Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Fetal, neonatal and maternal sequelae of birth weight and sex discordance among twin gestations Jahanfar, Shayesteh 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2016_september_jahanfar_shayesteh.pdf [ 3.44MB ]
Metadata
JSON: 24-1.0300362.json
JSON-LD: 24-1.0300362-ld.json
RDF/XML (Pretty): 24-1.0300362-rdf.xml
RDF/JSON: 24-1.0300362-rdf.json
Turtle: 24-1.0300362-turtle.txt
N-Triples: 24-1.0300362-rdf-ntriples.txt
Original Record: 24-1.0300362-source.json
Full Text
24-1.0300362-fulltext.txt
Citation
24-1.0300362.ris

Full Text

Fetal, neonatal and maternal sequelae of birth weight and sex discordance among twin gestations by Shayesteh Jahanfar Ph.D., University of New South Wales, 1995 M.Sc., University of Putra Malaysia, 2009  A THESIS SUBMITTED IN PARTIAL FULFILMENT OF  THE REQUIREMENTS FOR THE DEGREE OF   DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Health Care and Epidemiology) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) April 2016  © Shayesteh Jahanfar, 2016 ii  Abstract Birth weight discordance and sex discordance are two major predictors in identifying adverse fetal, neonatal and maternal outcomes in twin gestation. No study to date has comprehensively studied the role of these predictors together and in relation to placenta and cord. We analysed data from a large population-based sample of 10 years twin deliveries born in British Columbia and compared the result with a hospital-based sub-sample, taking chorionicity information into account. The unique aspect of our study is the use of a large population-based sample, generalized equation modeling, a wide range of confounding variables and analysing chorionicity and pathological aspects of placenta and cord in relation to growth and sex discordance. Aberrant growth among twins was related to unequal placenta sharing, existence of anastomosis between superficial vessels on the placental surface, length of umbilical cord and type of cord insertion. Fetal sex is an important predictor in placental findings including anastomosis, unequal placenta sharing, placental lesions and placental inflammation. Receiver operating characteristic analysis revealed that a threshold level of ≥30% had the optimal accuracy to detect perinatal mortality irrespective of chorionicity. Perinatal mortality and morbidity were also associated with growth and sex discordance. Early and late neonatal mortality were more likely in male infants from male-female twin pairs compared with females from female-female pairs. The predictive basis of growth discordance on stillbirth was dependent on fetal sex discordance, fetal growth, parity, gestational age and sizes of the twins.  Higher odds of adverse maternal outcomes were found for mothers carrying discordant growth twins compared to the reference category for the following conditions: preeclampsia, pregnancy iii  induced hypertension, preterm labor, premature rupture of membrane, prolonged preterm rupture of membrane, length of stay >3 days and cesarean section. Sex pairing is associated with postpartum length of stay >3 days, proteinuria, pregnancy induced hypertension, preeclampsia and cesarean section.   Given that chorionicity information is not available in most datasets, it is reasonable to use sex discordance as a proxy measure.      Twin pregnancies can benefit from reliance of clinicians on these data showing the importance of growth and sex discordance in prediction of adverse pregnancy outcomes.     iv  Preface This dissertation is based on secondary data collected by Perinatal Services British Columbia (PSBC) and pathology data from Children's & Women's Health Centre of British Columbia (C&W hospital). The work presented in this dissertation was conceived, analyzed and written by Shayesteh Jahanfar (S.J.). S.J. conceptualised the original idea behind this research project, performed the data analysis of the data and wrote the chapters. Study objectives/hypothesis, methodology, analysis and written chapters all benefited from the guidance and supervision of Dr. Eugenia Oviedo Joekes (E.O.J), Dr. Kenneth Lim (K.L.), and other committee members Dr. Martin T. Schechter (M.T.S.), and Dr. Patricia M. Spittal (P.M.S.).  This research was approved by the University of British Columbia Research Ethics Board (REB certificate number H11-03281), the Provincial Health Services Authority and PSBC (PSBC certificate number R2011008) and the Research Ethics Committee at C&W hospital.  List of publications stemmed from this work: 1. Jahanfar S., Lim, K. Oviedo-Joekes E. (2016) Optimal threshold for birth weight discordance, Journal of Perinatology, In press. This work is presented in Chapter 7. 2. Jahanfar S., Lim, K (2016), A systematic review of birth weight discordance, Fetal and Maternal Medicine Review 2016; 00:0 1–31. This work is presented in Chapter 4. 3. Jahanfar S., Lim K. Is birth weight discordance a good predictor of adverse perinatal outcomes? (2016) Fred Bryans Faculty Research Forum, Children and Women Hospital, Grandville Island, Vancouver, Canada. This work is partly presented in Chapter 7. 4. Jahanfar S., Lim K. A Canadian twin dataset (2015). International twin Registry Meeting, Osaka, Japan. This work is partly presented in Chapter 2. v  5. Jahanfar S. Ho J. Jaafar, S. H. Abraha I., Nisenblat V., Ellis U. Ultrasound for diagnosis of birth weight discordance in twin pregnancies. Cochrane systematic review. Review number 0880. A part of this work is presented in Chapter 1 and 13.  6. Jahanfar S., Lim, K. Oviedo-Joekes E. (2016) Stillbirth associated with birth weight discordance in twin gestations, two retrospective cohort studies. Canadian Society for Epidemiology and Biostatistics, 2016 National Student Conference, Manitoba. This work is presented in Chapter 8.  Contribution:  Jahanfar S.: Conceived the idea, analyzed the data, and wrote the manuscripts Lim K.: Reviewed articles and made clinical comments Oviedo-Joekes E.: Reviewed articles and made methodological comments.   vi  Table of Contents  Abstract .......................................................................................................................................... ii Preface ........................................................................................................................................... iv Table of Contents ......................................................................................................................... vi List of Tables ................................................................................................................................ xi List of Figures .............................................................................................................................. xv List of Abbreviations ................................................................................................................ xvii Acknowledgments .................................................................................................................... xviii Dedication .................................................................................................................................... xx CHAPTER 1 BACKGROUND, RATIONALE, AND OBJECTIVES ................................. 1 1.1 INTRODUCTION ......................................................................................................... 1 1.2 BACKGROUND ........................................................................................................... 2 1.2.1 BIRTH WEIGHT DISCORDANCE ....................................................................... 2 1.2.2 SEX DISCORDANCE ............................................................................................ 3 1.3 RATIONALE ................................................................................................................. 5 1.4 OBJECTIVES ................................................................................................................ 6 1.5 RESEARCH QUESTIONS ............................................................................................ 7 1.6 CONCEPTUAL FRAMEWORK .................................................................................. 9 1.7 OVERVIEW OF DISSERTATION ............................................................................... 9 CHAPTER 2 LITERATURE REVIEW................................................................................ 11 2.1 BIRTH WEIGHT DISCORDANCE............................................................................ 11 2.1.1 ADVERSE OUTCOMES ...................................................................................... 18 2.1.1.1 Adverse perinatal outcomes ............................................................................... 18 2.1.1.2 Adverse maternal outcomes ............................................................................... 37 vii  2.2 SEX DISCORDANCE ................................................................................................. 40 2.2.1 ADVERSE OUTCOMES ...................................................................................... 47 2.2.1.1 Adverse perinatal outcomes ............................................................................... 47 2.2.1.2 Adverse maternal outcomes ............................................................................... 51 2.3 STUDY SIGNIFICANCE ............................................................................................ 53 CHAPTER 3 METHODS ....................................................................................................... 55 3.1 OVERVIEW OF RESEARCH .................................................................................... 55 3.2 ETHICAL APPROVAL .............................................................................................. 55 3.3 COHORT DEFINITION .............................................................................................. 55 3.4 DATA SOURCES ........................................................................................................ 56 3.5 DEFINITIONS ............................................................................................................. 65 3.6 ANALYSIS PLAN ...................................................................................................... 72 CHAPTER 4 A SYSTEMATIC REVIEW OF BWD AND PERINATAL OUTCOMES 79 4.1 INTRODUCTION ....................................................................................................... 79 4.2 OBJECTIVES .............................................................................................................. 80 4.3 METHODS .................................................................................................................. 80 4.4 RESULTS .................................................................................................................... 83 4.5 DISCUSSION .............................................................................................................. 94 4.6 SUMMARY ............................................................................................................... 107 CHAPTER 5 BWD AND PLACENTAL PATHOLOGY .................................................. 108 5.1 INTRODUCTION ..................................................................................................... 108 5.2 OBJECTIVE .............................................................................................................. 108 5.3 METHODS ................................................................................................................ 108 5.4 RESULTS .................................................................................................................. 112 5.5 DISCUSSION ............................................................................................................ 121 viii  5.6 SUMMARY ............................................................................................................... 129 CHAPTER 6 FETAL SEX AND PLACENTAL PATHOLOGY ..................................... 131 6.1 INTRODUCTION ..................................................................................................... 131 6.2 OBJECTIVE .............................................................................................................. 131 6.3 METHODS ................................................................................................................ 131 6.4 RESULTS .................................................................................................................. 132 6.5 DISCUSSION ............................................................................................................ 141 6.6 SUMAMRY ............................................................................................................... 145 CHAPTER 7 OPTIMAL THRESHOLD FOR BIRTH WEIGHT DISCORDANCE, DOES KNOWLEDGE OF CHORIONICITY MATTER? .................................................. 147 7.1 INTRODUCTION ..................................................................................................... 147 7.2 OBJECTIVE .............................................................................................................. 148 7.3 METHODS ................................................................................................................ 148 7.4 RESULTS .................................................................................................................. 149 7.5 DISCUSSION ............................................................................................................ 162 7.6 SUMMARY ............................................................................................................... 167 CHAPTER 8 STILLBIRTH AND BIRTH WEIGHT DISCORDANCE......................... 168 8.1 INTRODUCTION ..................................................................................................... 168 8.2 OBJECTIVES ............................................................................................................ 169 8.3 METHODS ................................................................................................................ 169 8.4 RESULTS .................................................................................................................. 171 8.5 DISCUSSION ............................................................................................................ 180 8.6 SUMMARY ............................................................................................................... 187 CHAPTER 9 BWD AND ADVERSE PERINATAL OUTCOMES ................................. 188 9.1 INTRODUCTION ..................................................................................................... 188 ix  9.2 OBJECTIVE .............................................................................................................. 189 9.3 METHODS ................................................................................................................ 189 9.4 RESULTS .................................................................................................................. 191 9.5 DISCUSSION ............................................................................................................ 200 9.6 SUMMARY ............................................................................................................... 204 CHAPTER 10 FETAL SEX AND ADVERSE PERINATAL OUTCOMES ..................... 205 10.1 INTRODUCTION ..................................................................................................... 205 10.2 OBJECTIVE .............................................................................................................. 208 10.3 METHODS ................................................................................................................ 208 10.4 RESULTS .................................................................................................................. 209 10.5 DISCUSSION ............................................................................................................ 220 10.6 SUMMARY ............................................................................................................... 226 CHAPTER 11 ADVERSE MATERNAL OUTCOMES AND BWD .................................. 227 11.1 INTRODUCTION ..................................................................................................... 227 11.2 OBJECTIVE .............................................................................................................. 228 11.3 METHODS ................................................................................................................ 229 11.4 RESULTS .................................................................................................................. 231 11.5 DISCUSSION ............................................................................................................ 237 11.6 SUMMARY ............................................................................................................... 244 CHAPTER 12 FETAL SEX AND ADVERSE MATERNAL OUTCOMES ..................... 245 12.1 INTRODUCTION ..................................................................................................... 245 12.2 OBJECTIVE .............................................................................................................. 246 12.3 METHODS ................................................................................................................ 247 12.4 RESULTS .................................................................................................................. 248 12.5 DISCUSSION ............................................................................................................ 256 x  12.6 SUMMARY ............................................................................................................... 264 CHAPTER 13 RECOMMENDATIONS AND CONCLUSION ......................................... 265 13.1 SUMMARY OF STUDY FINDINGS ....................................................................... 265 13.2 STRENGTHS AND UNIQUE CONTRIBUTIONS ................................................. 268 13.3 RELEVANCE TO CLINICAL PRACTICE .............................................................. 270 13.4 LIMITATIONS .......................................................................................................... 272 13.5 CONCLUSIONS ........................................................................................................ 273 REFERENCES .......................................................................................................................... 276 Appendices ................................................................................................................................. 300   xi  List of Tables Table 2-1 Systematic review of papers for incidence of BWD .................................................... 14 Table 2-2 List of studies investigating clinical significance of BWD in twin pregnancy ............ 28 Table 3-1 Frequency of twin deliveries at BC hospitals, April 2000 to December 2010 ............. 57 Table 3-2 Pertinent data from three sources ................................................................................. 63 Table 3-3 Literature used for estimation of sample size ............................................................... 77 Table 4-1 Quality assessment of included articles by modified Dawn and Black tool ................ 85 Table 4-2 Description of the 40 articles reviewed ........................................................................ 86 Table 4-3 Risk ratio and 95%CI of perinatal outcomes in relation to BWD ................................ 88 Table 5-1 Comparing clinical characteristics of monochorionic and dichorionic twins in the cohorts born in C&W hospital (1493 pairs, n=2986) ................................................................. 112 Table 5-2 Comparing characteristics of placenta and cord by chorionicity for twins born in C&W hospital (1493 pairs, n=2986) ..................................................................................................... 114 Table 5-3 Comparisons of placenta and cord characteristics in groups with and without ≥20% BWD or ≥30% growth discordance for twin gestations registered at C&W hospital (1466 pairs, n=2932) ....................................................................................................................................... 