.80(1.12) 0.87 WLZ 0.86(1.14) 1.04(1.29) 0.82(1.11) 0.44 HCZ 0.63(1.03) 0.87(1.03) 0.58(1.03) 0.21 MUACZ 0.37(1.02) 0.38(1.44) 0.37(0.92) 0.99 Stunted 23(15.8) 4(16.7) 19(15.6) 1.00 Wasted 0(0.0) 0(0.0) 0(0.0) NA 6 months: (N=132) WAZ 0.18(1.19) 0.37(1.51) 0.15(1.13) 0.64 LAZ -0.53(1.03) -0.57(1.05) -0.53(1.03) 0.87 WLZ 0.75(1.29) 0.94(1.48) 0.71(1.25) 0.51 HCZ 0.76(1.22) 0.96(1.27) 0.73(1.10) 0.45 MUACZ 0.48(1.07) 0.32(1.70) 0.51(0.92) 0.64 Stunted 12(9.1) 3(15.0) 6(5.4) 0.39 Wasted 1(0.8) 0(0.0) 1(0.9) 1.00 93 6.4 Characterization of hospitalization events Seventeen HEU and eight HUU infants had a single infectious cause hospitalization and one HUU infant had two infectious cause hospitalizations before six months of age. These hospitalizations are presented in detail in section 6.4.1 with regards to the timing of hospitalizations, the type and severity of infectious events and important infant factors (feeding, nutrition and immunizations) at the time of hospitalization. Six infants hospitalized for non-infectious causes are presented briefly in section 6.4.2. 6.4.1 Infectious cause hospitalizations The median age at first infectious cause hospitalization or death was not significantly different in HEU and HUU infants (HEU 48.5 days (IQR 27.5, 135) vs. HUU 49.0 days (IQR 21, 77), p = 0.29). The time to the first infectious cause hospitalization or death was no different between HEU and HUU infants and is shown by Kaplan-Meier curve in Figure 6.5 (log-rank test p = 0.36). The majority of infectious cause hospitalizations occurred in the first three months of life in both groups, 58.8% (10/17) of HEU infant hospitalizations and 90% (9/10) of HUU infant hospitalizations (including one HUU infant death), with no difference at three months of age between the proportions of HEU and HUU infants hospitalized for an infection (Table 6.15). There was a noticeable difference to the pattern of HEU and HUU infant hospitalizations over time (Figure 6.6). Although not statistically significantly different, between 91 and 194 days old an additional seven (7.4%) HEU infants were hospitalized, compared to a single (1.2%) HUU infant (Fishers exact p=0.07). The median length of stay was two days longer in HEU than HUU infants (HEU 6 days (IQR 2,7) vs. HUU 4 days (IQR 2,4), p = 0.09), but not statistically significantly so. 94 Figure 6.5 Kaplan-Meier curves of time to first infectious cause hospitalization or death in HEU and HUU infants Table 6.15 Table of cumulative proportions of infants with infectious cause hospitalizations (including death) over time compared by HIV exposure group (all variables as number (%)) Infant age Total (N=176) HEU (N=94) HUU (N=82) 30 days 9 (5.1) 6 (6.4) 3 (3.7) 60 days 17 (9.7) 11 (11.7) 6 (7.3) 90 days 20 (11.4) 11 (11.7) 9 (11.0) 120 days 23 (13.1) 13 (13.8) 10 (12.2) 150 days 24 (13.6) 14 (14.9) 10 (12.2) 194 days 27 (15.3) 17 (18.1) 10 (12.2) 0 50 100 150 2000.800.850.900.951.00Infant age in daysProportion with hospitalization or deathHUUHEUP = 0.3695 Figure 6.6 Infant age at time of primary outcome compared by HIV exposure group Three (11.5%) of the 26 infants were hospitalized with mild-moderate events, 9 (34.6%) infants had a severe major infectious event and 14 (53.8%) infants had a very severe event (Table 6.16). All 26 infants with an infectious cause hospitalization had at least one event that met DAIDS criteria for a grade 3 event. None of the infants hospitalized required intensive care unit admission. Twenty two of the 26 hospitalized infants experienced a single infectious event during their hospitalization: 13 with lower respiratory tract infections (6 HEU, 7 HUU), five with diarrhoea (4 HEU, 1 HUU), two HEU infants with neonatal sepsis, one HEU infant with disseminated (including pulmonary) TB and one HUU infant with uncomplicated neonatal conjunctivitis. Three HEU infants experienced two infectious events during their hospitalization: two infants with diarrhoea and concurrent lower respiratory tract infection and one infant with post-neonatal Shigella sonnei septicaemia with diarrhoea. One HEU infant had three infectious events during the single hospitalization: neonatal sepsis with concurrent pneumonia and diarrhoea. The HUU infant hospitalized twice had a single lower respiratory tract infection event at both hospitalizations. No HEU infants had repeat hospitalizations before six months of age. The proportion of infants in the total cohort hospitalized at least once with a lower respiratory tract infection did not differ between HEU and HUU infants (10/94 (10.6%) HEU vs. 7/82 (8.5%) HUU, p = 0.83) (Table 6.16). Neither did the type of lower respiratory tract infection differ between HEU and HUU infants comprising six events of pneumonia (three each in HEU and HUU infants), nine events of bronchiolitis (five HEU infants and four HUU infants) and two events of TB (one confirmed and one probable, both in HEU infants). More HEU than HUU infants were hospitalized with a diarrhoeal episode (8/94(8.5%) HEU vs. 1/82(1.2%) HUU, Fishers exact p = 0.04) HUUHEU0 50 100 150 200Infant age in daysInfant group96 Table 6.16 Comparison of HEU and HUU infants by highest grade of severity and type of major infectious events (All variables are number (%)) Total N=176 HEU N=94 HUU N=82 P-value Grade: Mild-moderate 3(1.7) 2(2.1) 1(1.2) 1.00 Severe 9(5.1) 5(5.3) 4(4.9) 1.00 Very severe 14(8.0) 10(10.6) 4(4.9) 0.26 Type: LRTI 17(9.7) 10(10.6) 7(8.5) 0.83 Diarrhoea 9(5.1) 8(8.5) 1(1.2) 0.04 Other* 5(2.8) 4(4.3) 1(1.2) 0.37 HEU – HIV exposed uninfected; HUU – HIV unexposed uninfected; LRTI – lower respiratory tract infection; * Includes 3 HEU infants with presumed neonatal sepsis, 1 HEU infant with post-neonatal Shigella sonnei septicaemia, and 1 HUU infant with uncomplicated neonatal conjunctivitis At the time of hospitalization, three infants (two HEU, one HUU) were wasted (weight-for-length Z-score <-2) and four infants (three HEU, one HUU) were stunted (length-for-age Z-score < -2). Two-thirds (6/9) of HUU infants with an infectious cause hospitalization were breastfeeding at the time of hospitalization, compared to only 23.5% (4/17) of HEU infants. Weaning from breast milk may have directly preceded hospitalization in one HEU infant and two HUU infants. The single HUU infant with diarrhoea was still breastfed at the time of the event, compared to only one of eight HEU infants with diarrhoea still breastfed. The nine infants that experienced a diarrhoeal event all lived in households with access to a flush toilet connected to sewerage and piped water. One HEU infant and three HUU infants did not have confirmed up to date immunizations at the time of hospitalization. 6.4.2 Non-infectious cause hospitalizations Six infants were hospitalized for non-infectious events: three HEU infants, one each with uncomplicated neonatal jaundice without neonatal sepsis, congenital macrocephaly with neurodevelopmental delay and communicating hydrocephalus; three HUU infants, one each with accidental trauma, a history of apnoea without sepsis and poor weight gain without underlying infection. The HEU infant with congenital macrocephaly was born to an HIV-infected mother who initiated an EFV-containing cART regimen late during the first trimester of pregnancy and the infant with communicating hydrocephalus was born to an HIV-infected mother who initiated an EFV-containing cART regimen 4 years prior to pregnancy and was thus on EFV during the first trimester of pregnancy. Both mothers reported no use of alcohol, tobacco or other illegal drugs during pregnancy. 97 6.5 Determining the effect of HIV exposure on infectious morbidity To resolve the primary objective of determining whether HIV exposure places HEU infants at greater risk for infectious morbidity in the first six months of life, odds ratios for the primary and secondary outcomes calculated from multivariable logistic regression models are presented. To further understand the odds of infectious morbidity in HEU compared to HUU infants beyond the role of breastfeeding, a stratified analysis is presented comparing HEU and HUU infants within strata of breastfeeding or not breastfeeding at two weeks and six months of age. 6.5.1 Multivariable analysis The crude association between HEU infants and the primary outcome showed HEU infants to have a 48% increased risk (RiR 1.48, 95% CI 0.72, 3.06) and 59% greater odds (OR 1.59; 95% CI 0.69, 3.82) of at least one infectious cause hospitalization or death compared to HUU infants. However, the null hypothesis could not be rejected. Additional analyses adjusting for potential confounders, presented below, did not substantially alter the estimate of effect of this primary comparison. With reference to Table 6.17 and Table 6.18, each row represents a logistic regression model for the primary outcome, including the variable indicated in the row and the term for infant group (HEU relative to HUU). The first aOR column indicates the aOR for the variable indicated in the row, after adjusting for infant group; the second aOR column indicates the aOR for the effect of being an HEU infant after adjusting for the variable in the corresponding row. Increasing maternal age in years had a protective effect on infectious morbidity (adjusted OR (aOR) 0.94; 95% CI 0.86, 1.02), and strengthened the effect of being an HEU infant on the odds of infectious morbidity (Table 6.17). It was decided a priori that maternal CD4 count would not be adjusted for in the final models as this may be on the causal pathway between HIV exposure and infectious morbidity. The effect of maternal CD4 count on infant infectious morbidity was in the opposite direction to what was hypothesized and a higher maternal CD4 count increased the odds of infectious morbidity after adjusting for HIV exposure. This relationship persisted whether the absolute CD4 count was considered as a continuous variable or a categorical variable (>/<350 cells/µl or >/<500 cells/µl), and whether the log of the absolute CD4 count or the CD4 percentage was considered. Maternal smoking during pregnancy or postnatal body mass index did not play a role in determining infectious morbidity, neither did infant birth or immunization characteristics. The presence or absence of breastfeeding did alter the point estimates for the effect of being an HEU infant on the outcome at all time points. This remained so whether categorized as “exclusive breastfeeding”, “partial breastfeeding” or “no breastfeeding” at each visit, categorized as “any breastfeeding” or “no breastfeeding” at each visit or categorized by the last visit at which “any breastfeeding” was reported (Table 6.18). At all visits the point estimate for “no breastfeeding” in reference to “exclusive breastfeeding”, adjusted for being an HEU 98 infant, indicated that not breastfeeding was harmful. However, confidence intervals for the point estimate of not breastfeeding overlapped 1.0 at all time points and were wide. Table 6.17 Odds ratios for the associations between maternal and infant characteristics and the primary outcome, adjusted for HIV exposure aOR (95%CI) P-value HEU aOR (95%CI) P-value Maternal characteristics Maternal age in years 0.94 (0.86, 1.02) 0.14 1.82 (0.78, 4.48) 0.18 Multiparous 1.77 (0.62, 6.37) 0.33 1.49 (0.64, 3.60) 0.36 Gestation in weeks at 1st ANC visit 0.99 (0.92, 1.05) 0.70 1.56 (0.68, 3.77) 0.30 Number of ANC visits 1.00 (0.80, 1.24) 0.97 1.49 (0.64, 3.63) 0.36 Smoking during pregnancy 1.40 (0.30, 5.00) 0.62 1.54 (0.67, 3.74) 0.32 Delivery maternal CD4 Maternal CD4 count (cells/µl) 1.00 (1.00, 1.00) 0.009 2.22 (0.92, 5.74) 0.09 Maternal CD4 percent 1.02 (0.97, 1.08) 0.36 2.21 (0.74, 6.91) 0.16 Log maternal CD4 absolute 2.06 (0.90, 4.85) 0.09 2.03 (0.85, 5.12) 0.12 Maternal CD4 count > 500 2.94 (1.24, 7.05) 0.01 1.96 (0.83, 4.90) 0.14 BMI (kg/m2) postnatal 1.01 (0.92, 1.11) 0.84 1.58 (0.68, 3.81) 0.29 Infant characteristics Female 0.58 (0.25, 1.33) 0.20 1.57 (0.68, 3.79) 0.30 Gestational age in weeks 0.87 (0.66, 1.14) 0.30 1.50 (0.65, 3.63) 0.35 Birth weight in kg 1.26 (0.45, 3.59) 0.66 1.63 (0.71, 3.97) 0.26 Low birth weight 0.63 (0.03, 3.71) 0.68 1.61 (0.70, 3.81) 0.27 Immunizations not up to date at 6 months 0.95 (0.14, 3.92) 0.95 1.61 (0.61, 4.56) 0.35 ANC – antenatal clinic; aOR – adjusted odds ratio; BMI – body mass index 99 Table 6.18 Odds ratios for the association between infant feeding mode and the primary outcome, adjusted for HIV exposure aOR (95%CI) P-value HEU aOR (95%CI) P-value 2 weeks: 1.24 (0.38, 3.77) 0.70 Exclusive breastfeeding Reference Partial breastfeeding 0.91 (0.13, 3.82) 0.91 No breastfeeding 1.44 (0.49, 4.67) 0.52 2 months: 1.05 (0.39, 2.86) 0.93 Exclusive breastfeeding Reference Partial breastfeeding 0.33 (0.17, 1.93) 0.31 No breastfeeding 1.74 (0.65, 4.89) 0.28 4 months: 1.17 (0.48, 2.99) 0.74 Exclusive breastfeeding Reference Partial breastfeeding 0.48 (0.06, 2.65) 0.41 No breastfeeding 1.62 (0.53, 6.10) 0.43 6 months: 1.29 (0.52, 3.36) 0.59 Exclusive breastfeeding Reference Partial breastfeeding 0.67 (0.08, 14.53) 0.74 No breastfeeding 1.25 (0.20, 24.33) 0.84 No breastfeeding: (Reference “any” breastfeeding) 2 weeks 1.49 (0.49, 4.67) 0.51 1.25 (0.39, 3.77) 0.69 2 months 2.12 (0.78, 6.05) 0.15 1.04 (0.36, 2.94) 0.95 4 months 2.28 (0.84, 6.63) 0.11 1.05 (0.48, 3.58) 0.92 6 months 1.52 (0.54, 4.50) 0.44 1.28 (0.48, 3.58) 0.56 Last visit any breastfeeding reported: (Reference is 6 months) 1.23 (0.37, 3.79) 0.73 Never 1.97 (0.52, 8.88) 0.34 2 weeks 2.49 (0.62, 10.06) 0.19 2 months 1.56 (0.35, 6.50) 0.54 4 months 0.62 (0.03, 4.42) 0.68 6 months Reference HEU – HIV exposed uninfected 100 In the final adjusted analysis for the primary outcome of at least one infectious cause hospitalization or death, controlling for maternal age and any breastfeeding at two weeks of age (model A in Table 6.19) marginally reduced the odds ratio for HEU infants from 1.59 (95% CI 0.69, 3.82) to 1.45 (95% CI 0.44, 4.45). Controlling instead for any breastfeeding at six months of age also altered the odds ratio trivially to 1.47 (95% CI 0.54, 4.25) (model B Table 6.19). Maternal age may have a non-linear U-shaped relationship with infant morbidity and mortality, however including maternal age squared or maternal age cubed did not alter the effect estimate for HEU infants or improve the models. HEU infants did not have a definitively greater probability of any infectious cause hospitalization or death. The unadjusted effect of being an HEU infant on very severe infectious cause hospitalization or death was substantial with an OR of 1.83 (95% CI 0.62, 6.11) that increased further after adjusting for maternal age and any breastfeeding at two weeks of age (aOR 2.48 (95% CI 0.60, 10.19). Despite the large effect size the 95% confidence intervals include the null value and it is not possible to conclude a definitive difference between HEU and HUU infants (model A Table 6.19). Continuing to breastfeed until six months of age appears protective against very severe infectious cause hospitalization or death with an aOR for the effect of any breastfeeding at six months of 0.34 (95% CI 0.08, 1.13) and a reduction in the effect of being an HEU infant to an aOR of 1.37 (95% CI 0.39, 5.57) (model B Table 6.19). Due to the limited sample size and the low prevalence of outcome events it was not possible to include multiple measurements of breastfeeding status in a mixed effects model and breastfeeding status was ultimately modelled as a fixed effect at a single time point with similar results to the fixed effects logistic regression already presented. 101 Table 6.19 Logistic regression models for the odds of the primary and secondary outcomes in HEU infants relative to HUU infants Unadjusted OR(95%CI) Model A aOR(95% CI) Model B aOR(95% CI) Primary outcome – any infectious cause hospitalization or death in first 6 months of life HEU infant 1.59 (0.69, 3.82) 1.45 (0.44, 4.45) 1.47 (0.54, 4.25) Maternal age (years) 0.95 (0.87, 1.03) 0.94 (0.86, 1.02) 0.94 (0.86, 1.02) Any breastfeeding at 2 weeks 0.59 (0.26, 1.39) 0.70 (0.22, 2.06) -------------------- Any breastfeeding at 6 months 0.57 (0.22, 1.36) -------------------- 0.65 (0.21, 1.85) Secondary outcome 1 – severe infectious cause hospitalization or death HEU infant 1.42 (0.59, 3.59) 1.61 (0.48, 5.09) 1.31 (0.45, 4.02) Maternal age (years) 0.95 (0.86, 1.03) 0.94 (0.85, 1.02) 0.94 (0.86, 1.02) Any breastfeeding at 2 weeks 0.78 (0.32, 1.98) 0.98 (0.29, 3.08) -------------------- Any breastfeeding at 6 months 0.61 (0.22, 1.51) -------------------- 0.66 (0.21, 2.01) Secondary outcome 2 – very severe infectious cause hospitalization or death HEU infant 1.83 (0.62, 6.11) 2.49 (0.60, 10.19) 1.37 (0.39, 5.57) Maternal age (years) 0.92 (0.82, 1.02) 0.91 (0.80, 1.01) 0.91 (0.80, 1.01) Any breastfeeding at 2 weeks 0.78 (0.26, 2.36) 1.19 (0.30, 4.48) -------------------- Any breastfeeding at 6 months 0.34 (0.08, 1.13) -------------------- 0.37 (0.07, 1.52) aOR – adjusted odds ratio; CI – confidence interval; HEU – HIV exposed uninfected; OR – odds ratio Model A: the odds of the outcome in HEU compared to HUU infants adjusted for maternal age (continuous variable) and any breastfeeding at 2 weeks (binary variable) Model B: the odds of the outcome in HEU compared to HUU infants adjusted for maternal age (continuous variable) and any breastfeeding at 6 months (binary variable) 102 6.5.2 Stratified analysis A stratified analysis was performed to determine the effect of HIV exposure on the primary and both secondary outcomes when conditioned on the presence of any breastfeeding (Table 6.20). At two weeks there was only a single HUU infant not breastfeeding and thus it was not possible to model the relationship between HIV exposure and infectious morbidity in infants not breastfeeding at two weeks of age. Amongst infants not breastfeeding at six months of age there was no difference between HEU and HUU infants for the probability of the primary or secondary outcomes. Amongst only those infants receiving any breastfeeding at two weeks of age, HEU infants had an increased odds relative to HUU infants of very severe infectious cause hospitalization or death during the first six months of life (aOR 4.21, 95% CI 1.00, 19.22, p = 0.05). The odds of any infectious cause hospitalization or death and severe infectious cause hospitalization or death was also increased in HEU infants who were breastfed at two weeks, compared to HUU infants breastfed at two weeks. However, the confidence intervals for both these estimates were wide and crossed the null. Similarly amongst infants that were still breastfeeding at six months of age there tended to be a greater probability of infectious morbidity in HEU than HUU infants, but the estimates are imprecise with wide confidence intervals. Maternal age was retained in all stratified models. Confounding by maternal, household and infant factors was evaluated per stratum and there were no additional variables that substantially altered the point estimate of HEU infants within strata. 103 Table 6.20 Stratified analysis of the effect of HIV exposure on the primary and secondary outcomes, conditioned on breastfeeding and adjusted for maternal age Primary outcome Secondary outcome 1 Secondary outcome 2 n/N aOR (95% CI) n/N aOR (95% CI) n/N aOR (95% CI) No breastfeeding 2 weeks HEU 11/59 HUU 1/1 Model not possible HEU 8/59 HUU 1/1 Model not possible HEU 5/59 HUU 1/1 Model not possible 6 months HEU 15/80 HUU 4/25 1.34 (0.42, 5.19) HEU 12/80 HUU 4/25 1.06 (0.32, 4.18) HEU 9/80 HUU 3/25 1.04 (0.27, 5.09) Any breastfeeding 2 weeks HEU 6/35 HUU 9/81 1.96 (0.59, 6.22) HEU 6/35 HUU 8/81 2.16 (0.64, 7.01) HEU 5/35 HUU 4/81 4.21 (1.00, 19.22) 6 months HEU 2/14 HUU 6/57 1.82 (0.24, 10.2) HEU 2/14 HUU 5/57 2.07 (0.27, 12.0) HEU 1/14 HUU 2/57 3.75 (0.14, 6.0) aOR – adjusted odds ratio; HEU – HIV exposed uninfected; HUU – HIV unexposed uninfected Primary outcome: at least one infectious cause hospitalization or death <194 days old Secondary outcome 1: at least one severe infectious cause hospitalization or death <194 days old Secondary outcome 2: at least one very severe infectious cause hospitalization or death <194 days old 6.6 Summary of main findings Ninety four HEU and 82 HUU infants were followed to six months of age. Eighteen percent (17/94) of HEU and 12% (10/82) of HUU infants experienced a primary outcome event of at least one infectious cause hospitalization or death in the first six months of life (Chi-squared p = 0.38) for an unadjusted OR of 1.59 (95% CI 0.69, 3.82). In term infants born to mothers without major obstetric or medical morbidities, in similar socioeconomic and household circumstances and with an equivalent frequency of out-patient sick-clinic visits: • There is no evidence of a difference in infectious morbidity comparing HEU and HUU infants that never breastfed or stopped breastfeeding before six months of age; • Breastfed HEU infants compared to breastfed HUU infants may have a four times greater odds (aOR 4.21, 95% CI 1.00, 19.22) of very severe infectious cause hospitalization or death in the first six months of life. 104 Chapter 7 Results - In the antiretroviral therapy era, is HEU infant infectious morbidity associated with the extent of maternal HIV disease? The objective of this sub-group analysis was to determine whether previous or current severe maternal immune suppression was associated with HEU infant infectious morbidity by comparing infants of mothers on maternally indicated cART during pregnancy to HEU infants of mothers without immune suppression with a CD4 count of 350 cells/µl or greater and on VTP prophylaxis during pregnancy. A description of HIV-related factors is given for all HIV-infected mothers in the analytic cohort in section 7.1. Subsequently maternal, household and infant characteristics are compared firstly by maternal ARV regimen during pregnancy (maternally indicated cART or VTP prophylaxis) in section 7.2 to understand potential confounding differences between the two groups, and secondly by examining factors potentially related to the primary outcome of at least one infectious cause hospitalization or death in section 7.3. Finally, multivariable analysis to determine the effect of maternal ARV regimen in pregnancy on the odds of infectious cause hospitalization or death is presented in section 7.4. 7.1 Description of all HIV-infected mothers Ninety-four HIV-infected mothers and their HEU infants were included in this cohort study. Of the 94 HIV-infected mothers 51.1% (48/94) were diagnosed with HIV prior to pregnancy, and 85.1% (80/94) were considered to have WHO stage 1 (asymptomatic) disease at their first antenatal clinic visit. Fifty percent (47/94) were receiving maternally-indicated cART, 46.8% (44/94) received VTP prophylaxis and three mothers, two of whom were diagnosed with HIV at two weeks post-partum, received neither cART nor VTP prophylaxis. Of the 47 mothers on maternally-indicated cART, 20 (42.6%) mothers initiated cART prior to pregnancy. The median duration of cART by the time of delivery was 22 (IQR 14, 148) weeks with a minimum of two weeks and a maximum of 693 weeks. The majority (24/47, 51.1%) of mothers on cART during pregnancy were receiving the standard adult first line cART regimen at the time of TDF/3TC/EFV; 13 (27.7%) were receiving the alternative first line regimen of TDF/3TC/NVP; 7 (14.9%) were receiving alternative first line regimens including ZDV or d4T with 3TC and NVP or EFV, and 3 (6.4%) were on a second line cART regimen including LPV/r. In total, 39 (41.5%) of the 94 HIV-infected mothers were on a TDF containing regimen and their infants were in-utero TDF exposed, and 48 (51.1%) mothers were receiving ZDV, four in a cART combination and 44 on VTP prophylaxis and hence their infants were ZDV in-utero exposed. Of the 44 mothers on VTP prophylaxis, ZDV monotherapy was initiated during the first trimester in 3 (6.8%) mothers, the second trimester in 27 (61.4%) mothers and the third trimester in 14 (31.8%) mothers. The antenatal CD4 count was measured at a median of 19 (IQR 13, 105 26) weeks prior to delivery and did not differ whether the mothers were on cART or not. The median antenatal absolute CD4 count in HIV-infected mothers was significantly higher than the CD4 count at delivery (420 (IQR 284, 539) cells/µl vs. 343 (IQR 236, 501) cells/µl, p = 0.04). Of the 84 mothers on whom HIV viral load was measured at delivery, thirty seven percent (31/84) had undetectable delivery HIV viral loads below 40 copies/ml, 30/43 (69.8%) mothers on cART and 1/41 (2.4%) mothers not on cART. Fourteen (16.7%) of the 84 mothers with measured HIV viral load had delivery viral loads greater than 10 000 copies/ml, 3/43 (7.0%) mothers on cART and 11/41 (26.8%) mothers not on cART. The maximum viral load in mothers on cART was 23 580 copies/ml and the maximum viral load in mothers not on cART was 54 770 copies/ml. See Appendix D for complete data on all 94 HIV infected mothers and their HEU infants. 7.2 Sub-group: HEU infant comparison by maternally indicated cART or VTP prophylaxis Infants of mothers on maternally-indicated cART were compared to infants of mothers on VTP prophylaxis. This sub-group analysis excludes 5 HIV-infected mothers and their infants, three mothers that did not receive either cART or VTP prophylaxis during pregnancy and two mothers that received VTP prophylaxis but should have received cART for antenatal CD4 counts of below 350 cells/µl. Eighty-nine HIV-infected mothers and their HEU infants were included in this sub-group analysis, 47 HIV-infected mothers on maternally-indicated cART (further referred to as the cART group) and 42 HIV-infected mothers on VTP prophylaxis (further referred to as the VTP group). 7.2.1 Maternal characteristics Mothers in the cART group were significantly older at delivery compared to mothers in the VTP group, median 29.2 (IQR 26.2, 32.4) years and 27.4 (IQR 23.4, 30.4) years respectively (p = 0.04) (Table 7.1). Forty-six (97.9%) cART mothers and 37 (88.1%) VTP mothers were of black African race, with no significant differences in marital status or highest level of education between the two groups. Maternal income was significantly higher in the cART than the VTP group (ZAR 1600 (IQR 500, 2500) vs. ZAR 560 (IQR 100, 1500), p = 0.007). Gestational age at mothers first antenatal clinic visit and the number of antenatal clinic visits were similar in both groups, as was maternal body mass index measured at 2 weeks postnatal (Table 7.1). Significantly more cART mothers were diagnosed with HIV prior to pregnancy than VTP mothers (33/47 (70.2%) vs. 14/42 (33.3%), p = 0.001) (Table 7.2). As expected, median maternal antenatal and delivery absolute CD4 counts were significantly lower in cART mothers than VTP mothers, 332 (IQR 232, 420) 106 cells/µl compared to 571 (IQR 428, 623) cells/µl antenatally (p<0.001) and 313 (IQR 215, 459) cells/µl compared to 423 (IQR 288, 556) cells/µl at delivery (p = 0.02). During pregnancy 67.4% (31/46) of cART mothers had a CD4 count of less than 350 cells/µl. By delivery, 16 (38.1%) VTP mothers had CD4 counts that had dropped below 350 cells/µl. The absolute change in maternal CD4 count between the antenatal period and delivery differed significantly between the cART and VTP groups. Mothers on VTP prophylaxis experienced a median drop in CD4 count of 109 (IQR -238,+30) cells/µl between the antenatal period and delivery compared to mothers on cART who experienced a median increase in CD4 count of 17 (IQR -115,+143) cells/µl. A similar difference was reflected in the percentage change in the absolute CD4 count between the antenatal period and delivery. Maternal HIV viral load was significantly lower in cART mothers than VTP mothers. Seventy percent (30/43) of cART mothers had an undetectable HIV viral load at delivery compared to 2.6% (1/38) of VTP mothers (p<0.001). 