116 Table 5-4 Unadjusted and adjusted odds of anastomosis and unequal placenta sharing in relation to BWD>20% and >30%. ........................................................................................................... 117 Table 5-5 Regression analysis of cord length and type of cord insertion, beta, significance of  each term in the model, and 95% confidence interval. ............................................................... 117 Table 6-1 Comparing clinical characteristics of monochorionic and dichorionic twins born in C&W hospital (1493 pairs, n=2986) ........................................................................................... 133 Table 6-2 Characteristics of placenta and cord, overall and stratified by sex for twins born in C&W hospital (1493 pairs, n=2986) ........................................................................................... 134 Table 6-3 Regression analyses of pathology findings, comparing male and female twins born in C&W hospital (1493 pairs, n=2986) ........................................................................................... 135 Table 6-4 Association between sex pairing and placental /cord pathology findings in twins born at C&W hospital (1493 pairs, n=2986) ....................................................................................... 137 Table 6-5 Logistic regression analysis of anastomosis in twins born at C&W hospital (1493 pairs, n=2986) ............................................................................................................................. 138 xii  Table 6-6 Linear regression analyses of placenta weight and sizes in twins born at C&W hospital (1493 pairs, n=2986) ................................................................................................................... 139 Table 7-1 Distribution of perinatal loss, area under curve and 95% CI of twins born at C&W hospital (1493 pairs, n=2986) ..................................................................................................... 150 Table 7-2  Unadjusted diagnostic indices of selected thresholds of BWD for overall stillbirth and according to chorionicity in twins born in C&W hospital (1493 pairs, n=2986) ....................... 151 Table 7-3 Unadjusted diagnostic indices of selected thresholds of BWD for overall perinatal mortality and according to chorionicity in twins born in C&W hospital (1493 pairs, n=2986) . 152 Table 7-4 Unadjusted diagnostic indices of selected thresholds of BWD for overall perinatal morbidity and according to chorionicity in twins born in C&W hospital (1493 pairs, n=2986) 153 Table 7-5 Unadjusted diagnostic indices of selected thresholds of BWD for stillbirth, perinatal mortality and morbidity for twins born in BC (6328 pairs, n=12656) ....................................... 158 Table 7-6 Comparing survival analysis for perinatal mortality in two cohorts with and without chorionicity adjustment ............................................................................................................... 161 Table 8-1 Distribution of birth characteristics according to the percentage BWD levels in twins born in BC (n=6407) ................................................................................................................... 173 Table 8-2 Risk of stillbirth associated with BWD, among twins and in respect to fetal growth among twins born in BC (6407 pairs, n=12814)......................................................................... 174 Table 8-3 Risk of stillbirth associated with BWD, among twins and subcategories of parity among twins born in BC (6407 pairs, n=12814)......................................................................... 175 Table 8-4 Risk of stillbirth associated with BWD, among twins and subcategories of sex concordance among twins born in BC (6407 pairs, n=12814) ................................................... 176 Table 8-5 Risk of stillbirth associated with BWD, among twins and subcategories of preterm birth among twins born in BC (6407 pairs, n=12814) ................................................................ 176 Table 8-6 Risk of stillbirth associated with BWD, among twins and subcategories of individual twin sizes among twins born in BC (6407 pairs, n=12814) ........................................................ 177 Table 8-7 Risk of stillbirth associated with BWD below or equal and above 30% for twins born in C&W Hospital (1501 pairs, n=3002) ...................................................................................... 178 Table 8-8 Unadjusted and adjusted odds (95%CI) for stillbirth and BWD≥30% in two cohorts with and without chorionicity information ................................................................................. 179 Table 9-1  Characteristics of mothers who delivered in BC according to BWD (n=6328). ....... 192 xiii  Table 9-2 Overall adverse perinatal outcomes in relation to BWD in twins born in BC (6328 pairs, n=12656). .......................................................................................................................... 193 Table 9-3 Odds of perinatal mortality and morbidity associated with BWD in twins born in BC (6328 pairs, n=12656) ................................................................................................................. 195 Table 9-4 Odds (95%CI) of perinatal outcomes associated with BWD in respect to sex discordance in twins born in BC (6328 pairs, n=12656) ............................................................ 196 Table 9-5 Comparing perinatal outcomes in twins born in C&W hospital (1493 pairs, n=2986)..................................................................................................................................................... 197 Table 9-6 Unadjusted and adjusted OR (95%CI) of perinatal morbidity in twins born in C&W hospital (1493 pairs, n=2986) ..................................................................................................... 198 Table 9-7 Adjusted OR (95%CI) of perinatal outcomes in twins registered at BC (6328 pairs, n=12656) and C&W hospital (1493 pairs, n=2986) ................................................................... 200 Table 10-1 Maternal characteristics of twin gestations born in BC according to sex combinations (6321 pairs, n=12642) ................................................................................................................. 210 Table 10-2 Association between twin sex pairing and perinatal outcomes in twins born in BC (6328 pairs, n=12656) ................................................................................................................. 213 Table 10-3 Adjusted odds ratios and 95% confidence intervals of comparisons between females from female-female pairs with other sex-pairing categories from n twin gestations delivered in British Columbia, Canada from 2000-2010. ............................................................................... 215 Table 10-4 Adjusted odds ratios and 95% confidence intervals of comparisons between females and males from sex-discordant pairs born in BC (2167 pairs, n=4334) ..................................... 217 Table 10-5 Adjusted odds ratios and 95% confidence intervals of comparisons between females from female-female pairs with other sex-pairing categories from twin gestations born in C&W hospital (1493 pairs, n=2986) ..................................................................................................... 218 Table 10-6 P values for comparison between various combinations of sex discordance for twins born in C&W hospital (1493 pairs, n=2986) .............................................................................. 219 Table 11-1 Characteristics of mothers carrying growth-discordant and -concordant twins, delivered in BC (n=6328 twin gestations, 12656 twins) ............................................................ 232 Table 11-2 Comparing maternal outcomes for twin gestations registered in BC in relation to BWD≥30% (6328 gestations, n=12656 twins) ........................................................................... 233 xiv  Table 11-3 Multivariate analysis of fetal, neonatal and maternal outcomes of concordant (<30%) versus discordant twins with ≥30% BWD (6328 gestations, n=12656 twins) ........................... 234 Table 11-4 Multivariate analysis of maternal outcomes of concordant (<30%) versus discordant twins with ≥30% BWD among a subgroup of sex- discordant  twins (2167 gestations, n=4334)..................................................................................................................................................... 235 Table 11-5 Multivariate analysis of maternal outcomes of mothers with BWD (<30%) versus ≥30% BWD in mothers who delivered at C&W hospital (1493 gestations, n=2986) ................ 237 Table 11-6 Proportion of CD among twin gestations with and without BWD found by other studies. ........................................................................................................................................ 241 Table 12-1 Maternal characteristics of twin gestations according to sex combinations (6321 gestations) ................................................................................................................................... 249 Table 12-2 Comparing adverse maternal outcomes between male-male, female-female and male-female twin pairs using ANOVA and Post Hoc test ................................................................... 250 Table 12-3 Unadjusted and adjusted OR (95%CI) of maternal outcomes in relation to sex discordance in mothers delivered in BC (6321 gestations, n=12642) ........................................ 253 Table 12-4 Unadjusted and adjusted OR (95%CI) of maternal outcomes in relation to sex discordance in mothers delivered at C&W hospital (1493 gestations) ....................................... 255  xv  List of Figures Figure 1-1 Framework of the present study .................................................................................... 9 Figure 2-1 Inhibitory effect of 17β-HSD leads to respiratory distress ......................................... 43 Figure 4-1 Flow chart of included and excluded studies for a systematic review of BWD and perinatal outcomes ........................................................................................................................ 84 Figure 4-2  Meta-analysis for association between BWD and stillbirth ....................................... 89 Figure 4-3 Meta-analysis for association between BWD and early neonatal mortality ............... 90 Figure 4-4 Meta-analysis for association between BWD and late neonatal mortality ................. 90 Figure 4-5 Meta-analysis for association between BWD and congenital anomalies .................... 91 Figure 4-6 Meta-analysis for association between BWD and the fifth minute Apgar score <7 ... 91 Figure 4-7 Meta-analysis for association between BWD and NICU stay >2 days ....................... 92 Figure 4-8 Meta-analysis for association between BWD and newborn septicemia ..................... 92 Figure 4-9 Meta-analysis for association between BWD and newborn hypoglycaemia .............. 93 Figure 4-10 Meta-analysis for association between BWD and newborn anemia ......................... 93 Figure 4-11 Meta-analysis for association between BWD and respiratory distress syndrome .... 94 Figure 5-1 A comparison of fetal to placenta weight ratio between lighter discordant twins (n=236) and combined group of heavier co-twins and twins with concordant birth weight (n=1214) ...................................................................................................................................... 119 Figure 5-2 Relative frequency of unequal placenta sharing in birth weight discordant twin pairs registered in C&W hospital (1466 pairs, n=2932) ...................................................................... 120 Figure 6-1 Fetal to placenta weight ratio in sex pairing categories in twins born at C&W hospital (1493 pairs, n=2986) ................................................................................................................... 140 Figure 7-1 Positive likelihood ratio for ≥30% BWD threshold level in overall, MC and DC twins born among twins born in C&W hospital ................................................................................... 155 Figure 7-2 Positive likelihood ratio for ≥35% BWD threshold level in overall, MC and DC twins born among twins born in C&W hospital ................................................................................... 155 Figure 7-3 Kaplan-Meier analyses of perinatal loss for MC and DC twins born in C&W hospital (1493 pairs, n=2986) ................................................................................................................... 156 Figure 7-4 Positive likelihood ratio for ≥30% and ≥35% BWD threshold level among twins born in BC ........................................................................................................................................... 159 xvi  Figure 7-5 Cumulative survival function of perinatal loss for twins born in BC after adjustment for gestational age and sex discordance (6328 pairs, n=12656) ................................................. 160 Figure 8-1 Frequency of stillbirth (%)  in BWD categories with four layers for twins born in C&W hospital (1501 pairs, n=3002; p=0.01) ............................................................................. 178 Figure 9-1  Perinatal mortality and morbidity outcomes stratified by chorionicity in twins born in C&W hospital (1493 pairs, n=2986) ........................................................................................... 199 Figure 10-1 Birth weight of infants according to sex pairing ..................................................... 211    xvii  List of Abbreviations    ACOG American College of Obstetricians and Gynecologists  AGA Appropriate for gestational age BWD Birth weight discordance C&W hospital Children's & Women's Health Centre of British Columbia CPAP Continuous positive airway pressure CRL Crown rump length  CS Cesarean section CI DC Confidence interval Dichorionic GEE General estimating equation IPPV_ETT Intermittent positive pressure ventilation IVH Intraventricular hemorrhage IUGR Intrauterine growth restriction  LGA large for gestational age MC Monochorionic NICU Neonatal Intensive Care Unit  PIH Pregnancy induced hypertension PPROM Prolonged premature rupture of membrane PROM Premature rupture of membrane PSBC Perinatal Services British Columbia RDS Respiratory distress syndrome  ROC Receiving operating characteristic RR SOGC Relative risk Society of Gynecologists and Obstetricians of Canada SGA Small for gestational age  TTTs Twin to twin transfusion syndrome VCI Velamentous cord insertion      xviii  Acknowledgments First, I would like to express my sincere appreciation to Perinatal Services BC, BC Children's & Women's Health Centre and Population Services BC for providing access to the two most comprehensive datasets on twins born in British Columbia.  I am indebted to the twins and their families whose contributions have made this work possible.  I would like to extend my sincere gratitude to my committee members for their distinctive contributions to this work.  Assistant Professor Eugenia Oviedo–Joekes:  Your guidance elevated my spirit when I needed it the most.  Your direction on the rigour of research methodology has given me added confidence in this work.  As a mother of beautiful twins, Guadalupe and Lola, you have given me immense motivation.  As my thesis supervisor, you always encouraged me to follow the research agenda.  I thank you for your timely and invaluable support.  Associate Professor Kenneth Lim:  Your commitment to improving patient care has inspired me to learn a great deal in the area of fetal medicine.  As clinical supervisor, your contributions have been critical and precious. Without your strong clinical perspectives, this work would have not been possible.  Thank you very much for providing me with countless insights and opportunities.   Professor Martin Schechter:  Your contribution in the statistical analysis of this work, your advice and professional support are much appreciated.  Professor Patricia Spittal:  Your guidance and direction have taught me to be a strong researcher.  As the chair in our committee meetings, you taught me lessons in strength and vision.  Thank you all for your generous support and warm and welcoming demeanour.  xix  This research was generously funded by the UBC Faculty of Medicine and Canadian Institute of Health Research (MAH-115445) under the supervision of Professor K.S. Joseph. Finally, I would like to thank my husband, Dr. Saied Saiedi, my two sons, Reza and Ed, my parents, Mohammad and Mehrangiz, and my brothers (Ali, Shadman and Niknam) who have been incredibly supportive throughout the past few years.  To my family and friends, words cannot express my gratitude for all the unconditional love and sacrifices you have made.  May God bless you all.     xx  Dedication  For the Canadian twins and families   1  CHAPTER 1 BACKGROUND, RATIONALE, AND OBJECTIVES 1.1 INTRODUCTION The rate of twinning is increasing in Canada. The rate of multiple births increased from 2.2% in 1995 to 3.2% in 2008.1 This may be explained by the increased use of assisted reproductive technology among the aging maternal demographic.2,3  In British Columbia (BC), the average age of women having their first child increased from 21.5 in 2000 to 25.0 in 2008.4 Twin pregnancies are known to be linked to fetal, newborn and maternal complications.  Fetal and newborn complications include prematurity, low birth weight, intrauterine growth restriction, and congenital anomalies.5 In turn, preterm birth is associated with respiratory diseases, necrotizing enterocolitis and neurological handicaps.6 Maternal morbidities associated with twin gestation have also been studied by Walker et al. In a retrospective cohort study they compared 165188 singletons and 44674 multiple pregnancies in Canada from 1984 to 2000, and reported an increased risk of maternal complications, including preeclampsia (relative risk [RR] 2.8, 95%CI 2.7-2.9), myocardial infarction (RR 3.7,  95%CI 2.7-2.9), heart failure (RR 12.9, 95%CI 2.6-62.3), pulmonary edema (RR 7.1, 95%CI 4.5-11.2), venous thromboembolism (RR 2.6, 95%CI 2.0-3.5), caesarean section (CS) (RR 2.2, 95%CI 2.1-2.2), hysterectomy (RR 2.3, 95% CI 1.7-3.2), and blood transfusion (RR 1.7, 95% CI 1.1-2.5).7 Risk factors for adverse perinatal and maternal outcomes are beginning to emerge among twin pregnancies. Earlier studies suggest that discordance in birth weight and sex are predictors of morbidity and mortality specific to twin gestations. The present dissertation investigates the risk factors for adverse fetal, neonatal, and maternal outcomes specific to twin gestations.  2  1.2 BACKGROUND  1.2.1 BIRTH WEIGHT DISCORDANCE  Birth weight discordance (BWD) occurs when there is a disparity in birth weight between the larger and smaller infants of a twin set. This well-documented phenomenon occurs in 10-29% of twin pregnancies.8–10   The different BWD rates reported by investigators can be partly explained by the confusion in the literature over the definition of BWD.11 There are three alternative definitions of BWD currently in use.   The first definition regards BWD as ‘absolute discordance’ where the conclusive difference between smaller and larger twins is calculated.  In this case, the same degree of discordance is assigned to a twin pair that weigh 3000/2500 grams and to another pair that weigh 1500/1000 grams.9   The second definition calculates BWD as a percentage of the larger twin, that is, birth weight difference divided by the weight of the larger twin.12   The third definition, which is statistically driven, expresses birth weight differences either in terms of 95th percentiles or as the standard deviations from the mean of the twin birth weights. This definition has a limitation because large population-based studies have revealed that BWD does not have a normal distribution. Therefore, means and standard deviations do not accurately describe departures from a defined norm.11  None of the aforementioned definitions refers to the actual size of the twins, rendering it difficult to relate discordance to the probability of survival. 3  The mechanism of BWD has been attributed to unequal placental sharing, abnormalities of umbilical cord insertion,13 placental pathology, 14 and twin-to-twin transfusion syndrome (TTTs).7 To date, there has been little agreement in the literature on the impact of BWD on perinatal outcomes.11 Studies, small15–19 or large,20 retrospective16,18,19 or prospective5, hospital-based15,19–22 or population-based23 have produced widely disputed rates of perinatal mortality and morbidity.   The most frequent reported outcome of BWD in the literature is perinatal death. Some studies reported increased fetal and neonatal death rates among BWD twins as compared to concordant twins. 15,31,33,35 However, studies controlling for gestational age have failed to find an association with perinatal death.15,19  Other adverse perinatal outcomes reported to be associated with BWD include respiratory distress syndrome, congenital anomalies, hyperbilirubinemia, hypoglycemia, anemia, and longer neonatal intensive care unit (NICU) admission.16,19,20     Little attention has been given to the association between BWD and maternal outcomes. No population-based study has addressed the association between maternal outcomes and BWD. Among the limited number of hospital-based studies that have been conducted to date, very few have reported an association between BWD and maternal complications.22 Two such studies reported an increased risk of PIH, CS, and premature rupture of membranes.5,16   1.2.2 SEX DISCORDANCE Sex discordance is also associated with adverse perinatal and maternal outcomes. Sex discordance refers to the physiological sex differences of fetuses in a twin pregnancy.  4  It is theorised that the sheer presence of a male fetus in the uterus along with a female has a profound effect on the perinatal morbidity and mortality of the female fetus.25 Studies have shown that male hormones have two intra-uterine effects:  1. Female fetuses grow faster if they are in utero with a male co-twin.26  2. Androgen is found to have an inhibitory effect on fetal lung development.  In theory, this could explain the increased risk of respiratory morbidity found among male-female twin pairs compared to female-female twin pairs.27 The impact of fetal sex on maternal outcomes has also been investigated. Higher rates of preeclampsia are reported in twin versus singleton pregnancies (2.5, 95%CI 1.8-3.4)28 where increased levels of male-fetal antigen lead to preeclampsia. The pathogenesis of preeclampsia is attributed to the inability of the maternal immune system to respond favourably to paternally-derived antigenic components of the male fetus.29 Hyperemesis, often stated to be the result of high levels of human chorionic gonadotropin, is also considered to be associated with twinning and presence of a female in twin pregnancies.30 It is important, therefore, to obtain more information on fetal sex in twin gestation. Studies that have assessed sex discordance as a risk factor for adverse perinatal outcomes remain indecisive. Male fetal death among male-male concordant twins was found to be higher than31 among (dichorionic) male-female twins or identical32 to male mortality among male-female twins (with unknown chorionicity). In contrast, higher fetal death rates have been observed for males within male-female twins compared with male-male pairs with33 or without32 controlling for gestational age. Reported fetal/neonatal morbidity outcomes related to sex discordance 5  include prematurity and its complications (respiratory distress syndrome, pneumothorax, bronchopulmonary dysplasia).31,32  The literature on adverse maternal outcomes in relation to sex discordance remains sparse. The only study available to date34 obtained data from the Danish National Birth Registry, 1980 to 1996 (n=36674).  Considering singleton healthy pregnancies with males as the reference group, twin pregnancies had a higher risk of hyperemesis: female-female twins (2.4, 95%CI 1.8-3.2), male-female twins (2.4, 95%CI 1.8-3.4) and male-male twins (1.8, 95%CI 1.3-2.5). The data suggest that the presence of at least one female in the twin pair increased the risk of hyperemesis. The same author also investigated an association between sex discordance and preeclampsia. The risk of preeclampsia was not significantly associated with sex discordance: male-male twins (2.5, 95%CI 2.1-2.9), female-female twin pairs (2.4, 95%CI 2.1-2.9), and male-female twin pairs (2.2, 95%CI 1.9-2.7).  In contrast to these findings, a population-based study28 comparing female-female versus male-male twins reported a significantly higher risk of preeclampsia among mothers carrying female-female twins (7.6% vs. 3.6%, p=0.0005).  1.3 RATIONALE  Studies on birth weight and sex discordance to date are limited by insufficient statistical power and a lack of control for confounding factors. All BWD-related studies to date are hospital-based except for one Canadian study by Wen et al.23 However, that study excluded data for Ontario, and reported inconsistent data quality among various provinces.35  Adverse maternal outcomes of BWD have never been studied in a population-based study.  6  Studies on sex discordance are few.  Moreover, they do not adequately address the association between fetal sex and perinatal adverse outcomes in the context of various possible sex discordance combinations. Shinwell32 considered several possible sex discordance combinations but focused on very low birth weight babies only. The few studies of adverse maternal outcomes (hyperemesis and preeclampsia) to date  present inconsistent results, are underpowered,34 and do not control for potential confounders.29 No study was found which comprehensively assesses potential confounding factors such as history of stillbirth, congenital anomalies, preterm birth, or low birth weight in the association between birth weight/sex discordance and perinatal outcomes. The available studies also lack assessments of mother-related confounding factors such as history of CS, maternal weight gain during the current pregnancy, number of antenatal care visits, and chronic hypertension. The study described here explores the association between birth weight/sex discordance and adverse fetal, neonatal and maternal outcomes of twin gestations born in British Columbia, Canada from 2000 to 2010.  Moreover, this dissertation aims at investigating the association between chorionicity and placental pathology findings and BWD/sex discordance. Furthermore, an optimal threshold beyond which adverse perinatal outcomes are to be expected is investigated.  1.4 OBJECTIVES 1. To conduct a systematic review in the area of BWD and perinatal outcomes 2. To identify the role of chorionicity and placental pathology information on BWD  3. To identify the role of chorionicity and placental pathology information on fetal sex discordance 7  4. To identify the optimal threshold level of growth discordance in respect to chorionicity and sex discordance  5. To compare the fetal, neonatal and maternal outcomes of twin pregnancies with and without BWD 6. To compare the fetal, neonatal and maternal outcomes of twin pregnancies with and without sex discordance 1.5  RESEARCH QUESTIONS 1. How does a systematic review of the literature and meta-analysis inform us about the association between BWD and perinatal outcome? 2. What is the association of chorionicity and placental pathology information with BWD? 3. What is the association of chorionicity and placental pathology information with sex discordance? 4. What is the optimal threshold for growth discordance beyond which adverse perinatal outcomes are to be expected? Does chorionicity information change the threshold for growth discordance? Is sex discordance a good proxy for chorionicity?  5. Does the risk of stillbirth differ between twins with BWD at the optimal threshold level as compared to BWD below the optimal threshold level?  6. Does the risk of neonatal mortality and morbidity (as described below) differ between twins with BWD at the optimal threshold level compared to BWD below the optimal threshold level?  a. Neonatal mortality: early and late neonatal death b. Neonatal morbidity: short term and long term neonatal morbidities inclusive but not restricted to Apgar score of 1st and 5th minutes less than 7, length of stay in NICU more than 2 days, newborn septicemia, hypoglycemia, anemia, 8  intraventricular hemorrhage (IVH), pneumothorax, respiratory distress syndrome, retinopathy, composite perinatal morbidity outcomes  7. Do the risks of perinatal adverse outcomes (stillbirth, early and late neonatal outcomes) differ between twins with and without sex discordance?  8. Do the risks of adverse maternal outcomes (preeclampsia, PIH, preterm labour, premature rupture of membrane, cesarean section, gestational diabetes, length of stay in hospital more than 3 days, prolonged rupture of membrane) differ between twin gestations with BWD at the optimal threshold level compared to BWD below the optimal threshold level?  9. Do the risks of maternal adverse outcomes (as stated above) differ between twin gestations with and without sex discordance?   The study utilizes the British Columbia Perinatal Database Registry, enabling us to examine potential confounders including the following:  The history of previous adverse perinatal outcomes including stillbirth, congenital anomalies, preterm birth, and low birth weight   Maternal factors including weight gain during the current pregnancy and number of antenatal care visits  Pre-existing maternal conditions including CS, chronic hypertension and diabetes  Additionally, the study controls for sex discordance when investigating the role of BWD and vice versa.   9  1.6 CONCEPTUAL FRAMEWORK The potential relationships between two exposures (BWD and sex discordance) and adverse fetal, neonatal and maternal outcomes are shown in Figure 1-1, in which the horizontal arrow implies that chorionicity may influence BWD and sex discordance.          Figure 1-1 Framework of the present study  1.7 OVERVIEW OF DISSERTATION This dissertation is comprised of thirteen chapters addressing the research background, study methodology, a systematic review and the five objectives listed above, as well as a conclusion. The objectives are consecutively addressed in Chapters 4 to 12. These chapters were written as manuscripts envisioned for publication in peer-reviewed journals. Thus, there is some repetition in terms of study methods in each chapter. A brief reminder of the methodology specific to each chapter’s analytical approach is provided to improve readability and reduce the need for referring to the methods section (Chapter 3) to identify the specific analytical strategy. We have analyzed Chorionicity  Birth weight discordance  Mortality and Morbidity in Twin Pregnancy Sex discordance  Fetal adverse outcomes Neonatal Adverse outcomes Maternal adverse outcomes 10  data from a population-based dataset named Perinatal Services British Columbia (PSBC), collected from twin births across BC from 2000-2010. This dataset did not have pathology and chorionicity information. A subgroup of data with chorionicity information was therefore analyzed separately. This dataset is inclusive of twins born in C&W hospital from the same study period (2000-2010). Each chapter (5 to 12) provides analytical results on the PSBC datasets first, followed by results from the C&W study as a subgroup analysis. This order is reversed for Chapter 7: there the data set with chorionicity information is analyzed first, as this chapter answers the question of whether or not chorionicity is needed for identifying an optimal threshold for growth discordance. Finally, Chapter 13 offers a conclusion based on the research findings reported in Chapter 5 to 12, discusses the study’s strengths and limitations, and makes recommendations for clinical guidelines and policy directions.    11  CHAPTER 2 LITERATURE REVIEW This chapter sets out to provide a thorough literature review of the impact of two exposures, BWD and sex discordance, on fetal, neonatal and maternal outcomes.  First, the definition for each exposure is reviewed. Second, the incidence of the exposure is investigated through a systematic review of published articles. Third, the mechanism of the impact of the exposures on the outcomes is described. Finally, the chapter identifies the impact of the exposures on the fetal, neonatal and maternal outcomes.  The exposures, i.e., BWD and sex discordance, are evaluated separately.  A brief description of the interaction between these two exposures is given in the conclusion. 2.1 BIRTH WEIGHT DISCORDANCE Definition  BWD is a disparity in birth weight between the larger and smaller infants of a twin set.  For a number of reasons, it is important to define BWD clearly. First, several studies have challenged the clinical significance of BWD.  They argue that the birth weight of each twin, rather than the weight difference of the pair, is the most significant determinant of adverse fetal, neonatal, and maternal outcomes.36 Second, a clear understanding of BWD provides for objective, rather than arbitrary, cut-off levels for distinguishing normal versus abnormal growth discordance. For example, the relationship between BWD and fetal death might be interpreted differently based on various discordant values (10%, 20% or 30%) of twin birth weights.37,38 Third, there are three alternative definitions of BWD currently in use, as introduced in section 1.2.  12  The present dissertation adopts the definition incorporated in the clinical guidelines for analysis as suggested by the American College of Obstetricians and Gynecologists (ACOG)39 and the Society of Obstetricians and Gynecologists of Canada (SOGC).40 The guidelines facilitate inference and help avoid bias. They both consider the larger twin as the standard of growth and calculate BWD according to the following formula: BWD =(Larger estimated or actual weight – smaller estimated or actual weight)Larger estimate or actual weight× 100 The guidelines use both estimated and actual weights. The estimated growth discordance is defined as the difference in crown rump length (CRL) between twin pairs divided by the CRL of the larger twin, if the measurements are made in the first trimester through ultrasound.  During the second and third trimesters of pregnancy, however, a combination of the abdominal circumference and femur length is used to estimate fetal weight by means of the above formula.41  Retrospective cohort designs employ the actual weight measurements after birth, providing for a more reliable estimate of BWD. Incidence BWD is a well-documented phenomenon.13,15–17,36,42–44  The incidence of clinically significant growth discordance has been reported to lie between 8.0% and 58.2%. This broad range is due to variables such as country of origin, whether the target is population-based or hospital based,5,19 sample size (small or large),15–17,19,45 enrolment period (short or long),20,46 study design (prospective, retrospective cohort or case-control),5,47–49 the definition adopted, and the calculation method used to estimate BWD.  13  BWD incidence, among other factors, depends upon the target population chosen for the study. Because sound generalizability requires data on larger populations, quantitative population-based research provides a better foundation to explain the adverse outcomes of BWD. The present review of BWD incidence found 12 population-based vs. 29 hospital-based studies.  Larger sample sizes allow a superior estimate of the incidence of BWD. The sample sizes reported in the studies varied between 42 to 297155 twin pairs, including ten population studies with a size greater than 10000. None of the studies presented power calculations.  A longer period of enrolment increases the accuracy of the incidence reported. The longest enrolment periods were 12 years, reported by Wen et al. in Canada from 1986 to 1997,23 and 26 years, reported by Fraser et al. from the USA (1960-86).19 Both studies were conducted on secondary data sources. Twenty-two reports, including all the population-based studies, were published prior to 2000. In light of the increasing incidence of twin pregnancies due to new developments in assisted reproductive technology, new sets of incidence studies are of greater importance. Table 2-1 contains a summary of 40 publications, between 1986 and 2015, with more than 1.1 million twins indicating the incidence of BWD.    14  Table 2-1 Systematic review of papers for incidence of BWD   Author, year Country Number of twin pregnancies Population based Enrolment period BWD definition %* Incidence of BWD % Maternal age Mean(SD) Cohort studies Alam, 2010  Brazil 302 No 1998-2004 ≥20 26.5 29.1(6.2) Amaru, 2004 USA 1318 No 1992-2001 ≥20 16 - Appleton, 2008 Portugal 230 No 1989-2002 ≥20 23 29.7(4.7) Audibert, 2003 France 346 No 1996-1999 ≥20 22 31.9(4.4) Blickstein, 2004 USA 10683 Yes 1995-1997 ≥25 8.2 29.2(6.1) Branum, 2003 USA 128168 Yes 1995-1997 ≥15 26.5 - Breathnach, 2011+ Ireland 977 No 2007-2009 ≥15 31 33& Canpolat, 2006 Turkey 266 No 2000-2004 ≥20 20 - Cheung, 1995 Canada 122 No 1989-1992 ≥10 58.2 27.8(4.5) Cohen, 2002 France 2121 No 1984-1998 ≥30 39 30.5& Cooperstock, 2000 USA 9931 Yes 1978-1990 ≥30# - - Demissie, 2002 USA 297155 Yes 1995-1997 ≥20# 8.1 - Fakeye, 1986 Nigeria 622 No 1982-1983 ≥20 9 - Fraser, 1994 USA 1145 Yes 1960-1986 ≥25 13 - Graf, 1998 USA 213 No - ≥20 - - Harper, 2012 USA 1,145 No - ≥20 8 - Hollier, 1999 USA 1,370 No 1988-1996 ≥15# 29 - Hsieh, 1994 Taiwan 279 No 1986-1991 ≥15 39 - Jakobovits, 1992 Hungary 329 No 1975-1992 ≥15 15.2 - Kalish, 2005 USA 176985 Yes 1995-1998 ≥25 8.1 - 15  Author, year Country Number of twin pregnancies Population based Enrolment period BWD definition %* Incidence of BWD % Maternal age Mean(SD) Kato, 2006 Japan 25955 Yes 1995-1999 ≥25 9 - Kilic, 2006 Turkey 136 No 2003-2005 ≥15# 39.7 27.2(5.6) Kontopoulos, 2005 USA 170223 Yes 1995-1998 ≥20# 17 - Lee, 2007 Korea 251 No 2002-2004 ≥20 15 33.2(0.43) Lewi, 2008+ Belgium 202 No 2002-2007 ≥25 14 30.4(4) Mahony, 2009 Ireland 1166 No 1997-2006 ≥20 24.1 - Nawab, 2008 USA 1597 No 2001-2004 ≥20 24.7 - Patterson, 1990 USA 194 No 1982-1985 ≥20 26.2 - Queiros, 2010 Portugal 934 No 1994 to 2008 ≥25 9.2 - Severinski, 2004 Croatia 351 No 1993-2001 ≥20 15.1 - Smiljan, 2004 Croatia 351 No 1993-2001 ≥20 15.1 29.3(4.9) Szymanski, 2009 Croatia 280 No 2003-2009 ≥20 19 - Talbot, 1997 USA 119 No 1988-1995 ≥20 35 - Tan, 2005 USA 147262 Yes 1995-1997 ≥25 8.6 - Vergani, 2004 USA 335 No 1990-2000 ≥20 - 32.1(4.7) Vicotria, 2001 USA 382 No 1993-1995 ≥25 11 28.9(4.7) Wen, 2005 Canada 59034 Yes 1986-1997 ≥20# 17 - Wen, 2006 USA 147262 Yes 1995-1997 ≥25 - - Yalcin, 1998 Turkey 384 Yes 1993-1995 ≥20 32 25.8(-) Matched-case control study Lopriore, 2012 Netherland 94 No 2002-2011 ≥25 - - Sanghi, 2012 India 42 No 2004-2008 ≥15 - - * BWD was calculated by using the formula (larger twin weight - smaller twin weight)/larger twin weight*100;  + Prospective cohort study;  & Median;  # range from 10, 10-15, 15-20, 20-30 and >30 or other range;  SD: Standard deviation 16  Prospective cohort studies are optimal designs for estimating the incidence of BWD; only two such study designs were found.47,50 Other investigations included 36 retrospective cohort and two matched case-control studies.48,49 BWD incidence is confounded by maternal age as advanced age is associated with higher rates of adverse obstetric outcomes and preterm delivery.51 Only 13 studies reported maternal age, with an average maternal age of 29.7 years.  The definition used in the literature, based on selected guidelines (ACOG, SOGC), ranged from ≥10% to ≥30%. Among the investigations contained in Table 2-1, the range of BWD was as follows: BWD ≥15% in seven studies, BWD ≥20% in 20 studies, BWD ≥25% in 10 studies, and BWD ≥30% in two studies.  Effect mechanism  Several mechanisms have been proposed for growth discordance of fetuses exposed to the same intrauterine environment.52   For monochorionic (MC) twin pregnancies, the mechanism is explained through conditions such as twin to twin transfusion syndrome (TTTs) and intrauterine growth restriction (IUGR).  