107 Table 7.1 Maternal demographic & obstetric characteristics compared by maternal pregnancy ARV regimen (numeric variables as median (IQR), categorical variables as number (%)) Total N=89 Maternally indicated cART N=47 VTP Prophylaxis N=42 P-value Demographic characteristics Age in years 28.1(24.2,31.4) 29.2(26.2,32.4) 27.4(23.4,30.4) 0.04 Race: 0.10 Black African 83(93.3) 46(97.9) 37(88.1) Coloured 6(6.7) 1(2.1) 5(11.9) Marital Status: 0.27 Never married 59(66.3) 28(59.6) 31(73.8) Married 25(28.1) 17(36.2) 8(19.0) Widow/Separated/Divorced 5(5.6) 2(4.3) 3(7.1) Education: 0.42 No or any primary 6(6.7) 3(6.4) 3(7.1) Some secondary 56(62.9) 27(57.4) 29(69.0) Completed secondary 27(30.3) 17(36.2) 10(23.8) Monthly income (ZAR) 1060(280,2325) 1600(500,2500) 560(100,1500) 0.007 Health characteristics Primiparous 14(15.7) 7(14.9) 7(16.7) 1.00 Gestation at 1st antenatal visit (weeks) 20(16,26) 20(16,25) 22(16,26) 0.55 Number of antenatal visits 5(4,6) 5(4,6) 5(4,6) 0.85 BMI postnatal kg/m2 26.6(23.3,28.8) 26.8(22.9,28.4) 25.9(23.4,29.4) 0.53 BMI – body mass index; cART – combination antiretroviral therapy; VTP – vertical transmission prevention; ZAR – South African Rand 108 Table 7.2 Maternal HIV characteristics compared by maternal pregnancy ARV regimen (numeric variables as median (IQR), categorical variables as number (%)) Total N=89 Maternally indicated cART N=47 VTP Prophylaxis N=42 P-value HIV diagnosed during pregnancy 42(47.2) 14(29.8) 28(66.7) 0.001 Initiation of cART: Pre-pregnancy ---------------------- 20(42.6) ---------------------- First Trimester ---------------------- 3(6.4) ---------------------- Second Trimester ---------------------- 13(27.7) ---------------------- Third Trimester ---------------------- 11(23.4) ---------------------- Pregnancy cART regimen: TDF/3TC/EFV ---------------------- 24(51.1) ---------------------- First line non-TDF/3TC/EFV ---------------------- 20(42.6) ---------------------- Second line ---------------------- 3(6.4) ---------------------- Initiation of zidovudine prophylaxis: First Trimester ---------------------- ---------------------- 3(7.1) Second Trimester ---------------------- ---------------------- 27(64.3) Third Trimester ---------------------- ---------------------- 12(28.6) CD4 count Antenatal absolute CD4 cells/µl* 422(285,544) 332(232,420) 571(428,623) <0.0001 Delivery absolute CD4 cells/µl 344(246,502) 313(215,459) 423(288,556) 0.02 Delivery CD4 percent 26.1(21.6,32.5) 24.6(18.9,28.9) 29.1(22.9,34.4) 0.01 Antenatal CD4 category*: <0.0001 <350 cells/µl 31(35.2) 31(67.4) 0(0.0) 350-499 cells/µl 30(34.1) 9(19.6) 21(50.0) >500 cells/µl 27(30.7) 6(13.0) 21(50.0) Delivery CD4 category: 0.05 <350 cells/µl 46(51.7) 30(63.8) 16(38.1) 350-499 cells/µl 20(22.5) 8(17.0) 12(28.6) >500 cells/µl 23(25.8) 9(19.1) 14(33.3) Absolute change in CD4 count -40(-178,70) 17(-115,143) -109(-238, 30) 0.003 Percent change in CD4 count -7.8(-39.4,22.6) 7.6(-33.7,48.2) -18.9(-43.9,7.9) 0.004 HIV viral load# Category by copies/ml: <0.0001 <40 31(38.3) 30(69.8) 1(2.6) 40-999 24(29.6) 8(18.6) 16(42.1) 1000-9999 13(16.0) 2(4.7) 11(28.9) > 10 000 13(16.0) 3(7.0) 10(26.3) Log10 HIV viral load 2.01(1.59,3.08) 1.59(1.59,1.93) 2.98(2.01,3.81) <0.0001 cART – combination antiretroviral therapy; VTP – vertical transmission prevention; * 1 antenatal CD4 count missing in cART group; # 8 HIV viral loads missing (4 in each group) 109 7.2.2 Household characteristics Small differences existed in some household factors between the two groups: 53.2% (25/47) of cART group infant households compared to 31.0% (13/42) of VTP group infant households had access to water piped into the dwelling (p = 0.07) and electricity was available in 100% (47/47) of cART compared to 88.1% (37/42) of VTP group infant households (p = 0.02). There was no difference between the two groups in the type of house, access to a flush toilet connected to sewerage, the fuel source used for cooking or heating, and a host of household assets (Table 7.3). The number of rooms, the number of people living in the household and the distance of the house to the nearest primary health care clinic in minutes walked were no different between the cART and VTP groups. 110 Table 7.3 Household characteristics compared by maternal pregnancy ARV regimen (numeric variables as median (IQR), categorical variables as number (%), unless otherwise stated) Total N=89 Maternally indicated cART N=47 VTP Prophylaxis N=42 P-value Type of house: 0.39 Stand alone house 31(34.8) 19(40.4) 12(28.6) Apartment in apartment block 5(5.6) 4(8.5) 1(2.4) House/flat/room in backyard 4(4.5) 1(2.1) 3(7.1) Shack in backyard 33(37.1) 16(34.0) 17(40.5) Shack not in backyard 14(15.7) 6(12.8) 8(19.0) Other 2(2.2) 1(2.1) 1(2.4) Water supply: 0.07 Water piped into dwelling 38(42.7) 25(53.2) 13(31.0) Water piped into yard 47(52.8) 21(44.7) 26(61.9) Public tap 4(4.5) 1(2.1) 3(7.1) Sanitation: 1.00 Flush toilet (connected to sewerage) 87(97.8) 46(97.9) 41(97.6) Fuel (cook): 0.13 Electricity 83(93.3) 46(97.9) 37(88.1) Gas 3(3.4) 1(2.1) 2(4.8) Paraffin 3(3.4) 0(0.0) 3(7.1) Fuel (heat): 0.68 Electricity 34(38.2) 19(40.4) 15(35.7) Gas 1(1.1) 0(0.0) 1(2.4) Paraffin 47(52.8) 24(51.1) 23(54.8) Firewood/Other 7(7.9) 4(8.5) 3(7.1) Fuel (light): 0.02 Electricity 84(94.4) 47(100) 37(88.1) Paraffin 5(5.6) 0(0.0) 5(11.9) Number of rooms in house 2(1,3) 2(1,3) 2(1,3) 0.57 Number of people living in household 4(3,5) 4(3,5) 4(3,5) 0.64 Distance to clinic (minutes walked) 20(10,30) 15(10,30) 27.5(13.5,30) 0.20 Assets: Radio 51(57.3) 29(61.7) 22(52.4) 0.50 TV 74(83.1) 42(89.4) 32(76.2) 0.17 Computer 10(11.2) 5(10.6) 5(11.9) 1.00 Refrigerator 66(74.2) 37(78.7) 29(69.0) 0.42 Home phone 4(4.5) 3(6.4) 1(2.4) 0.62 Cell phone 88(98.9) 47(100) 41(97.6) 0.47 Bicycle 10(11.2) 6(12.8) 4(9.5) 0.74 Motorcycle or motor scooter 2(2.2) 2(4.3) 0(0.0) 0.50 cART – combination antiretroviral therapy; VTP – vertical transmission prevention 111 7.2.3 Infant characteristics HEU infants of mothers in the maternally indicated cART group compared to the VTP group did not differ in gestational age or birth weight (Table 7.4). Nor did they differ in the occurrence of infectious cause hospitalizations, 8 (17.0%) infants of cART mothers and 7 (16.7%) infants of VTP mothers were hospitalized for an infectious event. Fewer cART mothers, intended to exclusively breastfeed their babies than VTP mothers, 31.9% (15/47) vs. 47.6% (20/42), but this was not a statistically significant difference (p = 0.19) (Table 7.5). Corresponding with this, there was a tendency for fewer infants of cART mothers to be exclusively breastfed at two weeks of age than infants of VTP mothers (12/47, 25.5% vs. 18/42, 42.9%, p = 0.08), but none of these feeding differences were statistically significant. This feeding pattern was interrogated further, comparing mothers with undetectable HIV viral loads (below 40 copies/ml), in whom breastfeeding carries the lowest risk of HIV transmission, compared to mothers with detectable viral loads for the proportions of infants receiving any breastfeeding at two weeks. Significantly fewer infants of mothers with undetectable HIV viral loads were receiving any breastfeeding at two weeks (25.7% (9/35)), compared to 46.2% (18/39) of infants born to mothers with detectable HIV viral loads (p = 0.03). Significantly more infants in the VTP group had a grade 1 or more anaemia at two months of age compared to infants in the cART group (55.9% (19/34) VTP vs. 33.3% (14/42) cART, p = 0.02) (Table 7.6). Although more than one third (37.3% (25/67)) of HEU infants were still anaemic at six months of age this was no longer significantly different between the two groups. When infants were compared specifically as in utero ZDV-exposed (including the VTP infants and four infants of mothers on ZDV containing cART) or ZDV-unexposed significantly more ZDV-exposed infants had a grade 1 or more anaemia at two months compared to ZDV-unexposed HEU infants (Table D.7 in Appendix D). Considering HEU infant growth, infants in the cART group had mean WHO length-for-age Z-scores that were significantly lower than that of infants in the VTP group (Figure 7.1). This difference persisted from two weeks through to six months of age (Table 7.7). A significantly lower mean weight-for-age Z-score occurred at four and six months in the infants of the cART group (Figure 7.2). There was no difference in mean weight-for-length Z-score at any time point between infants of the cART or VTP group (Figure 7.3). The head circumference and mid upper arm circumference Z-scores were significantly lower at four months in the cART compared to the VTP group but this did not persist at six months of age. When infants were compared specifically as in-utero TDF-exposed and TDF-unexposed the pattern of a persistently lower length-for-age Z-score from two weeks through to six months in the TDF-exposed group persisted. Together with this, the TDF-exposed infants had significantly lower weight-for-age Z-scores than the TDF-unexposed group persistently from birth through to six months but equivalent weight-for-length Z-scores throughout the observation period. The head circumference and mid upper 112 arm circumference Z-scores were significantly lower at four months in the TDF-exposed compared to the TDF-unexposed group, but this difference did not remain at six months (Table D.8 in Appendix D). Table 7.4 Infant birth and healthcare characteristics compared by maternal pregnancy ARV regimen (numeric variables as median (IQR), categorical variables as number (%), unless otherwise stated) Total N=89 Maternally indicated cART N=47 VTP Prophylaxis N=42 P-value Male 43(48.3) 24(51.1) 19(45.2) 0.74 Gestational age in weeks – mean (SD) 38.8(1.5) 38.7(1.7) 38.9(1.3) 0.48 Birth weight in grams – mean (SD) 3110(379) 3077(397) 3147(358) 0.38 Low birth weight 6(6.7) 4(8.5) 2(4.8) 0.68 Immunizations not up to date at 6 months* 4(5.8) 3(7.9) 1(3.2) 0.81 All cause clinic visits per infant 6(5,7) 6(5,7) 6.5(5.3,7) 0.44 Sick clinic visits per infant 1(0,1) 1(0,1) 1(0,1.8) 0.43 Receiving CPT at 8 weeks 50(56.2) 30(63.8) 20(47.6) 0.19 Infectious cause hospitalization 15(16.9) 8(17.0) 7(16.7) 1.00 cART – combination antiretroviral therapy; CPT – cotrimoxazole preventive therapy; SD – standard deviation; VTP – vertical transmission prevention; * 20 unknown (9 in cART group, 11 in VTP group) 113 Table 7.5 Infant feeding characteristics compared by maternal pregnancy ARV regimen (all variables as number (%)) Total N=89 Maternally indicated cART N=47 VTP Prophylaxis N=42 P-value Intention to exclusively breastfeed 35(39.3) 15(31.9) 20(47.6) 0.19 Feeding mode at 2 weeks: (N=89) 0.08 Exclusive breastfeeding 30(33.7) 12(25.5) 18(42.9) Partial breastfeeding 3(3.4) 3(6.4) 0(0.0) No breastfeeding 56(62.9) 32(68.1) 24(57.1) Feeding mode at 2 months: (N=88) 0.10 Exclusive breastfeeding 22(25.0) 9(19.1) 13(31.7) Partial breastfeeding 4(4.5) 4(8.5) 0(0.0) No breastfeeding 62(70.5) 34(72.3) 28(68.3) Feeding mode at 4 months: (N=80) 0.22 Exclusive breastfeeding 10(12.5) 3(7.1) 7(18.4) Partial breastfeeding 7(8.8) 5(11.9) 2(5.3) No breastfeeding 63(78.8) 34(81.0) 29(76.3) Feeding mode at 6 months: (N=69) 1.00 Exclusive breastfeeding 4(5.8) 2(5.3) 2(6.5) Partial breastfeeding 7(10.1) 4(10.5) 3(9.7) No breastfeeding 58(84.1) 32(84.2) 26(83.9) cART – combination antiretroviral therapy; VTP – vertical transmission prevention 114 Table 7.6 Infant haemoglobin and anaemia compared by maternal pregnancy ARV regimen (numeric variables as mean (SD), categorical variables as number (%)) Total N=89 Maternally indicated cART N=47 VTP Prophylaxis N=42 P-value Mean haemoglobin (SD) 2 weeks (N=84) 14.2(2.0) 14.2(2.2) 14.1(1.8) 0.79 2 months (N=76) 10.5(0.9) 10.6(1.0) 10.3(0.9) 0.11 4 months (N=75) 11.1(0.9) 11.2(0.8) 11.0(1.0) 0.20 6 months (N=67) 11.2(0.9) 11.3(0.9) 11.2(0.9) 0.59 DAIDS anaemia grade 2 weeks: (N=84) 0.39 Normal 68(81.0) 34(79.1) 34(82.9) Grade 1 8(9.5) 3(7.0) 5(12.2) Grade 2 8(9.5) 6(14.0) 2(4.9) 2 months: (N=76) 0.02 Normal 43(56.6) 28(66.7) 15(44.1) Grade 1 22(28.9) 7(16.7) 15(44.1) Grade 2 10(13.2) 7(16.7) 3(8.8) Grade 3 1(1.3) 0(0.0) 1(2.9) 4 months: (N=75) 0.63 Normal 46(61.3) 26(66.7) 20(55.6) Grade 1 22(29.3) 10(25.6) 12(33.3) Grade 2 6(8.0) 3(7.7) 3(8.3) Grade 3 1(1.3) 0(0.0) 1(2.8) 6 months: (N=67) 0.83 Normal 42(62.7) 24(64.9) 18(60.0) Grade 1 21(31.3) 11(29.7) 10(33.3) Grade 2 3(4.5) 1(2.7) 2(6.7) Grade 3 1(1.5) 1(2.7) 0(0.0) cART – combination antiretroviral therapy; DAIDS – Division of AIDS; SD – standard deviation; VTP – vertical transmission prevention 115 Table 7.7 Infant anthropometric measurements compared by maternal pregnancy ARV regimen (numeric variables as mean (SD), categorical variables as number (%), unless otherwise stated) Anthropometric measurement Total N=89 Maternally indicated cART N=47 VTP Prophylaxis N=42 P-value Birth: (N=89) WAZ -0.35(0.84) -0.62(0.88) -0.19(0.79) 0.33 LAZ -0.47(1.86) -0.62(1.79) -0.47(1.96) 0.92 WLZ -0.23(1.72) -0.07(1.63) -0.13(1.82) 0.70 HCZ -0.36(1.26) -0.36(1.30) 0.10(1.24) 0.33 Stunted 16(18.0) 8(17.0) 8(19.0) 1.00 Wasted 9(10.1) 5(10.6) 4(9.5) 1.00 2 weeks: (N=89) WAZ -0.37(0.83) -0.49(0.86) -0.07(0.78) 0.13 LAZ -1.2(0.93) -1.3(0.85) -0.98(0.99) 0.04 WLZ 0.65(0.98) 0.69(0.99) 0.60(0.98) 0.68 HCZ 0.08(1.08) -0.02(1.17) 0.19(0.96) 0.35 Stunted 14(15.7) 8(17.0) 6(14.3) 0.78 Wasted 0(0.0) 0(0.0) 0(0.0) NA 2 months: (N=83) WAZ -0.31(1.04) -0.50(1.15) -0.09(0.85) 0.84 LAZ -1.17(1.05) 1.40(0.99) -0.90(1.08) 0.03 WLZ 1.12(1.14) 1.15(1.06) 1.08(1.24) 0.79 HCZ 0.30(1.14) 0.22(1.28) 0.39(0.97) 0.51 Stunted 18(21.7) 11(24.4) 7(18.4) 0.60 Wasted 0(0.0) 0(0.0) 0(0.0) NA 4 months: (N=77) WAZ -0.07(1.15) -0.36(1.14) 0.24(1.10) 0.02 LAZ -0.88(1.16) -1.24(1.03) -0.50(1.18) 0.004 WLZ 0.86(1.28) 0.83(1.22) 0.89(1.37) 0.83 HCZ 0.57(1.06) 0.31(1.09) 0.85(0.997) 0.02 MUACZ 0.32(1.08) 0.09(1.00) 0.57(1.13) 0.05 Stunted 13(16.9) 9(22.5) 4(10.8) 0.23 Wasted 0(0.0) 0(0.0) 0(0.0) NA 6 months: (N=69) WAZ 0.12(1.16) -0.15(1.12) 0.46(1.12) 0.03 LAZ -0.68(0.93) -0.90(0.85) -0.40(0.95) 0.03 WLZ 0.78(1.32) 0.60(1.30) 0.99(1.32) 0.21 HCZ 0.79(1.19) 0.61(1.13) 1.00(1.24) 0.19 MUACZ 0.40(1.08) 0.21(1.13) 0.63(0.98) 0.10 Stunted 7(10.1) 5(13.2) 2(6.5) 0.45 Wasted 1(1.4) 1(2.6) 0(0.0) 1.00 cART – combination antiretroviral therapy, HCZ – head-circumference Z-score; LAZ – length-for-age Z-score; MUACZ – mid-upper-arm-circumference Z-score; VTP – vertical transmission prevention, WAZ – weight-for-age Z-score; WLZ – weight-for-length Z-score 116 Figure 7.1 Infant WHO weight-for-age Z-score compared by maternal pregnancy ARV regimen Figure 7.2 Infant WHO length-for-age Z-score compared by maternal pregnancy ARV regimen 0 1 2 3 4 5 6−2−1012Age in monthsZ−scorecART groupVTP group* ** = P<0.050 1 2 3 4 5 6−2−1012Age in monthsZ−scorecART groupVTP group* * *** = P<0.05117 Figure 7.3 Infant WHO weight-for-length Z-score compared by maternal pregnancy ARV regimen 0 1 2 3 4 5 6−2−1012Age in monthsZ−scorecART groupVTP group118 7.3 Sub-group: HEU infants - comparison of predisposing characteristics by primary outcome Of the 89 infants included in this sub-group analysis 15 (16.9%) experienced the primary outcome of at least one infectious cause hospitalization in the first six months of life, there were no HEU infant deaths.. 7.3.1 Maternal characteristics Maternal age, education and income were similar in mothers of infants with and without an infectious cause hospitalization and there were no differences in parity, gestational age at first antenatal clinic visit or the number of antenatal clinic visits attended (Table 7.8). Similar proportions of mothers of infants with and without an infectious cause hospitalization were diagnosed with HIV during pregnancy (6/15 (40.0%) hospitalized vs. 36/74 (48.6%) not hospitalized, p = 0.58) and were receiving cART during pregnancy (8/15 (53.3%) hospitalized vs. 39/74 (52.7%) not hospitalized, p = 1.00). The timing of initiation of cART and the cART regimen during pregnancy were no different in mothers of infants with and without an infectious cause hospitalization (Table 7.9). ZDV prophylaxis was initiated more often in the third trimester in mothers of infants with the outcome (4/7 (57.1%)) compared to mothers of infants without the outcome (8/35 (22.9%)), but not statistically significantly so (p = 0.14). The median antenatal and delivery CD4 counts, the change in CD4 count as well as the delivery HIV viral load were no different between the two groups. Specifically there was no association between antenatal CD4 count below 350 cells/µl and an infectious cause hospitalization or death, 4/15 (26.7%) infants with the outcome vs. 27/73 (37.0%) infants without the outcome born to mothers with antenatal CD4 count below 350 cells/µl (p = 0.56). 7.3.2 Household characteristics Household characteristics were similar in both groups. The type of house, sanitation and possession of various assets did not differ between households of infants with and without the outcome (Table 7.10). Only availability of electricity differed significantly between households of infants with the outcome (15/15 (100%)) compared to households of infants without the outcome (69/74 (93.2%), p = 0.02). 119 Table 7.8 Maternal demographic and obstetric characteristics compared by HEU infant outcome group Primary outcome is at least 1 infectious cause hospitalization or death before 195 days old (numeric variables as median (IQR), categorical variables as number (%)) BMI – body mass index; ZAR – South African Rand Total N=89 Infants hospitalized (N= 15) Infants not hospitalized (N= 74) P-value Age in years 28.1(24.2,31.4) 27.9(22.7,32.0) 28.1(24.3,31.1) 0.79 Education: 0.22 No or any primary 6(6.7) 1(6.7) 5(6.8) Some secondary 56(62.9) 12(80.0) 44(59.5) Completed secondary 27(30.3) 2(13.3) 25(33.8) Monthly income (ZAR) 1060(280,2325) 1050(140,1750) 1080(280,2400) 0.43 Primiparous 14(15.7) 2(13.3) 12(16.2) 1.00 Gestation at 1st antenatal visit (weeks) 20(16,26) 20(16.5,26) 20(16,25) 0.83 Number of antenatal visits 5(4,6) 5(4,6) 5(4,7) 0.51 BMI postnatal kg/m2 26.6(23.3,28.8) 27.0(23.4,30.3) 26.6(23.0,28.5) 0.58 120 Table 7.9 Maternal HIV characteristics compared by HEU infant outcome group (numeric variables as median (IQR), categorical variables as number (%)) Total N=89 Infants hospitalized (N= 15) Infants not hospitalized (N= 74) P-value cART during pregnancy 47(52.8) 8(53.3) 39(52.7) 1.00 Initiation of cART (N=47) 0.98 Pre-pregnancy 20(42.6) 4(50.0) 16(41.0) First Trimester 3(6.4) 0(0.0) 3(7.7) Second Trimester 13(27.7) 2(25.0) 11(28.2) Third Trimester 11(23.4) 2(25.0) 9(23.1) Pregnancy cART regimen (N=47) 0.75 TDF/3TC/EFV 24(51.1) 4(50.0) 20(51.3) First line non-TDF/3TC/EFV 20(42.6) 4(50.0) 16(41.0) Second line 3(6.4) 0(0.0) 3(7.7) Initiation of zidovudine prophylaxis (N=42) 0.14 First Trimester 3(7.1) 1(14.3) 2(5.7) Second Trimester 27(64.3) 2(28.6) 25(71.4) Third Trimester 12(28.6) 4(57.1) 8(22.9) CD4 count Antenatal absolute CD4 cells/µl* 422(285,544) 467(349,642) 411(285,521) 0.23 Delivery absolute CD4 count cells/µl 344(246,502) 308(218,601) 350(249,487) 0.96 Delivery percent 26.1(21.6,32.5) 29.7(22.3,33.6) 25.6(21.7,31.6) 0.31 Antenatal CD4 count category*: 0.38 <350 cells/µl 31(35.2) 4(26.7) 27(37.0) 350-499 cells/µl 30(34.1) 4(26.7) 26(35.6) >500 cells/µl 27(30.7) 7(46.7) 20(27.4) Delivery CD4 count category: 0.04 <350 cells/µl 46(51.7) 9(60.0) 37(50.0) 350-499 cells/µl 20(22.5) 0(0.0) 20(27.0) >500 cells/µl 23(25.8) 6(40.0) 17(23.0) Absolute change in CD4 count -40(-178,70) -53(-207,31) -40(-175,96) 0.45 Percent change in CD4 count -7.8(-39.4,22.6) -5.7(-47.2,11.9) -14.4(-38.4,29.8) 0.26 Delivery HIV viral load#: Category in copies/ml (%) 0.84 <40 copies/ml 31(38.3) 5(38.5) 26(38.2) 40-999 copies/ml 24(29.6) 5(38.5) 19(27.9) 1000-9999 copies/ml 13(16.0) 2(15.4) 11(16.2) > 10 000 copies/ml 13(16.0) 1(7.7) 12(17.6) Log10 HIV viral load 2.01(1.59,3.08) 1.72(1.60,2.88) 2.05(1.59,3.21) 0.44 ARV – antiretroviral, cART – combination antiretroviral therapy; * 1 antenatal CD4 count missing in no outcome; # 8 HIV viral loads missing (2 in group with the outcome, 6 in group without the outcome) 121 Table 7.10 Household characteristics compared by HEU infant outcome group (numeric variables as median (IQR), categorical variables as number (%)) Total N=89 Infants hospitalized (N= 15) Infants not hospitalized (N= 74) P-value Type of house: 0.39 Stand alone house 31(34.8) 4(26.7) 27(36.5) Apartment in apartment block 5(5.6) 1(6.7) 4(5.4) House/flat/room in backyard 4(4.5) 0(0.0) 4(5.4) Shack in backyard 33(37.1) 8(53.3) 25(33.8) Shack not in backyard 14(15.7) 1(6.7) 13(17.6) Other 2(2.2) 1(6.7) 0(0.0) Water supply: 0.07 Water piped into dwelling 38(42.7) 9(60.0) 29(39.2) Water piped into yard 47(52.8) 5(33.3) 42(56.8) Public tap 4(4.5) 1(6.7) 3(4.1) Sanitation: 1.00 Flush toilet (connected to sewerage) 87(97.8) 15(100) 72(97.3) Flush toilet (septic tank) 2(2.2) 0(0.0) 2(2.7) No toilet facilities 0(0.0) 0(0.0) 0(0.0) Fuel (cook): 0.13 Electricity 83(93.3) 15(100) 68(91.9) Gas 3(3.4) 0(0.0) 3(4.1) Paraffin 3(3.4) 0(0.0) 3(4.1) Fuel (heat): 0.68 Electricity 34(38.2) 5(33.3) 29(39.2) Gas 1(1.1) 0(0.0) 1(1.4) Paraffin 47(52.8) 9(60.0) 38(51.4) Firewood/Other 7(7.9) 1(6.7) 6(8.1) Fuel (light): 0.02 Electricity 84(94.4) 15(100) 69(93.2) Paraffin 5(5.6) 0(0.0) 5(6.8) Number of rooms in house 2(1,3) 1(1,2) 2(1,3) 0.57 Number of people living in household 4(3,5) 4(3,6) 4(3,5) 0.64 Distance to clinic (minutes walked) 20(10,30) 20(12.5,30) 20(10,30) 0.20 Assets: Radio 51(57.3) 7(46.7) 44(59.5) 0.50 TV 74(83.1) 12(80.0) 62(83.8) 0.17 Computer 10(11.2) 2(13.3) 8(10.8) 1.00 Refrigerator 66(74.2) 12(80.0) 54(73.0) 0.42 Home phone 4(4.5) 1(6.7) 3(4.1) 0.62 Cell phone 88(98.9) 14(93.3) 74(100) 0.47 Bicycle 10(11.2) 3(20.0) 7(9.5) 0.74 Motorcycle or motor scooter 2(2.2) 0(0.0) 2(2.7) 0.50 122 7.3.3 Infant characteristics HEU infants with and without the primary outcome were of similar mean gestational age (38.5 weeks (SD 1.5) vs. 38.8 weeks (SD 1.5) respectively) and birth weight (3073 grams (SD 380) vs. 3118 grams (SD 381) respectively) (Table 7.11). Of the 69 infants in follow-up at six months, four (5.8%) infants did not have complete immunizations at six months of age, one (8.3%) infant with an infectious cause hospitalization and 3 (5.3%) infants without a hospitalization. Infants with the outcome had a median of two (IQR 1,3) sick-clinic visits compared to a median of zero (IQR 0,1) sick clinic visits in infants without the outcome (p<0.001). Only 60% (50/83) of infants were receiving trimethoprim-sulphamethoxazole prophylaxis at eight weeks of age but this did not differ significantly between infants with and without an infectious cause hospitalization. There was no significant difference in infant feeding mode, and approximately two-thirds of infants with and without the outcome were not breastfeeding at two weeks of age (Table 7.12). However, there was a consistent tendency for hospitalized HEU infants to less often be breastfed at two, four and six months than HEU infants who were not hospitalized. Table 7.11 Infant birth and healthcare characteristics compared by infant outcome group (numeric variables as median (IQR), categorical variables as number (%), unless otherwise stated) Total N=89 Infants hospitalized (N= 15) Infants not hospitalized (N= 74) P-value Male 43(48.3) 6(40.0) 37(50.0) 0.67 Gestational age in weeks – mean (SD) 38.8(1.5) 38.5(1.5) 38.8(1.5) 0.29 Birth weight in grams – mean (SD) 3110(379) 3073(380) 3118(381) 0.68 Low birth weight 6(6.7) 1(6.7) 5(6.8) 1.00 Immunizations not up to date at 6 months* 4(5.8) 1(8.3) 3(5.3) 0.28 All cause clinic visits per infant 6(5,7) 6.5(6,7) 6(5,7) 0.44 Sick clinic visits per infant 1(0,1) 2(1,3) 0(0,1) <0.001 Receiving CPT at 8 weeks# 50(60.2) 7(53.8) 43(61.4) 0.60 CPT – cotrimoxazole preventive therapy; * 20 unknown (3 in group with the outcome, 17 in group without the outcome); # 6 unknown (2 in group with the outcome; 4 in group without the outcome) 123 Table 7.12 Infant feeding characteristics compared by HEU infant outcome group (categorical variables as number (%)) Total N=89 Infants hospitalized (N= 15) Infants not hospitalized (N= 74) P-value Intention to exclusively breastfeed 35(39.3) 7(46.7) 28(37.8) 0.66 Feeding mode at 2 weeks: (N=89) 1.00 Exclusive breastfeeding 30(33.7) 5(33.3) 25(33.8) Partial breastfeeding 3(3.4) 0(0.0) 3(4.1) No breastfeeding 56(62.9) 10(66.7) 46(62.2) Feeding mode at 2 months: (N=88) 0.36 Exclusive breastfeeding 22(25.0) 2(13.3) 20(27.4) Partial breastfeeding 4(4.5) 0(0.0) 4(5.5) No breastfeeding 62(70.5) 13(86.7) 49(67.1) Feeding mode at 4 months: (N=80) 0.43 Exclusive breastfeeding 10(12.5) 1(7.1) 9(13.6) Partial breastfeeding 7(8.8) 0(0.0) 7(10.6) No breastfeeding 63(78.8) 13(92.9) 50(75.8) Feeding mode at 6 months: (N=69) 1.00 Exclusive breastfeeding 4(5.8) 0(0.0) 4(7.0) Partial breastfeeding 7(10.1) 1(8.3) 6(10.5) No breastfeeding 58(84.1) 11(91.7) 47(82.5) 124 7.4 Determining the effect of maternal ARV regimen during pregnancy on HEU infant risk for infectious morbidity To resolve the final objective of determining whether HEU infants born to mothers on maternally-indicated cART during pregnancy have a greater odds of infectious morbidity in the first six months of life compared to HEU infants born to mothers on VTP prophylaxis, multivariable logistic regression was used. The crude association between HEU infants of mothers on cART compared to HEU infants of mothers on VTP prophylaxis showed no difference between the two groups in the odds of infectious morbidity in the first six months of life (OR 1.03, 95% CI 0.33, 3.20) (Table 7.13). Controlling for maternal factors including age, parity and timing of maternal HIV diagnosis did not alter this relationship. Maternal CD4 count was not adjusted for as the maternal groups were defined partly on antenatal CD4 count. Infant birth characteristics and trimethoprim-sulphamethoxazole prophylaxis received at two months of age were also not associated with an infectious cause hospitalization and did not alter the effect of maternal cART or VTP prophylaxis. Although receipt of any breastfeeding at all time points appears to reduce the odds of infectious morbidity, controlling for any breastfeeding did not alter the relationship between maternal ARV group and infant infectious morbidity within this sub-group of HEU infants. When considering the secondary outcomes of at least one severe or at least one very severe infectious cause hospitalization the odds ratios also remained close to 1. In this sub-group of HEU infants, there was no evidence that infants born to mothers on maternally indicated cART were at greater risk for all infectious morbidity or severe infectious morbidity compared to infants born to mothers on VTP prophylaxis. 