In the case of dichorionic (DC) twin pregnancies, IUGR and placenta pathology can cause growth discordance. Abnormal cord insertion and insufficient placenta implantation in the uterus wall are other proposed mechanisms for BWD. These mechanisms are explained as follows. TTTs occurs in 5-38% of monochorionic twins.53,54  This condition, threatening both twins, is due to an unequal exchange of blood, nutrients, and oxygen between the twins through their shared placenta. The postulated mechanism behind TTTs is hypo-perfusion of one twin vs. 17  hyper-perfusion of the other.55  For one twin, the blood that leaves the fetus through vascular anastomoses is not replaced by an equal amount of blood coming back to the fetus, producing an anaemic, smaller twin known as the ‘donor’. For the other twin, the amount of blood received through the veins, which comes back from the placenta, is higher than the amount of blood sent in the arteries, thereby producing a hypervolemic twin known as the ‘recipient’. The recipient is at risk of cardiovascular volume overload, which thickens the heart muscle and can lead to heart failure.7  The donor is at risk of malnutrition with severe growth restriction, cardiovascular collapse, and malformations due to compression. If one twin dies in the uterus, the other is in immediate danger of losing considerable blood volume across the connecting vessels into the dead twin. Blood pressure in the surviving twin drops suddenly, causing heart attack or stroke.56  IUGR, associated with both DC and MC twins, occurs because of unbalanced placenta sharing.14  In IUGR, unlike TTTs, the growth of one twin is normal while that of the other is compromised. The smaller fetus has a smaller share of the placenta. Therefore, it receives less blood, nutrients, and oxygen. This phenomenon is referred to as ‘selective’ IUGR.52 Placental pathology in MC twins is localized to one placenta only. This explains another mechanism of BWD.52 A hemodynamic imbalance in vascular-thrombotic lesions of the placenta leads to a malnourished twin-pair.57,58  In addition, BWD my occur due to insufficient placental attachment to the uterus wall because of large stretched uterus in twin pregnancies.42  If trophoblasts, i.e., unit-cells of the placenta, do not invade the uterus wall deeply enough, the integrity of the placenta could be compromised at the implantation site.  Bleker et al. studied the placental index, defined as the ratio of placenta weight to newborn weight, and associated BWD with poor placental development.37 18  The type of umbilical cord insertion (central, marginal or velamentous) influences birth weight in both mono- and dichorionic twins.38,59  An increased risk of abnormal cord insertion is reported in multiple versus singleton pregnancies.60 The BWD risk is 50%  higher in MC pregnancies with velamentous umbilical cord insertion.60 A pregnancy with non-central insertion of the umbilical cord (marginal or velamentous) is 33% more likely to exhibit BWD > 20%61 compared to one with central cord insertion. 2.1.1 ADVERSE OUTCOMES 2.1.1.1 Adverse perinatal outcomes Twin gestations when compared to singletons are associated with higher perinatal mortality and morbidity. The higher rates of complications have been attributed to BWD by some16,18,20,21,23,62 but not by others.15,19  The most frequent reported BWD outcome is perinatal mortality. Perinatal mortality refers to still birth (intrauterine fetal death), as well as early (0-7 days), late (8-28 days) and overall (0-28 days) neonatal mortalities.  Other adverse perinatal outcomes that have been reported to be associated with BWD include respiratory distress syndrome,5,19,20 congenital anomalies,5,63 Apgar score <7 at 5 minutes,16,63 IVH,5,19 hypoglycaemia,19 anemia,19 hemorrhage,5 and longer neonatal intensive care unit admission.16,20,64  Perinatal outcomes and BWD Erkkola et al.18, one of the earliest publications in this area, investigated the impact of BWD on perinatal death among twins and emphasised the clinical importance of BWD. They studied 460 twin pregnancies delivered at the University Central Hospital of Turku, Finland, from 1970 to 1981. BWD was defined as ≥25% discordance between twin pairs calculated from the weight of 19  the larger twin.  Twin pairs with <25% discordance were considered as the reference group. Perinatal mortality was defined as all stillbirths plus newborns that died within seven completed neonatal days (early neonatal mortality). The perinatal mortality rate was found to be significantly higher in the twins with ≥25% discordance (9.7%) compared to the reference group (3.7%, p < 0.01).  The risk was even greater when stillbirths alone were compared. In the group with weight differences of 25% or more, the risk of fetal death was 6.5 times greater when compared to twins with the lower weight difference. The study’s strength was the exclusion of twin pairs with maceration, which indicated that the deaths had occurred more than a week before delivery, as these cases make the estimation of concordant growth difficult. The limitations of this study include an insufficient number of perinatal deaths (39 out of 920 twins), use of hospital cases, and failure to control for potential confounding variables such as gestational age. The authors concluded that a divergent growth pattern in a twin pregnancy carried an elevated risk of stillbirth, and that the clinical significance of BWD largely depended on the unclear definitions of normal and discordant growth. Perinatal outcomes and actual birth weight Unlike the study by Erkkola et al., Patterson et al.15 challenged the clinical importance of BWD and argued that the actual birth weight of each twin within the pair, rather than the difference between the two twins, is the most important determinant of outcome.  The authors estimated the frequency of BWD in three gestational age categories: 26-32, 33-36 and 37-42 weeks of gestation. BWD was defined as the difference between the larger and smaller twin divided by the birth weight of the larger twin multiplied by 100. This retrospective cohort study examined 194 sets of live-born twins born between 1982-1985 at the University of Texas Health Science Center, San Antonio at 26 weeks or greater of gestational age.  Neonatal mortality was defined as 20  death within the first 28 days of life in either infant (overall neonatal mortality). Subjects were categorized into six groups based on the BWD percentage: ≤5%, 6-10%, 11-15%, 16-20%, 21-25% and >25%. The frequencies of congenital anomalies were 6%, 9%, 8%, 4%, 8%, and 9%, respectively.  The corresponding values for perinatal morbidity were 23%, 13%, 17%, 23%, 36%, and 21%, respectively. The frequency of neonatal mortality for each group was 6%, 2%, 3%, 4%, 8%, and 4%, respectively.    Perinatal outcomes adjusted for gestational age Taking gestational age into account, the ‘mean ± SD’ of BWD was reported by Patterson et al. as 7.6 ± 5.9% for 26 to 32 weeks of gestation, 9.5 ± 9.5% for 33 to 36 weeks, and 11.4 ± 8.5% for 37 to 42 weeks (p =0.07). The occurrence of morbidity (p<0.001), neonatal death (p<0.001), and congenital anomalies (p<0.06) increased with prematurity.  However, neither perinatal morbidity (p=0.64) nor neonatal death (p=0.84) nor congenital anomalies (p=0.98) was significantly related to the level of discordance. These data demonstrated that prematurity may present a greater threat to twins than does BWD.  The authors concluded that BWD itself is a poor predictor of adverse outcomes and that BWD is much less important than prematurity.  The conclusion presented by Patterson et al. may be challenged for a number of reasons.  1. The number of adverse outcomes among different BWD levels was very small. For instance, with samples of ranging from 11 to 66 subjects for the BWD categories, there was insufficient power (40%) to find a significant difference in outcomes based on prevalence of the variables studied, e.g., neonatal death.  2. Only live-birth twins were included in the study as the weight of stillbirths was found to be unreliable.  The study was therefore unable to determine fetal death as an outcome.  21  3. The study was limited to twins born at 26 weeks of gestation and above, and thus did not capture twins with significant discordance.  Below 26 weeks, the divergence of twin weight is more likely to be due to congenital anomalies or TTTs.62    4. Confounders other than gestational age were not assessed.  Perinatal outcomes adjusted for IUGR In addition to gestational age, IUGR can confound the impact of BWD on perinatal mortality.  A study by Fraser et al.19 addressed both of these confounding variables. This hospital-based study from Soroka Medical Center, Israel, compared neonatal outcomes between 133 discordant versus 1012 concordant twin pairs between 1960 and 1986.  BWD was defined as ≥25% discordance based on the weight of the heavier twin. IUGR was defined as birth weight <10th percentile of the sex-specific birth weight for singleton births for the given gestational age. Discordant twins had an elevated risk of early neonatal death (2.0, 95%CI 1.1-3.4) and stillbirth (6.4, 95%CI 2.9-9.0) in comparison with concordant twins.  Subjects were then divided into four groups: Discordant twins with (n=118) and without (n=148) IUGR and concordant twins with (n=354) and without (n=1670) IUGR.  The reference group was the concordant group without IUGR in all comparisons.  The odds of stillbirth were found to be the highest for discordant twins with IUGR (7.2, 95%CI 2.9-17.2). The noticeably wide CI range is of concern in this comparison. In multiple logistic regression analyses (adjusting for gestational age, BWD, and IUGR), low birth weight was the single consistent factor associated with elevated risks of mortality. Consistent with the findings by Patterson et al., Fraser et al. concluded that birth weight itself is of great importance.  22  Fraser et al. also found that some indicators of morbidity decreased in BWD twins in comparison with concordant twins (respiratory distress syndrome [0.3, 95%CI 0.0-1.8], hyperbilirubinemia [0.8, 95%CI 0.3-2.0], septicemia [0.7, 95%CI 0.1-3.3]), while other indicators increased in discordant twins (still birth [4.7, 95%CI 2.4-9.0], early neonatal deaths [2.0, 95%CI 1.1-3.4], hypoglycaemia [3.1, 95%CI 1.2-7.8], anemia [1.2, 95%CI 0.4-3.5], and IVH [1.2, 95%CI 0.1-9.4]). The study lacked the power to estimate the outcomes observed. For instance, to obtain a statistically significant association with the observed difference of 3.5% (4.7% concordant vs. 1.2% discordant) in respiratory distress syndrome, the study had less than 30% power. The analysis considered only gestational age, IUGR, and birth weight itself as the confounding factors. Additionally, the authors mentioned a lack of data for neonatal death despite their primary intention to measure the outcome.  Perinatal outcomes and twin size Unlike Patterson et al. and Fraser et al., Hollier et al.’s20 retrospective study reported a positive association between BWD and stillbirth even after controlling for the gestational age. They also showed that greater BWD was associated with stillbirth. The authors assessed intra-pair birth weight differences among 1370 consecutive twin pairs delivered at Parkland Hospital, Dallas, Texas, during 1988-1996, among pregnant women who had live births or fetal death within 7 days of delivery. Discordance was defined as the intra-pair difference in birth weight expressed as a percentage of the larger twin’s weight. Intra-pair birth weight differences were stratified into the following six BWD groups: 14% or less, 15-20%, 21-25%, 26-30%, 31-40%, and 41% or more using the larger infant as the reference. Infants were defined as small for gestational age (SGA) when their birth weight was less than the 10th percentile for twins according to Cohen.65  A higher rate of induced preterm delivery (70%) was found among those with a BWD more than 23  41%, compared with those who had less than a 14% BWD (15%, p<0.01). SGA (45% vs. 4%, p<0.001), intensive care admission (55% vs. 15%, p<0.001), and respiratory distress syndrome (41% vs. 12%, p<0.001) were also related to high (≥41%) vs. low (≤14%) BWD.  Moreover, fetal anomalies (3% for ≤14% vs. 18% for ≥41%, p<0.001) and stillbirth (4% for ≤14% vs. 24% for ≥41%, p<0.001) were found to be more common with increased discordance from <14% up to ≥41% (p<0.001).  As mentioned earlier, SGA was categorized according to Cohen (1997), a criterion based on pregnancies induced by ovulation-induction techniques.  This categorization may not be generalizable to spontaneously conceived pregnancies. In addition, unspecified numbers of neonates were excluded if one or both were stillborn, or if the mother smoked, had a significant chronic illness, or was prescribed any regular medication.   Perinatal outcomes adjusted for infant sex and birth order The studies described so far have shown that the impact of BWD on perinatal adverse outcomes differed with gestational age, IUGR, and birth weight of the smaller twin.  In a population-based study, Demissie et al.24 estimated the risk of adverse perinatal outcomes separately in the larger and smaller twins while controlling for additional confounding variables, such as infant sex and birth order. The retrospective population-based study was carried out among 148578 twin pairs in the United States between 1995 and 1997. BWD was calculated as the intra-pair difference in birth weight expressed as a percentage of the larger twin’s weight. SGA was defined as birth weight below the 10th percentile for gestational age and based on a normogram of all twin births delivered in the United States during the period 1995-1997. This definition, different from that used by Hollier et al.,20 is appropriate. The stillborn fetus rate increased progressively with increasing BWD for smaller and larger twins.  Compared with the category of <5% discordance, the adjusted odds ratio (OR) for a stillborn fetus associated with 5–9%, 10–19%, 20–29%, 30–24  39%, and >40% discordance were 0.81 (95% CI 0.6- 1.1), 1.41 (95% CI 1.1- 1.8), 1.74 (95% CI 1.3- 2.4), 3.06 (95% CI 2.2- 4.2), and 4.29 (95% CI 3.1- 6.0), respectively, for smaller twins.  The same trend was observed for larger twins: the corresponding ORs were 0.78 (95% CI 0.6, 1.1), 1.26 (95% CI 0.9, 1.7), 1.77 (95% CI 1.3, 2.5), 3.38 (95% CI 2.3, 4.9), and 2.91 (95% CI 1. 9, 4.5) when compared to the <5% BWD category. The authors concluded that accounting for several important confounding factors (the actual birth weight of each twin, gestational age, birth order and infant sex) BWD was associated with an increased risk of stillborn fetus in both the smaller and larger twins. Although the study sample size was the largest among all investigations reporting on BWD and perinatal outcomes, it has a number of limitations.   First, gestational age was based on the last menstrual period (LMP) date, as adopted from the birth certificate. This measure is not accurate as it relies on the mother’s memory, which is prone to recall bias. Furthermore, records with missing LMP dates were imputed when a valid month or year was reported. This procedure introduces noticeable bias.66    Second, cases with TTTs were not excluded from the analysis, introducing a potential bias in the study that leads to overestimation of BWD because the larger twin may be hydrops, an inappropriate standard for judging the extent of discordancy.   Third, chorionicity was not adjusted for. Adjustment for chorionicity is important because once there is death of one twin in a MC twin gestation, the death of the second may no longer be an independent event due to the shared placenta.   Fourth, the study failed to control for maternal age, which is a significant determinant of birth weight.67   Fifth, the study, and others introduced earlier, included twins with stillbirths in the analysis which may bias the result as a dead fetus that is retained in utero for a long period of time, 25  may cause growth discordance. To void bias in this situation, a study would need to precisely predict the time of fetal death or exclude cases with single-stillbirth (where only one co-twin is alive and the other is dead) with signs of maceration. Our research has excluded cases with single stillbirth to avoid such bias.  Perinatal outcomes adjusted for sex discordance and maternal age Wen et al.23 considered sex discordance and maternal age as confounding variables, adjusting for a larger number of variables compared to previous studies. The investigation examined the relationship between discordance degree and adverse perinatal outcomes using data from Statistics Canada, Canadian Birth Database, and Canadian Mortality Database from 1986 to 1997.  Fetal death was defined as stillbirth with birth weight ≥500g or gestational age ≥20 weeks. Neonatal death was defined as a live birth that died prior to the 28th day. Out of 59034 twin births, 53% had 0-9% BWD, 30% had 10-19% discordance, 11% had 20-29% discordance, and 6% showed >30% discordance. The fetal death rate was 0.95, 1.26, 3.14, and 11.10% for each group, respectively. Corresponding rates for neonatal death were 1.90, 2.32, 3.05 and 8.86%, respectively.  This study was a breakthrough in design because of its use of population-based data from across Canada, controlling for many potential confounding factors (birth year, maternal age, birth order within the twin set, parity, infant sex, sex discordance, birth weight for gestational age, and gestational age), and employing a conservative definition for stillbirth. However, it suffers the following limitations:   1. Similar to other studies, the study did not adjust for single stillbirth, maternal education and maternal weight before pregnancy or at admission.  These variables are shown to be important determinants of birth weight.68–70   26  2. Ontario births were excluded from the analysis due to documented problems with data quality.   3. Records from 1985 births were excluded from the final analysis because the ‘birth order’ of the twin was missing in a large proportion of the records for that year.  Notwithstanding these limitations, the study found that the BWD-related risk of perinatal death was greater in smaller twins than in larger twins. It also determined that the BWD-related risk of perinatal death varied substantially at different gestational ages.  These findings are consistent with the only other population-based study, Demissie et al. described above.24  Both investigations highlight the importance of twin size when interpreting the impact of BWD on perinatal outcomes.  Perinatal outcomes adjusted for chorionicity Another study that compared the impact of BWD on adverse perinatal outcomes and considered the gestational age and size of the twins was conducted by Amaru et al.64 Conducted in Mount Sinai Medical Center, New York, this retrospective study also took into account chorionicity and growth restriction. Only pregnancies with two live-born infants born at 24 weeks or later were included. Discordance was defined as a difference in birth weight of 20% or greater based on the heavier twin. SGA was defined as <10th percentile using a singleton distribution as the reference group based on Brenner’s criteria (1976). Four groups were compared: 1. Appropriate for gestational age (AGA) and concordant (n=586), 2. AGA discordant (n=70), 3. SGA concordant (n=254), and 4. SGA discordant (n=138).  The ORs adjusted for SGA status showed a higher risk for discordant twins in comparison with concordant twins for the following outcomes: neonatal death (2.5, 95%CI 0.7-9.2), hyperbilirubinemia (1.4, 95%CI 1.3-1.7), respiratory distress 27  syndrome (1.4, 95%CI 0.8-2.4), and neonatal intensive care unit admission (1.9, 95%CI 1.6-2.2).  The study also showed that perinatal mortality and morbidity was much higher in MC twins compared to DC twins because MC placentation was independently associated with neonatal death (4.1, 95%CI 1.2-14.2). A major strength of the study is its incorporation of placentation, maternal age, race, smoking, and alcohol and drug use in the regression analysis.  Adjusting for sensitive data, such as substance use, is unique to this study, as this type of information is difficult to acquire. However, several methodological flaws of the study are worthy of mention.    First, the use of singleton norms to define AGA and SGA introduced a measure bias because twins are known to demonstrate lower weights than singletons after 30 weeks of gestation.71     Second, the study was limited to twins born at 24 weeks of gestation or more capturing a wider group of twins with significant discordance. Below 26 weeks, divergence in twin weight is more likely to be due to congenital anomalies or TTTs.62      Third, as stated by the authors, the study was underpowered to assess the association of growth discordance and neonatal mortality within the grouping of growth restriction.  Fourth, chorionicity determination in twin gestation is an important issue in the risk assessment of twin pregnancies. Twins who are connected to a single placenta MC are more prone to adverse perinatal outcomes compared to those with one placenta each DC.72  Although the distinction between MC and DC twins can be made by ultrasound exam in the first and second trimesters, these screenings are not free from bias.73  Table 2-2 summarizes the characteristics of studies that considered chorionicity when investigating the impact of BWD on perinatal outcomes or used a surrogate measure instead of chorionicity (e.g., analysis of unlike-sex twins). The former studies determined chorionicity 28  using various methods including pathology reports, ultrasound screening during pregnancy, and fetal sex determination in ultrasound.   