125 Table 7.13 Odds ratios for the associations between maternal and infant characteristics and the primary outcome, adjusted for maternal ARV regimen aOR (95%CI) P-value cART group aOR (95%CI) P-value Maternal characteristics Maternal age in years 0.97 (0.87, 1.08) 0.64 1.09 (0.35, 3.54) 0.88 Multiparous (Reference group is primiparous) 1.26 (0.29, 8.70) 0.78 1.02 (0.33, 3.09) 0.97 Gestation in weeks at 1st ANC visit 1.00 (0.91, 1.10) 0.97 1.03 (0.33, 3.21) 0.96 Number of ANC visits 0.94 (0.69, 1.29) 0.71 1.03 (0.34, 3.24) 0.95 BMI (kg/m2) postnatal 1.04 (0.91, 1.19) 0.58 1.08 (0.35, 3.41) 0.89 HIV diagnosis during pregnancy (Reference group is pre-pregnancy) 1.49 (0.45, 5.25) 0.85 0.89 (0.26, 3.00) 0.52 Log10 HIV viral load 0.73 (0.34, 1.38) 0.36 0.85 (0.24, 3.10) 0.81 HIV viral load category > 40 copies/ml (Reference is < 40 copies/ml) 0.74 (0.18, 3.01) 0.68 0.96 (0.23, 3.86) 0.96 Infant characteristics Gestational age in weeks 0.88 (0.62, 1.27) 0.47 0.99 (0.32, 3.11) 0.99 Birth weight in kg 0.73 (0.16, 3.22) 0.68 1.00 (0.32, 3.15) 0.98 Low birth weight (<2500g) 0.98 (0.05, 6.79) 0.99 1.03 (0.33, 3.21) 0.96 Trimethoprim-sulphamethoxazole at 2 months 0.62 (0.19, 1.93) 0.41 1.11 (0.36, 3.55) 0.86 Receiving any breast milk (Reference is no breast milk) at: 2 weeks 0.82 (0.23, 2.59) 0.74 1.00 (0.33, 3.16) 0.99 2 months 0.30 (0.04, 1.20) 0.13 0.97 (0.31, 3.08) 0.96 4 months 0.24 (0.01, 1.33) 0.18 1.00 (0.32, 3.17) 0.99 6 months 0.41 (0.02, 2.37) 0.41 1.05 (0.34, 3.30) 0.93 ANC – antenatal clinic; aOR – adjusted odds ratio; cART – combination antiretroviral therapy 126 Chapter 8 Discussion HEU infants experience numerous exposures during early life that could theoretically increase their vulnerability to infectious diseases during infancy compared to HIV unexposed infants. We understand little though, about how HEU infants compare to HUU infants experiencing similar social, environmental and nutritional constraints to their health. As explored in Chapter 1, evidence directly comparing infectious morbidity in HEU and HUU infants, particularly in Africa where the majority of HEU infants are born, is limited (Table 1.1). The avoidance or attenuation of breastfeeding in HIV exposed infants has certainly played a role in exacerbating infant morbidity and mortality as have poor birth outcomes (preterm birth, low birth weight, small for gestational age) associated with maternal HIV. Poverty and compromised socio-economic circumstances may further concentrate risk for infectious and other morbidities amongst children in HIV-affected households. Perturbations of the HEU infant immune system have been observed but have not been investigated in association with infectious morbidity. Through the work presented in this dissertation we have sought to understand whether a difference in risk remains, beyond the morbidity associated with universal risk factors for infant morbidity. In this context should HEU infants still experience a greater risk for infectious morbidity, further investigation of associations between immunological aberrations and infectious morbidity would be warranted. Interrogation of the clinical pattern of infectious morbidity observed could aid understanding of the potential mechanism of vulnerability and define a focus for further immunologic investigations within the vastness of the infant immune system. This discussion proceeds firstly with an overview of the mothers, households and infants in the study in section 8.1, so as to understand how comparable HEU and HUU infants were in terms of potential confounders of infectious morbidity. This is followed by consideration of similarities, differences and explanations for the infectious morbidity observed in HEU and HUU infants in this cohort as well as compared to the existing understanding in the literature of HEU infant infectious morbidity in sections 8.2 and 8.3. A closer look is then taken at the HIV-infected mothers and their HEU infants and whether infectious morbidity risk within HEU infants is associated with advanced maternal HIV requiring cART in section 8.4. Observations with regards to potential effects of in utero ARV exposure are discussed in section 8.5. The chapter concludes with discussion of the challenges, limitations and implications of this study in sections 8.6 and 8.7. 127 8.1 The mothers, households and infants in this study To pursue the primary study objective of comparing infectious morbidity in the first six months of life between HEU and HUU infants, this study was designed to reduce potential confounding by maternal obstetric or other co-morbidities, poor infant birth outcomes and dissimilar household and socioeconomic environments. This aim of the study design was largely met. The only differences in maternal characteristics were that HIV-infected mothers in this cohort were significantly older and more often multiparous than HIV-uninfected mothers, a known trend when comparing HIV-infected and HIV-uninfected pregnant women (38,41,42). Older maternal age has been associated with lower neonatal and infant mortality and could be a confounder to infant infectious morbidity (209,210). The two groups of pregnant women otherwise were quite similar in terms of demographic and social factors and reflected the community in which the study was set. HIV-infected and HIV-uninfected mothers had similar obstetric courses, with no mothers experiencing major obstetric morbidities that could have been associated with infant morbidity. Maternal syphilis and TB occurred infrequently in this cohort, and there was no statistically significant difference between smoking or alcohol use during pregnancy in HIV-infected and HIV-uninfected women, all factors that could be associated with an increase in infant infectious morbidity. HEU and HUU infants experienced similar household circumstances as well as access to municipal services. The typical house in the cohort was a two roomed informal shack with access to piped water, a flush toilet and electricity, occupied by four people and was less than a 30 minute walk away from the nearest primary healthcare clinic. The similarity in the household and social environments experienced by both HEU and HUU infants is reassuring that potentially confounding socioeconomic differences between the two groups were limited by the study design. Importantly, restricting enrolment to well-defined low socioeconomic neighbourhoods and frequency matching HIV-infected and HIV-uninfected mothers on rac/ethnicity ensured similar household circumstances and maternal social habits respectively. The most important difference between HEU and HUU infants was the proportion of infants who were breastfed: two thirds of HEU infants compared to a single HUU infant were never breastfed. The infants in this cohort were of term gestation, of normal birth weights and displayed appropriate uptake of immunizations by six months of age. The important clinical differences between HEU and HUU infants were an unsustained lower haemoglobin and a shorter length-for-age in HEU infants that are discussed further in relation to HEU infant in-utero ARV exposure in section 8.5. Essentially this cohort of HEU and HUU infants and their mothers were well matched except for potential confounding differences of older age in HIV-infected mothers and a lower proportion of breastfed HEU infants. These differences were taken into account in the analysis to determine the probability of infectious cause hospitalization or death in HEU compared to HUU infants in the first six months of life. 128 8.2 Infectious morbidity in HEU compared to HUU infants In the absence of an existing diagnostic classification system that met the needs of our study, a study specific tool was designed for outcome classification according to the type and severity of the infectious cause hospitalizations. The PIET-R was evaluated for precision and accuracy through a separate sub-study of 50 hospitalization events. There are certainly limitations to the PIET-R classification, but importantly the results of the evaluation indicated that neither the occurrence of diarrhoea and LRTI events nor the severity of events would have been over-classified. Firm conclusions to resolve the primary objective of whether a difference in infectious morbidity exists between HEU and HUU infants cannot be drawn from the results of this study. Of note, though, is the consistency in the observed pattern of increased severity of infectious events in HEU infants with a greater proportion of these infants experiencing infectious cause hospitalizations, very severe hospitalizations and a longer length of hospital stay, despite a similar rate of sick-clinic visits to HUU infants. There is evidence from this cohort of three findings: 1) that even in the presence of breastfeeding HEU infants could have a four times greater odds of a very severe infectious cause hospitalization compared to HUU infants; 2) infectious morbidity in HEU infants that are not optimally breastfed was no different to that experienced by HUU infants also not optimally breastfed; and 3) a low maternal CD4 count was not associated with infectious morbidity in this cohort. Discussion follows, in the rest of section 8.2, on the burden of infectious morbidity observed in the cohort and the pattern of infectious morbidity, specifically the types of infections, severity and timing of the events and how this compares to previous studies. The CD4 dynamics observed as well as the influence of breastfeeding on infectious morbidity are also discussed. 8.2.1 All cause infectious morbidity HEU and HUU infants in this cohort had a similar rate of all-cause sick clinic visits (unadjusted RaR 0.82, 95% CI 0.58, 1.16), indicating not only that the incidence of infections was similar, but also that both groups of infants accessed health care services in an equivalent manner with similar attention paid to infant symptoms by caregivers. Despite this similar rate of sick-clinic visits, 18% of all HEU infants, experienced a primary outcome event of at least one infectious cause hospitalization or death in the first six months of life compared to 12% of all HUU infants. The 18% hospitalization rate observed in HEU infants corresponds to that documented in Botswanan and South American and Caribbean HEU infants in the first 6 months of life, 16% and 17.5% respectively (39,51). In contrast the rate we observed in HUU infants of 12% was somewhat higher than the 7% observed in HUU infants in the PROMISE EBF study conducted in Kwa-Zulu Natal, South Africa and 6% of HUU infants hospitalized in the first six months of life in the Botswanan Mashi study (39,211). The similar rate of HEU hospitalization but higher rate of HUU hospitalization in our cohort compared to other similar cohorts is reassuring to some extent that the HUU 129 hospitalization rate was not underestimated and thus the difference in the rate of hospitalization between HEU and HUU infants in our cohort was less likely to have been overestimated. The unadjusted association between HIV exposure and the primary outcome of at least one infectious cause hospitalization or death showed HEU infants in our cohort to have a 48% greater risk (RiR 1.48, 95% CI 0.72, 3.06) of infectious cause hospitalization or death than HUU infants. This is very similar to a recently published Mozambican prospective cohort study that included 153 HIV-uninfected infants and observed an unadjusted RaR of 1.51 (95% CI 0.71, 3.18) for hospital admission in HEU compared to HUU infants in the first year of life (44). As in our cohort, there was no difference in out-patient sick-clinic visits between HEU and HUU infants. Moreover, in the presence of a similar rate of all acute lower respiratory tract infections, HEU infants in the Mozambican study tended to have a greater risk for clinically severe pneumonia than HUU infants (RaR 5.60, 95% CI 0.67, 46.55, p = 0.056). These findings corroborate the pattern seen in our South African infants of a similar overall incidence of infections, but a tendency to more often require hospitalization and experience more severe disease than HUU infants in this region. 8.2.2 Lower respiratory tract infectious morbidity Lower respiratory tract infections predominated in this cohort and were the most frequent reason for a sick-clinic visit and for hospitalization in both HEU and HUU infants. LRTIs accounted for 51% of hospitalizations and occurred at least once in 11% of HEU and 9% of HUU infants. A number of studies have noted the burden of respiratory tract infections in HEU infants, but without an HUU control group to compare to it has not been possible to understand whether these HEU infants are experiencing the morbidity expected of all infants in their context or something in excess of that (10,50,169,170). Although the most common reason for admission of HEU infants in our cohort was for an LRTI, this was no different to the proportion of HUU infants requiring hospitalization for LRTIs. This same observation was made comparing HEU and HUU Botswanan infants, in whom there was no difference in pneumonia episodes in the first six months of life (39). They did however differ by 24 months of age with a greater proportion of HEU infants having experienced a pneumonia episode by this age. The Drakenstein Child Health Study (DCHS), an ongoing birth cohort study in the Western Cape Province about 50 km from Kraaifontein, where our study was set, has recently observed an incidence of pneumonia in HIV-uninfected infants, including both HIV exposed and unexposed, much higher than previous modelled estimates of pneumonia incidence for low-middle income countries and for South Africa specifically (47). This highlights the limitations in using national or regional rates as well as modelled estimates in making comparisons and the need for direct comparison of infants with and without whatever the relevant exposure is under study, from the same context to understand differences in morbidity outcomes. In the DCHS, HEU infants had a marginally significantly greater risk for all 130 pneumonia in the first year of life, RiR 1.62 (95% CI 1.01, 2.61) following adjustment for maternal smoking and any formula feeding before six months of age (47). HEU infants however, were at substantially higher risk for severe pneumonia with an adjusted RiR of 4.04 (95% CI 1.51, 10.8). This is in keeping with our observation that HEU infants are vulnerable to a greater severity but not necessarily an increased frequency of common childhood infectious diseases. There is evidence now from a number of studies to support that while HEU infants experience a substantial burden of respiratory tract infections as reflected by the incidence of disease, this burden is no different to HUU infants in the same setting in Southern Africa (39,47). However, as we have shown, these two groups of infants could differ in the severity of the LRTIs, an observation that deserves further evaluation and is elaborated on in 8.2.4. 8.2.3 Diarrhoeal morbidity Although LRTIs were the most frequent reason for hospitalization in both HEU and HUU infants in our cohort, a difference between HEU and HUU infants was observed in terms of diarrhoeal disease. Significantly more HEU infants were hospitalized at least once with any diarrhoeal disease than HUU infants, 9% vs. 1% respectively. The rate of diarrhoea was generally low in our cohort with only 10% of infants having a sick-clinic visit for diarrhoea compared to more than 30% of infants before six months of age experiencing any diarrhoea episode in the similar Botswanan Mashi cohort (39). This difference may be a reflection of the municipal services and infrastructure available to the infant households in our cohort where 93% of households had a safe piped water source and 98% of households had a flush toilet connected to sewerage. The number of infants with diarrhoea in our cohort was too small to interrogate further the association between diarrhoea specifically and the absence of breastfeeding in the HEU infants, which may be the most obvious explanation for this difference between the two groups. However, the rate of diarrhoea-associated sick clinic visits was no different in HEU and HUU infants (RaR 1.13, 95% CI 0.42, 3.25) indicating that the frequency of diarrhoea was similar in HEU and HUU infants but HEU infants experienced diarrhoea events of greater severity that significantly more often required hospitalization. If the deprivation of the protective effects of breastfeeding and the risk of bacterial contamination of formula milk were the major drivers of this difference it might be expected that the frequency of diarrhoeal episodes, in general, would be increased in HEU infants as represented by an increased rate of diarrhoea-associated sick clinic visits. That was not the case in this cohort. 8.2.4 Pattern of infectious morbidity: the timing and severity of infectious events The temporal pattern of infectious cause hospitalizations as well as the severity of infectious events tended to differ between HEU and HUU infants. Three-quarters of all the primary outcome events (infectious cause hospitalization or death) had occurred within the first three months of life with no difference by three months of age in the proportion of HEU and HUU infants that had experienced a primary outcome event. Between three and six months of age the two groups diverged with a continuation of infectious cause hospitalizations in HEU infants but a tapering of hospitalizations in HUU 131 infants. HEU infants also tended to experience a greater severity of infections on the background of similar overall frequency of events. The all-cause sick-clinic visit rate did not differ between HEU and HUU infants, but HEU breastfed infants had a greater probability of hospitalization with very severe infectious events than HUU breastfed infants. There is existing evidence in the literature to support this pattern of timing and severity, but these links have not previously been explicitly made or explored. A discussion of how the current study’s findings fit within the existing evidence of the timing and severity of infections in HEU infants follows in sections 8.2.4.1 and 8.2.4.2 respectively. 8.2.4.1 Timing of infectious morbidity The pattern of no difference in infectious morbidity in the neonatal and early infant period, but rather an increased risk in HEU infants during mid-infancy, is in keeping with the pattern of infant mortality observed in the ZVITAMBO study (37). In ZVITAMBO the highest absolute mortality in HEU infants occurred in the neonatal period, but this was no different to the mortality experienced by HUU infants during the neonatal period. The relative increase in HEU mortality was greatest in the period from two to six months of life, reaching 5.5 (95% CI 3.9, 7.9) times that of HUU infants (37). In keeping with this pattern that greater HEU infant risk becomes apparent in mid-infancy and not during the neonatal period are the findings from a large South African neonatal sepsis prevention trial (42). HEU infants were found to have equivalent rates of neonatal sepsis, in the first 28 days of life, to HUU infants. The prevalence of vaginal colonization of the common causative organisms of neonatal sepsis was similar in HIV-infected and HIV-uninfected mothers, that could explain why the risk of early onset neonatal sepsis in HEU and HUU neonates was no different (42). South African and Belgian HEU neonates do have an increased risk of specifically Group B Streptococcus (GBS) infection compared to HUU neonates (173,174). However, the greatest difference in risk between HEU and HUU infants in both South Africa and Belgium was for late-onset (7 to 90 days of life) as opposed to early onset (0 to 7 days of life) GBS. There is evidence to support that in Southern Africa where risk for neonatal sepsis is high in all infants, there is no difference in the rate of all-cause neonatal sepsis between HEU and HUU infants (42,63). However there may well be a greater risk for infectious morbidity and mortality in the post-neonatal and middle infancy period and for some specific Streptococcal infections (34,37,174). Our findings contribute to this evidence. It is postulated that this pattern may be in keeping with two described immunological aberrations in HEU infants: 1) poor placental transfer of maternal antibodies from HIV-infected women to their foetus in utero and 2) delayed adaptive immune system development in HEU infants. These are discussed in detail in section 8.3. 132 8.2.4.2 Severity of infectious morbidity Early studies observed a greater relative difference in infant mortality than in infant morbidity or hospitalization in HEU compared to HUU infants (37–39). Additionally, Southern African HEU infants have a substantially increased risk for severe community acquired pneumonia but only a marginally increased risk for all community acquired pneumonia (44,47). Two studies looking at outcome 48 hours after commencing treatment for severe pneumonia, one in South Africa and the other in Botswana, found that HEU infants more often failed empiric antibiotic therapy for pneumonia and had higher in-hospital mortality than HUU infants (35,40). This further speaks to the possibility of more severe disease in HEU than HUU infants. The current study is the first to have tried to explore more than just the incidence of infections, but has attempted to understand a possible difference in severity of infections between HEU and HUU infants. Our observation that HEU infants have a higher probability of very severe infections, defined as infectious events with signs of a severe event persisting for at least 48 hours, supports these trends described in the existing literature. This pattern potentially indicates that it is not more frequent exposure to common infectious pathogens, but possibly an altered immunological or inflammatory response to the pathogens that results in more severe manifestations of disease in HEU infants. In relation to the more severe respiratory infections, an alternative potential explanation is that HEU infants more often experience infections due to atypical pathogens that do not adequately respond to empiric antibiotic therapy for community acquired pneumonia (196). There are numerous published case reports of HEU infants with Pneumocystis pneumonia and unpublished reports of confirmed CMV pneumonitis in HEU infants (40,162,163) (personal communication Dr H. Rabie & Dr. A. Engelbrecht, 2015). In spite of the limitations of isolated case reports, they do indicate that HEU infants can be infected with unusual pathogens that cause very severe respiratory disease. Their vulnerability to these atypical organisms could be due to a combination of increased exposure to opportunistic organisms and altered immune response (21,163). 8.2.5 The relationship between maternal CD4 count and infant infectious morbidity Two observations in relation to maternal CD4 count in both HIV-infected and HIV-uninfected mothers in our cohort require further consideration: 1) the lower than expected CD4 count in HIV-uninfected mothers and 2) the absence of an association between a low maternal CD4 count and infant infectious morbidity in either HEU or HUU infants. These two observations are discussed. The median CD4 count of HIV-uninfected mothers in this cohort was lower than expected at 467 (IQR 363, 675) cells/µl. Twenty two percent (18/82) of HIV-uninfected mothers in our cohort had a CD4 count of below 350 cells/µl at delivery. This is below the 2.5th percentile lower bound of the normal CD4 reference range for South African HIV-uninfected women that is 500 cells/ µl (109). The CD4 count has been observed to drop below 350 cells/µl in the absence of HIV-infection in African adults 133 (37,109,212,213). CD4 count is also known to decline during pregnancy and rebound again post-partum, possibly due to haemodilution of pregnancy or due to a required maternal state of relative immune suppression to ensure foetal allograft survival and avoid rejection of foreign foetal tissue by the maternal immune system (214,215). Compared to Zimbabwean and Botswanan HIV-uninfected pregnant women the HIV-uninfected women in our cohort still displayed markedly lower CD4 counts at delivery, Zimbabwean and Botswanan women having a mean CD4 count respectively of 782 (95%CI 759, 805) cells/µl at delivery and 626 (95% CI 599, 653) cells/µl antenatally (212,216). The HIV-uninfected mothers in our cohort had CD4 counts even lower than those documented for HIV-uninfected peri-partum women in neighbouring countries. There is no obvious explanation for this in our cohort. In the absence of CD4 reference ranges for South African HIV-infected and HIV-uninfected pregnant women, further study and understanding of this observation could be valuable. It was observed in both HEU and HUU infants in our cohort that a low maternal CD4 count at delivery was not associated with infectious cause hospitalization or death. More than half of the infants that experienced a primary outcome event were born to mothers with a CD4 count of 500 cells/µl or greater and no HUU infants born to HIV-uninfected mothers with an unusually low CD4 count of below 350 cells/µl experienced an infectious cause hospitalization or death. The absolute CD4 count can not be interpreted as an equivalent representation of immune function in HIV-infected and HIV-uninfected women in this cohort. HIV-infected women have a chronic infectious disease that specifically targets CD4 lymphocytes while HIV-uninfected women do not have an immune compromising disease and are generally healthy. A CD4 count of below 350 cells/µl as an isolated measure in an HIV-uninfected women of otherwise good health cannot be equated to a CD4 count of below 350 cells/µl in an HIV-infected women in whom it signifies immune suppression. As illustrated by our study, 22% of HIV-uninfected mothers had a CD4 count of below 350 cells/µl, and these mothers and their infants were healthy, showing that measurement of a single parameter cannot fully represent immune function. It was decided a priori that maternal CD4 count could lie on the causal pathway between HIV exposure and infant infectious morbidity and thus would not be adjusted for in multivariable models of this association. Adjusting for CD4 count in multivariable models including both HIV-infected and HIV-uninfected women may adjust for different indicators in each group of women. Despite CD4 count having a statistically significant effect on the odds of infectious cause hospitalization or death, the a priori analytic strategy was maintained. The observation in this cohort that a higher, as opposed to a lower, maternal CD4 count was associated with increased infectious morbidity in both HEU and HUU infants is probably a spurious finding in HUU infants, but may be important in HEU infants and is discussed in detail in section 8.4. It is at least reassuring that within HIV-uninfected women, the infants of mothers with an unusually low CD4 count were not compromised in any obvious manner. 134 8.2.6 Infectious morbidity risk beyond suboptimal breastfeeding This study did not set out to determine the effect of breastfeeding on HEU infant infectious morbidity. Rather, the primary objective was to determine whether there is a greater probability of infectious morbidity in HEU infants relative to HUU infants through HIV-specific pathways besides those related to avoidance of breastfeeding and other universal infant risk factors. This required that the association between breastfeeding and infectious morbidity be measured and appropriately adjusted for in analysis. Despite our imperfect quantification of the effect of breastfeeding, the stratified analysis conditioned on the presence of any breastfeeding at two weeks proved valuable in understanding whether HIV-specific factors other than breastfeeding could have an influence on HEU infant infectious morbidity. HEU infants in our cohort who did receive any breastfeeding for a similar duration to HUU infants who received any breastfeeding, had a substantially greater odds of very severe infectious morbidity compared to their HUU breastfed peers (aOR 4.21, 95% CI 1.00, 19.22, p = 0.05). Thus, in the context of similar breast milk exposure in both HEU and HUU infants a mechanism other than the lack of protection afforded by breast milk should be considered for the increased probability of very severe infectious morbidity seen in HEU infants. Breastfeeding is protective regardless of infant HIV exposure, and it is clearly established in the literature that avoidance or attenuation of breastfeeding has been detrimental to HEU infant health and survival in Southern Africa (32,70,75–77). In our cohort, irrespective of HIV exposure, all infants who were not breastfed experienced a greater probability of infectious morbidity than infants who were either exclusively or partially breastfed at all evaluated time points. It is recognized that the estimates of this effect were equivocal in this study, but the point estimates are in keeping with what would be expected based on prior evidence of the detrimental effect of not breastfeeding (70). Part of the challenge in understanding the magnitude of the effect of breastfeeding on reducing infectious morbidity is that ultimately, in statistical models, breastfeeding is represented simplistically. The models do not take into account the entirety of breast milk exposure including duration, quantity and quality of breast milk. Few studies have had sufficient size to consider infant feeding according to all the WHO defined categories (exclusive breastfeeding, predominant breastfeeding, partial breastfeeding and no breastfeeding), or been able to accurately incorporate the time spent in each of these categories in multivariable models (41). Added to this is that infant feeding choices and the counselling surrounding infant feeding in the context of a high HIV prevalence has undergone many changes over the last two decades as infant feeding policies have changed (24). At the time of our study, the Western Cape Province was transitioning away from providing free infant formula to HIV-infected mothers who chose to breastfeed, to promoting breastfeeding as the feeding mode of choice for all infants, with antiretroviral prophylaxis to HIV-exposed breastfeeding infants. 135 The observation that fewer HEU infants of mothers with undetectable HIV viral loads were breastfed than HEU infants of mothers with detectable HIV viral loads is concerning. It may indicate that VTP guidelines are not reaching their intended goal. In terms of HIV transmission risk breastfeeding is safest in mothers who have a suppressed HIV viral load and whose infants are exposed to the lowest possible viral dose (217). One of the possible benefits of expanding cART availability in South Africa to all HIV-infected pregnant women irrespective of CD4 count, is that HIV-infected mothers can breastfeed with less fear of HIV transmission to their infants (113). It is possible that mothers on cART with more advanced disease chose not to breastfeed due to constraints on their physical capacity to breastfeed. However, HIV-infected mothers in our cohort did not experience major comorbidities and none required hospitalization during their infants first six months of life. The majority of HIV-infected people once on cART, and particularly once their HIV viral load is suppressed, experience a general improvement in physical health (218). This study represents a single community of HIV-infected women, that in most respects is very similar to many of the HIV-infected communities in South Africa. Although this single community observation can not be generalized to all HIV-infected women in South Africa, it signifies that we do not fully understand how the infant feeding messages provided by primary healthcare workers are understood and applied in broader community contexts with many other competing priorities that are taken into consideration in a mothers decision to breastfeed. This study cannot comment on the impact of infant feeding choices on HEU infant infectious morbidity in South Africa, and has likely done an injustice to the true effect of breastfeeding simplistically represented in the statistical models. It is fully recognized that breastfeeding is to the benefit of HEU infants and has a large role to play in reducing their excess morbidity and mortality. We have evidence though that supports that pathways other than deprivation of breast milk require consideration and study to comprehensively understand HEU infant risk for infectious morbidity. 8.3 Potential immunologic explanations for HEU infant infectious morbidity The difference in infectious morbidity between HEU and HUU infants in this cohort was apparent in the severity and timing of infectious events. Barring HIV exposure, these two groups of infants were very similar on factors representing a number of pathways to infectious morbidity, specifically their socioeconomic environment, maternal obstetric and general health, birth outcomes, immunization and nutritional status. Breastfeeding, or the lack thereof in HEU infants did explain some of their increased infectious morbidity, but not all of it. Three possible immunologic explanations for the pattern of infectious morbidity observed are discussed. These include 1) the poor transplacental transfer of maternal antibodies; 2) delayed functional infant immune development and 3) breast milk immunologic quality in HIV-infected mothers. 136 8.3.1 Deficient transplacental antibody transfer It is well established that the active transfer of maternal antibodies across the placenta is deficient in HIV-infected women but this has not yet been directly associated with an increased risk for infections in HIV exposed infants (125,127,128). Maternally acquired antibody levels are highest at birth and wane with time to reach a nadir by approximately six months of age (172). By this stage the normally developing infant immune system has already gained capacity to respond to infectious threats and infants have received most of their primary series of vaccinations. Although HEU infants may start life with lower levels of maternally acquired antibodies than HUU infants, these levels may still initially be sufficient to afford protection during the neonatal period. The initial quantity of maternal antibodies received by the foetus determines the duration of passive protection provided (219). It is possible that as maternally acquired antibody levels wane they reach the nadir earlier in HEU infants than HUU infants and before the HEU infant immune system has developed sufficient competence. This would leave these infants with a critical period of vulnerability during the post-neonatal period when maternally acquired antibodies have declined but infants have not yet developed their own immune competence. HEU infants respond as well, if not better, to vaccination than HUU infants in terms of the quantity of vaccine specific antibodies they produce (128,131). However, little is known about the functional capacity of these antibodies and whether they afford equivalent protection as found in HUU infants . Deficient maternal antibody transfer in HIV-infected women is associated with a high HIV viral load but not with a low CD4 count (127). One of the characteristics of the dysfunctional immune activation experienced by HIV-infected people not yet on cART is a non-specific hypergammaglobulinaemia (107). It has been postulated that in the presence of this non-specific hypergammaglobulinaemia there may be competition between the non-specific IgG antibodies and the beneficial specific IgG antibodies for the placental receptors (hFcRN) that actively transport IgG across the placenta from the maternal to the foetal side (125). If this is so, maternal cART, by controlling HIV viral replication and non-specific hypergammaglobulinaemia, may aid in improving transplacental transfer of maternal antibodies. What is known about maternal-foetal antibody transfer in HIV-infected women comes from the pre-antiretroviral therapy era in women largely not on cART. The effect of maternal cART on the dynamics of maternal-foetal antibody transfer has not yet been studied. 8.3.2 Delayed functional immune development Whether HIV exposed or not, the neonatal period holds the greatest absolute risk for infectious morbidity of any period during life, due in part to the immature nature of the newborn immune system (172). Under appropriate circumstances, the neonatal and infant immune system gains in function and competence and can better protect against infectious insults with time. It could be that at birth both HEU and HUU infants have equally immature immune systems with the same risk for neonatal infections. While HUU infants undergo normal immune maturation in the early months of life this maturation may be temporarily delayed or altered in HEU infants. Functional rather than quantitative differences in cells of the innate 137 immune system (monocytes and dendritic cells) have been observed in HEU compared to HUU infants specifically in response to bacterial antigens in the first two months of life (116). At 2 weeks of age HEU infants demonstrated a lower proportion of B-lymphocytes, cells that link the innate with the adaptive immune system (116). A more pronounced loss of adaptive immune function was observed at 14 weeks than at 6 weeks of age in HEU compared to HUU infants (220). Despite demonstrating robust T-lymphocyte counts, HEU infants experienced a profound loss of function of their T-lymphocytes as demonstrated by their restricted or absent cytokine production in response to BCG and Bordetella pertussis vaccine-antigen-specific stimulation (220). Significant loss of T-lymphocyte function remained at 14 weeks, even after controlling for birth weight, gestational age and breastfeeding and was not associated with maternal CD4 count in HEU infants. This deficit in adaptive immune function during the post-neonatal period could be in keeping with a diverging trajectory of HEU and HUU infant immune development as infancy progresses. An important next step in delineating the potential mechanism of HEU infant infectious morbidity will be to look at organism specific causes of infectious morbidity, their timing and their association with specific immune deficits. 8.3.3 Breast milk immunologic quality Breastfeeding HEU infants in our cohort still had a greater probability of very severe infectious morbidity than similarly breastfed HUU infants. This finding has been seen elsewhere and scientists have pondered whether the immunologic quality of HIV-infected mothers breast milk is inferior and thus the immunological protection afforded to their infants less than that for HUU infants (39,221). A case-control sub-study nested in the Mashi study looked at the immunologic parameters of breast milk in relation to HEU infant infectious morbidity (39). There was no difference in measured breast milk immunologic parameters between HEU infants who experienced a severe respiratory tract infection or episode of diarrhoea compared to control HEU infants without these infections. There was no difference in immunoglobulin levels to common respiratory and enteric pathogens, innate immune factors of lactoferrin and lysozyme or in breast milk cytokines measured. Comparing the breast milk of HIV-infected and HIV-uninfected mothers, again there was largely no difference in the measured immunologic factors between these two groups. However, this study looked at quantitative rather than qualitative functions of immunologic factors. While the amount of these factors might be similar, functionality of these protective factors in HIV-infected women warrants further study. A sub-study of ZEBS has recently evaluated the role of human milk oligosaccharides (HMOs) in HEU mortality (221). Human breast milk is rich in HMOs that act as prebiotics to promote healthy gastrointestinal tract flora and provide other beneficial immune modulating functions in breastfed infants (222). In ZEBS higher breast milk concentrations of some specific HMOs was protective against infant mortality while breastfeeding. Breastfeeding was only associated with a reduced mortality in HEU infants receiving breast milk with high concentrations of fucosylated-HMOs. HEU infants receiving breast milk poor in fucosylated-HMOs had no survival advantage compared to infants that had stopped breastfeeding (221). ZEBS did not have an HIV-138 unexposed infant control group to compare to. However, constituents of breast milk could explain the differences observed between breastfed HEU and HUU infants in our cohort. 8.4 The extent of maternal HIV disease and HEU infant infectious morbidity The purpose of the final objective was to determine whether the association of greater infectious morbidity observed in HEU infants born to mothers with advanced HIV during the pre-cART era has endured in the era of expanding availability of cART for pregnant women (10,31,38). As of December 2014 South Africa implemented WHO Option B+ for prevention of vertical HIV transmission, that recommends cART for all pregnant women irrespective of CD4 count (113). Thus, our study was the last opportunity to compare infants of mothers on maternally-indicated cART to infants of mothers without advanced immune suppression and not yet on cART. Infants were compared specifically for infectious cause hospitalizations or death according to whether the mother was on maternally-indicated cART or VTP prophylaxis during pregnancy. Important differences existed between mothers on maternally-indicated cART and mothers on VTP prophylaxis. Mothers on cART had more often been diagnosed with HIV prior to pregnancy than mothers on VTP prophylaxis. Antenatal and delivery CD4 counts and HIV viral load at delivery were substantially lower in mothers on cART compared to mothers on VTP prophylaxis. Although mothers on VPT prophylaxis experienced a profoundly larger drop in CD4 count by the time of delivery. Finally, mothers on cART were significantly older with a substantially higher income prior to the birth of the baby, than mothers on VTP prophylaxis. There were no other differences in demographic or obstetric factors. Infants of mothers on cART more often lived in households with electricity, but the households were otherwise quite similar. The infants in each group were equivalent in terms of gestational age, birth weight, immunization uptake and number of all-cause and sick clinic visits. These differences were considered in determining whether there was a difference in the probability of infectious morbidity between HEU infants of mothers on maternally-indicted cART and HEU infants of mothers on VTP prophylaxis. There was no evidence for a difference in the probability of an infectious cause hospitalization in the first six months of life between HEU infants of mothers on maternally-indicated cART or on VTP prophylaxis during pregnancy (OR 1.03, 95% CI 0.33, 3.20). Controlling for maternal demographic and obstetric factors did not alter this relationship, neither did controlling for timing of maternal diagnosis of HIV. Also there was no difference in infectious morbidity between the two groups of infants after controlling for receipt of cotrimoxazole preventive therapy or any breastfeeding. When considering the secondary outcomes of at least one severe or at least one very severe infectious cause hospitalization there was no 139 difference between the infants of mothers on maternally indicated cART compared to infants of mothers on VTP prophylaxis. Our findings of no difference in infectious morbidity by maternal ARV regimen contrast with observations of greater HEU infant mortality and morbidity associated with advanced maternal HIV in the pre-cART era (10,31,38). The provision of cART to HIV-infected mothers could have remarkably altered the in-utero experience of their HIV-exposed infants and may have had an impact on infant health. Potential explanations include: 1) the contrasting maternal states of improving immune function in cART mothers and declining immune function in VTP mothers; 2) rapid control of dysfunctional immune activation in cART mothers compared to ongoing immune activation in VTP mothers; 3) the reduced risk for opportunistic infections in mothers experiencing immune reconstitution on cART and 4) the positive impact of the healthcare system for cART mothers and their infants. Each explanation is explored below. 8.4.1 Maternal disease stabilisation versus disease progression Whether the rate of change of the CD4 count during pregnancy, as opposed to the absolute CD4 count, might influence infant well-being was considered. Although mothers on VTP in our cohort did not have CD4 depletion early in pregnancy, they did experience a significant decline in CD4 count during pregnancy. Mothers on cART experienced a rise in CD4 count between the antenatal period and delivery, whereas mothers on VTP prophylaxis experienced a median decline of more than 100 cells/µl. This CD4 decline is unlikely to be solely due to the expected pregnancy-associated CD4 decline irrespective of HIV-infection and the change in CD4 count in the VTP mothers in this cohort was of greater magnitude than that observed in other HIV-infected African pregnant women (215). In two studies CD4 count was lower by 51 (95% CI 30,72) cells/µl and 70 (95% CI 51,90) cells/µl during pregnancy and post-partum periods respectively (216,223). In our cohort 38% of HIV-infected mothers on VTP had a CD4 decline to below 350 cells/µl by delivery, making them eligible for cART when their infant was born. Therefore, VTP mothers may have undergone measurable disease progression during pregnancy compared to disease stabilisation or improvement in mothers on cART. However, we found no evidence for an association between infant infectious morbidity and the maternal antenatal CD4 count, delivery CD4 count or change in CD4 count that would support this explanation. The CD4 count is only one part of the immune system affected by HIV and using a quantitative measure of CD4 cells does not fully represent the impact of HIV on maternal immune function. There may be unmeasured maternal immune parameters, other than CD4 cell quantitative depletion, that mediate the influence of maternal HIV disease progression on infant immune function and infectious morbidity risk (108,224). 140 8.4.2 Maternal immune suppression versus immune activation As described in section 1.4.1.1, HIV-infection impacts the HIV-infected person through two contrasting pathways: firstly, immune suppression with long term consequences even after initiation of cART, and secondly, dysfunctional immune activation that responds rapidly to suppression of HIV replication due to cART (104,108,110). Although quantitative increases in CD4 T-lymphocytes occur in response to effective cART, complete normalization of the T-lymphocyte phenotype with total resolution of immune suppression is rare (114,115,225). Suppression of HIV replication by cART on the other hand, rapidly reduces the immune activation from B-lymphocytes responsible for hypergammaglobulinaemia (226). Mothers on cART in this cohort were heterogeneous in terms of antenatal CD4 count, two-thirds of the sample comprising mothers who were still severely immune compromised with CD4 counts below 350 cells/µl. In approximately one-third of mothers, CD4 counts had already risen to above 350 cells/µl in response to cART. Although quantitatively the CD4 count may recover once cART has been initiated, T-lymphocyte homeostasis may never be fully restored if cART is initiated when the CD4 cell count is below 350 cells/µl (115,225). Thus, mothers on cART with CD4 counts above 350 cells/µl were likely to differ immunologically to mothers on VTP prophylaxis who had not yet reached the same state of immune suppression despite having similar CD4 counts. Eighty percent of mothers on cART had suppressed or nearly suppressed HIV viral loads and likely had less immune activation than mothers on VTP prophylaxis with minimal viral suppression. The mothers on cART may still experience long term consequences of severe immune suppression in the presence of minimal dysfunctional immune activation. In contrast, the mothers on VTP prophylaxis do not yet have severe immune suppression but do have ongoing dysfunctional immune activation. These are likely two different states of immune dysregulation and may affect the foetus differently. A consequence of the dysfunctional immune activation experienced by HIV-infected people is a non-specific increase in B-lymphocyte activity and consequent immunoglobulin production. Where immunoglobulins are usually directed towards specific pathogens, this state of hypergammaglobulinaemia in HIV-infected people is non-specific and does not provide protection against pathogens. HEU infants may have immunological consequences from in-utero exposure to maternal B-lymphocyte dysregulation and hypergammaglobulinaemia. HEU infants have higher non-specific IgG levels than HIV unexposed infants that are positively correlated with maternal pregnancy IgG through to five years of age (133). That this correlation between maternal and HEU child IgG persisted until 5 years of age suggests that HEU infants exposed to HIV-infected mothers with unrestricted immune activation, undergo a fundamental and lasting change in immune function early in life (133). The infant IgG levels are not correlated with maternal CD4 count or maternal HIV viral load, indicating that there is something else about the maternal immune system perturbations influencing infant IgG levels. In our cohort, there was no association 141 between maternal CD4 count or HIV viral load with infectious morbidity, but further immune function of the mothers and infants was not explored. Finally, it is possible that immunological factors that previously accentuated HEU infant morbidity and mortality risk in mothers at the lower end of the CD4 spectrum are now being addressed by cART and that the effect of maternal immune suppression is becoming less. During the pre-cART era the mothers with the most advanced HIV, whose infants had the greatest risk for morbidity and mortality, had the lowest CD4 counts and the highest HIV viral loads. These infants in the pre-ART era were exposed to in utero environments altered by both severe immune suppression and marked immune activation, both potentially contributing to their poor outcomes. In the era of maternally-indicated cART and VTP prophylaxis, and in our cohort, infants were exposed to in-utero environments altered by either immune suppression or immune activation respectively. The unrestricted viral replication-induced immune activation present in mothers on VTP prophylaxis may be of similar or greater consequence to the developing foetus as the long term effects of immune suppression in the mothers receiving cART. This could explain why infectious morbidity in infants of mothers qualifying for VTP prophylaxis was no less than that of infants born to mothers on maternally-indicated cART. 8.4.3 Improved maternal physical well-being HIV-infected people in the pre-cART era experienced a great burden of opportunistic infections due to untreated severe immune suppression (227). In the current era, the risk of opportunistic infections has been reduced in mothers with low CD4 counts on cART with less exposure to their infants in-utero and thereafter to these opportunistic pathogens. Cytomegalovirus (CMV) is an important co-pathogen in HIV-infected severely immune suppressed adults. Infant CMV exposure can occur in-utero and during breastfeeding. Maternal CD4 count of below 200 cells/µl was independently associated with congenital CMV in HIV-exposed infants in Cape Town, after adjusting for maternal cART or VTP prophylaxis (18). CMV altered T-cell differentiation in Gambian infants and a higher breast milk CMV viral load was associated with lower weight and length-for-age Z-scores in Malawian HEU infants (182,228). Although these CMV induced alterations have not been studied in direct association with wider infectious morbidity in HEU infants, they demonstrate that CMV infection has substantial consequences in infants. European HEU infants have a higher prevalence of congenital CMV than HUU infants, although this is being reduced with expanding maternal cART (181). In this way maternally-indicated cART may correct the previously observed difference in risk for morbidity and mortality in HEU infants by reducing the burden of maternal opportunistic infections, particularly CMV. Maternal and infant CMV in our cohort was not evaluated further for this dissertation but is the topic of a sub-study in progress. 142 Mothers on cART were in regular monthly or bi-monthly follow-up in a primary healthcare clinic system. HIV-infected mothers not yet immune suppressed and not qualifying for maternally-indicated cART, were seen less often postnatally, at most every six months at the primary healthcare clinic for a repeat CD4 count. Many were in no postnatal follow-up at all. There may be unmeasured benefits for both the mother on cART and her infant in being well connected to a healthcare system with opportunities for general health information provision, advice seeking and access to formal and informal support services that the mothers on VTP prophylaxis may not have experienced. 8.5 Potential effects of ARV exposure observed in HEU infants This study was not designed to evaluate the safety or consequences for HEU infants of antiretroviral exposure. There were however three noteworthy observations within HEU infants deserving further discussion: 1) the association between a lower infant haemoglobin and exposure to ZDV prophylaxis; 2) a persistently lower length-for-age Z-score in TDF-exposed infants than if mothers were only on VTP prophylaxis; and 3) the observation of two congenital central nervous system defects in infants exposed to first trimester EFV. 8.5.1 Anaemia in zidovudine-exposed HEU infants HEU infants compared to HUU infants had a significantly lower mean haemoglobin (Hb) at two months of age, but did not differ significantly in the proportion of infants with a DAIDS grade 1 or more anaemia. This pattern was mirrored within the HEU infant sub-group analysis. The HEU infants born to mothers on VTP prophylaxis had a lower mean Hb at two months of age and significantly more infants had a DAIDS grade 1 or higher anaemia than those of mothers on cART. Mothers on VTP prophylaxis all received ZDV during pregnancy which has been associated with anaemia in infants (137,138). This observation was sustained when all infants exposed to ZDV, including those of mothers on ZDV for VTP prophylaxis as well as those of mothers on a ZDV containing cART regimen, were compared to all HEU infants not exposed to ZDV. The most likely explanation for the difference in Hb in HEU compared to HUU infants at two months of age is due to ZDV-associated anaemia in the ZDV-exposed HEU infants. Whether or not ZDV-associated anaemia puts HEU infants at increased risk for infectious morbidity has not been established. In this cohort the infant Hb was significantly lower at four months of age in infants with an infectious cause hospitalization or death compared to infants without an infectious cause hospitalization or death. Within HEU infants, the mean Hb at four months was also significantly lower in infants with the primary outcome compared to those without the primary outcome. The majority (23 of 26) of infectious cause hospitalizations had occurred by four months of age. The association seen between infants with the primary outcome and anaemia at four months of age is unlikely to be causally related to the infectious cause hospitalizations but more likely as a consequence of the infection resulting in hospitalization. 143 Although we have evidence for a ZDV-associated anaemia in HEU infants this is unlikely to be causally associated with an increased risk for infectious morbidity in this cohort. 