Table 2-2 List of studies investigating clinical significance of BWD in twin pregnancy Author Year Study design Chorionicity Definition of Discordance Chorionicity determination Amaru64 2004 Retrospective cohort Yes 20% Pathology report Alam 74 2009 Prospective cohort Yes ≥20 Histopathology, first-trimester ultrasound , discordant fetal sex Appleton16 2007 Retrospective cohort Yes ≥20% Unknown Hartley75 2002 Population-based retrospective No ≥25% Not available on birth certificate, suggested to use unlike-sex twins as proxy Demissie.24 2002 Retrospective cohort No 10% for smaller twin, 20% for larger twin Not available on matched data Redman.76 2002 Retrospective cohort Yes ≥30% Medical records by separate placentas, neonates discordant for sex, gross and histologic pathology assessment Victoria52 2001 Retrospective cohort Yes >25% (severe discordant), 5-25% (mild discordance) Pathology report Hollier20 1999 Retrospective cohort No ≥25% Twin pairs of the same sex were evaluated as a surrogate for monochorionic twins Lanni77 1998 Retrospective cohort No 90th percentile Study of like-sex and unlike- sex pairs recommended Erkkola18 1985 Retrospective cohort No ≥25% Not considered  Many studies of discordance did not separate twins by chorionicity.18–20,78  Furthermore, studies that took chorionicity into consideration showed inconsistent results in terms of its effect on perinatal outcomes.  A twin study involving term and preterm twins found that the worst outcome in MC twins was mediated by the greater risk of severe prematurity among MC rather than DC twins.52  Moreover, BWD was significantly larger among MC compared to DC twins. These findings led the authors 29  to conclude that mono-chorionicity was at the root of both extreme prematurity and inter-twin discordance. However, the rates of adverse perinatal outcomes such as fetal death (4/36 vs. 3/44, p=0.69) and NICU stay >30 days (11/36 vs. 7/44, p=0.2) among cases with severe weight discordance (i.e., >25%) were not significantly different between MC and DC twins. Consistent with this result, Dubé et al. found that a BWD >25% is equally present in MC and DC gestations (24/280 vs. 47/728 p=0.3).79  A prospective cohort study of 356 twin gestations accounted for chorionicity by ultrasound assessment of the dividing membrane, neonatal sex, and examination of the placenta at birth. Multivariate analysis showed that BWD was associated with adverse outcomes independent of chorionicity.80  Alam et al.74 studied the early neonatal morbidity and mortality in 151 twin pregnancies with growth discordance at the Multiple Pregnancy Unit of Sao Paulo University Hospital between 1998 and 2004.  Chorionicity was defined in 84.2% of cases (n=127) by histopathology, in 6.6% (n=10) with first-trimester ultrasound, in 2.6% (n=4) using the presence of distinct placental sites at ultrasound scan, and in 6.6% (n=10) based on discordant fetal sex of the cases. An assessment of both concordant and discordant groups using chorionicity showed a worse outcome in the MC concordant sub-group. In concordant twin pregnancies, MC cases presented a lower gestational age (34.3±3.2 weeks vs. 36.2±2.7 weeks, p=0.004) and weight at birth (2067±582 grams vs. 2334±501 grams, p=0.002) and a longer period of hospital stay (5.5 days vs. 3 days, p=0.002) than DC cases. This could be explained by intrinsic deficits secondary to the unique placental angioarchitecture of these cases.  30  Another prospective cohort study of 1028 twin pairs in Ireland used series of ultrasound (from 2007 to 2009) to determine chorionicity. Chorionicity was determined through standard ultrasound criteria including placental number, identification of lambda or T sign, inter-twin membrane thickness, and fetal sex at the first visit. Subsequent correlation was sought with placental pathologic examination. Fortnightly growth scans were performed from 16 weeks of gestation until delivery for MC twin pairs and from 24 weeks until delivery in DC pregnancies.47 Perinatal mortality, individual morbidity, and composite perinatal morbidity were increased with BWD exceeding 18% for DC pairs (hazard ratio 2.2, 95% CI 1.6–2.9) and 18% for MC twins without TTTs (hazard ratio 2.6, 95% CI 1.6–4.3). Appleton et al.16 conducted a retrospective cohort study of 230 twin gestations that ended at 34 weeks of gestation or later. They attempted to study the impact of chorionicity but encountered a low number of MC twins (15 in discordant twins and 35 in concordant twins), preventing them from extended evaluation based on placentation. The comparison between the growth concordant (<20%; n=176) and discordant groups (n=54) did not indicate statistically significant differences in the parameters under study, including hypertensive disorders of pregnancy, gestational age at birth, route of delivery, reason for termination of pregnancy, Apgar scores, birth weights, admission to neonatal intensive care unit, significant morbidity, malformations found at birth, and perinatal mortality.  Inconsistency in the findings on the significance of chorionicity was not the only reason that led some researchers to ignore chorionicity in the investigation of BWD and perinatal outcomes. The main reasons were data inaccessibility and inability to link chorionicity data with adverse perinatal outcomes.  31  Chorionicity can be incorporated in twin studies provided that  a. Birth registers have access to the electronic patient records,  b. The twins are traceable by a unique identification number, and  c. The chorionicity information is included in the electronic medical records.  Unfortunately, these conditions are not met in many registries. According to a study conducted in America81, chorionicity-related ICD-9_CM codes did not exist in 2007. Beginning in fiscal year 2011, coders were able to specify if a twin was from a monochorionic-monoamniotic, monochorionic-diamniotic, or dichorionic-diamniotic pregnancy by using ICD-9-CM codes V91.01, V91.02, and V91.03, respectively.82  To compensate for the lack of chorionicity information, a European study group attempted to investigate whether BWD can be used as a chorionicity indicator.83 The study sample consisted of 4060 monozygotic live-born twin pairs from the East Flanders Prospective Twin Survey containing data on zygosity and chorionicity at birth. For discordancy levels of <10%, 10–15%, 15–20%, 20–25% and > 25%, the ORs (95% CI) were 1.16 (0.91–1.47), 1.38 (1.05–1.80), 2.13 (1.51–3.01), 2.73 (1.73–4.29) and 2.81 (2.81–4.35), respectively. BWD was considered to be 42.2% sensitive and 72.8% specific for determining the chorionicity. The investigators maintained that although a higher discordancy level resulted in higher ORs in MC twins, BWD level can only be used as a poor proxy for chorionicity.  This conclusion highlights the need to assess and record chorionicity data in obstetrical units. Other studies that have used sex-discordant twins as a surrogate measure for chorionicity are listed in Table 2.2.   32  Perinatal outcomes adjusted for prenatal care Another interesting confounding factor that hinders understanding of the impact of BWD on perinatal outcomes is whether or not the mother receives routine prenatal care during pregnancy.  Mazhar et al.5 controlled for booking status during pregnancy, an important variable in the health of mothers and babies living in developing countries. This prospective study compared 96 discordant twin infants with 410 concordant twins in a local tertiary hospital in Pakistan from 2005 to 2007.  Discordance was defined as >20% difference between the birth weight of the heavier and lighter twin divided by the heavier twin’s weight multiplied by 100.  Higher frequencies of stillbirth (10.4% vs. 4.3%, p<0.03), preterm birth (68% vs. 25%; p<0.001), respiratory distress syndrome (12.5% vs.3%; p<0.001), malformation (8.3% vs. 2%, p<0.001), IVH (5.2% vs.3.1%, p<0.007), lower frequencies of sepsis (2% vs. 3.1%, p<0.009) and anemia (0% vs. 3.6%, p<0.001) were observed for discordant twins in comparison with concordant twins.  One infant among the discordant twin pairs (1%), as opposed to 10 among the concordant twins (2.4%), expired in the neonatal period (p<0.39). The study did not present a definition for neonatal death and lacked the power (23%) to estimate a significant difference for neonatal mortality. Inclusion criteria were limited to infants who required admission to hospital in the early neonatal period. This means that late neonatal mortality was not accounted for. Such a limitation in reporting neonatal mortality is typical of hospital-based studies. This is expected because late neonatal mortality is not as easy to monitor as early neonatal deaths. Late neonatal death might be recorded in a setting different from where the baby is born. Therefore, the overall mortality rate, which also includes late neonatal death, can be reported inaccurately. A clear definition of early, late or overall neonatal mortality should be provided to avoid confusion.   33  Perinatal outcomes of near term pregnancies Appleton et al.16 investigated the clinical importance of “near term” weight discordance in twin pregnancies with both AGA and SGA newborns. This retrospective study included 230 twin pregnancies (from Hospital Universitá Rio de Santa Maria, Portugal) which ended at ≥34 weeks of gestation.  Discordance was defined as an inter-twin birth weight difference ≥20% calculated from the heavier newborn. The discordant pairs were subdivided into two groups: 1. Both twins were AGA (AGA/AGA), and 2. One twin was SGA (SGA/AGA). The two groups were compared to each other and to the control group of concordant pairs. A comparison between discordant twins (n=54) and concordant twins (n=176) was made. There were significantly higher frequencies of neonatal morbidity (2.7, 95%CI 1.4-5.3) and admissions to the neonatal intensive care unit (5.7, 95%CI 3.2-10.0) among the discordant twins. When the discordant subgroups [AGA/AGA (n=24) and AGA/SGA (n=30)] were compared to the concordant group (n=176), this difference, particularly in the AGA/SGA group, persisted. A greater need for intensive neonatal care was seen both in AGA/AGA (3.6, 95%CI 1.6-7.8) and AGA/SGA pairs (7.9, 95%CI 4.1-15.5) in discordant twins when compared with concordant twins. These findings support the view that when a discordant pair includes an SGA newborn, it is expected to observe increased morbidity and greater need for NICU admission. Despite raising an important clinical point, the study findings are to be interpreted in light of several limitations. The study did not have adequate power (<80%) to look into the secondary outcomes.  Parity, maternal age, and method of conception were analyzed as study outcomes rather than being adjusted for as confounding factors.    34  Perinatal outcomes of preterm birth Unlike the study by Appleton et al. which focused on “near term” weight discordance in twin pregnancies, Cheung et al.63 restricted the analysis of BWD to “preterm” birth.  A small sample size of 122 live born twin sets, delivered between 25 and 34 completed weeks of gestation at St. Joseph’s Health Center in London (Ontario, Canada), was retrospectively examined. The study purpose was to determine the predictive role of discordant fetal growth in relation to subsequent neonatal outcomes for different degrees of BWD in preterm twin gestations.  Twin sets were stratified into five categories according to percentage differences in infant birth weight: <10%, 10% to 15%, 15% to 20%, 20% to 30%, and >30%.  Preterm twin gestations with >30% discordance were found to be associated with a higher incidence of stillbirth (25%, p<0.006), congenital abnormalities (37.5%, p<0.001), 5th minutes Apgar score of <7 (33.3%, p<0.003), and small for gestational age infants (31.8%, p<0.0001) compared with all other groups. Although the >30% discordant group infants had a significantly higher incidence of periventricular leukomalacia (16.7, p<0.01), no differences were found among groups for the incidence of other prematurity-related morbidities such as respiratory distress syndrome, IVH, and neonatal jaundice. The target population were preterm infants, and no difference in the incidence of prematurity-related complications among the five groups was found.  Consequently, the direct relationship between adverse outcomes and gestational age was unsubstantiated.  More importantly, a birth weight difference of >30% in twin gestations was associated with a higher incidence of SGA infants (31.8%) and lower frequency of 5th minutes Apgar scores >7 (33.3%) compared to twins with <10% discordance. Thus, the authors hypothesised that the perinatal adverse outcome could be associated with impaired intrauterine growth and not with TTTs. The incidence of suspected TTTs cases, as defined by an inter-twin hemoglobin difference of 35  >5gm/dl, was similar among groups and did not explain the higher incidence of infant death in the >30% discordant group. This argument is invalid because haemoglobin measurement cannot be the identifier for TTTs and any combination of weight and haemoglobin-hematocrit discordance could occur in the monochorionic twin.84 An optimum study design should exclude cases of TTTs.  What is the threshold of BWD? There is a controversy surrounding the clinical importance of fetal growth discordance.  The threshold of BWD at which clinicians should be alerted is even more vigorously questioned. Various BWD thresholds ranging from 10% to 30% have been proposed as clinically significant (Table 2-1).  The cut-off values in the studies referred to in Table 2-1 were chosen subjectively. None of the cut-off values was based on the observed distribution of birth weight discordancy in the reference population.  Conventionally, a measure that is beyond the 90th percentile of the population value is considered abnormal. Lanni et al.85 adopted such a population to obtain a cut value for BWD by studying a cohort of 849 Italian twin pairs from 49 maternity hospitals located in 14 out of 20 regions.  The overall sample did not differ from the population in terms of twinning rate (0.99% vs. 1.1%, 95%CI 0.9-1.1) and prematurity rate (7.9% vs. 6.5%, 95% CI 6.2-9.9). The gestational age was estimated by ultrasound at the 16th week. They proposed 23.9% and 29.2% as the cut-off values for the definition of mild and severe BWD, respectively, based on the 90th and 95th percentiles of the population distribution after adjusting for sex. The clinical validity of this 36  method is debatable as they did not use any adverse outcomes as a reference point to define the cut-off values.  Some differences in birth weight between infants can be expected in twin gestations as a normal variation between twins. Cheung et al.63 sought to identify the threshold of BWD at which a significant difference in perinatal and neonatal outcomes was more likely to occur.  They categorized twins based on BWD (<10%, 10% to 15%, 15% to 20%, 20% to 30%, and >30%) and obtained the adverse outcome within each category. As high adverse perinatal outcomes in preterm twin gestations appeared at the >30% BWD level, the authors concluded that the usage of 30% birth weight difference is optimal and clinically relevant when defining discordance in preterm twin gestation and in identifying infants at risk for adverse outcomes. Similar to Cheung et al., Cooperstock et al.86 suggested a cut-off value of >30% as clinically significant. The study was done in Missouri from 1978 to 1990 in discordant (n=2255) and concordant twins (n=5851). The authors compared the risk of very preterm births (<32 weeks of gestation) and excluded births between 32 and 37 weeks of gestation, a period considered to have mild preterm births. Twins within this category (32 to 37 weeks) endure an increased rate of fetal and neonatal death compared with term births.84 Thus, defining a cut-off point for all pregnancies based on twin pairs with very preterm birth is not optimal.  Demissie et al.,24 in the population-based study described above, used a set of detailed cut-off values (0-4%, 5-9%, 10-19%, 20-29%, 30-39%, and 40%).  They used the Matched Multiple Birth File to describe the role of intra-pair BWD on fetal and neonatal mortality for the larger and the smaller twin in a twin pair. A discordance of 10% or more conferred a significant increase in the stillborn fetus rate for smaller twins (1.4, 95%CI 1.1, 1.8), whereas this threshold was 20% or more for larger twins (1.8, 95%CI 1.3, 2.5) compared with <5% BWD. Wen et al. used broader cut-off values (0-9%, 37  10-19%, 20-29%, and >30%) and found a significantly increased fetal and neonatal mortality with a BWD of 20-29%, and a dramatically increased mortality with a BWD of >30%.  The discussion of the optimal cut-off levels is important as it is still unclear whether greater birth weight disparity represents a transition from normal variation to growth deviation through a certain critical discordance level, or whether small and large discordance values represent distinct entities. The current thesis aims to shed light on both the impact of BWD on adverse perinatal outcomes and the optimal cut-off level for BWD.   2.1.1.2 Adverse maternal outcomes Only a few studies have addressed the association between BWD and maternal outcomes, none of which was population-based. Among the hospital-based studies, six studies did report an increased risk of preterm birth, pregnancy-induced hypertension, CS, and premature rupture of membranes, as explained below.  A retrospective study in India included 150 women with twin pregnancy ≥28 weeks of gestation, delivered at Siriraj Hospital at Mahidol University, from 2003 to 2004.87 Discordance was defined as ≥20% difference in birth weight based on the heavier twin.  Baseline maternal characteristics and obstetrical history analyzed included maternal age, occupation, income, education, parity, mode of delivery, chorionicity, mean placenta weight, sex, and mean of gestational age. No variable was found to be associated with discordancy (p>0.05). However, discordant twins were delivered at an earlier gestational age compared to concordant twins (34.9 + 3.2 and 36.2 + 2.4 weeks, p =0.037). This study did not find any significant association 38  between BWD and maternal medical complications (8.6% in discordant group vs. 16.5% in concordant group, p=0.290).  This investigation analysed medical complications in aggregate, did not define the complications, and lacked adequate power (as low as 32%) to identify a significant difference between groups for medical complications given that there was a 7.9% difference between concordant and discordant twins.  The impact of BWD on preterm birth was also studied by Cookerstock et al.88 Twin deliveries were abstracted from birth certificates in Missouri, USA, to obtain data on 9931 pregnancies at ≥17 weeks of gestation registered between 1978 and 1990. Gestational age on the birth certificate was calculated using LMP. Cases were stratified according to BWD into <30%, 30%-40%, and ≥40%. The rate of preterm delivery prior to 32 weeks of gestation was 9.5% with a discordance <30%, 13.2% with a discordance of 30%-40% and 34.1% with a discordance ≥40% (p<0.001). Compared with twins with <30% BWD, the risk of preterm birth was significantly higher for twins with ≥40% BWD at <32 weeks of gestation (9.5, 95%CI 2.3-4.5) in a multivariate model. The model was adjusted for black race, small for gestational age, unmarried, teenage mother, number of male fetuses, like fetal sex, education <12 years, nulliparity, and cigarette smoking. Risk of preterm birth was also increased among twin pregnancies with BWD (1.3, 95%CI 0.4-3.9) in the previously introduced study by Appleton et al.16 They also reported higher risks of pregnancy-induced hypertension (1.9, 95%CI 0.9-4.2) and iatrogenic preterm birth (1.10, 95%CI 0.6-2.1) for discordant twins in comparison with mothers of concordant twins. Although such 39  differences were not statistically significant, insufficient power was observed for analysis of pregnancy-induced hypertension (24%), iatrogenic preterm birth (13%), and CS (54%).   A higher rate of CS was also reported in a hospital-based retrospective study by Kilic et al.89 who followed 136 preterm deliveries admitted to the intensive care unit (ICU) from 2003-2005 in Ondojuz Mayis University Hospital, Turkey.  The twins were categorized as discordant if the birth weight difference of twin pairs was more than 15%. Birth by CS was significantly higher in discordant twins (56.1% vs. 77.8%) in comparison with concordant twins (p <0.05). Although the guidelines for selective CS may vary in different countries, a higher rate of CS due to discordance is of major clinical importance.  