8.5.2 Shorter length in tenofovir-exposed HEU infants HEU infants in our cohort had significantly lower mean length-for-age Z-scores (LAZ) than HUU infants from two weeks of age persisting to six months of age. This pattern was again mirrored in the HEU sub-group comparison, but this time the HEU infants of mothers on maternally-indicated cART during pregnancy had significantly lower mean LAZ than HEU infants of mothers on VTP prophylaxis. This association with a persistently lower LAZ was even stronger when infants specifically exposed to TDF were compared to all HEU infants not exposed to TDF. There was no association in this cohort with a lower LAZ or stunting and infectious morbidity in the total cohort of HEU and HUU infants, nor in the sub-group of HEU infants compared by TDF-exposure. Although not a stated objective of this study, this was the last opportunity to compare TDF-exposed and TDF-unexposed HEU infants. As South Africa has now moved to cART for all pregnant women irrespective of CD4 count, most HIV-infected pregnant women now receive a TDF-containing triple ARV regimen whether for maternal therapy or for prophylaxis. That a shorter length in cART and specifically TDF-exposed HEU infants was present already at two weeks of age and was seen in the context of appropriate and equivalent weight-for-length Z-scores, suggests that this difference in length-for-age was less likely due to long-term nutritional compromise, as is often the cause of stunting and reduced length-for-age. Rather it was possibly as a result of an in-utero effect or early postnatal effect. Eighty three percent of mothers on cART were receiving a TDF containing regimen during pregnancy. TDF is known to alter bone metabolism resulting in reduced bone mineral density in HIV-infected children treated with TDF (142,143). Due to this observation in HIV-infected children the potential for bone toxicity with TDF in-utero exposure has been a concern, particularly in relation to birth growth outcomes. However, this concern in terms of compromised growth at birth has not been borne out in previous US or African cohorts of TDF in-utero exposed HEU infants (144–146). In a recent cross-sectional evaluation of bone mineral content measured by dual-energy X-ray absorptiometry at two weeks of age, TDF-exposed neonates had a significantly lower bone mineral content than TDF-unexposed neonates following control for confounders (229). Our findings concur with this pattern of normal growth at birth in TDF-exposed HEU infants. Yet, by two weeks of age a difference in LAZ became apparent in TDF-exposed HEU infants in our cohort and was sustained until last follow-up at six months. Two US cohort studies have also observed later growth differences in TDF-exposed HEU infants in the presence of equivalent birth growth parameters (144,145). The Perinatal HIV/AIDS Cohort Study (PHACS) in the US, observed a significantly lower LAZ at 1 year of age in TDF-exposed HEU infants (144). The International Maternal Paediatric Adolescent AIDS Clinical Trials Group (IMPAACT P1025) observed a lower weight-for-age Z-score at six months despite similar 144 birth weight (145). PHACS did not have interval observations between birth and one year of age to directly compare to the early onset of lower length-for-age in our cohort and IMPAACT P1025 looked only at differences in weight at birth and six months and not at length. In contrast to the US cohorts and our cohort, the only previous African study to consider the potential of bone toxicity in TDF-exposed HEU infants found no important differences in weight or length up to four years of age in 62 infants with no TDF in-utero exposure compared to 111 infants exposed to TDF in-utero for at least 90% of pregnancy days (146). With expansion of cART to all HIV-infected pregnant South African women, irrespective of CD4 count, almost all HIV-exposed infants will now be TDF in-utero exposed. This pattern of potentially delayed consequences for long bone growth despite having no growth deficits at birth certainly requires further longitudinal study of TDF-exposed HIV exposed infants and children. In the foreseeable future HEU infants will always be exposed to antiretroviral drugs to prevent vertical HIV infection. Should this association between TDF in utero exposure and shorter length growth be observed in larger cohorts, available alternatives to TDF should be considered for first line cART in pregnant women. 8.5.3 Central nervous system defects in HEU infants with first trimester efavirenz exposure Two HEU infants in this cohort born to mothers receiving EFV containing cART regimens during their first trimester of pregnancy were identified with congenital central nervous system (CNS) abnormalities, one with congenital macrocephaly and neurodevelopmental delay and another with communicating hydrocephalus of unknown aetiology. Since the report of CNS and facial teratogenic defects in monkeys following first trimester EFV exposure at levels similar to human exposure, and isolated case-reports of human CNS defects following first trimester EFV exposure, there has been concern regarding the safety of including EFV as part of first-line cART regimens for use in women of child-bearing age (147,156). Small studies vary in their observations of the risk associated with EFV exposure (230–232). Reassurance is provided through ongoing prospective surveillance conducted by the Antiretroviral Pregnancy Registry (APR) that has followed 825 cases of first trimester EFV exposure up to July 2014 and found no risk (147). This represents sufficient cases to exclude a doubling in the risk of all birth defects compared to second or third trimester EFV exposure. Two cases of CNS defects have been reported in the 825 first trimester EFV exposure cases and monitoring in this regard is ongoing in the APR. The majority of cases reported in the APR are from North America. A more widely generalizable systematic review and meta-analysis included the APR cases as well as United Kingdom, Northern Ireland, West and Southern African cohorts, totalling 1437 cases of first trimester EFV exposure (157). By meta-analysis, the relative risk of any birth defects due to EFV first trimester exposure compared to other ARV first trimester exposure was 0.85 (95% CI 0.61, 1.20). Our small cohort study was not designed to systematically identify congenital anomalies and no conclusions can be drawn about an association between first trimester EFV exposure and the risk for CNS or other birth defects. Fifty percent of mothers 145 on cART in this cohort were on an EFV containing regimen. Since the conduct of this study EFV use in South African women of child-bearing age has expanded tremendously with the adoption nationally of an EFV containing triple ARV regimen for first-line antiretroviral therapy in all pregnant women irrespective of CD4 count. Essentially all South African HIV exposed newborns will also be EFV exposed in the future, warranting a structured prospective pharmacovigilance system to ensure the safety of these highly exposed infants. Alongside this, establishment of national systematic surveillance of congenital anomalies in South Africa, will provide population rates of congenital anomalies, currently lacking, with child health benefits beyond only understanding of ARV exposure risks. 8.6 Challenges and limitations The study experienced implementation challenges of inadequate enrolment in conjunction with substantial early attrition and lower outcome event rates than anticipated. Analytic challenges were experienced in adequately controlling for differences in breastfeeding. Reasons for these challenges, their impact on the study and additional limitations are considered below. 8.6.1 Implementation challenges Only eighty percent of the enrolment target was reached (264 of a targeted 330 mother-infant pairs) following extension of the enrolment duration from nine to 12 months. Budget constraints limited further extension of enrolment. Of the 260 HIV-uninfected infants eligible for the study, only two-thirds returned at two weeks of age to continue in follow-up. Although efforts were made to sensitize pregnant women about the study at their antenatal clinic visits so that they had time prior to delivery to consider their participation, timing of enrolment at delivery possibly placed undue pressure on mothers to agree to participation. This high rate of early attrition was greatest at the start of the study and coincided with a new study team eager to fulfill enrolment targets. The study also commenced during the rainy winter season in Cape Town which was possibly an environmental deterrent to attending the first study visit with a two week old infant. Every effort was made by the study team to make attendance of study visits as convenient as possible for mothers and their infants including later pick-up times at the most convenient location for mothers. Retention was subsequently better once mothers and infants returned for the two week visit. Similar early attrition has been observed as a challenge in other South African observational cohort studies with peripartum enrolment (233,234). This marked early attrition could have introduced selection bias into the comparison between the HEU and HUU infants retained in the cohort should there have been important differences related to risk for infectious morbidity between HEU and HUU infants that did not return to continue in the study. Similar proportions of HEU (32%) and HUU (36%) infants were lost following enrolment and according to 146 available maternal and infant baseline characteristics the infants lost were no different to those retained. Thus, it is reassuring that the infants included in the analytic cohort are representative of the initial sample of infants that included almost 20% of the population of infants born at Kraaifontein MOU. Fewer primary outcome events, infectious cause hospitalizations or deaths, occurred than originally anticipated. This could be due to a combination of failure to identify all outcome events or improving infant health in South Africa. However, we think it unlikely that we failed to identify outcome events. Through the use of the province wide unique healthcare identification number issued to every infant at birth, we were able to link all our study participants, including those lost to face-to-face follow-up, with the Western Cape Province electronic hospital administrative system as well as the mortality database. This allowed for complete determination of all hospitalizations that occurred within the Western Cape Province. We were unable to later re-establish direct contact with seven (7.4%) HEU and seven (8.5%) HUU infants who were not linked to any hospitalization or mortality events through the Western Cape Province databases. It is possible that these infants could have migrated out of the province and hospitalizations or deaths in these infants could have been missed. There is no obvious reason though that the rate of hospitalization or death observed in the rest of the cohort would have been different in these 14 infants and the identification of a single infant death before six months of age is in keeping with our a priori estimated infant mortality rate in this cohort. A more likely explanation for the unexpected decrease in events is improved infant health outcomes. South Africa has experienced a steady reduction in infant and under-5 mortality during the past five years since the introduction of an effective vertical HIV transmission prevention programme and enhancement of the infant vaccination schedule in 2009 (55,57). This positive trend in child survival has specifically impacted on pneumonia and diarrhoea morbidity experienced by all infants in South Africa (57). This may account for the substantially lower rate of infectious cause hospitalizations observed in this cohort study than that observed in the initial pilot study conducted between 2009 and 2011. The likelihood of a lower outcome rate was taken into account in the sample size calculations for the study However, the resource constraints limited the size of the sample initially enrolled. This combined with difficulties with early retention of participants already discussed, left the study under-powered for the primary objective. 8.6.2 Analytic challenges Due to a local infant feeding policy in transition during the study design phase, it was decided not to enrol infants according to intended feeding mode but to deal with these differences in analysis. This presented a major analytic challenge in how to most appropriately adjust through regression for the influence of breastfeeding so as to achieve the primary objective of determining the effect of HIV exposure unconfounded by the effect of breastfeeding. The analysis in this regard was limited further by working 147 with a small sample size, as well as the distribution of breastfeeding that discriminated almost exclusively according to HIV exposure. Stratified analysis proved helpful to overcome some of these challenges. Many longitudinal studies, that consider child health outcomes, collect detailed infant feeding information at multiple visits. The hope in collecting this rich feeding detail is that it will be possible to accurately reflect the infant feeding reality and its effect on the outcome of interest in statistical models. Despite collection of rich detail, infant feeding status is often reduced to a single variable of simplified categories at one time point (47,211,235). Multilevel models, otherwise known as mixed or random effects models, have an advantage over single level logistic regression in that repeated measurements of breastfeeding status at multiple time points can be included in the model simultaneously. The correlated nature of the repeated breastfeeding measures within individuals is taken into consideration so as not to underestimate the standard error for the effect of the variable with repeated measurements (236). Multilevel models can also take into account varying effects of breastfeeding at different time points (237). This is a desirable manner in which to incorporate maximal feeding information to approximate reality in a statistical model. A mixed effects model still requires an adequate sample size to include the number of terms representing infant feeding at various time points in the model. And even with mixed effects models there is also the limitation that unmeasured factors influencing the mothers choice to breastfeed could be associated with the infant outcome and remain as residual confounders. Due to our limited sample size and limited number of outcome events we were not able to make use of mixed effects models and were constrained to dichotomizing feeding at a single time point for inclusion in the fixed effect regression models. Breastfeeding status at two weeks of age discriminated almost completely according to HIV exposure, with only a single HUU infant not breastfeeding at two weeks of age compared to two thirds of HEU infants not breastfeeding at two weeks. The single never breastfed HUU infant experienced a very severe infectious cause hospitalization compared to 5 of 59 (8.5%) never breastfed HEU infants. This gave an artificial OR of close to zero for HEU relative to HUU never breastfed infants and explains why the unstratified aOR for very severe infectious cause hospitalization or death is lower than the stratified estimate for the stratum breastfeeding at two weeks (aOR 2.5 vs. 4.2). The single observation in the HUU non-breastfeeding category requires that the unstratified estimates adjusted for breastfeeding status at two weeks must be interpreted with caution. With the limitations in controlling for breastfeeding in the unstratified logistic regression estimates, stratification provided an alternative approach to deal with confounding by breastfeeding on the effect of HIV exposure. Stratified analyses conditioned on the presence or absence of any breastfeeding separately at two weeks and at six months were performed. While it was not possible to compare HEU and HUU infants never breastfed at two weeks due to the single HUU infant in this stratum, some information could be gleaned from how HEU breastfed and HUU breastfed infants compared to each 148 other. This helped to identify that there was a greater probability of very severe infectious cause hospitalization or death amongst HEU infants even when breastfed compared to HUU also breastfed infants. The stratified analysis conditioned on the presence or absence of breastfeeding at six months revealed the possibility that the effect of HIV exposure could be modified by breastfeeding. Amongst infants breastfeeding at six months there was possibly a greater probability of infectious morbidity in HEU compared to HUU infants. On the other hand, HIV exposure was not associated with a greater probability of infectious morbidity when comparing HEU and HUU infants both not breastfeeding at six months, both groups experienced equally high probabilities of infectious morbidity. When effect modification is present it is not appropriate to present the unstratified estimates of effect (238). Unfortunately the confidence intervals around these estimates do not allow firm conclusions about the presence or absence of effect modification by breastfeeding on HIV exposure, and it is possible that the appearance of effect modification is spurious due to the imprecision of the estimates. As such, recognizing the limitations in the unstratified and stratified aOR estimates, both have been presented. 8.6.3 Additional study limitations The inadequate sample size hinders any definitive conclusions that can be drawn from this study with regards to differences in overall risk for infectious morbidity in HEU and HUU infants. We may not have identified a substantial difference between HEU and HUU infants due to inclusion of the lowest risk HIV-infected mothers and lowest risk infants. This was done with intent to determine whether there is evidence for pathways other than the universal pathways that determine infectious morbidity risk in all infants, both HIV exposed and unexposed. The deficiencies in the sample size were countered, somewhat, by the well matched infant groups that limited confounding by differences in birth outcomes (preterm birth and low birth weight), household and socioeconomic factors. It is recognized that even though infants were well matched on measured socioeconomic factors there is always the potential for unmeasured differences and residual confounding. Follow-up for this analysis was limited to six months of age. Due to provincial administrative reasons linkage with the provincial mortality registry was only possible until December 2013, when the youngest infants in our cohort would have been six months of age. This made complete outcome determination impossible beyond this point. Resource considerations were an additional factor limiting further intensive follow-up of the cohort. Long term follow-up of HEU infants is certainly valuable to better understand this vulnerable group. However, to understand the difference in infectious morbidity between HEU and HUU infants, the period of greatest relative risk in HEU infants observed in the large early studies was before six months of age (37,38). An increase in HEU infant morbidity and mortality has been observed beyond six months of age associated with weaning (32,75). Eighty five percent of HEU infants in our cohort had already weaned from breast milk by six months of age, and in this cohort infectious morbidity was not 149 clearly linked to weaning, making an increase in morbidity associated with weaning beyond six months of age unlikely in this particular context. It would have been advantageous to have conducted aetiological investigations during infectious cause hospitalizations that could further our understanding of the specific vulnerabilities and mechanisms of HEU infant infectious morbidity risk. Such resources were not available and the sample size was insufficient to probe further into associations with specific types of infections (diarrhoea, LRTI, bacterial sepsis) or specific causative organisms. 8.7 Recommendations and implications 8.7.1 The importance of an appropriate control group In Southern Africa, where the vast majority of infants are vulnerable to high rates of morbidity and mortality irrespective of HIV exposure, and where income inequality is marked, comparison of HEU infant outcomes to population rates of disease or other reference standards can be misleading (52,55). As an example, we used the WHO child growth standards to calculate gender and age appropriate Z-scores of the various anthropometric indices measured, for comparison between HEU and HUU infants (78). Studies of growth and nutrition in HEU African infants have correctly concluded that HEU infants have high rates of growth faltering in comparison to the WHO growth standards, but this has not helped to understand how they compare to HUU children experiencing the same constraints on growth during childhood (80–82). The WHO growth standards are not designed to be a reference comparison that can be generalized to all child populations across the world. Rather, they are designed as standards of growth achieved by healthy children and what should be strived for to achieve children of maximal health (239). Thus when these growth standards are applied in countries in Southern and Eastern Africa, where malnutrition and infectious disease burdens are high, up to 50% of children may be regarded as stunted according to the WHO LAZ standard (79). We observed a difference in LAZ in HEU compared to HUU infants, HEU infants were shorter possibly as a result of cART or specifically TDF exposure. However, HUU infants were also shorter than the WHO standards with mean LAZ well below the WHO standard mean. The difference between HEU and HUU infants was much smaller than the difference between HEU infants and the WHO standard, and without the appropriate HUU comparison group alarming conclusions could have been drawn about the extent of HEU growth impairment due to HIV exposure. This example underscores the importance of having an appropriate HUU control group to compare HEU outcomes to in the context of universally high rates of infant morbidity. 150 8.7.2 Looking beyond incidence of morbidity We chose infectious cause hospitalizations as the primary outcome because it is a more objective outcome event than reported sick-clinic visit rates. But even hospitalization rates could be biased if healthcare professional decision making with regards to hospital admission is systematically different in one group than another, particularly due to concern amongst caregivers and healthcare professionals about the high risk nature of HEU infants. Using the PIET-R severity grading we were able to have certainty around the need for hospital admission in both groups of infants based on the extent of their presenting symptoms and that one group was not being over hospitalized compared to the other. Using the DAIDS grading would not have given us this clarity; all 26 infants with an infectious cause hospitalization had at least one event graded as a DAIDS grade 3 event, no events were graded as grade 4 (life-threatening) and the one death received a grade 5. The pattern observed in this cohort of a greater severity of infectious morbidity in HEU infants would have also been missed using only the DAIDS grade. The PIET-R, that graded hospitalization events as mild-moderate, severe or very severe, allowed for a more detailed understanding of the differences in infectious morbidity between HEU and HUU infants. To move forward in understanding differences in infectious morbidity risk in HEU infants, it would be constructive for future studies in this field to reach consensus on the use of standardized objective infectious event outcome definitions. This could aid comparison as well as collaboration across geographic settings. The Brighton Collaboration is a successful example in this regard, achieving global standardization of definitions for the evaluation and surveillance of adverse events following immunizations (188). 8.7.3 Implications Although the estimate from this study of increased infectious morbidity risk in HEU infants is imprecise, it raises the possibility that there could be a strong association between HIV exposure and infectious morbidity. The unadjusted association between HIV exposure and the primary outcome showed HEU infants to have a 48% greater risk for any infectious cause hospitalization or death compared to HUU infants in the first 6 months of life (RiR 1.48, 95% CI 0.72,3.06). A similar effect size was seen in a recent Mozambican cohort (44). HIV exposure without HIV infection could account for up to one-third of infectious cause hospitalizations in HEU infants. At a population level this translates to 12.5% of infectious cause hospitalizations in all HIV-uninfected infants under six months of age could be as a consequence of being born to an HIV-infected mother. In South Africa this could account for an excess of 35 000 infant hospitalizations per year. Even if the true RiR is only 1.1, being an HEU infant could still account for approximately 3% of infant infectious cause hospitalizations and an excess of 8100 infant hospitalizations per year in South Africa. This highlights the potential contribution of HIV exposure, without HIV-infection, through both universal and HEU-unique pathways, to infant morbidity and motivates for further dedicated study of HEU infants. See Appendix E for calculation of above attributable and population attributable fractions. 151 Chapter 9 Conclusion In this cohort study of term HEU and HUU infants experiencing similar social and household circumstances, breastfeeding HEU infants had a substantially greater probability of very severe infectious morbidity than breastfeeding HUU infants. HEU infants tended to experience a greater severity of infectious morbidity in the context of a similar frequency of infections, the greatest relative difference between HEU and HUU infants occurring between three and six months of age. Within HEU infants, and in the context of maternally-indicated cART, the severity of maternal HIV disease measured by CD4 count or HIV viral load was not associated with infant infectious morbidity outcomes. Despite its limitations, this study has helped to identify a more specific pattern to HEU infectious morbidity not previously described in the absence of appropriate HUU control groups and discriminating outcome measurement tools. The observed pattern could be in keeping with mechanisms related to HEU-unique exposures including deficient transplacental acquisition of maternal antibodies, delayed functional immune development in HEU infants or altered immunologic quality of HIV-infected maternal breast milk. It is fully recognized that enhanced efforts to improve uptake of sustained breastfeeding in South African HIV exposed infants are required. We suggest that two further avenues warrant exploration to potentially improve HEU infant outcomes: 1) evaluating the quality of breast milk in HIV-infected mothers and whether it differs to HIV-uninfected mothers; if so to identify ways that breast milk quality could be enhanced 2) to understand whether the observed deficit in maternally derived antibodies in HEU infants is specifically associated with infectious morbidity; if so can the most vulnerable stage during infancy be defined and will maternal cART or an augmented infant or maternal vaccination schedule be able to ameliorate this vulnerability? In addition a long-term sustainable pharmacovigilance surveillance system urgently needs to be established in South Africa to understand the effects of in-utero antiretroviral exposure in 30% of South African infants, children and future adults. To definitively understand HEU infant infectious morbidity risk, substantially larger studies with appropriate HUU infant comparison groups are going to be necessary. The field would benefit from enhanced collaboration amongst researchers to achieve the quality of evidence required to improve HEU infant outcomes in Southern Africa. Agreement on standardized infectious morbidity outcome measurement would allow for collaboration of multiple cohorts while still pursuing their own unique individual hypotheses and objectives. In designing further studies and looking for possible interventions to secure HEU infant well-being, researchers need to be cognisant of the vulnerabilities that all infants and children in Southern Africa experience, whether HIV exposed or not. Interventions that are found to improve health in HEU infants may very well also improve outcomes for HUU infants. The bigger picture of improving health and well-being for all infants in this setting needs to be kept in mind, and one group 152 should not be heavily prioritized to the detriment of the other. If future clinical trials should be implemented to test interventions for improvement of HEU infant outcomes, strong consideration should be given to inclusion of HIV unexposed infants, as they too could derive benefit from the interventions under study. HEU infants comprise almost one third of the South African infant population and their health and wellbeing, beyond avoiding HIV-infection, deserve a more prominent position in the local and international HIV research agenda. 153 References 1. Keele BF, Van Heuverswyn F, Li Y, Bailes E, Takehisa J, Santiago ML, et al. Chimpanzee reservoirs of pandemic and nonpandemic HIV-1. Science (80). 2006;313(5786):523–6. 2. UNICEF. The state of the worlds children: excluded and invisible. 2006. 3. Hankin C, Thorne C, Peckham C, Newell M-L. The health and social environment of uninfected infants born to HIV-infected women. AIDS Care. 2004 Apr;16(3):293–303. 4. Forbes JC, Alimenti AM, Singer J, Brophy JC, Bitnun A, Samson LM, et al. A national review of vertical HIV transmission. AIDS. 2012 Mar 27;26(6):757–63. 5. Joint United Nations Programme on HIV/AIDS (UNAIDS). GLOBAL REPORT: UNAIDS report on the global AIDS epidemic 2013. 2013. 6. Kuhn L, Stein ZA. Mother-to-infant HIV transmission: timing, risk factors and prevention. Paediatr Perinat Epidemiol. 1995 Jan;9(1):1–29. 7. Joint United Nations Programme on HIV/AIDS (UNAIDS). UNAIDS World AIDS Day Report 2011: Faster Smarter Better. 2011. 8. Goga AE, Dinh TH, Jackson DJ. Evaluation of the Effectiveness of the National Prevention of Mother-To-Child Transmission (PMTCT) Programme on Infant HIV Measured at Six Weeks Postpartum in South Africa, 2010. South African Medical Research Council, National Department of Health of South Africa, PEPFAR / US Centers for Disease Control and Prevention. 2012. 9. South African National Department of Health. The 2011 National Antenatal Sentinel HIV and Syphilis Prevalence Survey. 2012. 10. Kuhn L, Kasonde P, Sinkala M, Kankasa C, Semrau K, Scott N, et al. Does severity of HIV disease in HIV-infected mothers affect mortality and morbidity among their uninfected infants? Clin Infect Dis. 2005 Dec 1;41(11):1654–61. 11. Filteau S. The HIV-exposed, uninfected African child. Trop Med Int Heal. 2009 Mar;14(3):276–87. 12. Mofenson LM, Watts DH. Safety of pediatric HIV elimination: the growing population of HIV- and antiretroviral-exposed but uninfected infants. PLoS Med. 2014 Apr;11(4):e1001636. 154 13. Bunders MJ, Van Hamme JL, Jansen MH, Boer K, Kootstra NA, Kuijpers TW. Fetal exposure to HIV-1 alters chemokine receptor expression by CD4 + T cells and increases susceptibility to HIV-1. Sci Rep. 2014;(4):6690. 14. World Health Organisation. Consolidated Guidelines on The Use of Antiretroviral Drugs for Treating and Preventing HIV Infection. 2013. 15. Pillay T, Sturm AW, Khan M, Adhikari M, Moodley J, Connolly C, et al. Vertical transmission of Mycobacterium tuberculosis in KwaZulu Natal: impact of HIV-1 co-infection. Int J Tuberc Lung Dis. 2004;8(1):59–69. 16. Lawn SD, Bekker L-G, Middelkoop K, Myer L, Wood R. Impact of HIV Infection on the Epidemiology of Tuberculosis in a Peri-Urban Community in South Africa: the Need for Age-Specific Interventions. Clin Infect Dis. 2006;42(1 April):1040–7. 17. Duryea EL, Sánchez PJ, Sheffield JS, Jackson GL, Wendel GD, McElwee BS, et al. Maternal Human Immunodeficiency Virus Infection and Congenital Transmission of Cytomegalovirus. Pediatr Infect Dis J. 2010;29(10):915. 18. Manicklal S, Van Niekerk AM, Kroon SM, Hutto C, Novak Z, Pati SK, et al. Birth prevalence of congenital cytomegalovirus among infants of HIV-infected women on prenatal antiretroviral prophylaxis in South Africa. Clin Infect Dis. 2014;58(10):1467–72. 19. Chen JY, Ribaudo HJ, Souda S, Parekh N, Ogwu A, Lockman S, et al. Highly active antiretroviral therapy and adverse birth outcomes among HIV-infected women in Botswana. J Infect Dis. 2012 Dec 1;206(11):1695–705. 20. Sibiude J, Warszawski J, Tubiana R, Dollfus C, Faye A, Rouzioux C, et al. Premature delivery in HIV-infected women starting protease inhibitor therapy during pregnancy: role of the ritonavir boost? Clin Infect Dis. 2012 May;54(9):1348–60. 21. McNally LM, Jeena PM, Lalloo UG, Nyamande K, Gajee K, Sturm AW, et al. Probable mother to infant transmission of Pneumocystis jiroveci from an HIV-infected mother to her HIV-uninfected infant. AIDS. 2005;19(14):1548–9. 22. Cotton MF, Schaaf HS, Lottering G, Weber HL, Coetzee J, Nachman S. Tuberculosis exposure in HIV-exposed infants in a high-prevalence setting. Int J Tuberc Lung Dis. 2008;12(2):225–7. 155 23. Moore DP, Schaaf HS, Nuttall J, Marais BJ. Childhood tuberculosis guidelines of the Southern African Society for Paediatric Infectious Diseases. South African J Epidemiol Infect. 2009;24(3):57–68. 24. Humphrey JH. The risks of not breastfeeding. J Acquir Immune Defic Syndr. 2010 Jan;53(1):1–4. 25. Marais BJ, Esser M, Godwin S, Rabie H, Cotton MF. Poverty and Human Immunodeficiency Virus in Children: A View from the Western Cape, South Africa. Ann N Y Acad Sci. 2008 Jan;1136:21–7. 26. Joint Learning Initiative on Children and HIV/AIDS. Home truths. Facing the facts on children, AIDS and poverty. UNICEF. 2009. 27. Newell M-L, Coovadia H, Cortina-Borja M, Rollins N, Gaillard P, Dabis F. Mortality of infected and uninfected infants born to HIV-infected mothers in Africa: a pooled analysis. Lancet. 2004;364(9441):1236–43. 28. Thea DM, St. Louis ME, Atido U, Kanjinga K, Kembo B, Matondo M, et al. A prospective study of diarrhea and HIV-1 infection among 429 Zairian infants. N Engl J Med. 1993; 29. Spira R, Lepage P, Msellati P, Van de Perre P, Leroy V, Simonon A, et al. Natural History of Human Immunodeficiency Virus Type 1 Infection in Children: A Five-Year Prospective Study in Rwanda. Pediatrics. 1999;104(e56):1–9. 30. Taha TE, Graham SM, Kumwenda NI, Broadhead RL, Hoover DR, Markakis D, et al. Morbidity Among Human Immunodeficiency Virus-1-Infected and -Uninfected African Children. Pediatrics. 2000 Dec 1;106(6):e77. 31. Chatterjee A, Bosch RJ, Hunter DJ, Fataki MR, Msamanga GI, Fawzi WW. Maternal disease stage and child undernutrition in relation to mortality among children born to HIV-infected women in Tanzania. J Acquir Immune Defic Syndr. 2007 Dec 15;46(5):599–606. 32. Kourtis AP, Wiener J, Kayira D, Chasela C, Ellington SR, Hyde L, et al. Health outcomes of HIV-exposed uninfected African infants. AIDS. 2013 Mar 13;27(5):749–59. 33. The Kesho Bora Study Group. Eighteen-month follow-up of HIV-1-infected mothers and their children enrolled in the Kesho Bora study observational cohorts. J Acquir Immune Defic Syndr. 2010;54(5):533–41. 156 34. Von Mollendorf C, von Gottberg A, Tempia S, Meiring S, de Gouveia L, Quan V, et al. Increased risk and mortality of invasive pneumococcal disease in HIV-exposed-uninfected infants under 1 year of age in South Africa, 2009-2013. Clin Infect Dis. 2015;60(9):1346–56. 35. Kelly MS, Wirth KE, Steenhoff AP, Cunningham CK, Arscott-Mills T, Boiditswe SC, et al. Treatment Failures and Excess Mortality Among HIV-Exposed, Uninfected Children With Pneumonia. J Pediatric Infect Dis Soc. 2014; doi 10.1093/jpids/piu092 36. Brahmbhatt H, Kigozi G, Wabwire-Mangen F, Serwadda D, Lutalo T, Nalugoda F, et al. Mortality in HIV-infected and uninfected children of HIV-infected and uninfected mothers in rural Uganda. J Acquir Immune Defic Syndr. 2006;41(4):504–8. 37. Marinda E, Humphrey JH, Iliff PJ, Mutasa K, Nathoo KJ, Piwoz EG, et al. Child mortality according to maternal and infant HIV status in Zimbabwe. Pediatr Infect Dis J. 2007 Jun;26(6):519–26. 38. Koyanagi A, Humphrey JH, Ntozini R, Nathoo K, Moulton LH, Iliff P, et al. Morbidity Among Human Immunodeficiency Virus-exposed But Uninfected, Human Immunodeficiency Virus-infected, and Human Immunodeficiency Virus-unexposed Infants in Zimbabwe Before Availability of Highly Active Antiretroviral Therapy. Pediatr Infect Dis J. 2011;30(1):45–51. 39. Shapiro RL, Lockman S, Kim S, Smeaton L, Rahkola JT, Thior I, et al. Infant morbidity, mortality, and breast milk immunologic profiles among breast-feeding HIV-infected and HIV-uninfected women in Botswana. J Infect Dis. 2007;196(4):562–9. 40. McNally LM, Jeena PM, Gajee K, Thula SA, Sturm AW, Cassol S, et al. Effect of age, polymicrobial disease, and maternal HIV status on treatment response and cause of severe pneumonia in South African children: a prospective descriptive study. Lancet. 2007 Apr 28;369(9571):1440–51. 41. Rollins NC, Ndirangu J, Bland RM, Coutsoudis A, Coovadia HM, Newell M-L. Exlcusive Breastfeeding, Diarrhoeal Morbidity and All-Cause Mortality in Infants of HIV-Infected and HIV Uninfected Mothers: An Intervention Cohort Study in KwaZulu Natal, South Africa. PLoS One. 2013;8(12):e81307. 42. Cutland CL, Schrag SJ, Zell ER, Kuwanda L, Buchmann E, Velaphi SC, et al. Maternal HIV infection and vertical transmission of pathogenic bacteria. Pediatrics. 2012 Sep;130(3):e581–90. 157 43. Landes M, van Lettow M, Chan AK, Mayuni I, Schouten EJ, Bedell RA. Mortality and Health Outcomes of HIV-Exposed and Unexposed Children in a PMTCT Cohort in Malawi. PLoS One. 2012;7(10):e47337. 44. Moraleda C, de Deus N, Serna-Bolea C, Renom M, Quintó L, Macete E, et al. Impact of HIV exposure on health outcomes in HIV-negative infants born to HIV-positive mothers in Sub-Saharan Africa. J Acquir Immune Defic Syndr. 2014;65(2):182–9. 45. Slogrove A, Reikie B, Naidoo S, De Beer C, Ho K, Cotton M, et al. HIV-exposed uninfected infants are at increased risk for severe infections in the first year of life. J Trop Pediatr. 2012;58(6):505–8. 46. Marquez C, Okiring J, Chamie G, Ruel TD, Achan J, Kakuru A, et al. Increased Morbidity in Early Childhood Among HIV-exposed Uninfected Children in Uganda is Associated with Breastfeeding Duration. J Trop Pediatr. 2014;60(6):434–41. 47. Le Roux DM, Myer L, Nicol MP, Zar HJ. Incidence and severity of childhood pneumonia in the first year of life in a South African birth cohort: the Drakenstein Child Health Study. Lancet Glob Heal. 2015;3(2):e95–103. 48. Zaba B, Whitworth J, Marston M, Nakiyingi J, Ruberantwari A, Urassa M, et al. HIV and Mortality of Mothers and Children. Epidemiology. 2005 May;16(3):275–80. 49. Adler C, Haelterman E, Barlow P, Marchant A, Levy J, Goetghebuer T. Severe Infections in HIV-Exposed Uninfected Infants Born in a European Country. PLoS One. 2015;10(8):e0135375. 50. Paul ME, Chantry CJ, Read JS, Frederick MM, Lu M, Pitt J, et al. Morbidity and Mortality During The First Two Years of Life Among Uninfected Children Born to Human Immunodeficiency Virus Type 1-Infected Women. Pediatr Infect Dis J. 2005 Jan;24(1):46–56. 51. Mussi-Pinhata MM, Freimanis L, Yamamoto AY, Korelitz J, Pinto JA, Cruz MLS, et al. Infectious disease morbidity among young HIV-1-exposed but uninfected infants in Latin American and Caribbean countries: the National Institute of Child Health and Human Development International Site Development Initiative Perinatal Study. Pediatrics. 2007 Mar;119(3):e694–704. 52. UNICEF. The State of The World’s Children 2015 Country Statistical Information [Internet]. The State of The World’s Children 2015: Reimagine the Future: Innovation for every child. 2015 [cited 158 2015 Jul 30]. Available from: http://www.data.unicef.org/resources/the-state-of-the-world-s-children-report-2015-statistical-tables 53. Leatt A. Income poverty in South Africa. In: Monson J, Hall K, Smith C, Shung-King M, editors. South African Child Gauge. Cape Town; 2006. p. 25. 54. Hall K, Woolard I. Children and inequality: An introduction and overview. In: Hall K, Woolard I, Lake L, Smith C, editors. South African Child Gauge. Cape Town; 2012. p. 35. 55. Kerber KJ, Lawn JE, Johnson LF, Mahy M, Dorrington RE, Phillips H, et al. South African child deaths 1990–2011: have HIV services reversed the trend enough to meet Millenium Development Goal 4? AIDS. 2013;27(16):2637–48. 56. Bourne DE, Thompson M, Brody LL, Cotton M, Draper B, Laubscher R, et al. Emergence of a peak in early infant mortality due to HIV / AIDS in South Africa. AIDS. 2009;23(1):101–6. 57. Madhi SA, Bamford L, Ngcobo N. Effectiveness of pneumococcal conjugate vaccine and rotavirus vaccine introduction into the South African public immunisation programme. South African Med J. 2014;104(Suppl 1). 58. South African National Department of Health. Interim Report of the Committee on Morbidity and Mortality in Children Under 5 Years (CoMMiC). 2012. 59. Groome MJ, Page N, Cortese MM, Moyes J, Zar HJ, Kapongo CN, et al. Effectiveness of monovalent human rotavirus vaccine against admission to hospital for acute rotavirus diarrhoea in South African children: a case-control study. Lancet. 2014;14(November):1096–104. 60. Von Gottberg A, de Gouveia L, Tempia S, Quan V, Meiring S, von Mollendorf C, et al. Effects of Vaccination on Invasive Pneumococcal Disease in South Africa. N Engl J Med. 2014;371(20):1889–99. 61. Wang H, Liddell CA, Coates MM, Mooney MD, Levitz CE, Schumacher AE, et al. Global, regional, and national levels of neonatal, infant, and under-5 mortality during 1990-2013: a systematic analysis for the Global Burden of Disease Study 2013. Lancet. 2014;384(September 13):957–79. 62. Statistics South Africa. Recorded live Births 2011. Pretoria; 2012. 63. Lawn JE, Cousens S, Zupan J. 4 million neonatal deaths: When? Where? Why? Lancet. 2005;365(March 5):891–900. 159 64. Black RE, Cousens S, Johnson HL, Lawn JE, Rudan I, Bassani DG, et al. Global, regional, and national causes of child mortality in 2008: a systematic analysis. Lancet. 2010 Jun 5;375(9730):1969–87. 65. Sania A, Spiegelman D, Rich-Edwards J, Okuma J, Kisenge R, Msamanga G, et al. The contribution of preterm birth and intrauterine growth restriction to infant mortality in Tanzania. Paediatr Perinat Epidemiol. 2014 Jan;28(1):23–31. 66. Patel K, Shapiro DE, Brogly SB, Livingston EG, Stek AM, Bardeguez AD, et al. Prenatal protease inhibitor use and risk of preterm birth among HIV-infected women initiating antiretroviral drugs during pregnancy. J Infect Dis. 2010 Apr 1;201(7):1035–44. 67. Sofeu CL, Warszawski J, Ateba Ndongo F, Penda IC, Tetang Ndiang S, Guemkam G, et al. Low birth weight in perinatally HIV-exposed uninfected infants: observations in urban settings in Cameroon. PLoS One. 2014 Jan;9(4):e93554. 68. Wei R, Msamanga GI, Spiegelman D, Hertzmark E, Baylin A, Manji K, et al. Association Between Low Birth Weight and Infant Mortality in Children Born to Human Immunodeficiency Virus 1-Infected Mothers in Tanzania. Pediatr Infect Dis J. 2004 Jun;23(6):530–5. 69. Slyker JA, Patterson J, Ambler G, Richardson BA, Maleche-Obimbo E, Bosire R, et al. Correlates and outcomes of preterm birth, low birth weight, and small for gestational age in HIV-exposed uninfected infants. BMC Pregnancy Childbirth. 2014 Jan;14:7. 70. WHO Collaborative Study Team. Effect of breastfeeding on infant and child mortality due to infectious diseases in less developed countries: a pooled analysis. WHO Collaborative Study Team on the Role of Breastfeeding on the Prevention of Infant Mortality. Lancet. 2000 Feb 5;355(9202):451–5. 71. Nduati R, John G, Mbori-Ngacha D, Richardson B, Overbaugh J, Mwatha A, et al. Effect of breastfeeding and formula feeding on transmission of HIV-1: a randomized clinical trial. J Am Med Assoc. 2000 Mar 1;283(9):1167–74. 72. Mbori-Ngacha D, Nduati R, John G, Reilly M, Richardson B, Mwatha A, et al. Morbidity and Mortality in Breastfed and Formula-Fed Infants of HIV-1–Infected Women. J Am Med Assoc. 2001;286(19):2413–20. 160 73. Thior I, Lockman S, Smeaton LM, Shapiro RL, Wester C, Heymann SJ, et al. Breastfeeding plus infant zidovudine prophylaxis for 6 months vs formula feeding plus infant zidovudine for 1 month to reduce mother-to-child HIV transmission in Botswana: a randomized trial: the Mashi Study. J Am Med Assoc. 2006 Aug 16;296(7):794–805. 74. Kuhn L, Aldrovandi GM, Sinkala M, Kankasa C, Semrau K, Mwiya M, et al. Effects of early, abrupt weaning on HIV-free survival of children in Zambia. N Engl J Med. 2008;359(2):130–41. 75. Kuhn L, Sinkala M, Semrau K, Kankasa C, Kasonde P, Mwiya M, et al. Elevations in mortality associated with weaning persist into the second year of life among uninfected children born to HIV-infected mothers. Clin Infect Dis. 2010 Feb 1;50(3):437–44. 76. Onyango-Makumbi C, Bagenda D, Mwatha A, Omer SB, Musoke P, Mmiro F, et al. Early Weaning of HIV-Exposed Uninfected Infants and Risk of Serious Gastroenteritis: Findings from Two Perinatal HIV Prevention Trials in Kampala, Uganda. J Acquir Immune Defic Syndr. 2010 Sep 25;53(1):20–7. 77. Kafulafula G, Hoover DR, Taha TE, Thigpen M, Li Q, Fowler MG, et al. Frequency of gastroenteritis and gastroenteritis-associated mortality with early weaning in HIV-1-uninfected children born to HIV-infected women in Malawi. J Acquir Immune Defic Syndr. 2010 Jan;53(1):6–13. 78. De Onis M, Blossner M. The World Health Organization Global Database on Child Growth and Malnutrition: methodology and applications. Int J Epidemiol. 2003 Aug 1;32(4):518–26. 79. Black RE, Allen LH, Bhutta ZA, Caulfield LE, de Onis M, Ezzati M, et al. Maternal and child undernutrition: global and regional exposures and health consequences. Lancet. 2008 Jan 19;371(9608):243–60. 80. Arpadi S, Fawzy A, Aldrovandi GM, Kankasa C, Sinkala M, Mwiya M, et al. Growth faltering due to breastfeeding cessation in uninfected children born to HIV-infected mothers in Zambia. Am J Clin Nutr. 2009;90:344–53. 81. McGrath CJ, Nduati R, Richardson BA, Kristal AR, Mbori-Ngacha D, Farquhar C, et al. The Prevalence of Stunting Is High in HIV-1–Exposed Uninfected Infants in Kenya. J Nutr. 2012;142:757–63. 161 82. Owor M, Mwatha A, Donnell D, Musoke P, Mmiro F, Allen M, et al. Long-term follow-up of children in the HIVNET 012 perinatal HIV prevention trial: five-year growth and survival. J Acquir Immune Defic Syndr. 2013 Dec 15;64(5):464–71. 83. Sherry B, Embree JE, Mei Z, Ndinya-Achola JO, Njenga S, Muchunga ER, et al. Sociodemographic characteristics, care, feeding practices, and growth of cohorts of children born to HIV-1 seropositive and seronegative mothers in Nairobi, Kenya. Trop Med Int Heal. 2000 Oct;5(10):678–86. 84. Isanaka S, Duggan C, Fawzi WW. Patterns of postnatal growth in HIV-infected and HIV-exposed children. Nutr Rev. 2009;67(6):343–59. 85. Filteau S, Baisley K, Chisenga M, Kasonka L, Gibson RS. Provision of micronutrient-fortified food from 6 months of age does not permit HIV-exposed uninfected Zambian children to catch up in growth to HIV-unexposed children: a randomized controlled trial. J Acquir Immune Defic Syndr. 2011 Feb;56(2):166–75. 86. Patel D, Bland R, Coovadia H, Rollins N, Coutsoudis A, Newell M-L. Breastfeeding, HIV status and weights in South African children: a comparison of HIV-exposed and unexposed children. AIDS. 2010 Jan 28;24(3):437–45. 87. Muhangi L, Lule SA, Mpairwe H, Ndibazza J, Kizza M, Nampijja M, et al. Maternal HIV infection and other factors associated with growth outcomes of HIV-uninfected infants in Entebbe, Uganda. Public Health Nutr. 2013;16(9):1548–57. 88. Ramokolo V, Lombard C, Fadnes LT, Doherty T, Jackson DJ, Goga AE, et al. HIV Infection, Viral Load, Low Birth Weight, and Nevirapine Are Independent Influences on Growth Velocity in HIV-Exposed South African Infants. J Nutr. 2014;144:42–8. 89. Briand N, Traisathit P, Karnchanamayul V, Hansudewechakul R, Ngampiyasakul C, Bhakeecheep S, et al. Growth of Human Immunodeficiency Virus-Uninfected Children Exposed to Perinatal Zidovudine for the Prevention of Mother-to-Child Human Immunodeficiency Virus Transmission. Pediatr Infect Dis J. 2006;25(4):325–32. 162 90. Powis KM, Smeaton L, Ogwu A, Lockman S, Dryden-Peterson S, van Widenfelt E, et al. Effects of in utero antiretroviral exposure on longitudinal growth of HIV-exposed uninfected infants in Botswana. J Acquir Immune Defic Syndr. 2011 Feb;56(2):131–8. 91. Marais BJ, Gie RP, Schaaf HS, Hesseling AC, Obihara CC, Starke JJ, et al. The natural history of childhood intra-thoracic tuberculosis: A critical review of literature from the pre-chemotherapy era. Int J Tuberc Lung Dis. 2004;8(4):392–402. 92. Shiri T, Auranen K, Nunes MC, Adrian P V, van Niekerk N, de Gouveia L, et al. Dynamics of pneumococcal transmission in vaccine-naive children and their HIV-infected or HIV-uninfected mothers during the first 2 years of life. Am J Epidemiol. 2013 Dec 1;178(11):1629–37. 93. Bekker A, Slogrove AL, Schaaf HS, Du Preez K, Hesseling AC. Determinants of tuberculosis treatment completion among newborns in a high-burden setting. Int J Tuberc Lung Dis. 2014;18(3):335–40. 94. Cranmer LM, Kanyugo M, Jonnalagadda SR, Lohman-Payne B, Sorensen B, Maleche Obimbo E, et al. High prevalence of tuberculosis infection in HIV-1 exposed Kenyan infants. Pediatr Infect Dis J. 2014 Apr;33(4):401–6. 95. Madhi SA, Nachman S, Violari A, Kim S, Cotton MF, Bobat R, et al. Primary Isoniazid Prophylaxis against Tuberculosis in HIV-Exposed Children. N Engl J Med. 2011;365(1):21–31. 96. Almeida VDC, Negrini BVM, Cervi MC, Isaac MDL, Mussi-Pinhata MM. Pneumococcal Nasopharyngeal Carriage Among Infants Born to Human Immunodeficiency Virus-infected Mothers Immunized With Pneumococcal Polysaccharide Vaccine During Gestation. Pediatr Infect Dis J. 2011 Jun;30(6):466–70. 97. Black V, Brooke S, Chersich MF. Effect of Human Immunodeficiency Virus Treatment on Maternal Mortality at a Tertiary Center in South Africa: A 5-Year Audit. Obstet Gynecol. 2009;114(2):292–9. 98. Sherr L, Cluver LD, Betancourt TS, Kellerman SE, Richter LM, Desmond C. Evidence of impact: health, psychological and social effects of adult HIV on children. AIDS. 2014;28(Suppl 3):S251–9. 99. De Maria A, Cirillo C, Moretta L. Occurrence of Human Immunodeficiency Virus Type 1 (HlV-l) -Specific Cytolytic T Cell Activity in Apparently Uninfected Children Born to HIV-1- Infected Mothers. J Infect Dis. 1994;170(5):1296–9. 163 100. Rowland-Jones SL, Nixon DF, Aldhous MC, Gotch F, Ariyoshi K, Hallam N, et al. HIV-specific cytotoxic T-cell activity in an HIV-exposed but uninfected infant. Lancet. 1993;341(April 3):860–1. 101. Clerici M, Saresella M, Colombo F, Fossati S, Sala N, Bricalli D, et al. T-lymphocyte maturation abnormalities in uninfected newborns and children with vertical exposure to HIV. Blood. 2000;96(12):3866–71. 102. Coffin JM. Virology of AIDS: 1990. AIDS. 1990;4(Suppl 1):S1–8. 103. Sela M. Immunology in AIDS. AIDS. 1990;4(Suppl 1):S9–14. 104. De Milito A, Nilsson A, Titanji K, Thorstensson R, Reizenstein E, Narita M, et al. Mechanisms of hypergammaglobulinemia and impaired antigen-specific humoral immunity in HIV-1 infection. Blood. 2004 Mar 15;103(6):2180–6. 105. Anderson RM, Medley GF. Epidemiology of HIV infection and AIDS: incubation and infectious periods, survival and vertical transmission. AIDS. 1988;2 Suppl 1:S57–63. 106. Fournier A-M, Baillat V, Alix-Panabieres C, Fondere J-M, Merle C, Segondy M, et al. Dynamics of spontaneous HIV-1 specific and non-specific B-cell responses in patients receiving antiretroviral therapy. AIDS. 2002;16(13):1755–60. 107. Lane HC, Masur H, Edgar LC, Whalen G, Rook AH, Fauci AS. Abnormalities of B-cell activation and immunoregulation in patients with the acquired immunodeficiency syndrome. N Engl J Med. 1983;309:453–8. 108. Appay V, Sauce D. Immune activation and inflammation in HIV-1 infection: causes and consequences. J Pathol. 2008;214:231–41. 109. Lawrie D, Coetzee LM, Becker P, Mahlangu J, Stevens W, Glencross DK. Local reference ranges for full blood count and CD4 lymphocyte count testing. South African Med J. 2009;99(4):243–8. 110. HIV/AIDS Programme. WHO case definitions of HIV for surveillance and revised clinical staging and immunological classification of HIV-related disease in adults and children. World Health Organisation. 2007. 111. Fox MP, Brooks DR, Kuhn L, Aldrovandi G, Sinkala M, Kankasa C, et al. Role of breastfeeding cessation in mediating the relationship between maternal HIV disease stage and increased child mortality among HIV-exposed uninfected children. Int J Epidemiol. 2009;38(2):569–76. 164 112. South African National AIDS Council. CLINICAL GUIDELINES: PMTCT (Prevention of Mother-to- Child Transmission) National Department of Health, South Africa. South African National Department of Health. 2010. 113. South African National Department of Health. National consolidated guidelines for the prevention of mother-to-child transmission of HIV (PMTCT) and the management of HIV in children, adolescents and adults. 2014. 114. Rallón N, Sempere-Ortells JM, Soriano V, Benito JM. Central memory CD4 T cells are associated with incomplete restoration of the CD4 T cell pool after treatment-induced long-term undetectable HIV viraemia. J Antimicrob Chemother. 2013;68(11):2616–25. 115. Ndumbi P, Gillis J, Raboud J, Cooper C, Hogg R, Montaner J, et al. Characteristics and determinants of T-cell phenotype normalization in HIV-1-infected individuals receiving long-term antiretroviral therapy. HIV Med. 2014;15(3):153–64. 116. Reikie BA, Adams RCM, Leligdowicz A, Ho K, Naidoo S, Ruck CE, et al. Altered innate immune development in HIV-exposed uninfected infants. J Acquir Immune Defic Syndr. 2014;66(3):245–55. 117. Velilla PA, Montoya CJ, Hoyos A, Moreno ME, Chougnet C, Rugeles MT. Effect of intrauterine HIV-1 exposure on the frequency and function of uninfected newborns’ dendritic cells. Clin Immunol. 2008 Mar;126(3):243–50. 118. Borges-Almeida E, Milanez HMBPM, Vilela MMS, Cunha FGP, Abramczuk BM, Reis-Alves SC, et al. The impact of maternal HIV infection on cord blood lymphocyte subsets and cytokine profile in exposed non-infected newborns. BMC Infect Dis. 2011 Jan;11(1):38. 119. Legrand FA, Nixon DF, Loo CP, Ono E, Chapman JM, Miyamoto M, et al. Strong HIV-1-specific T cell responses in HIV-1-exposed uninfected infants and neonates revealed after regulatory T cell removal. PLoS One. 2006 Jan;1(1):e102. 120. Kuhn L, Meddows-Taylor S, Gray G, Tiemessen C. Human immunodeficiency virus (HIV)-specific cellular immune responses in newborns exposed to HIV in utero. Clin Infect Dis. 2002 Jan 15;34(2):267–76. 165 121. Kakkar F, Lamarre V, Ducruet T, Boucher M, Valois S, Soudeyns H, et al. Impact of maternal HIV-1 viremia on lymphocyte subsets among HIV-exposed uninfected infants: protective mechanism or immunodeficiency. BMC Infect Dis. 2014;14(1):236. 122. Bunders M, Thorne C, Newell ML. Maternal and infant factors and lymphocyte, CD4 and CD8 cell counts in uninfected children of HIV-1-infected mothers. AIDS. 2005;19(10):1071–9. 123. Hygino J, Vieira MM, Guillermo L V, Silva-Filho RG, Saramago C, Lima-Silva AA, et al. Enhanced Th17 phenotype in uninfected neonates born from viremic HIV-1-infected pregnant women. J Clin Immunol. 2011;31(2):186–94. 124. Vigano A, Saresella M, Schenal M, Erba P, Pacentini L, Tornaghi R, et al. Immune activation and normal levels of endogenous antivirals are seen in healthy adolescents born of HIV-infected mothers. AIDS. 2007;21(2):245–8. 125. De Moraes-Pinto MI, Almeida ACM, Kenj G, Filgueiras TE, Tobias W, Santos AMN, et al. Placental transfer and maternally acquired neonatal IgG immunity in Human Immunodeficiency Virus Infection. J Infect Dis. 1996;173(5):1077–84. 126. De Moraes-Pinto MI, Verhoeff F, Chimsuku L, Milligan PJM, Wesumperuma L, Broadhead RL, et al. Placental antibody transfer: influence of maternal HIV infection and placental malaria. Arch Dis Childhood Fetal Neonatal Ed. 1998;79:F202–5. 127. Farquhar C, Nduati R, Haigwood N, Sutton W, Mbori-Ngacha D, Richardson B, et al. High maternal HIV-1 viral load during pregnancy is associated with reduced placental transfer of measles IgG antibody. J Acquir Immune Defic Syndr. 2005;40(4):494–7. 128. Jones CE, Naidoo S, De Beer C, Esser M, Kampmann B, Hesseling AC. Maternal HIV infection and antibody responses against vaccine-preventable diseases in uninfected infants. J Am Med Assoc. 2011 Feb 9;305(6):576–84. 129. Dangor Z, Kwatra G, Izu A, Adrian P, van Niekerk N, Cutland CL, et al. HIV-1 Is Associated With Lower Group B Streptococcus Capsular and Surface-Protein IgG Antibody Levels and Reduced Transplacental Antibody Transfer in Pregnant Women. J Infect Dis. 2015;212:453–62. 130. Madhi SA, Cutland CL, Kuwanda L, Weinberg A, Hugo A, Jones S, et al. Influenza Vaccination of Pregnant Women and Protection of Their Infants. N Engl J Med. 2014;371(10):918–31. 166 131. Reikie BA, Naidoo S, Ruck CE, Slogrove AL, de Beer C, la Grange H, et al. Antibody responses to vaccination among South African HIV-exposed and unexposed uninfected infants over the first two years of life. Clin Vaccine Immunol. 2013 Oct 31;20(1):33–8. 132. Madhi SA, Izu A, Violari A, Cotton MF, Panchia R, Dobbels E, et al. Immunogenicity following the first and second doses of 7-valent pneumococcal conjugate vaccine in HIV-infected and -uninfected infants. Vaccine. 2013;31(5):777–83. 133. Bunders M, Pembrey L, Kuijpers T, Newell M-L. Evidence of impact of maternal HIV infection on immunoglobulin levels in HIV-exposed uninfected children. AIDS Res Hum Retroviruses. 2010 Sep;26(9):967–75. 