Mazhar et al.5 also reported a higher frequency of CS for discordant twin gestations (29%) compared to 21% among the concordant twin pregnancies. The retrospective study examined 253 mothers who delivered from January 2005 to December 2007 in a Maternal Child Health Center in Islamabad, Pakistan.  BWD was defined as >20% twin birth weight difference based on the heavier twin.  Similar to Appleton et al.’s study, pregnancy-induced hypertension was found to be significantly more frequent in the BWD group compared to those with concordant birth weight (22% vs. 12%, p<0.004). The authors argued that in Pakistan (where many women deliver at home) the older, registered, grand-multiparous and hypertensive women were more likely to deliver discordant twins. This could lead to overestimation of PIH as more women with complications deliver in a hospital setting in search of better care for their twin conception. Complications associated with PIH, such as eclampsia and preeclampsia, have not been studied by Mazhar et al.5 or other investigators.  40  The study by Demissie et al. 24 described earlier also reported higher rates for PIH among mothers of discordant twins compared with concordant ones. Considering <5% BWD as a reference category, the proportion of mothers with PIH associated with 5–9%,10–19%, 20–29%, 30–39%, and >40% BWD were 7.3%, 8.0%, 9.8%, 10.7% and 9.5%, respectively. The proportion of mothers with chronic hypertension, placental abruption, and CS also increased with increasing BWD. Two issues with the study findings are as follows. First, the authors acknowledged the inaccuracy of birth certificate records leading to underreporting of pregnancy complications. Second, no clear definition of PIH or chronic hypertension was provided.    2.2 SEX DISCORDANCE  Definition Adverse perinatal and maternal outcomes are associated not only with BWD but also with sex discordance, defined as the physiological sex differences of fetuses in a twin pregnancy.  Incidence There are two types of twins: monozygotic (identical) and dizygotic (non-identical). Identical twins always have the same sex while non-identical twins can have the same sex (sex concordant) or the opposite sex (sex discordant). The incidence rate of identical twins is 3.5 per 1000 births while non-identical twin births range from 6.7/1000 births in Japan to a high of 40/1000 births in Nigeria.90 The probability of conceiving non-identical twins depends on maternal age at the time of conception, the history of multiples on the mother’s side of the family, race, parity and the use of assisted reproductive technology. Approximately one half of same-sex multiples are identical and one half are non-identical.   41  Effect mechanism on perinatal outcomes It has been hypothesized that the mere presence of a male fetus in the uterus along with a female has a profound effect on the perinatal morbidity and mortality of the female fetus.25 Studies have shown that male hormones have two intra-uterine effects: 1. Female fetuses grow faster if they are in utero with a male co-twin. 2. Androgens are found to inhibit fetal lung development.91   Animal studies Animal studies92,93 have postulated that androgens are transferred between fetuses across membranes, and that female fetuses in discordant twin pairs vs. female co-twins have higher concentrations of serum testosterone.  A study in Austin, Texas93 compared the morphological, physiological and behavioural characteristics of 250 sixty-day-old female mice.  Two groups of female mice were compared: females who grow between two male fetuses in the uterus (2M-females) and females not adjacent to males in the uterus (0M-females).  2M-female mice were found to be more masculinized morphologically and behaviourally relative to 0M-females.  They had a significantly larger anogenital space (distance from the anus to the genitalia), and were significantly more aggressive (measured by urine marking). In addition, prenatal androgen treatment has been shown to increase birth weight and growth rate in lambs.94 The androgen treatment involved a single 2-ml intramuscular injection of 100 mg/ml testosterone propionate, which was administered at 30, 40, 50, and 60-day postcoitus. Acute prenatal androgen treatment caused an overall increase in the mean birth weight of 14.6% (p < 0.05), with a 13.2% increase in male lambs (n=7) and an 18.2% increase in female lambs (n=6).   42  Human studies In humans, androgens have been shown to exert a positive effect on fetal growth, as absence of androgen can adversely affect birth weight.95 Boys are heavier than girls at term birth. A study by de Zagher et al.95 compared birth weights of boys and girls with a group of children who were suffering from androgen insensitivity syndrome. These children with a 46XY karyotype and androgen insensitivity syndrome were found to have a birth weight close to that of girls. These findings support the hypothesis that the difference in birth weight between boys and girls is generated by androgen action.  Based on these animal and human studies, it has been postulated that female fetuses would grow faster if they are in utero with a male co-twin. Glinianaia et al.26 studied 1087 sex-discordant twin pairs and 1089 concordant pairs, identified from a population-based Medical Birth Registry in Norway during 1967-1974. The mean birth weight of females from sex discordant pairs was 2684±15 grams compared to 2647±19 grams in females from concordant pairs (p=0.06).26  The finding suggests that the birth weight of females is influenced by the presence of a male co-twin. This finding was not due to difference in gestational age between the two groups. Moreover, androgen is found to have an inhibitory effect on lung development. The levels of gene 17β-hydroxysteroid dehydrogenase (17β-HSD) expression at 15 to 17 days of gestational age in six mice were explored (Figure 2-1).27  This androgen has inhibitory effects on maturation of Type II pneumocyte (PTII) cells in the lungs that in turn leads to surfactant deficiency and respiratory distress in neonates. A delayed expression of 17β-HSD level was observed for male animals in comparison with female (>17.5 weeks vs. 17 weeks of gestational age).  It was hypothesized that this delay is based on the delay of PTII cell maturation observed for males.  According to this hypothesis, the presence of androgen in the lungs could result in a delay in 43  lung maturation for the female fetus. In theory, this could explain the increased risk of respiratory morbidity found among male-female twin pairs compared to female-female twin pairs (1.3, 95%CI 1.1–1.7).31                                                          Inhibition of surfactant secretion Figure 2-1 Inhibitory effect of 17β-HSD leads to respiratory distress  Some investigations that address the pathophysiology underlying poorer outcomes in pregnancies with male fetuses are outcome-specific. For instance, a higher incidence of prematurity in male births compared to female births is attributed to an increased susceptibility to infection in mothers carrying a male pregnancy96,97 and a higher rate of pregnancy-related complications.31  Effect mechanism on maternal outcome Preeclampsia Findings for singleton and multiple pregnancies suggest there is a link between fetal sex and the pathophysiology of preeclampsia.  In singleton pregnancies, several studies have recognised the association of fetal sex with preeclampsia and pregnancy-induced hypertension. An analysis of over 52000 singleton pregnancies at Hershey Medical Centre, Pennsylvania, showed that early gestational blood pressure was significantly higher in preeclamptic women with male than with female fetuses.98 In addition, in singleton pregnancies, Toivanen and Hirvonen99 reported that the ratio of males to 17β-HSD  Lung cell   44  females in babies born to mothers with pregnancy-induced hypertension was 1.24 and the ratio increased to 1.72 depending on hypertension severity.99 On the other hand, Hsu et al.100 found a predominance of female fetuses in preeclamptic pregnancies compared with normotensive pregnancies (p = 0.043) in preterm infants.   In multiple pregnancies, Makhseed et al. 101 found no difference in the male/female ratio between normotensive mothers and preeclamptic mothers. This contradicts the findings of Sheiner et al. 102 in a retrospective study of 108995 deliveries assessing the complications associated with fetal sex. Women carrying male fetuses had higher rates of gestational diabetes mellitus (1.1, 95%CI 1.01-1.12), fetal macrosomia (2.0, 95% CI 1.8-2.1), cord prolapse (1.3, 9S% CI 1.1-1.6), and CSs (1.1, 95% CI 1.06-1.16) compared with women carrying female fetuses.  The effect mechanism of sex on preeclampsia has been discussed by Shiozaki et al. 29 Three types of pathogenesis are hypothesized for the risk of preeclampsia in male-male and female-female twins: fetal histocompatibility antigen (HY) theory, increased immune-incompatibility between mother and fetus, and the increased level of fetal antigen in twin pregnancies. These hypotheses are explained following a description of the study.   A Perinatal Research Network in Japan consisting of 125 tertiary perinatal centers contributed data between 2001 and 2005.29  There were 241672 singleton and 20050 twin births.  Twins were divided into subgroups consisting of monochorionic diamniotic and dichorionic diamniotic twins.  Monochorionic diamniotic twins were further categorized into female-female and male-male twins while dichorionic diamniotic twins were divided into three groups: male-female, male-male and female-female twins.  Three theories were studied.  45  Fetal histocompatibility antigen theory  The first hypothesis suggested that increased levels of fetal HY antigen, a male tissue specific antigen, lead to the pathogenesis of preeclampsia, such that the higher the number of male fetuses, the higher the HY antigen. In turn, increased HY antigen will raise the incidence of preeclampsia. If this were the case, the male-male twins should have the highest preeclampsia rate, the female-female twins would present the lowest rate, and the male-female twins are expected to have an intermediate rate. Unlike what was predicted, the study found that the female-female twins had the highest preeclampsia rate (5.2%), male-female (opposite-sex) twins had an intermediate rate (4.6%), and the male-male twins had the lowest rate (3.9%). Thus, the results did not support the fetal histocompatibility antigen theory.  Increased immune-incompatibility between mother and fetus  The second theory speculated that immune-incompatibility between mother and fetus contributes to the pathogenesis of preeclampsia.  Immune-incompatibility between mother and fetus will be higher in twin pregnancies with two sets of different genes (dizygotic twins) compared to monozygotic pregnancies carrying twins with a similar pool of genes.  If this theory was true, the incidence of preeclampsia should be higher in dichorionic diamniotic twins than in monochorionic diamniotic twins because all the monochorionic diamniotic twins are derived from monozygotic eggs (with an identical gene pool) and 80-90% of the dichorionic diamniotic twins are derived from dizygotic twins (with two sets of different gene pools).  In other words, having two sets of different gene pools would trigger more immune-incompatibility between the mother and the fetus than two sets of similar gene pools. The study demonstrated that the incidence of preeclampsia was similar in both types of twins (dichorionic diamniotic twins, 4.5% 46  vs. monochorionic diamniotic twins, 4.2%, p> 0.05). Therefore, the analysis was not substantiated.  Theory of maternal exposure to different amounts of fetal antigens  The third theory, also called fetal antigen theory, proposed that an increased amount of fetal antigen leads to the pathogenesis of preeclampsia. If this were the case, the incidence of preeclampsia should be twice as high in twins compared to singleton pregnancies. The study found that the occurrence of preeclampsia increased by 1.5-1.6 times compared with singleton pregnancies. Therefore, the fetal antigen theory was supported. Hyperemesis The influence of fetal sex on hyperemesis was explored by Verberg et al.103 who suggested that female sex and twinning are associated with higher Human Chorionic Gonadotropin (hCG) levels,30 often stated as the most likely cause of hyperemesis. It was argued that the highest incidences of hyperemesis occur when hCG is at its peak level (8 to 11 weeks of gestation). Hyperemesis has a higher incidence in conditions considered to be associated with elevated hCG levels, i.e., molar pregnancies and twinning.103  In a prospective study in Bergen, Norway, Steier et al.30 compared 57 women with twin pregnancies with 66 (29 males and 37 female) singleton pregnancies. The twins were split into three groups: male-male, male-female, and female-female, each group containing 19 cases.  hCG was measured from maternal venous blood 2-10 minutes after delivery. a. In singleton pregnancies, hCG level was significantly higher in maternal blood if the fetus was female versus male (12,000 IU/Liter vs. 8000 IU/Liter; p<0.001). 47  b. In twin pregnancies, concentration of hCG in maternal blood at delivery was almost twice as high in female-female twins (28000 IU/Liter; p<0.001) and in male-female twins (27500 IU/Liter; p<0.005) in comparison with male-male twins (130000 IU/Liter).  c. The ratio of “maternal hCG/placental weight” was highest in twin pregnancies when one or both infants were female (p<0.001).  The findings indicate a sex-related fetal influence on hCG concentration in maternal serum. This will in turn lead to hyperemesis. It is important to obtain information about the mechanism/physiological basis under which fetal sex, especially in twins, influences hCG during pregnancy via further research. To the best of my knowledge, the mechanism of effect on the impact of sex discordance on maternal outcomes such as diabetes, premature rupture of membrane, prolonged rupture of membrane, preterm birth and CS has not been discussed in the literature. 2.2.1 ADVERSE OUTCOMES 2.2.1.1 Adverse perinatal outcomes As stated earlier, there are conflicting findings in the literature pertaining to the impact of sex discordance on adverse perinatal outcomes.  Male death among male-male concordant twins was found to be higher than 31 or identical32 to male mortality among male-female twins. In contrast, higher fetal death rates have been observed for males within male-female twins compared with male-male with33 or without32 adjusting for gestational age. Fetal and neonatal morbidity outcomes related to sex discordance include prematurity and its complications (respiratory distress syndrome, pneumothorax, and bronchopulmonary dysplasia).31,91 Higher morbidities among sex discordant twins could be due to the fact that discordant sex twins are almost always 48  dichorionic, where as same sex twins could be monochorionic, thus, have more complications related to chorionicity rather than sex discordance. Current dissertation sets to examine this hypothesis.   In the following sections, the impact of twin sex discordance on perinatal and maternal outcomes is reviewed in the context of the following sex-pairing categories:   a. Males versus females within the sex discordant twin set b. Sex discordant twins versus concordant twins (male-male or female-female)  c. Males in sex discordant twins versus males in concordant twins  d. Females in sex discordant twins versus females in concordant twins e. Males from concordant twins versus females from concordant twins Males vs. females within the sex discordant twin set Donaldson et al. 33 collected data from the obstetrical departments of 12 US teaching hospitals from 1950 to 1960. Neonatal mortality in 796 sets of sex discordant twins was compared between males and females.  Higher fetal death rates for the males were observed (10.05% males vs. 6.53% females, p<0.01).  A problem with this study was that males were designated with a specific punch code while females were designated by the absence of a code.  The authors regarded this as a limitation because if the sex was not recorded, it might have been analyzed as ‘female’. Similarly, the incidence of mortality was compared between male versus female within the sex discordant twin pairs by Shinwell et al.91  This study was done in Israel among 2448 very low birth weight (500 to 1500 grams) infants of 24-34 weeks of twin gestation. There were 478 pairs of sex discordant twins.  Once adjustment for gestational age was made, male sex was associated 49  with increased odds for mortality (1.36, 95%CI 1.05-1.77) in comparison with females. No other study with a larger sample size has compared other perinatal morbidities between male and female discordant twins. Sex discordant twins vs. concordant twins  Rydhstroem et al.104 compared fetal mortality in Sweden between 1973 and 1996 in relation to gestational duration. The data was collected from the Medical Birth Registry, National Board of Health and Welfare, Stockholm.  This large database included infant mortality up to 1 year of age for 52658 twins. The data were stratified for parity, maternal age, year of delivery and delivery unit.  Fetal death was defined as a stillborn of at least 28 completed gestational weeks.  Sex concordant twins had a significantly higher OR for fetal death than sex discordant twins (4.3, 95%CI 3.8-5.1 vs. 2.5, 95%CI 1.8-3.6) when compared with singleton pregnancies. This study was limited by lack of chorionicity data, as well as incomplete ascertainment of twins, in that data was missing for 3-4% of twin pregnancies according to the national health statistics records. Gestational duration was only confirmed by ultrasound after 1980 for 98% of pregnancies. Thus, between 1973 to 1980 (7 years) gestational age was determined by clinical judgment only, not verified by ultrasound screening.  Another retrospective study of all dichorionic twin pregnancies was accomplished at the Robin Medical Center in Israel between 1995 and 2006.31 The sample included 2704 twin pregnancies out of which there were 1878 sex discordant twins and 826 concordant twins (including 436 female-female pregnancies and 390 male-male pregnancies). The risk of preterm delivery at less than 31 weeks of gestation was highest in the male-male group (9.2%), intermediate in the male-female group (7.5%), and lowest in the female-female group (5.5%).  Taking female-female as 50  the reference group, the OR for prematurity was found to be 1.7 (95%CI 1.2-2.6) for the male-male group and 1.4 (95%CI 1.1-1.9) for the male-female group.   These findings suggest that the presence of a male in sex discordant pregnancies was associated with an increased risk of prematurity compared with female-female pregnancies. The authors believe that this could be due to longer gestational age in female-female pregnancies. The main limitation of this study, as recognised by the authors, is exclusion of a large (but unspecified) number of pregnancies from the study due to insufficient data on zygosity of twins.  Similar to all previous hospital-based studies, and unlike population based studies, this study lacks generalizability. It is worth noting that unlike the study of Rydhstroem et al., in which singleton pregnancy was used as the reference group, this study used female-female concordant twins. Cooperstock et al.88 also compared the risk of preterm birth in sex-discordant twins (n=2255) and concordant twins (n=5851: 2935 female-female, 2916 male-male). The investigation was conducted in Missouri between 1978 and 1990.  The risk of preterm birth before 35 weeks of gestation was highest in the male-male group (20.2%), lowest in the female-female group (15.7%), and intermediate in sex discordant twins (17.9%, p<0.001).  This study like previous ones did not take chorionicity into consideration but noted that importance of such knowledge when analysing this type of data. Males in sex discordant twins vs. males in concordant twins  The investigation by Shinwell et al.91 on very low birth weight preterm twins showed that the male mortality rate was identical in discordant and concordant (male-male) twin pairs (21.6%).  Similar rates were observed for retinopathy of prematurity (7.1% vs. 7.9%, p>0.05) and IVH (14.4% vs. 14.6%, p>0.05) when compared with males in discordant twin sets versus males in 51  concordant twins. However, discordant twins had lower risks for pneumothorax (9.7% vs. 11.7%, p<0.05), periventricular leukomalacia (6.8% vs. 9.8%, p<0.01), respiratory distress syndrome (73.6% vs. 76.6%, p=0.007), and bronchopulmonary dysplasia (11.9% vs. 13.7%, p=0.04). Females in sex discordant twins vs. females in concordant twins  Shinwell et al.91 also reported a marginally higher female mortality rate in sex discordant twins in comparison with that of concordant twin pairs (18.2% vs. 16.1%).  