134. Gandhi M, Mwesigwa J, Aweeka F, Plenty A, Charlebois E, Ruel TD, et al. Hair and plasma data show that lopinavir, ritonavir, and efavirenz all transfer from mother to infant in utero, but only efavirenz transfers via breastfeeding. J Acquir Immune Defic Syndr. 2013;63(5):578–84. 135. Heidari S, Mofenson L, Cotton MF, Marlink R, Cahn P, Katabira E. Antiretroviral drugs for preventing mother-to-child transmission of HIV: a review of potential effects on HIV-exposed but uninfected children. J Acquir Immune Defic Syndr Defic Syndr. 2011 Aug 1;57(4):290–6. 136. Kenny J, Musiime V, Judd A, Gibb D. Recent advances in pharmacovigilance of antiretroviral therapy in HIV-infected and exposed children. Curr Opin HIV AIDS. 2012 Jul;7(4):305–16. 137. Le Chenadec J, Mayaux M-J, Guihenneuc-Jouyaux C, Blanche S. Perinatal antiretroviral treatment and hematopoiesis in HIV-uninfected infants. AIDS. 2003 Sep 26;17(14):2053–61. 138. Dryden-Peterson S, Shapiro RL, Hughes MD, Powis K, Ogwu A, Moffat C, et al. Increased Risk of Severe Infant Anemia After Exposure to Maternal HAART, Botswana. J Acquir Immune Defic Syndr. 2011;56(5):428–36. 139. Levy A, Fraser D, Rosen SD, Dagan R, Deckelbaum RJ, Coles C, et al. Anemia as a risk factor for infectious diseases in infants and toddlers: Results from a prospective study. Eur J Epidemiol. 2005 Jan;20(3):277–84. 140. Van Rompay KKA, Brignolo LL, Meyer DJ, Jerome C, Tarara R, Spinner A, et al. Biological Effects of Short-Term or Prolonged Administration of 9-[2-(Phosphonomethoxy) Propyl] Adenine 167 (Tenofovir) to Newborn and Infant Rhesus Macaques. Antimicrob Agents Chemother. 2004;48(5):1469–87. 141. Van Rompay KKA, Durand-Gasselin L, Brignolo LL, Ray AS, Abel K, Cihlar T, et al. Chronic administration of tenofovir to rhesus macaques from infancy through adulthood and pregnancy: Summary of pharmacokinetics and biological and virological effects. Antimicrob Agents Chemother. 2008;52(9):3144–60. 142. Gafni RI, Hazra R, Reynolds JC, Maldarelli F, Antonella N, Decarlo E, et al. Tenofovir Disoproxil Fumarate and an Optimized Background Regimen of Antiretroviral Agents as Salvage Therapy: Impact on Bone Mineral Density in HIV-Infected Children. Pediatrics. 2006;118(3):e711–8. 143. Hussain S, Khayat A, Tolaymat A, Rathore MH. Nephrotoxicity in a child with perinatal HIV on tenofovir, didanosine and lopinavir/ritonavir. Pediatr Nephrol. 2006 Jul;21(7):1034–6. 144. Siberry GK, Williams PL, Mendez H, Seage GR, Jacobson DL, Hazra R, et al. Safety of tenofovir use during pregnancy: early growth outcomes in HIV-exposed uninfected infants. AIDS. 2012 Jun 1;26(9):1151–9. 145. Ransom CE, Huo Y, Patel K, Scott GB, Watts HD, Williams P, et al. Infant growth outcomes after maternal tenofovir disoproxil fumarate use during pregnancy. J Acquir Immune Defic Syndr. 2013 Dec 1;64(4):374–81. 146. Gibb DM, Kizito H, Russell EC, Chidziva E, Zalwango E, Nalumenya R, et al. Pregnancy and Infant Outcomes among HIV-Infected Women Taking Long-Term ART with and without Tenofovir in the DART Trial. PLoS Med. 2012;9(5):e1001217. 147. Antiretroviral Pregnancy Registry. Summary and advisory committee consensus. 2014. 148. Poirier MC, Divi RL, Al-Harthi L, Olivero OA, Nguyen V, Walker B, et al. Long-term mitochondrial toxicity in HIV-uninfected infants born to HIV-infected mothers. J Acquir Immune Defic Syndr. 2003 Jun 1;33(2):175–83. 149. Brogly SB, Ylitalo N, Mofenson LM, Oleske J, Van Dyke R, Crain MJ, et al. In utero nucleoside reverse transcriptase inhibitor exposure and signs of possible mitochondrial dysfunction in HIV-uninfected children. AIDS. 2007 May 11;21(8):929–38. 168 150. Shah I. Lactic acidosis in HIV-exposed infants with perinatal exposure to antiretroviral therapy. Ann Trop Paediatr. 2009 Dec;29(4):257–61. 151. Hernàndez S, Morén C, López M, Coll O, Cardellach F, Gratacós E, et al. Perinatal outcomes, mitochondrial toxicity and apoptosis in HIV-treated pregnant women and in-utero-exposed newborn. AIDS. 2012 Feb 20;26(4):419–28. 152. Jitratkosol MHJ, Sattha B, Maan EJ, Gadawski I, Harrigan PR, Forbes JC, et al. Blood mitochondrial DNA mutations in HIV-infected women and their infants exposed to HAART during pregnancy. AIDS. 2012 Mar 27;26(6):675–83. 153. Brogly S, Williams P, Seage GR, Van Dyke RB. In utero nucleoside reverse transcriptase inhibitor exposure and cancer in HIV-uninfected children: an update from the Pediatric AIDS Clinical Trials Group 219 an 219C cohorts. J Acquir Immune Defic Syndr. 2006;41(4):535–6. 154. Le Doare K, Bland R, Newell M-L. Neurodevelopment in Children Born to HIV-Infected Mothers by Infection and Treatment Status. Pediatrics. 2012;130(5):e1326–44. 155. Wilkinson JD, Williams PL, Leister E, Zeldow B, Shearer WT, Colan SD, et al. Cardiac biomarkers in HIV-exposed uninfected children. AIDS. 2013 Apr 24;27(7):1099–108. 156. Saitoh A, Hull AD, Franklin P, Spector SA. Myelomeningocele in an infant with intrauterine exposure to efavirenz. J Perinatol. 2005 Aug;25(8):555–6. 157. Ford N, Calmy A, Mofenson L. Safety of efavirenz in the first trimester of pregnancy: an updated systematic review and meta-analysis. AIDS. 2011 Nov 28;25(18):2301–4. 158. Coovadia HM, Brown ER, Fowler MG, Chipato T, Moodley D, Manji K, et al. Efficacy and safety of an extended nevirapine regimen in infant children of breastfeeding mothers with HIV-1 infection for prevention of postnatal HIV-1 transmission (HPTN 046): a randomised, double-blind, placebo-controlled trial. Lancet. 2012 Jan 21;379(9812):221–8. 159. Powis KM, Kitch D, Ogwu A, Hughes MD, Lockman S, Leidner J, et al. Increased risk of preterm delivery among HIV-infected women randomized to protease versus nucleoside reverse transcriptase inhibitor-based HAART during pregnancy. J Infect Dis. 2011 Aug 15;204(4):506–14. 160. Madhi SA, Cutland C, Ismail K, O’Reilly C, Mancha A, Klugman KP. Ineffectiveness of Trimethoprim-Sulfamethoxazole Prophylaxis and the Importance of Bacterial and Viral 169 Coinfections in African Children with Pneumocystis carinii Pneumonia. Clin Infect Dis. 2002;35(1 November):1120–6. 161. Chintu C, Mudenda V, Lucas S, Nunn A, Lishimpi K, Maswahu D, et al. Lung diseases at necropsy in African children dying from respiratory illnesses: a descriptive necropsy study. Lancet. 2002 Sep 28;360(9338):985–90. 162. Morrow BM, Hsaio N-Y, Zampoli M, Whitelaw A, Zar HJ. Pneumocystis pneumonia in South African children with and without human immunodeficiency virus infection in the era of highly active antiretroviral therapy. Pediatr Infect Dis J. 2010 Jun;29(6):535–9. 163. Slogrove AL, Cotton MF, Esser MM. Severe infections in HIV-exposed uninfected infants: clinical evidence of immunodeficiency. J Trop Pediatr. 2010 Apr;56(2):75–81. 164. Dow A, Kayira D, Hudgens M, Van Rie A, King CC, Ellington S, et al. Effects of cotrimoxazole prophylactic treatment on adverse health outcomes among HIV-exposed, uninfected infants. Pediatr Infect Dis J. 2012 Aug;31(8):842–7. 165. Coutsoudis A, Pillay K, Spooner E, Coovadia HM, Pembrey L, Newell M. Routinely available cotrimoxazole prophylaxis and occurrence of respiratory and diarrhoeal morbidity in infants born to HIV-infected mothers in South Africa. South African Med J. 2005;95(5):339–45. 166. Coutsoudis A, Kindra G, Esterhuizen T. Impact of cotrimoxazole prophylaxis on the health of breast-fed, HIV-exposed, HIV-negative infants in a resource-limited setting. AIDS. 2011 Sep 10;25(14):1797–9. 167. Moodley D, Reddy L, Mahungo W, Masha R. Factors associated with coverage of cotrimoxazole prophylaxis in HIV-exposed children in South Africa. PLoS One. 2013 Jan;8(5):e63273. 168. Groenewald P, Berteler M, Bradshaw D, Coetzee D, Cornelius K, Daniels J, et al. Western Cape Mortality Profile 2010. South African Medical Research Council. 2013. 169. Mussi-Pinhata MM, Motta F, Freimanis-Hance L, de Souza R, Szyld E, Succi RCM, et al. Lower respiratory tract infections among human immunodeficiency virus-exposed, uninfected infants. Int J Infect Dis. 2010 Sep;14 Suppl 3:e176–82. 170 170. Mwiru R, Spiegelman D, Hertzmark E, Duggan C, Msamanga G, Aboud S, et al. Nutritional predictors of acute respiratory infections among children born to HIV-infected women in Tanzania. J Trop Pediatr. 2013 Jun;59(3):203–8. 171. Santos RP, Tristram D. A Practical Guide to the Diagnosis, Treatment, and Prevention of Neonatal Infections. Pediatr Clin North Am. 2015;62(2):491–508. 172. Schelonka RL, Infante AJ. Neonatal Immunology. Semin Perinatol. 1998;22(1):2–14. 173. Epalza C, Goetghebuer T, Hainaut M, Prayez F, Barlow P, Dediste A, et al. High incidence of invasive group B streptococcal infections in HIV-exposed uninfected infants. Pediatrics. 2010 Sep;126(3):e631–8. 174. Cutland CL, Schrag SJ, Thigpen MC, Velaphi SC, Wadula J, Adrian P V, et al. Increased Risk for Group B Streptococcus Sepsis in Young Infants Exposed to HIV, Soweto, South Africa. Emerg Infect Dis. 2015;21(4):638–45. 175. McAlmon KR. Necrotizing Enterocolitis. In: Cloherty JP, Eichenwald EC, Stark AR, editors. Manual of Neonatal Care. Edition 5. Philadelphia: Lippincott Williams & Wilkins; 2004. p. 643–5. 176. Desfrere L, de Oliveira I, Goffinet F, el Ayoubi M, Firtion G, Bavoux F, et al. Increased incidence of necrotizing enterocolitis in premature infants born to HIV-positive mothers. AIDS. 2005;19(14):1487–93. 177. Karpelowsky JS, van Mil S, Numanoglu A, Leva E, Millar AJW. Effect of maternal human immunodeficiency virus status on the outcome of neonates with necrotizing enterocolitis. J Pediatr Surg. Elsevier Inc.; 2010 Feb;45(2):315–8. 178. Arnold M, Moore SW. HIV exposure does not worsen outcome in stage III necrotizing enterocolitis with current treatment protocols. J Pediatr Surg. 2012 Apr;47(4):665–72. 179. Taron-Brocard C, Le Chenadec J, Faye A, Dollfus C, Goetghebuer T, Gajdos V, et al. Increased Risk of Serious Bacterial Infections Due to Maternal Immunosuppression in HIV-Exposed Uninfected Infants in a European Country. Clin Infect Dis. 2014;59(9):1332–45. 180. D’Agaro P, Burgnich P, Comar M, Dal Molin G, Bernardon M, Busetti M, et al. HHV-6 is frequently detected in dried cord blood spots from babies born to HIV-positive mothers. Curr HIV Res. 2008 Sep;6(5):441–6. 171 181. Guibert G, Warszawski J, Le Chenadec J, Blanche S, Benmebarek Y, Mandelbrot L, et al. Decreased risk of congenital cytomegalovirus infection in children born to HIV-1-infected mothers in the era of highly active antiretroviral therapy. Clin Infect Dis. 2009 Jun 1;48(11):1516–25. 182. Meyer SA, Westreich DJ, Patel E, Ehlinger EP, Kalilani L, Lovingood R V, et al. Postnatal cytomegalovirus exposure in infants of antiretroviral-treated and untreated HIV-infected mothers. Infect Dis Obstet Gynecol. 2014 Jan;2014:989721. 183. Roxby AC, Atkinson C, Asbjörnsdóttir K, Farquhar C, Kiarie JN, Drake AL, et al. Maternal valacyclovir and infant cytomegalovirus acquisition: a randomized controlled trial among HIV-infected women. PLoS One. 2014 Jan;9(2):e87855. 184. Provincial Government of the Western Cape. Regional Development Profile City of Cape Town Working paper. 2013. 185. Provincial Government of the Western Cape, Department of Health. HIV and syphilis prevalence in the Western Cape: results of the 2009 HIV and syphilis antenatal provincial and sub-district surveys. 2011. 186. Larsen E, Lunding S, Helleberg M, Katzenstein T, Nordly S, Weis N, et al. Hospitalizations among uninfected children exposed or unexposed to HIV - a nationwide cohort. Abstract 877. Conference on Retroviruses and Opportunistic Infections. Seattle, Washington; 2015. 187. National Institute of Allergy and Infectious Diseases Division of AIDS. Division of AIDS Table for Grading the Severity of Adult and Pediatric Adverse Events. 2009. 188. Kohl KS, Gidudu J, Bonhoeffer J, Braun MM, Buettcher M, Chen RT, et al. The development of standardized case definitions and guidelines for adverse events following immunization. Vaccine. 2007 Aug 1;25(31):5671–4. 189. ICH Harmonised Tripartite Guideline. Guideline for good clinical practice E6(R1). ICH Harmonised Tripartite Guideline. 1996. 190. Nielsen-Saines K, Watts DH, Joao EC, Pilotto JH, Gray G, Theron G, et al. Infectious morbidity, mortality, growth of HIV-exposed uninfected, formula-fed infants enrolled in NICHD/HPTN 040/PACTG 1043. Abstract O_11. 3rd International Workshop on HIV Pediatrics. Rome, Italy; 2011. 172 191. Gidudu J, Sack DA, Pina M, Hudson MJ, Kohl KS, Bishop P, et al. Diarrhea: case definition and guidelines for collection, analysis, and presentation of immunization safety data. Vaccine. 2011 Jan 29;29(5):1053–71. 192. World Health Organisation. The treatment of diarrhoea: A manual for physicians and other senior health workers. 4th ed. Geneva, Switzerland: WHO Press; 2005. 3-43 p. 193. Centers for Disease Control and Prevention. Case Definitions for Infectious Conditions Under Public Health Surveillance. Morb Mortal Wkly Rep. 1997;46(No.RR-10):1–55. 194. World Health Organisation, UNICEF. Integrated management of childhood illness for hig HIV settings: Chart Booklet. 2008. 195. South African National Department of Health. Strategic plan for maternal, newborn, child and women’s health (MNCWH) and Nutrition in South Africa. 2012. 196. Zar HJ, Jeena P, Argent A, Gie R, Madhi S. Diagnosis and management of community-acquired pneumonia in childhood--South African Thoracic Society Guidelines. South African Med J. 2005 Dec;95(12 Pt 2):977–90. 197. Green RJ, Zar HJ, Jeena PM, Madhi SA, Lewis H. South African guideline for the diagnosis, management and prevention of acute viral bronchiolitis in children. South African Med J. 2010 May;100(5):320–5. 198. International Union Against Tuberculosis and Lung Disease. Desk-guide for diagnosis and management of TB in children. Paris, France; 2010. 199. Byrt T. How Good Is That Agreement!? Epidemiology. 1996;7(5):561. 200. Cohen J. A coefficient of agreent for nominal scales. Educ Psychol Meas. 1960;XX(1):37–46. 201. Feinstein AR, Cicchetti D V. High agreement but low kappa: I. The problems of two paradoxes. J Clin Epidemiol. 1990;43(6):543–9. 202. Byrt T, Bishop J, Carlin JB. Bias, prevalence and kappa. J Clin Epidemiol. 1993 May;46(5):423–9. 203. Looney SW, Hagan JL. Statistical methods for assessing biomarkers and analyzing biomarker data. Handbook of Statistics 27: Epidemiology and Medical Statistics. Amsterdam: Elsevier B.V.; 2008. p. 115–6. 173 204. Sauro J, Lewis JR. Estimating Completion Rates from Small Samples Using Binomial Confidence Intervals: Comparisons and Recommendations. Proc Hum Factors Ergon Soc Annu Meet. 2005;49(24):2100–3. 205. Steyn K, Bradshaw D, Norman R, Laubscher R, Saloojee Y. Tobacco use in South Africans during 1998: the first demographic and health survey. J Cardiovasc Risk. 2002;9(3):161–70. 206. Petersen Williams P, Jordaan E, Mathews C, Lombard C, Parry CDH. Alcohol and Other Drug Use during Pregnancy among Women Attending Midwife Obstetric Units in the Cape Metropole, South Africa. Adv Prev Med. 2014;2014:871427. 207. Kruger M, Rosenkranz B. Health Research Ethics Committee Guideline: paediatric blood volume for research purposes. Cape Town: Stellenbosch University; 2010. p. 1–2. 208. World Health Organisation. WHO Anthro (version 3.2.2, January 2011) and macros [Internet]. 2011 [cited 2015 Jun 12]. Available from: http://www.who.int/childgrowth/software/en/ 209. Kozuki N, Lee AC, Silveira MF, Victora CG, Adair L, Humphrey J, et al. The associations of birth intervals with small-for-gestational-age, preterm, and neonatal and infant mortality: a meta-analysis. BMC Public Health. 2013;13(Suppl 3):S2. 210. Nabongo P, Verver S, Nangobi E, Mutunzi R, Wajja A, Mayanja-Kizza H, et al. Two year mortality and associated factors in a cohort of children from rural Uganda. BMC Public Health. 2014;14(314):1–9. 211. Doherty T, Jackson D, Swanevelder S, Lombard C, Engebretsen IMS, Tylleskär T, et al. Severe events in the first 6 months of life in a cohort of HIV-unexposed infants from South Africa: Effects of low birthweight and breastfeeding status. Trop Med Int Heal. 2014;19(10):1162–9. 212. Bussmann H, Wester CW, Masupu K V, Peter T, Gaolekwe SM, Kim S, et al. Low CD4+ T-lymphocyte Values in Human Immunodeficiency Virus-Negative Adults in Botswana. Clin Diagn Lab Immunol. 2004;11(5):930–5. 213. Crampin AC, Mwaungulu FD, Ambrose LR, Longwe H, French N. Normal Range of CD4 Cell Counts and Temporal Changes in Two HIV Negative Malawian Populations. Open AIDS J. 2011;5:74–9. 174 214. Chama CM, Morrupa JY, Abja UA, Kayode A. Normal CD4 T-lymphocyte baseline in healthy HIV-negative pregnant women. J Obstet Gynaecol (Lahore). 2009;29(8):702–4. 215. Towers C V, Rumney PJ, Ghamsary MG. Longitudinal study of CD4+ cell counts in HIV-negative pregnant patients. J Matern Neonatal Med. 2010;23(10):1091–6. 216. Hargrove JW, Humphrey JH. Mortality among HIV-positive postpartum women with high CD4 cell counts in Zimbabwe. AIDS. 2010 Jan;24(3):F11–4. 217. Townsend CL, Cortina-Borja M, Peckham CS, de Ruiter A, Lyall H, Tookey PA. Low rates of mother-to-child transmission of HIV following effective pregnancy interventions in the United Kingdom and Ireland, 2000-2006. AIDS. 2008;22(8):973–81. 218. Jelsma J, Maclean E, Hughes J, Tinise X, Darder M. An investigation into the health-related quality of life of individuals living with HIV who are receiving HAART. AIDS Care. 2005;17(5):579–88. 219. Van den Berg JP, Westerbeek EAM, van der Klis FRM, Berbers GAM, Van Elburg RM. Transplacental transport of IgG antibodies to preterm infants: A review of the literature. Early Hum Dev. 2011;87(2):67–72. 220. Kidzeru EB, Hesseling AC, Passmore J-AS, Myer L, Gamieldien H, Tchakoute CT, et al. In-utero exposure to maternal HIV infection alters T-cell immune responses to vaccination in HIV-uninfected infants. AIDS. 2014;28(10):1421–30. 221. Kuhn L, Kim H-Y, Hsiao L, Nissan C, Kankasa C, Mwiya M, et al. Oligosaccharide Composition of Breast Milk Influences Survival of Uninfected Children Born to HIV-Infected Mothers in Lusaka, Zambia. J Nutr. 2015;145:66–72. 222. Bode L. Human milk oligosaccharides: Every baby needs a sugar mama. Glycobiology. 2012;22(9):1147–62. 223. Heffron R, Donnell D, Kiarie J, Rees H, Ngure K, Mugo N, et al. A prospective study of the effect of pregnancy on CD4 counts and plasma HIV-1 RNA concentrations of antiretroviral-naive HIV-1-infected women. J Acquir Immune Defic Syndr. 2014;65(2):231–6. 224. Huson MAM, Grobusch MP, van der Poll T. The effect of HIV infection on the host response to bacterial sepsis. Lancet Infect Dis. 2014;15(1):95–108. 175 225. Ndumbi P, Falutz J, Pai NP, Tsoukas CM. Delay in cART initiation results in persistent immune dysregulation and poor recovery of T-cell phenotype despite a decade of successful HIV suppression. PLoS One. 2014;9(4):1–8. 226. Morris L, Binley JM, Clas BA, Bonhoeffer S, Astill TP, Kost R, et al. HIV-1 antigen-specific and -nonspecific B cell responses are sensitive to combination antiretroviral therapy. J Exp Med. 1998 Jul 20;188(2):233–45. 227. Kaplan JE, Hanson D, Dworkin MS, Frederick T, Bertolli J, Lindegren ML, et al. Epidemiology of Human Immunodeficiency Virus-Associated Opportunistic Infections in the United States in the Era of Highly Active Antiretroviral Therapy. Clin Infect Dis. 2000;30 Suppl 1:S5–14. 228. Miles DJC, van der Sande M, Jeffries D, Kaye S, Ismaili J, Ojuola O, et al. Cytomegalovirus infection in Gambian infants leads to profound CD8 T-cell differentiation. J Virol. 2007;81(11):5766–76. 229. Siberry GK, Jacobson DL, Kalkwarf HJ, Wu JW, DiMeglio LA, Yogev R, et al. Lower newborn bone mineral content associated with maternal use of tenofovir disoproxil fumarate during pregnancy. Clin Infect Dis. 2015;Jun 9 (epub advanced access). 230. Bera E, McCausland K, Nonkwelo R, Mgudlwa B, Chacko S, Majeke B. Birth defects following exposure to efavirenz-based antiretroviral therapy during pregnancy: a study at a regional South African hospital. AIDS. 2010;24(2):283–9. 231. Knapp KM, Brogly SB, Muenz DG, Spiegel HML, Conway DH, Scott GB, et al. Prevalence of congenital anomalies in infants with in utero exposure to antiretrovirals. Pediatr Infect Dis J. 2012 Feb;31(2):164–70. 232. Williams PL, Crain MJ, Yildirim C, Hazra R, Van Dyke RB, Rich K, et al. Congenital Anomalies and In Utero Antiretroviral Exposure in Human Immunodeficiency Virus–Exposed Uninfected Infants. JAMA Pediatr. 2015;169(1):48. 233. Venkatesh KK, de Bruyn G, Marinda E, Otwombe K, van Niekerk R, Urban M, et al. Morbidity and mortality among infants born to HIV-infected women in South Africa: Implications for child health in resource-limited settings. J Trop Pediatr. 2011;57(2):109–19. 176 234. Jones SA, Sherman GG, Varga CA. Exploring socio-economic conditions and poor follow-up rates of HIV-exposed infants in Johannesburg, South Africa. AIDS Care. 2005;17(May):466–71. 235. Koyanagi A, Humphrey JH, Moulton LH, Ntozini R, Mutasa K, Iliff P, et al. Effect of early exclusive breastfeeding on morbidity among infants born to HIV-negative mothers in Zimbabwe. Am J Clin Nutr. 2009;89:1375–82. 236. Gelman A, Hill J. Data Analysis Using Regression and Multilevel/Hierarchical Models. New York, NY: Cambridge University Press; 2007. p. 1-8 237. Gelman A, Hill J. Data Analysis Using Regression and Multilevel/Hierarchical Models. New York, NY: Cambridge University Press; 2007. p. 245-246 238. Szklo M, Nieto FJ. Epidemiology Beyond The Basics. Edition 3. Burlington, MA: Jones & Bartlett Learning; 2014. p. 211 239. Turck D, Michaelsen KF, Shamir R, Braegger C, Campoy C, Colomb V, et al. World Health Organization 2006 Child Growth Standards and 2007 Growth Reference Charts: A Discussion Paper by the Committee on Nutrition of the European Society for Pediatric Gastroenterology, Hepatology, and Nutrition. J Pediatr Gastroenterol Nutr. 2013;57(2):258–64. 177 Appendices Appendix A PIET-R case-definitions Table A.1 Table of PIET-R case-definitions Classification Definition of case Definition of severe case A. Lower respiratory tract infections 1. Pneumonia History of cough or difficulty breathing PLUS tachypnea for age or oxygen required WITHOUT wheeze At least 1 of • Lower chest wall indrawing or nasal flaring • Not able to drink or breastfeed • Vomiting everything • Convulsions during this illness • Lethargic or unconscious • Oxygen saturation < 92% on room air • Any oxygen required 2. Bronchiolitis History of cough or difficulty breathing PLUS wheeze or evidence of hyperinflation on physical exam At least 1 of • Tachypnea for age • Lower chest wall indrawing or nasal flaring • Not able to drink or breastfeed • Vomiting everything • Convulsions during this illness • Apnea on history or witnessed • Lethargic or unconscious • Oxygen saturation < 92% on room air • Any oxygen required 3. Laryngo-tracheobronchitis (LTB) History of cough or difficulty breathing PLUS stridor At least 1 of • Stridor on inspiration and expiration • Not able to drink or breastfeed • Lethargic or unconscious • Oxygen saturation < 92% on room air 4. Tuberculosis Started TB treatment in hospital OR Close TB contact PLUS chest X-ray suggestive of pulmonary TB OR Bacteriological confirmation on sputum, gastric aspirate, biopsy OR A close TB contact PLUS 2 of • Persistent cough >14/7 • Persistent fever >14/7 • Unsatisfactory weight gain • Fatigue, reduced playfulness Hospital diagnosed extrapulmonary/miliary TB OR At least 1 of • Tachypnea for age PLUS lower chest wall indrawing or oxygen required • Headache, neck stiffness, drowsiness, irritability, convulsions • Hepatosplenomegaly • Peripheral oedema • Distended abdomen with or without ascites • Angulation of the spine/gibbus 178 Classification Definition of case Definition of severe case B. Diarrhoeal disease 1. Acute diarrhoea Liquid stools (more unformed than usual) with increased number of stools for < 14 days At least 2 of • Lethargic or unconscious • Sunken eyes • Not able to drink or drinking poorly • Skin pinch takes > 2 seconds to return 2. Persistent diarrhoea Liquid stools (more unformed than usual) with increased number of stools for > 14 days At least 2 of • Lethargic or unconscious • Restless or irritable • Sunken eyes • Not able to drink, drinking poorly, drinking eagerly, thirsty • Skin pinch takes > 1 second to return 3. Dysentery Any blood in the stool At least 1 of • Age < 12 months • Any dehydration present C. Skin and mucocutaneous infections 1. Measles Fever AND diffuse maculopapular rash PLUS 1 of • Cough • Coryza • conjunctivitis At least 1 of • Pneumonia (as previously defined) • LTB (as previously defined) • Diarrhoea (as previously defined) • Any general danger sign 2. Varicella zoster Fever AND diffuse vesicular rash At least 1 of • Pneumonia (as previously defined) • LTB (as previously defined) • Diarrhoea (as previously defined) Any general danger sign 3. Non-specific viral exanthem Fever AND rash AND doesn’t meet criteria for measles or varicella Any general danger sign 4. Stomatitis Erythema AND oral mucosal ulceration (lips, gingiva, tongue) At least 1 of • Stridor • Unable to eat orally 5. Bacterial skin infection At least 1 of • Impetigo (diffuse pustular eruption) • Abscess Any general danger sign 179 Classification Definition of case Definition of severe case D. Invasive bacterial infections 1. Neonatal sepsis Hospital diagnosis < 28 days a. Confirmed – blood or CSF culture positive b. Presumed – no positive culture, treated for neonatal sepsis with a minimum of 7 days of intravenous antibiotics All considered as severe 2. Post-neonatal sepsis Hospital diagnosis > 28 days a. Confirmed – blood culture positive b. Presumed – no positive culture, treated for bacterial sepsis with a minimum of 7 days of intravenous antibiotics All considered as severe 3. Urinary tract sepsis Hospital diagnosis PLUS sterile urine culture positive for urinary tract pathogen All considered as severe 4. Meningitis Hospital diagnosis PLUS abnormal CSF findings All considered as severe 5. Septic arthritis Specialist diagnosis All considered as severe 6. Osteomyelitis Specialist diagnosis All considered as severe 7. Pyomyositis Specialist diagnosis All considered as severe E. Congenital infections 1. Congenital tuberculosis Paediatrician diagnosis All considered as severe 2. Congenital cytomegalovirus Paediatrician diagnosis All considered as severe 3. Congenital syphilis Paediatrician diagnosis All considered as severe 4. Neonatal herpes simplex Paediatrician diagnosis All considered as severe 5. Other Paediatrician diagnosis All considered as severe 180 Appendix B PIET-R Evaluation B.1 PIET-R Hospitalization Event Gold Standard Source Document Hospitalization event ID: _____________ Infant age: __________ completed months Date of admission: __________________ (dd/mon/yyyy) Date of discharge: __________________ (dd/mon/yyyy) Confirmation of eligibility criteria: 1=Yes 3=No 77=Unknown 1. Infant between 1 and 12 months of age 2. Hospitalized in general paediatric ward 3. HIV-infection excluded 4. No previous hospitalization for this infant included in evaluation Gold Standard Diagnosis: 1=Mild-Mod 2=Severe 4=Not Present 1. Lower respiratory tract infection: a. Pneumonia..................................................... b. Bronchiolitis.................................................... c. Tuberculosis................................................... 2. Diarrhoea: a. Acute diarrhoea................................................. b. Persistent diarrhoea.......................................... v v v v v v v v v v v v v v v v v v v v v v v v v v v vvvvvvvvvvvvvvvvvvvvvvvvvvv181 B.2 PIET-R: Abstraction Source Document Hospitalization event ID: _______________ Abstractors ID: ________________________ Section A: Doctors Problem List ________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Section B: History 1=Yes < 14d 2=Yes > 14d 3=No 77=Unknown 1. Cough ................................................................ 2. Fever .................................................................. 3. Liquid stools........................................................ 1=Yes 3=No 77=Unknown 4. Difficulty breathing............................................................................. 5. Apnoea episodes................................................................................... 6. Not able to drink or breastfeed.......................................................... 7. Not able to eat orally.......................................................................... 8. Vomiting everything.......................................................................... 9. Bloody stools..................................................................................... 10. Convulsions during this illness.......................................................... 11. Fatigue or reduced playfulness........................................................... 12. O2 required during transfer from clinic............................................. 13. Weight loss or unsatisfactory weight gain......................................... 14. Close TB contact.................................................................................. 182 Section C: Physical Examination 1=Yes 3=No 77=Unknown 1. O2 saturation < 92% on pulse oximeter.............................................. 2. Any form of O2 (NPO2, NCO2, HB O2, NCPAP, ETT) ........................ 3. Lower chest indrawing, recession, nasal flare.................................... 4. Unable to drink orally, require NG, OG or IV fluid............................... 5. Vomiting everything............................................................................ 6. Convulsions (witnessed) during admission........................................ 7. Acutely unwell, lethargic, unconscious............................................... 8. Apnoea episodes witnessed during admission..................................... 9. Skin pinch > 2 sec (CRT > 2 sec)...................................................... 10. Tachypnea for age............................................................................. 11. Persistently irritable > 2 days.............................................................. 12. Cervical lymph node mass > 2x2cm.................................................. 13. Bilateral pitting oedema, dorsum hands or feet.................................. 1=Yes 3=No 77=Unknown 14. Hydration normal...................................................................................... If No indicate abnormality, otherwise continue with 15. a) Sunken eyes, dry mucous membs or sunken fontanel ........... b) Skin pinch 1-2 sec (CRT 1-2 sec) or reduced skin turgor............ c) Not able to drink........................................................................ d) Drinking poorly......................................................................... e) Drinking eagerly/thirsty........................................................... f) Other (specify)_____________________________________ 1=Yes 3=No 77=Unknown 15. Musculoskeletal exam normal................................................................. If No indicate abnormality, otherwise continue with 16. a) Gibbus or spinal angulation....................................................... b) Other (specify)_____________________________________ 1=Yes 3=No 77=Unknown 183 16. Respiratory exam normal.......................................................................... If No indicate abnormality, otherwise continue with 17. a) Reduced cardiac dullness or hyperresonance........................ b) Liver ptosis (upper border in 7th ICS or below)........................ c) Wheeze not further described................................................. d) Wheeze bronchodilator responsive........................................... e) Wheeze bronchodilator unresponsive / low monophonic......... f) Other (specify)_____________________________________ 1=Yes 3=No 77=Unknown 17. Abdominal exam normal......................................................................... If no indicate abnormality, otherwise continue with 25. a) Abdominal distension.................................................................. b) Hepatomegaly............................................................................. c) Splenomegaly.............................................................................. d) Ascites........................................................................................ e) Other (specify)_____________________________________ 1=Yes 3=No 77=Unknown 18. Central Nervous system exam normal...................................................... If No indicate abnormality, otherwise continue with Section D. a) Level of consciousness reduced/altered...................................... b) Neck stiffness............................................................................... c) Other (specify)_____________________________________ Section D: Hospital diagnoses and Chest X-Ray findings 1=Yes 3=No 77=Unknown 1. Hospital diagnosed pulmonary TB, started on treatment........................... 2. Hospital diagnosed extrapulmonary or miliary TB, started on treatment... 3. Abnormal Chest X-Ray suggestive of TB (see instructions) 4. Abnormal Chest X-Ray with hyperinflation (see instructions) v v v v v v v v v v v v v v v v v v v v v v v v v v v v v 184 B.3 PIET-R: Final Classification and Grade Source Document Hospitalization event ID: ____________ Abstractors ID: ____________________ PIET-R Classification & Grade: 1=Mild-Mod 2=Severe 3=Ungraded 4=Not Present 1. Lower respiratory tract infection: a. Pneumonia................................. b. Bronchiolitis............................... c. Tuberculosis.............................. 2. Diarrhoea: a. Acute diarrhoea........................... b. Persistent diarrhoea.................... 3. Unclassified.......................................... v v v v v v v v v v v v v v v v v v v 185 B.4 PABAK descriptive classification according to Byrt Table B.1 PABAK descriptive classification (199) Kappa Interpretation 0.93-1.00 Excellent agreement 0.81-0.92 Very good agreement 0.61-0.80 Good agreement 0.41-0.60 Fair agreement 0.21-0.40 Slight agreement 0.01-0.20 Poor agreement 0.00 or less No agreement 186 Appendix C Study variables Table C.1 Table of variables Variable Type Source Response options Maternal variables – all mothers HIV status Binary Maternal obstetric record & confirmed with rapid HIV test at 2 week visit 1=HIV-infected; 2=HIV-uninfected Age Numeric continuous Maternal baseline interview Years/days; calculated from date of birth Race Nominal Maternal baseline interview 1=Black African, 2=Coloured, 3=Asian or Indian, 4=White, 5=Other Language Nominal Maternal baseline interview 1=Afrikaans, 2=English, 3=Xhosa, 4=Zulu, 5=Other Marital status Nominal Maternal baseline interview 1 = never married, 2 = married, 3 = widowed/separated/divorced Maternal education Nominal Maternal baseline interview 1 = no or any primary, 2 = some secondary, 3 = completed secondary Monthly income in ZAR Numeric continuous Maternal baseline interview Range 0-20 000 Primiparous Binary Maternal obstetric record 1 = Yes, 2 = No Gestational age at first antenatal visit Numeric integer Maternal obstetric record Number of completed weeks Number of antenatal clinic visits Numeric integer Maternal obstetric record Range 0-12 Antenatal syphilis status Nominal Maternal obstetric record Positive; Negative; Indeterminate Obstetric complications Nominal Maternal obstetric record 1 = gestational hypertension, 2 = gestational diabetes, 3 = antepartum haemorrhage, 4 = postpartum haemorrhage TB treatment during pregnancy Binary Maternal baseline interview 1 = Yes, 2 = No TB treatment postnatal Binary Maternal interview at each visit 1 = Yes, 2 = No 187 Variable Type Source Response options Smoked during pregnancy Ordinal Maternal baseline interview 1 = never (did not smoke on any day during pregnancy), 2 = occasionally (smoked but not every day during pregnancy), 3 = daily (smoked every day or most days during pregnancy) Alcohol use during pregnancy Ordinal Maternal baseline interview 1 = never, 2 = only once during the pregnancy, 3 = once or twice a month, 4 = about once a week, 5 = 2-3 times a week, 6 = 4 or more times a week Illicit drug use during pregnancy Binary Maternal baseline interview 1 = Yes, 2 = No Previous still births Numeric integer Maternal baseline interview Range 0-5 Previous child death (<5 years old) Numeric integer Maternal baseline interview Range 0-5 Delivery CD4 absolute count in cells/µl Numeric continuous Laboratory result Range 0-3000 Delivery CD4 % Numeric continuous Laboratory result Range 0%-60% Feeding intention at delivery Nominal Maternal baseline interview 1 = only breast milk, 2 = only formula milk, 3 = combination of breast milk, formula milk or solids Body mass index in kg/m2 Numeric continuous Measured at 2 week visit Range 12-40 Maternal HIV specific variables Timing of HIV diagnosis Binary derived from date of HIV diagnosis and baby's date of birth Maternal baseline interview 1 = pre-pregnancy, 2 = pregnancy Timing of initiation of cART Ordinal derived from start date of cART and baby's date of birth Maternal baseline interview 1 = pre-pregnancy, 2 = 1st trimester, 3 = 2nd trimester, 4 = 3rd trimester, 5 = None Timing of initiation of ZDV Ordinal derived from start date of ZDV and baby's date of birth Maternal obstetric record 1 = 1st trimester, 2 = 2nd trimester, 3 = 3rd trimester, 4 = None ARV type during pregnancy Nominal Maternal obstetric record 1 = ZDV, 2 = cART, 3 = None WHO HIV stage during pregnancy Nominal Maternal obstetric record 1 = stage 1, 2 = stage 2, 3 = stage 3, 4 = stage 4 Antenatal CD4 absolute count Numeric continuous Maternal obstetric record Range 0-3000 188 Variable Type Source Response options Delivery HIV viral load Numeric continuous Laboratory result Range <40 – 7 million cART regimen Nominal - derived from report of individual ARVs received Maternal interview at each visit 1 = \"None\", 2 = \"TDF/3TC/NVP\", 3 = \"TDF/3TC/EFV\", 4 = \"TDF/3TC/LPVr\", 5 = \"d4T/3TC/NVP\", 6 = \"ZDV/3TC/NVP\", 7 = \"d4T/3TC/EFV\", 8 = \"ZDV/3TC/EFV\", 9 = \"ZDV/3TC/LPVr\", 10 = \"Other\" cART regimen line Binary - derived from report of individual ARVs received Maternal interview at each visit 1 = first line (all NNRTI based regimens), 2 = second line (TDF/3TC/LPVr; ZDV/3TC/LPVr) Maternal cART adherence Nominal Maternal interview at each visit 1 = Very poor, 2 = Poor, 3 = Fair, 4 = Good, 5 = Very Good, 6= Excellent Household variables Type of house Nominal Maternal baseline interview 1 = stand alone house, 2= apartment in apartment block, 3 = house/flat/room in backyard, 4 = shack in backyard, 5 = shack not in back yard, 6 = other Water supply Nominal Maternal baseline interview 1 = water piped into dwelling, 2 = water piped into site/yard, 3 = public tap, Sanitation Nominal Maternal baseline interview 1 = flush toilet (connected to sewage), 2 = flush toilet (with septic tank), 3 = no facility/bush/field Fuel for cooking Nominal Maternal baseline interview 1 = electricity, 2 = gas, 3 = paraffin Fuel for heating Nominal Maternal baseline interview 1 = electricity, 2 = gas, 3 = paraffin, 4 = firewood, 6 = other Fuel for light Nominal Maternal baseline interview 1 = electricity, 2 = paraffin Number of rooms in the house Numeric integer Maternal baseline interview Range 1-10 Number of people in the household Numeric integer Maternal baseline interview Range 1-20 189 Variable Type Source Response options Distance to clinic in minutes Numeric integer Maternal baseline interview Range 1-60 Household assets Binary Maternal baseline interview - answered Yes or No for household possession of each asset Radio, TV, Computer, Fridge, Home phone, Cell phone, Bicycle, Motorcycle, Car or truck, Donkey or horse, Sheep goats or cattle, Daytime caregiver Nominal Maternal baseline interview 1 = Mother, 2 = Mother's husband/partner, 3 = Child's father if not mother's partner, 4 = Grandmother, 5 = Other family, 6 = Other non-family Daycare attendance Binary Maternal baseline interview 1 = Yes, 2 = No Child support grant received Binary Maternal interview at 6 month visit 1 = Yes, 2 = No Infant variables Gender Binary Birth record Male; Female Gestational age at birth Numeric integer Birth record Number of completed weeks Birth weight in grams Numeric continuous Birth record Range 2000g-4500g Birth length in cm Numeric continuous Birth record Range 40cm-55cm Birth head circumference in cm Numeric continuous Birth record Range 30cm-50cm Immunizations up to date Binary Road to health book at each visit 1 = Yes, 2 = No Current TB contact Binary Caregiver interview at each visit 1 = Yes, 2 = No Currently on TB prophylaxis Binary Caregiver interview at each visit 1 = Yes, 2 = No Currently receiving TB treatment Binary Caregiver interview at each visit 1 = Yes, 2 = No Number of all-cause clinic visits Numeric integer Caregiver interview at each visit Range 0-5 Number of sick clinic visits Numeric integer Caregiver interview at each visit Range 0-5 Symptoms for sick clinic visit Nominal Caregiver interview at each visit 1 = cough, 2 = Fever, 3 = Diarrhoea, 4 = Poor feeding, 5 = Poor growth, 6 = Other Current feeding mode Nominal - derived from more detailed feeding categories Caregiver interview at each visit 1 = exclusive breastfeeding, 2 = partial breastfeeding, 3 = mixed breastfeeding Weight in kg Numeric continuous Measured at each visit Range 2.00kg-12.00kg Length in cm Numeric continuous Measured at each visit Range 40cm – 100cm 190 Variable Type Source Response options Head circumference in cm Numeric continuous Measured at each visit Range 30cm-60cm Haemoglobin in g/dl Numeric continuous Measured at each visit Range 5g/dl-20g/dl HIV-PCR (HEU infants only) Binary Laboratory result, 2 and 6 months Positive; Negative HIV-ELISA (HUU infants only) Binary Laboratory result 6 month visit Reactive; Non-reactive Hospitalization variables Number of hospitalizations Numeric integer Abstraction (PIET-R CRF) Range 0-5 Age at hospitalization Numeric continuous Abstraction (PIET-R CRF) Days, calculated from dates of admission and birth Length of stay Numeric integer Abstraction (PIET-R CRF) Days, calculated from dates of admission and discharge Infectious event type Nominal Abstraction (PIET-R CRF) 1 = Pneumonia, 2 = Tuberculosis - confirmed, 3 = Tuberculosis - probable, 4 = Bronchiolitis, 5 = Acute diarrhoea, 6 = Persistent diarrhoea, 7 = Dysentery, 8 = Non-specific viral exanthem, 9 = Invasive bacterial infection - Neonatal, 10 = Invasive bacterial infection - Post-neonatal Infectious event severity grade (PIET-R) Binary Abstraction (PIET-R CRF) Mild-moderate; Severe; Very severe Infectious event severity grade (DAIDS) Ordinal Abstraction (PIET-R CRF) 1,2,3,4 Non-infectious event type Nominal Abstraction (PIET-R CRF) 1 = Neonatal Jaundice, 2 = Acute malnutrition, 3 = Chronic malnutrition, 4 = Chronic non-infectious condition , 5 = Other, 191 Appendix D Subgroup of all HEU infants – additional tables Table D.1 Maternal characteristics compared by HEU infant outcome group (Primary outcome is at least 1 infectious cause hospitalization or death before 195 days old; numeric variables as median(IQR), categorical variables as number (%)) Total N=94 Infants with the outcome N=17 Infants without the outcome N=77 P-value Age in years 27.8(23.8,31.1) 27.9(23.2,32.0) 27.8(24.2,30.7) 0.54 Primiparous 16(17.0) 2(11.8) 14(18.2) 0.73 Gestation at 1st antenatal visit (weeks) 20(16,26) 20(16,26) 20(16,26) 0.96 Number of antenatal visits 5(4,6) 5(4,6) 5(4,6) 0.49 Intention to exclusively breastfeed 37(39.4) 8(47.1) 29(37.7) 0.66 BMI postnatal kg/m2 26.0(23.1,28.7) 25.8(23.2,29.8) 26.3(23.1,28.6) 0.91 HIV diagnosed prior to pregnancy 48(51.1) 10(58.8) 38(49.4) 0.23 Pregnancy antiretroviral regimen: 0.15 cART for maternal indication 47(50.0) 8(47.1) 39(50.7) Zidovudine prophylaxis 44(46.8) 7(41.2) 37(48.1) None 3(3.2) 2(11.8) 1(1.3) Timing of cART initiation: 1.00 Pre-pregnancy 20(42.6) 4(50.0) 16(41.0) First trimester 3(6.4) 0(0.0) 3(7.7) Second trimester 13(27.7) 2(25.0) 11(28.2) Third trimester 11(23.4) 2(25.0) 9(23.1) Pregnancy cART regimen: 0.77 TDF/3TC/EFV 24(51.1) 4(50.0) 20(51.3) First line non-TDF/3TC/EFV 20(32.6) 4(50.0) 16(41.0) Second line 3(6.4) 0(0.0) 3(7.7) Timing of zidovudine initiation: 0.21 First trimester 3(6.8) 1(14.3) 2(5.4) Second trimester 27(61.4) 2(28.6) 25(67.6) Third trimester 14(31.8) 4(57.1) 10(27.0) 3TC – lamivudine; BMI – body mass index; cART – combination antiretroviral therapy; EFV – efavirenz; TDF – tenofovir disoproxil fumarate; WHO – World Health Organisation 192 Table D.2 Maternal CD4 and HIV viral load compared by HEU infant outcome group (Primary outcome is at least 1 infectious cause hospitalization or death before 195 days old; numeric variables as median(IQR), categorical variables as number (%)) Total N=94 Infants with the outcome N=17 Infants without the outcome N=77 P-value CD4 count Antenatal absolute count* 420(284,539) 458(384,640) 410(284,515) 0.19 Antenatal CD4 count category*: 0.40 <350 33(36.3) 4(25.0) 29(38.7) >350 58(63.7) 12(75.0) 46(61.3) Delivery absolute count 343(236,501) 308(202,694) 346(246,486) 0.87 Delivery percent 26.1(21.1,32.4) 29.7(18.9,33.8) 25.5(21.4,30.8) 0.35 Delivery CD4 count category: 0.60 <350 49(52.1) 10(58.8) 39(50.7) >350 45(47.9) 7(41.2) 38(49.4) Delivery HIV viral load# Category: 0.67 <40 31(36.9) 5(35.7) 26(37.1) 40-999 25(29.8) 6(42.9) 19(27.1) 1000-9999 14(16.7) 2(14.3) 12(17.1) >10000 14(16.7) 1(7.1) 13(18.6) Log10 HIV viral load 2.1(1.6,3.2) 1.87(1.59,2.9) 2.1(1.6,3.3) 0.42 *3 missing antenatal CD4 counts; 10 missing HIV viral loads, 3 in outcome group and 7 in no outcome group 193 Table D.3 Infant characteristics compared by HEU infant outcome group (Primary outcome is at least 1 infectious cause hospitalization or death before 195 days old; numeric variables as median(IQR), categorical variables as number (%), unless otherwise stated) Total N=94 Infants with the outcome N=17 Infants without the outcome N=77 P-value Male 46(48.9) 8(47.1) 38(49.4) 1.00 Gestational age – mean(SD) 38.7(1.5) 38.41(1.4) 38.78(1.5) 0.35 Birth weight in grams – mean(SD) 3118(375) 3084(371) 3125(377) 0.68 Low birth weight 6(6.4) 1(5.9) 5(6.5) 1.00 Immunizations not up to date at 6 months 5(6.9) 1(5.9) 4(5.2) 0.48 Total all-cause clinic visits 6(5,7) 6(5,7) 6(5,7) 0.50 Total sick clinic visits 1(0,1) 2(1,3) 1(0,1) <0.001 Receiving CPT at 8 weeks 51(54.3) 7(41.2) 44(57.1) 0.35 Excellent adherence to CPT at 8 weeks 49(96.1) 7(100) 42(95.5) 1.00 CPT – cotrimoxazole preventive therapy 194 Table D.4 Infant feeding characteristics compared by HEU infant outcome group (Primary outcome is at least 1 infectious cause hospitalization or death before 195 days old; all variables as number (%)) Total N=94 Infants with the outcome N=17 Infants without the outcome N=77 P-value Feeding mode at 2 weeks: (N=94) 1.00 Exclusive breastfeeding 32(34.0) 6(35.3) 26(33.8) Partial breastfeeding 3(3.2) 0(0.0) 3(3.9) No breastfeeding 59(62.8) 11(64.7) 48(62.3) Feeding mode at 2 months: (N=92) 0.32 Exclusive breastfeeding 23(25.0) 2(12.5) 21(27.6) Partial breastfeeding 4(4.4) 0(0.0) 4(5.3) No breastfeeding 65(70.7) 14(87.5) 51(67.1) Feeding mode at 4 months: (N=84) 0.45 Exclusive breastfeeding 11(13.1) 1(6.7) 10(14.5) Partial breastfeeding 7(8.3) 0(0.0) 7(10.1) No breastfeeding 66(78.6) 14(93.3) 52(75.4) Feeding mode at 6 months: (N=73) 1.00 Exclusive breastfeeding 4(5.5) 0(0.0) 4(6.7) Partial breastfeeding 7(9.6) 1(7.7) 6(10.0) No breastfeeding 62(84.9) 12(92.3) 50(83.3) 195 Table D.5 Infant haemoglobin and anaemia compared by HEU infant outcome group (Primary outcome is at least 1 infectious cause hospitalization or death before 195 days old; numeric variables as mean(SD), categorical variables as number (%)) Total N=94 Infants with the outcome N=17 Infants without the outcome N=77 P-value Mean Haemoglobin (SD) 2 weeks (N=87) 14.2(2.0) 14.3(2.1) 14.2(2.00) 0.88 2 months (N=80) 10.5(0.9) 10.4(0.9) 10.5(0.9) 0.92 4 months (N=79) 11.1(1.0) 10.6(0.8) 11.2(1.0) 0.04 6 months (N=71) 11.2(0.9) 10.9(0.9) 11.3(0.9) 0.20 DAIDS anaemia grade 2 weeks: (N=87) 0.14 Normal 71(81.6) 10(71.4) 61(83.6) Grade 1 8(9.2) 2(14.3) 6(8.2) Grade 2 8(9.2) 2(14.3) 6(8.2) 2 months: (N=80) 0.78 Normal 45(56.3) 8(61.5) 37(55.2) Grade 1 24(30.0) 4(30.8) 20(29.9) Grade 2 10(12.5) 1(7.7) 9(13.4) Grade 3 1(1.25) 0(0.0) 1(1.5) 4 months: (N=79) 0.37 Normal 47(59.5) 6(42.9) 41(63.1) Grade 1 23(29.1) 6(42.9) 17(26.2) Grade 2 7(8.9) 1(7.1) 6(9.2) Grade 3 2(2.5) 1(7.1) 1(1.5) 6 months: (N=71) 0.12 Normal 44(62.0) 5(41.7) 39(65.0) Grade 1 22(31.0) 6(50.0) 16(26.7) Grade 2 4(5.6) 0(0.0) 5(8.3) Grade 3 1(1.4) 1(8.3) 0(0.0) DAIDS – Division of AIDS; SD – standard deviation 196 Table D.6 Infant anthropometric measurements compared by HEU infant outcome group (Primary outcome is at least 1 infectious cause hospitalization or death before 195 days old; numeric variables as mean(SD), categorical variables as number (%)) Total N=94 Infants with the outcome N=17 Infants without the outcome N=77 P-value Birth: (N=176) WAZ -0.40(0.83) -0.47(0.81) -0.38(0.83) 0.70 LAZ -0.57(1.87) -0.80(2.31) -0.52(1.77) 0.64 WLZ -0.24(1.72) -0.34(1.70) -0.21(1.73) 0.79 HCZ -0.16(1.24) -0.04(1.20) -0.19(1.26) 0.65 Stunted 17(18.1) 3(17.7) 14(18.2) 1.00 Wasted 10(11.4) 2(11.8) 8(10.4) 2 weeks: (N=176) WAZ -0.40(0.81) -0.56(0.92) -0.36(0.79) 0.42 LAZ -1.19(0.93) -1.19(0.89) -1.20(0.94) 0.96 WLZ 0.65(0.98) 0.36(1.09) 0.72(0.95) 0.23 HCZ 0.10(1.08) 0.08(0.66) 0.10(1.15) 0.93 Stunted 15(16.0) 3(17.7) 12(15.6) 1.00 Wasted 0(0.0) 0(0.0) 0(0.0) NA 2 months: (N=162) WAZ -0.34(1.03) -0.63(1.42) -0.28(0.94) 0.39 LAZ -1.19(1.04) -1.21(.38) -1.18(0.97) 0.93 WLZ 1.10(1.12) 0.68(1.05) 1.18(1.12) 0.12 HCZ 0.28(1.15) 0.11(1.21) 0.31(1.15) 0.23 Stunted 18(20.7) 2(14.3) 16(21.9) 0.72 Wasted 0(0.0) 0(0.0) 0(0.0) NA 4 months: (N=146) WAZ -0.09(1.14) -0.31(1.45) -0.04(1.07) 0.51 LAZ -0.91(1.40) -1.11(1.21) -0.86(1.13) 0.49 WLZ 0.85(1.26) 0.77(1.32) 0.86(1.26) 0.82 HCZ 0.57(1.06) 0.67(1.11) 0.55(1.05) 0.70 MUACZ 0.28(1.08) -0.14(1.41) 0.37(0.98) 0.20 Stunted 14(17.3) 4(26.7) 10(15.2) 0.28 Wasted 0(0.0) 0(0.0) 0(0.0) NA 6 months: (N=132) WAZ 0.09(1.17) -0.11(1.32) 0.13(1.13) 0.54 LAZ -0.71(0.93) -0.88(0.91) -0.68(0.94) 0.49 WLZ 0.76(1.30) 0.64(1.35) 0.79(1.30) 0.71 HCZ 0.78(1.18) 0.89(1.29) 0.75(1.16) 0.73 MUACZ 0.38(1.09) -0.20(1.50) 0.51(0.96) 0.12 Stunted 8(11.0) 3(23.1) 5(8.3) 0.29 Wasted 1(1.4) 0(0.0) 1(1.7) 1.00 HCZ – head-circumference Z-score; LAZ – length-for-age Z-score; MUACZ – mid-upper-arm-circumference Z-score; WAZ – weight-for-age Z-score; WLZ – weight-for-length Z-score 197 Table D.7 Infant haemoglobin and anaemia compared by in-utero ZDV-exposure (Numeric variables as mean(SD), categorical variables as number (%)) Total N=94 ZDV-exposed N=48 ZDV-unexposed N=46 P-value Mean haemoglobin (SD) 2 weeks (N=87) 14.2(2.0) 14.2(1.8) 14.3(2.2) 0.78 2 months (N=80) 10.5(0.9) 10.3(0.9) 10.6(1.0) 0.17 4 months (N=79) 11.1(1.0) 10.9(1.0) 11.2(0.9) 0.24 6 months (N=71) 11.2(0.9) 11.3(1.0) 11.2(0.9) 0.54 DAIDS anaemia grade 2 weeks: (N=87) 0.04 Normal 71(81.6) 37(80.4) 34(82.9) Grade 1 8(9.2) 7(15.2) 1(2.4) Grade 2 8(9.2) 2(4.3) 6(14.6) 2 months: (N=80) 0.03 Normal 45(56.3) 18(46.2) 27(65.9) Grade 1 24(30.0) 17(43.6) 7(17.1) Grade 2 10(12.5) 3(7.7) 7(17.1) Grade 3 1(1.3) 1(2.6) 0(0.0) 4 months: (N=79) 0.40 Normal 47(59.5) 22(53.7) 25(65.8) Grade 1 23(29.1) 14(34.1) 9(23.7) Grade 2 7(8.9) 3(7.3) 4(10.5) Grade 3 2(2.5) 2(4.9) 0(0.0) 6 months: (N=71) 0.88 Normal 44(62.0) 23(65.7) 21(58.3) Grade 1 22(31.0) 10(28.6) 12(33.3) Grade 2 4(5.6) 2(5.7) 2(5.6) Grade 3 1(1.4) 0(0.0) 1(2.8) DAIDS – Division of AIDS; SD – standard deviation; ZDV - zidovudine 198 Table D.8 Infant anthropometric measurements compared by in-utero TDF-exposure (All variables as mean (SD)) Total N=94 TDF-exposed N=39 TDF-unexposed N=55 P-value Birth: (N=176) WAZ -0.40(0.83) -0.57(0.89) -0.28(0.76) 0.12 LAZ -0.57(1.87) -0.85(1.72) -0.38(1.96) 0.22 WLZ -0.24(1.72) -0.19(1.38) -0.26(1.92) 0.85 HCZ -0.16(1.24) -0.24(1.34) -0.09(1.18) 0.59 2 weeks: (N=176) WAZ -0.40(0.81) -0.59(0.92) -0.26(0.70) 0.06 LAZ -1.19(0.93) -1.47(0.85) -1.00(0.94) 0.01 WLZ 0.65(0.98) 0.68(1.01) 0.63(0.97) 0.81 HCZ 0.10(1.08) -0.03(1.18) 0.19(1.00) 0.36 2 months: (N=162) WAZ -0.34(1.03) -0.61(1.21) -0.13(0.82) 0.04 LAZ -1.19(1.04) -1.49(1.00) -0.95(1.06) 0.02 WLZ 1.10(1.12) 1.11(1.05) 1.10(1.17) 0.96 HCZ 0.28(1.15) 0.15(1.28) 0.37(1.05) 0.40 4 months: (N=146) WAZ -0.09(1.14) -0.44(1.21) 0.15(1.04) 0.03 LAZ -0.91(1.4) -1.31(0.98) -0.64(1.18) 0.006 WLZ 0.85(1.26) 0.78(1.18) 0.89(1.33) 0.69 HCZ 0.57(1.06) 0.25(1.12) 0.79(0.96) 0.03 MUACZ 0.28(1.08) -0.02(1.02) 0.48(1.08) 0.04 6 months: (N=132) WAZ 0.09(1.17) -0.19(1.17) 0.31(1.13) 0.07 LAZ -0.71(0.93) -0.92(0.82) 0.55(0.99) 0.08 WLZ 0.76(1.30) 0.56(1.33) 0.93(1.28) 0.24 HCZ 0.78(1.18) 0.64(1.19) 0.89(1.17) 0.36 MUACZ 0.38(1.09) 0.14(1.20) 0.58(0.98) 0.10 HCZ – head-circumference Z-score; LAZ – length-for-age Z-score; MUACZ – mid-upper-arm-circumference Z-score; TDF – tenofovir disoproxil fumarate; WAZ – weight-for-age Z-score; WLZ – weight-for-length Z-score 199 Appendix E Population effect of HIV exposure South Africa has an infant population of approximately one million, 300 000 of these being HEU infants (8,62). Considering the proportion of infants hospitalized for an infectious cause in this study (18.1% of HEU and 12.2% of HUU), it is estimated that approximately 139 700 HIV-uninfected infants are hospitalized in the first six months of life. Using the unadjusted risk ratio of 1.48 (95% CI 0.72, 3.06) for an infectious cause hospitalization or death in HEU relative to HUU infants, the attributable fraction in HEU infants was 32.4% (Table E.1). Almost one third of infectious cause hospitalizations in HEU infants could be as a result of being born to an HIV-infected mother, irrespective of the mechanism through which HIV exposure increases risk for infectious morbidity. The population attributable fraction at a risk ratio of 1.48 and HIV exposure prevalence of 30% would be 12.6%. Approximately one eighth of all infectious cause hospitalizations in South African HIV uninfected infants under six months of age could be as a result of consequences associated with being born to an HIV-infected mother in the absence of infant HIV-infection. In absolute numbers, of the estimated 139 700 HIV-uninfected infants with an infectious cause hospitalization in the first six months of life, roughly 17 600 infants could be hospitalized as a consequence specifically of being born to an HIV-infected mother, after removing the baseline risk for infectious cause hospitalizations in all HIV-uninfected infants. Alternative risk ratios and their corresponding attributable fractions and population attributable fractions are shown in Table E.1. Should the true RiR be as low as 1.1 the attributable fraction and population attributable fraction would be 9.1% and 2.9% respectively. Should the true RiR be closer to 3, the upper bound of the 95% CI, the attributable fraction and population attributable fraction could be as high as 66.7% and 38.2% respectively. Table E.1 Attributable fraction and population attributable fraction for estimates of the risk ratio for infectious cause hospitalization or death in HEU relative to HUU infants in South Africa (Observed risk ratio in this study indicated in bold) Risk ratio Attributable fraction (%) Population attributable fraction (%) Excess hospitalizations infants 0-6 months Annual excess hospitalizations 1.10 9.1 2.9 4 050 8 100 1.30 23.1 8.3 11 600 23 200 1.48 32.4 12.6 17 600 35 200 3.00 66.7 38.2 53 300 106 600 "@en ; edm:hasType "Thesis/Dissertation"@en ; vivo:dateIssued "2016-02"@en ; edm:isShownAt "10.14288/1.0220504"@en ; dcterms:language "eng"@en ; ns0:degreeDiscipline "Population and Public Health"@en ; edm:provider "Vancouver : University of British Columbia Library"@en ; dcterms:publisher "University of British Columbia"@en ; dcterms:rights "Attribution-NonCommercial-NoDerivs 2.5 Canada"@* ; ns0:rightsURI "http://creativecommons.org/licenses/by-nc-nd/2.5/ca/"@* ; ns0:scholarLevel "Graduate"@en ; dcterms:title "The pattern and pathways of infectious morbidity in South African HIV exposed uninfected infants"@en ; dcterms:type "Text"@en ; ns0:identifierURI "http://hdl.handle.net/2429/55603"@en .