No significant differences were found between the two groups for any other measured outcomes. Males from concordant twins vs. females from concordant twins Yet again, in the Shinwell et al.91 study, males from concordant (male-male) twin pairs had a higher rate of respiratory distress syndrome (1.6, 95%CI 1.1-2.3), pneumothorax (1.9, 95%CI 1.3-2.9), and bronchopulmonary dysplasia (1.6, 95%CI 1.0-2.5) in comparison with concordant females (female-female).   2.2.1.2 Adverse maternal outcomes The literature on adverse maternal outcomes with respect to sex discordance is rare. Publications are available for two categories:  Singleton pregnancies versus sex concordant and discordant multiple pregnancies.  Male-male vs. female-female in concordant twins  Singleton pregnancies versus sex concordant and discordant multiple pregnancies A population-based study34 obtaining data from the Danish National Birth Registry from 1980 to 1996 studied the impact of sex discordance on maternal outcome.  Considering singleton healthy 52  pregnancies with males as the reference group (n=36674), twin pregnancies had higher risks of hyperemesis: female-female twins (2.4, 95%CI 1.8-3.2), male-female twins (1.8, 95%CI 1.8-3.4). The data suggested that the presence of more than one female in the twin pair increased the risk of hyperemesis. The same author also investigated an association between sex discordance and preeclampsia. The risk of preeclampsia was also significantly associated with sex discordance: male-male twins (2.5, 95%CI 2.1-2.9), female-female twin pairs (2.5, 95%CI 2.1-2.9), and male-female twin pairs (2.2, 95%CI 1.9-2.7). Similarly, another study found that the rates of both gestational hypertension and preeclampsia were significantly higher among women with twin gestations than among those with singleton gestations (2.0, 95%CI 1.6-2.6 and 2.6, 95%CI 2.0-3.4, respectively).28 The study population consisted of healthy women with either singleton (n=2946) or twin gestations (n=684) who were enrolled in two separate multicenter trials of low-dose aspirin for prevention of preeclampsia.  The Japanese population-based study29 described earlier also found that the incidence of preeclamptic mothers with a female fetus was significantly higher than those with a male fetus (3.8% vs. 3.2%. p<0.0001) in singleton pregnancies.   Male-male vs. female-female in concordant twins In the same Japanese study, comparison of male-male with female-female twins in sex concordant twins yielded a significantly higher risk of preeclampsia among mothers carrying female-female twins (7.6% vs. 3.6%, p=0.0005). Pre-existing hypertension was not taken into consideration as a potential confounder. Within the scope of literature relating to sex discordance and adverse maternal outcomes, no studies to date have comprehensively assessed potential confounding factors such as history of 53  chronic hypertension, diabetes mellitus, and maternal age. Furthermore, the impact of fetal sex on maternal outcomes in several combinations of sex pairing has not been investigated.   2.3 STUDY SIGNIFICANCE In order to overcome the above-mentioned shortcomings in the literature, the present thesis examines whether or not birth weight and sex discordance are predictive of fetal, neonatal and maternal outcomes. Current dissertation aims at analysing a large population-based database of 10 years’ worth of twin data, controlling for all the confounding variables noted in the literature to this date. These are as inclusive of gestational age, IUGR, twin size, infant sex, sex discordance, maternal age, chorionicity, number of prenatal visits and preterm birth. Adding to this list, our study adjusted the data according to the following variables: history of stillbirth, congenital anomalies, preterm birth, or low birth weight in the association between birth weight/sex discordance and perinatal outcomes. The available studies also lack assessments of mother-related confounding factors such as history of CS, maternal weight gain during the current pregnancy, number of antenatal care visits, and chronic hypertension. Access to information on placenta pathology for a subgroup makes it possible to investigate the potential role of placenta numbers, pathological characteristics of placenta and cord in relation to both BWD and sex discordance. The clinical significance of actual twin birth weight differences is controversial. Current research explores the outcomes associated with BWD. This, in turn, will allow discussion on the clinical twin birth weight differences.   Furthermore, the threshold for significant intra-pair BWD is also disputed. The current investigation will specify the BWD threshold at which a significant difference in perinatal and 54  maternal outcomes are more likely to occur. The threshold will help timely antenatal detection of adverse outcomes through serial ultrasound, which is needed for adequate antenatal surveillance and timely delivery.  Fear of fetal death often prompts clinicians to intervene by performing cesarean section, leading to complications of prematurity. The findings could be utilised to improve or revise the clinical guidelines on timing of cesarean section and serial ultrasound screening. The ultrasound scanning could focus on determination of fetal sex, which if substantiated as an important indicator for adverse outcomes can be of great clinical value.  Association between BWD and adverse perinatal and maternal outcomes has long led to iatrogenic premature pregnancy terminations, prematurity-related complications, and greater prevalence of cesarean section.16 The decision to perform cesarean section is based on the existence of BWD alone and in the absence of other indications is not justified if BWD has no significant impact on the adverse outcomes.       55  CHAPTER 3 METHODS 3.1 OVERVIEW OF RESEARCH A retrospective population-based cohort study was designed to address the study objectives, using data related to twin pregnancies registered in British Columbia (BC) between April 2000 and December 2010. This study period was selected because full provincial data collection of Perinatal Services BC commenced on April 1st, 2000.  De-identified data were abstracted from perinatal data collection. Information on the explanatory variables (fetal, neonatal and maternal outcomes) was collected from the mother suite, mother postpartum suite, baby newborn suite and baby transfer readmission. Twin pregnancies were included in the data if mothers and babies were linked, babies had calculated estimated gestational age at birth ≥20 weeks, no baby had a termination code, and the mother did not have a termination procedure.  3.2 ETHICAL APPROVAL This research was approved by the University of British Columbia (UBC) Research Ethics Board (certificate number H11-03281). On July 30, 2014, an amendment to the study was approved by UBC-Children’s and Women’s Health centre of British Columbia Research Ethics Board (UBC C&W REB) to collect chorionicity data (certificate number: CW14-0220/H11-03281). 3.3 COHORT DEFINITION The study cohort comprised mothers with twin pregnancies with greater or equal to 20 weeks of gestation. The units of analysis were mothers when comparing maternal outcomes, and twins when comparing fetal and neonatal outcomes.   56  3.4 DATA SOURCES The British Columbia Perinatal Database Registry is a population-based registry which contains information on approximately 99% of births in the province of British Columbia.  Data is collected from over 60 hospitals as well as births occurring at home attended by registered midwives.1 Annually, over 45000 births are recorded. Current study includes provincial data from 1st April 2000 to 31st December 2010.105 The information in the registry is compiled from standardized forms completed by clinicians. The validity of the data is ensured by quality control measures including built-in warnings in the data entry software and periodic data screening.106 According to PSBC, the highest frequency of twin births belonged to C&W hospital. Table 3-1 shows the number of twin births delivered in hospitals across the province from 1st April 2000 to December 31st 2010. .                                                  1 Perinatal Services BC (2011): British Columbia Perinatal Data Registry. Population Data BC. Data Extract. PSBC (2014). http://www.perinatalservicesbc.ca/health-professionals/data-surveillance/perinatal-data-registry   57  Table 3-1 Frequency of twin deliveries at BC hospitals, April 2000 to December 2010    Total Twin Deliveries† by Calendar Year of Baby Date of Birth Institution Name Institution ID 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Total 100 Mile District General Hospital 708 0 0 0 0 0 <5 0 0 0 0 0 <5 Abbotsford Regional Hospital & Cancer Centre 609 0 0 0 0 0 0 0 0 12 36 34 82 BC Women's Hospital & Health Centre 104 130 156 168 150 144 168 161 183 154 184 179 1,777 Bulkley Valley District Hospital 903 7 5 <5 5 <5 5 <5 <5 0 <5 0 47 Burnaby Hospital 130 13 14 14 12 12 13 12 21 20 25 15 171 Campbell River Hospital 508 <5 <5 <5 <5 <5 <5 <5 <5 <5 <5 <5 <55 Cariboo Memorial Hospital 406 <5 6 8 9 7 5 <5 9 <5 6 <5 <65 Chilliwack General Hospital 601 10 5 12 11 14 13 16 14 12 10 <5 <122 Cowichan District Hospital 203 <5 <5 <5 0 <5 <5 6 6 <5 8 6 <56 Dawson Creek and District Hospital 704 <5 <5 <5 <5 0 0 <5 7 <5 5 <5 42 East Kootenay Regional Hospital 756 6 8 <5 <5 <5 10 7 <5 6 6 11 <74 Elk Valley Hospital 753 0 0 0 0 0 0 0 0 0 0 0 0 Fort St. John Hospital 701 9 5 5 6 6 <5 <5 <5 <5 5 6 <62 GR Baker Memorial Hospital 705 <5 0 <5 <5 <5 0 0 0 0 0 0 <20 Golden & District General Hospital 409 0 0 0 <5 0 <5 <5 <5 0 0 0 <20 Kelowna General Hospital 302 18 12 20 20 21 22 26 24 27 20 21 231 Kootenay Boundary Regional Hospital 801 8 <5 <5 <5 <5 5 5 0 <5 5 7 <60 Kootenay Lake Hospital 651 0 <5 <5 <5 <5 <5 <5 5 <5 5 <5 <50 Langley Memorial Hospital 115 12 17 11 19 17 25 23 23 21 21 17 206 Lions Gate Hospital 112 20 19 21 26 19 24 30 28 19 21 22 249 Matsqui-Sumas-Abbotsford General Hospital 603 11 21 19 16 16 25 23 21 14 0 0 166 Mills Memorial Hospital 912 <5 7 <5 <5 <5 <5 5 <5 7 <5 5 <59 Nanaimo Regional General Hospital 501 15 9 23 18 18 19 26 22 21 23 24 218 Peace Arch Hospital 131 6 9 8 8 <5 8 10 8 0 <5 10 <77 Penticton Regional Hospital 303 7 6 9 9 9 7 13 9 7 5 8 89 Port Hardy Hospital 510 0 0 <5 0 0 0 0 0 0 0 0 <5  58  Institution Name Institution ID 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Total Powell River General Hospital 111 0 <5 <5 0 <5 <5 <5 <5 <5 <5 <5 <45 Prince Rupert Regional Hospital 902 5 <5 <5 <5 <5 <5 <5 <5 <5 <5 <5 <55 Queen Victoria Hospital 402 0 0 0 0 0 0 0 0 0 0 <5 <5 Richmond Hospital 121 8 8 7 10 10 8 9 12 15 10 16 113 Ridge Meadows Hospital 604 <5 8 <5 7 <5 6 8 13 9 7 6 <79 Royal Columbian Hospital 109 55 47 63 58 71 63 55 80 64 71 78 705 Royal Inland Hospital 401 10 19 8 17 18 20 28 25 24 24 21 214 Shuswap Lake General Hospital 404 <5 0 0 0 0 <5 0 0 0 0 0 <10 Squamish General Hospital 128 0 0 0 0 <5 0 0 0 0 0 <5 <10 St. John Hospital 702 <5 <5 <5 <5 <5 0 0 0 0 <5 <5 <35 St. Joseph's General Hospital 502 7 9 <5 5 9 <5 <5 8 9 7 6 <75 St. Mary's Hospital 113 0 0 0 <5 <5 0 <5 0 <5 <5 <5 <30 St. Paul's Hospital 102 13 13 27 21 28 16 22 21 24 25 23 233 Stuart Lake Hospital 717 0 0 0 0 0 0 0 0 0 0 0 0 Surrey Memorial Hospital 116 36 42 51 44 33 49 48 52 45 57 70 527 University Hospital of Northern British Columbia 703 9 14 24 18 26 17 24 29 23 27 24 235 Vernon Jubilee Hospital 301 7 13 7 15 10 16 12 8 11 19 10 128 Victoria General Hospital 202 42 38 47 59 44 58 47 48 60 63 58 564 West Coast General Hospital 851 0 <5 <5 0 <5 <5 <5 <5 <5 <5 0 <40 Wrinch Memorial Hospital 901 0 <5 0 0 <5 0 0 0 0 0 0 <10 Total British Columbia   489 530 585 585 569 635 646 696 641 713 704 6793     59  C&W hospital by far has the highest frequency of twin births. Pathology data was collected from this hospital.  Strengths of the Perinatal Service British Columbia data  The data for the present investigation was obtained by PSBC. Some strengths of this data are as follows: Province-wide: PSBC-data is a province-wide perinatal database completed for the purpose of evaluating perinatal outcomes, care processes and resources to improve maternal, fetal, and newborn care. The capture rate is 99% of births in the province of British Columbia. Comprehensive: PSBC-data contains comprehensive information on demographic, clinical, and some risk factor data for mothers, readmitted mothers, newborns, stillbirths, and readmitted neonates. The following explains why the data is comprehensive.  Data is routinely collected from obstetrical facilities as well as from births occurring at home attended by BC registered midwives. The collected data from across BC are then imported into the central Perinatal Data Registry. Centrally-installed software in hospitals is utilized to send data periodically to the provincial database. The Canadian Institute for Health Information and British Columbia Vital Statistics Agency complement the data elements.  The scope of hospital data collection includes a mother suite, postpartum suite, baby newborn suite and baby transfer/readmission suite. Data collection starts by filling in the delivery episode. One mother suite is completed per pregnancy.  One postpartum suite for each surgical day care or inpatient admission is filled after the mother suite is completed. Similarly, one newborn suite is completed for each newborn or stillborn, followed by one baby transfer/readmission suite for each admission. Twin data contains one mother suite and two newborn/stillborn suites. The  60  information on time of birth and time of placenta expulsion is different between twins. Therefore, the data is entered twice: once for each twin. Twins are named as “Twin A” and “Twin B” based on birth order. Each baby will be given an identifier number, linking the baby to the mother and his/her twin.  Standard tool:  Strength of PSBC-data is the use of a standard tool for data collection. First, the standard software validates the data in that it identifies specific errors or omissions, if any, in the mother record. Once the validation is successful, the data is linked to a previously entered newborn or baby transfer/readmission record. Each step of data entry is checked for field completion before the coder proceeds to the next record. The admission number assigned by the facility to identify the specific “mother admission” is used for linkage purposes.  Guideline for coders: A detailed definition of terms is provided as a guideline for coders to ensure quality.107 Free access for UBC graduate students: The BC Perinatal Data registry goals are to aggregate and report on perinatal events, care processes, and outcomes at the provincial, regional and community levels. A part of their mandate is to support perinatal health services research aimed at improving delivery of patient care by providing researchers with access to information from a wealth of information at their disposal.  The current thesis is gratefully benefiting from this free access. The PSBC-data as it was made available for the present investigation, however, suffers from a major limitation discussed in the following section. Limitation of PSBC-data  The collection of registry data is not planned for any specific study objective. Therefore, the data lacks some properties significant for health researchers. For instance, chorionicity information is  61  not collected by PSBC, although chorionicity determination in twin gestation is a vital aspect of risk evaluation for twin pregnancies (see Chapters 2 and 4 for the importance of chorionicity).  Although PSBC has been active in determining clinically relevant data, the collection of chorionicity information is not considered a priority yet.  According to PSBC procedures, justification has to be provided for adding clinically relevant information to the current standard forms.  In September 2011, the BC Patient Safety and Quality Council produced a report on medical imaging that made a recommendation to develop or adopt and promulgate standards for obstetrical ultrasound assessments performed in community and tertiary facilities.108 PSBC coordinated the development of standards outlining the minimum requirement of an obstetric ultrasound assessment. Information on chorionicity is collected by sonographers based on the recommended guidelines.  However, the data has not been systematically processed by PSBC. Additionally, ICPD codes pertaining to chorionicity are not activated yet. Chorionicity information is also available in pathology along with other important micro- and macro-pathological findings of placenta and cord. None of these are collected by PSBC.  It is also not possible to study small number of monozygotic twins who are dichorionic in nature using PSBC data as zygosity and chorionicity information is missing from PSBC dataset. Furthermore, PSBC can benefit from medical genetic information obtained from single birth babies suffering from major chromosomal abnormalities.  Due to the above-mentioned limitation, chorionicity information could not be accessed via the PSBC database. An inquiry was made on June 2014 to PSBC to obtain the number of twin pregnancies by year in all the hospitals at BC. It was found that C&W hospital has the highest  62  frequency of twin births. Chorionicity data was available in the pathology laboratory. Data on chorionicity, along with other pathological data, were collected and analysed from C&W hospital from 1st April 2000 to 31st December 2010. To compensate for the lack of chorionicity in the whole database, sex discordance was used as a proxy measure of chorionicity. A comparison between the overall data and the subsample is made at the end of each chapter when presenting the fetal, neonatal and maternal outcomes.  Justification for collection of chorionicity information  Current investigation will provide scientific evidence for the necessity of chorionicity examination at the delivery room and collection of such data in order to assess the risks for twin gestations.  British Columbia Health Authorities and Hospitals provide supplementary screens entitled “institution fields” to collect data which are not available on files to be included in the PSBC database.  These fields are considered under jurisdiction of Health Authorities and Hospitals and might be used for historical purposes. However, the Health Authority does not guarantee the data quality of these fields. For a new field to be activated and supported by PSBC for quality, a request is to be made formally to PSBC. Once the new field is formally approved by PSBC, these fields will be activated as required and will be mandatory to complete. If chorionicity data is to be added to PSBC, such a request has to be justified. The results of this dissertation might provide evidence that warrant collection of such information in future on the recommendation of  other institutions such as the Society of Obstetricians and Gynaecologists of Canada.    63  Data provided by PSBC The data used in this research was provided in three text-formatted files including de-identified information for mothers of twins, birth-episode information for twin infants in the cohort, and information on labour and delivery (Table 3-2). The latter file included a maternal and fetal identification code which was used to link the three files.  The data dictionary is attached as  Appendix I.  Table 3-2 Pertinent data from three sources  Mothers of Twins Twin Infants Labour and Delivery Study ID for mother Study ID for twins Study ID for mother Study ID for twins Study ID for mother  The chorionicity data was added for a sub-sample of 3182 twins (1591 twin gestations). This data was obtained from the C&W hospital pathology department, linked by PSBC and stored in a secure research environment at Population Data BC. The pathology data included the following variables: number of chorions, number of amnions, placenta weight (grams), placenta length, width and thickness (cm), disk division percentage, cord length (cm), cord distance from margin, membrane, and other cord (cm), chorionic villi inflammation, chorionitis, placental anastomosis, placenta shape and cord insertion type. A subgroup analysis was done using this data and the outcome variables. Initial data management The data files were initially read into Excel for checking and recoding. Adverse perinatal and maternal outcomes, coded with ICD codes 9 or 10, were combined. Variables related to  64  congenital anomalies were combined to create a single variable entitled “any congenital anomalies”.  A variable was created by taking the difference between the birth weight of the larger twin and the smaller twin, divided by larger twin weight and multiplied by 100; see Equation 1:  Equation 1: 𝐿𝑎𝑟𝑔𝑒𝑟 𝑡𝑤𝑖𝑛 𝑤𝑒𝑖𝑔ℎ𝑡−𝑆𝑚𝑎𝑙𝑙𝑒𝑟 𝑡𝑤𝑖𝑛 𝑤𝑒𝑖𝑔ℎ𝑡𝐿𝑎𝑟𝑔𝑒𝑟 𝑡𝑤𝑖𝑛 𝑤𝑒𝑖𝑔ℎ𝑡× 100    This variable was named “birth weight discordance” or BWD.  Another variable entitled “sex discordance” was created using the sex variables for twin 1 and twin 2. Twins were named discordant if one twin was female while the other was male (male-female). Twins were considered concordant for sex if both twins were males (male-male) or females (female-female). Sex concordant twins were further categorized to male-male and female-female twins. Finally, a new variable was created with three categories including the following codes: FF for female-female, MM for male-male and MF for male-female. Each individual was also identified in a separate variable entitled ‘sex-pairing variable’ where the sex of an individual twin in relation to the pair she/he belongs to was identified. Four categories were defined as follows: female of a female-female pair, male of a male-male pair, female of a male-female pair and male of a male-female pair. Data linkage PABC linked mothers’ records to birth-episode information for twin infants using the mother’s and twins’ unique identifier. These data were also linked to information on labour and delivery. BWD and sex-discordant variables were added to the database. The data were then imported to  65  SPSS version 22.0 (Chicago, IL) and used in the process of estimating the impact of BWD and sex-discordance on fetal, neonatal and maternal outcomes.   For a sub-sample of twins born at C&W hospital, data on chorionicity was obtained and linked using Patient Health Number (PHN), maternal and baby’s ID and date of birth.  This task was done by PSBC by a hired data steward so that de-identified data would remain anonymous to the researcher.  3.5 DEFINITIONS The outcomes of interest were categorised as fetal, neonatal and maternal outcomes as follows:  a. Fetal: stillbirth b.  Neonatal: early and late neonatal death, congenital anomalies, Apgar score of five minutes less than 7, length of stay in NICU more than 2 days, newborn septicaemia, hypoglycaemia, anemia, IVH, pneumothorax, respiratory distress syndrome, retinopathy  c.  Maternal: preeclampsia, PIH, preterm labour, premature rupture of membrane, cesarean section, gestational diabetes, length of stay in hospital more than 3 days, prolonged rupture of membrane  Definitions used to record the outcome variables were adopted from BC Perinatal Services Data Registry Reference manual105 and explained below.  Stillbirth was defined as the complete expulsion or extraction of fetus from the mother at least 20 weeks’ pregnancy or after attaining a weight of at least 500 grams, of a product of conception in which, after the expulsion or extraction, there is no breathing, beating of heart, pulsation of the umbilical cord or unmistakeable movement of voluntary muscle.  The variable contained four  66  categories: fetal death after onset of labour, prior to onset of labour, unknown timing and death after live birth.  Early neonatal death was defined as death of the baby within seven days of birth, while neonatal death was attributed to babies’ death within 28 days of birth.  Both of these variables included inpatient death and death identified through linkage with Vital Statistics. Major congenital anomalies refer to any structural or functional anomalies, displayed in the fetus or the baby that renders serious medical, surgical or cosmetic consequences and may have required intervention. The anomalies considered for the current study include anomalies of nervous, respiratory, circulatory and cardiovascular systems, anomalies of heart, and chromosomal abnormalities. International classification of disease (ICD9) codes 740.0 to 759.9 and ICD10 codes Q00.0 to Q99.9 were used to record hospital discharge abstracts as well as hospital readmissions and transfers during the first day of life.  Apgar scores at 1st and 5th minute after birth were provided as continuous variables which were recoded as Apgar score of “less than 7 (out of 10)” or “7 or more”. The score was defined as the total score between 0 to 10, based on each of five criteria (skin colour, heart rate, reflex irritability, muscle tone, respiration) on a scale of 0 to 2, recoded at 1st and/or 5th minutes of birth. NICU days were inclusive of the total number (in whole numbers) of days the baby was in NICU level 2 and 3, characterized by high-dependency neonatal care, such as total parenteral nutrition, mechanical ventilation, major surgery or extracorporeal membrane oxygenation.109  A new variable was created as length of stay in NICU more than 2 days. Short-term morbidities and mortalities are linked with use of NICU resources for longer than 2 days.110 NICU stay was  67  defined as those who stayed longer than 2 days based on the median duration of stay in Canadian NICUs.  Newborn septicaemia was defined as bacterial infection of the blood in a neonate younger than four weeks of age. Newborn hypoglycaemia or syndrome of “infant of a diabetic mother” was defined as a blood glucose concentration ≤ 2.6 mmol/L.111 Newborn anemia referred to a hematocrit level of less than 45% in a term infant. A decreasing hematocrit, reticulocytopenia, bone marrow hypoplasia, and endogenous erythropoietin concentrations are inappropriately low relative to the degree of anemia.112 Intraventricular haemorrhage (IVH) grade 1 is defined as intraventricular haemorrhage in 10% of ventricular space while IVH grade 2 is defined as intraventricular haemorrhage in 10-50% of ventricular space. IVH grade 3 is defined as IVH in more than 50% of ventricular space; IVH grade 4 includes parenchymal bleeding.12 We analysed this data collectively. Pneumothorax is considered to be an accumulation of air or gas in the pleural space, which may occur spontaneously or as a result of trauma or a pathological process, or be introduced deliberately.32 Respiratory distress syndrome (RDS) is defined as the need for mechanical ventilation for more than 24 hours with supportive radiographic features.  Mild RDS is defined as the need for mechanical ventilation for less than 24 hours with supportive radiographic features.12   68  Retinopathy is defined as follows. Stage 1: Demarcation line separates avascular from vascularized retina; Stage 2: Ridge arising in region of demarcation line; Stage 3: Extra retinal fibrovascular proliferation /neovascularization; Stage 4: Partial retinal detachment; Stage 5: Total retinal detachment.113 For the purpose of analysis in this thesis, all stages were combined as stratified categories were only available for ICPD10 codes and not for ICPD 9 codes.  Pregnancy induced hypertension (PIH) is diagnosed by the physician when the mother is found to have gestational hypertension during the current pregnancy.  Preeclampsia is defined using two variables: PIH and proteinuria. If a mother is diagnosed by a physician to have both PIH and positive proteinuria (the presence of urinary protein in amounts exceeding concentrations more than 1g per litre or 1+ on a urine dipstick), she is diagnosed as preeclamptic.  Preterm labour is defined as a labour which occurs from 20 to 36 completed weeks gestation (140 days to 258 days), counting from the first day of the last menstrual period. Premature rupture of membrane is defined as rupture of the membrane of the amniotic sac and chorion more than one hour before the onset of labor. Cesarean section delivery is a surgical procedure to deliver the baby who is ≥20 completed weeks gestation (140 days and greater). Gestational diabetes mellitus is diagnosed by a physician during routine antenatal care using 3 hours’ glucose tolerance test. In these cases, diabetes has occurred during pregnancy, and the condition was not pre-existing.        69  Maternal length of stay in hospital more than 3 days is calculated by deducting the admission date from the discharge date. Prolonged rupture of membrane is calculated by PSBC from time and date of rupture of membrane and is defined as rupture of membrane longer than 18 hours.  Gestational age (GA) in completed weeks was calculated based on an algorithm considering last menstrual period (LMP), ultrasound, newborn exam, and information from the maternal chart. The algorithm adopted from the Society of Obstetricians and Gynaecologists of Canada (SOGC) guideline was as follows:  If LMP2 was recorded and there was no ultrasound (US), GA was calculated from LMP.  If LMP was recorded and there was no early ultrasound, but a clinical exam of baby gave a GA at least 3 weeks different than LMP, GA from the clinical exam was used.  If LMP was recorded and was equal to GA in weeks from US results at <14 weeks, GA was calculated from LMP. If estimates were not equal, GA was estimated from US findings.  If LMP was recorded and within 1 week of GA in weeks from US at 14-20 weeks, GA was estimated from LMP. If the difference was more than 1 week, GA was calculated from the US report.  If LMP was recorded and within 2 weeks of GA in weeks from US at 21-24 weeks, GA was estimated from LMP. If the difference was more than 2 weeks, GA from the US report was used.                                                  2: Only LMP estimates between 15 and 45 weeks and ultrasound estimates between 17 and 43 weeks are considered valid for the purposes of the calculation  70   If LMP was not recorded but GA from US <25 weeks was recorded, GA from US was recorded.  If LMP and early US were not recorded, GA from the newborn clinical exam was recorded.  If LMP, early US, and newborn clinical exam were not recorded, GA from chart documentation was recorded.  If all the information were missing or out of range, GA was considered missing. Preterm is defined as babies born alive before 37 weeks of pregnancy are completed. There are sub-categories of preterm birth, based on gestational age:  extremely preterm (<28 weeks)  very preterm (28 to <32 weeks)  moderate to late preterm (32 to <37 weeks).114,115 Large for gestational age (LGA) is an indication of high prenatal growth rate. LGA is defined as a weight that lies above the 90th percentile for that gestational age using twin data. Small for gestational age (SGA) refers to an infant whose birth weight was below the 10th percentile for the appropriate gestational age.  Appropriate for gestational age (AGA) describes a fetus or newborn infant whose size is within the normal range for his or her gestational age. Pathology data was inclusive of chorionicity information; placenta weight, height, width and thickness; cord length; cord distance from the margin, membrane and other cord; type of cord insertion (marginal, central, velamentous); chorionic villi inflammation; chorionitis; placenta anastomosis; placenta shape (circular, oval, others). In a histological examination of the placenta, vascular-thrombotic lesions (infarction, chorangioma, subchorial fibrin deposition, and retro placental hematoma) were recorded. Arteries were identified as vessels that are  71  situated superficial in relation to the veins. A composite score was created for all of the pathological lesions for the purpose of analysis. Anastomosis between fetal vessels was defined as the presence of an impaired vessel from one twin feeding an area drained by the co-twin.  Injection studies were performed in fresh specimen to identify unidirectional arteriovenous shunt(s) between donor and recipient.  Placenta’s maternal and fetal inflammatory response corresponding to chorionitis, chorioamnioitis, and chorionic villi inflammation were also collected in a composite variable. Four composite scores were created for placenta and cord related outcomes inclusive of the following:  Composite of Placenta lesions: Placenta infarction, placenta abnormalities, chorangioma, hematoma and placenta morphological and functional anomalies  Composite of inflammation in utero: Chorioamnioitis, chorionic villi inflammation, chorionitis, membrane inflammation and other abnormalities of amnion and chorion  Composite of issues with cord: Number of cord vessels less than 3, compression of umbilical cord problems, and other unspecified conditions related to cord.  Placenta other: Placenta previa, placenta abruptio, placenta accreta, and other types of placenta separations that causes hemorrhage Chorionicity refers to the number of chorions.   Monochorionic twins share the same chorion or placenta and are therefore monozygous or identical. These twins are always of the same sex. Monochorionic twins can be either monochorionic-monoamniotic, which means that they share both a placenta and a single  72  amniotic sac, or mono-chorionic-diamniotic, which means that they share a placenta but there is a separating membrane between them, thus they are in two separate amniotic sacs.  Dichorionic twins develop within their own chorion. Thus, dichorionic twins are twins with two separate placentas. They can be either the result of a single egg splitting (monozygotic), or they can be the result of two separate fertilized eggs (dizygotic). Dichorionic twins always have two amniotic sacs (diamniotic). Equal placental sharing was defined as 40% to 60% of the placenta attributed to each twin. We chose this range because preliminary data revealed that twins with 40/60 sharing and 50/50 sharing had similar gestational ages at delivery and degree of BWD.116 Unequal placental sharing was defined as 1 twin receiving blood from more than 60% of the placenta. 3.6 ANALYSIS PLAN Statistical analysis was performed using SPSS software (IBM SPSS Statistics for Windows, Version 22.0, and released 2013. Armonk, NY: IBM Corp.). A p-value of <0.05 was considered statistically significant. Data analyses involved descriptive, bivariate and multivariate analysis.   Descriptive analysis  Distributions of birth characteristics were described in four categories:  1) Overall sample,  2) Within BWD categories based on the optimal threshold,    3) Within sex concordant/discordant twins, and 4) Within the subsample of twins born at C&W hospital, analysis of pathology information.   73  Variables described were inclusive of socio-demographic characteristics of the mother (maternal age, maternal education, drug and alcohol use during current pregnancy), parity, gestational age, sex of infant, sex discordance, birth weight, and smoking habits prior to, during and after pregnancy. In addition, the distributions of adverse perinatal and maternal outcomes were described in the same fashion for the overall data and for the subsample of data from C&W hospital stratified by chorionicity. Bivariate and multivariate analysis Next, the crude and adjusted odds ratios (95% CI) of fetal, neonatal, and maternal adverse outcomes were estimated for both exposures (birth weight and sex discordance) in separate models (for a long list of confounding variables discussed in the literature, see Chapter 1 and 2). For sex discordance, the odds were estimated for 5 different comparisons: males versus females within the sex discordant twin set, sex discordant twins versus concordant twins, males in sex discordant twins versus males in concordant twins, females in sex discordant twins versus females in concordant twins, and males from concordant twins versus females from concordant twins. These comparisons are presented according to significance and relevance. Factors that were tested for their role as confounders for fetal and neonatal mortality and morbidity were as follows: maternal age, chorionicity, history of stillbirth, congenital anomalies, preterm birth, or low birth weight, number of antenatal care visits during current pregnancy, weight gain during pregnancy, birth order, parity, infant sex and gestational age. Factors that were tested for their role as confounders for maternal morbidity included previous cesarean section and maternal medical history (chronic hypertension and pre-existing diabetes).   74  The independent associations among potential confounders, key explanatory variables, and outcome variables were reported as unadjusted odds ratios with 95% CI. Socio-demographic characteristics were tested and the statistically significant variables were retained. In sequential models, one potential confounding variable was tested at a time.  Retaining the most statistically significant variable, the remaining variables were entered separately in sequential models, and then the process was repeated until all variables had been tested. Adjusted odds ratios for the final model (and 95% CI) were reported.   Generalized estimating equation modelling Twins provide naturally matched pairs or clusters with within-twin-pair and between-pair effects. This type of data needs a specialized standard regression model that reflects the paired structure of the data, which induces correlation between twins. Generalized estimating equations (GEE) procedures that accommodate correlated clustered data were used in our data analysis to build up such models. GEE models include separate regression coefficients for within-twin-pair and between-pair effects. To describe the models used in the analysis, some notations are introduced here. Y denotes a binary outcome (e.g. still birth) and X the covariate of immediate interest (e.g. BWD), with i used to index twin pairs and j=1, 2 to index individual twins within pairs. Thus, Yij, Xij, represents the outcome and covariate value of the jth twin of the ith pair, respectively. The expected value of Yij, given the value of Xij, is denoted by E(Yij). An example of the data view in a sample database is shown below. i j X Y 1 1 24 0 1 2 36 1 2 1 16 1 2 2 14 1  75  Receiving operating characteristic analysis We used receiving operating characteristic (ROC) analysis as a tool to select the optimal threshold level for BWD beyond which adverse perinatal outcomes would be expected. Cut off values of BWD> 5%, >10%, >15%, >20%, >25%, >30% and >35% were tested to obtain the best threshold level. Following establishment of an optimal cut-off for BWD, the prospective risk of perinatal mortality and morbidity above and below the optimal cut-off was calculated. Positive likelihood ratio was used to estimate test accuracy. LR+ values >10 were considered to provide a “conclusive” increase in the likelihood of disease while scores 5-10 and <5  were categorized as “moderate” and “minimal” increases in likelihood of disease, respectively.117  Survival analysis Survival trends from 20 weeks’ gestation until delivery according to different degrees of BWD (≥30% and ≥35%) were assessed using time-to-event analysis (Kaplan–Meier), in which duration of gestation was used as the time scale and stillbirth/perinatal mortality as the event. Survival data were plotted as cumulative percentage without event and compared by the log-rank test. Finally, cox proportional hazard analysis was carried out in order to take confounders such as chorionicity and sex-discordance into account for the estimation of survival rate among those with and without growth discordance. Sample Size  About 40000 deliveries occur in British Columbia every year. The rate of twinning is increasing in Canada.118 It is estimated that there would be 12520 twin births from 2000-2010 delivered in British Columbia. The sample sizes required when Type I error = 0.05, Type II error = 0.2 (80% power), and hypothesis are two sided are given in    76  Table 3-3. The outcome that requires the largest sample is congenital anomalies. It was estimated that 1471 twin sets were needed for each comparison group to provide 80% power to detect a 60% relative difference from 0.05, based on the literature, to 0.08 in congenital anomalies at an alpha level of 0.05.  Chorionicity data was available for a sub-sample of population from C&W hospital. The estimated number of twin gestations delivered at the hospital from 2000-2010 was about 1800. In a twin gestation, the probability of having a dichorionic twin is 2/3 and the probability of having a monochorionic twin is 1/3.119 Thus, based on a ratio of (r=n1/n2 =dichorionic/monochorionic) 2/1, the adjusted unequal sample size for dichorionic twins was 900 for dichorionic twins and 450 for monochorionic twins.  Based on the literature120 for the monochorionic group, it was estimated that 102 twin sets are needed to obtain 80% power to detect 12.7% relative difference from 5.7 (BWD of <5%) to 18.4 (BWD of >20%). The corresponding sample size for the dichorionic group was 182 given the relative difference of 9.5% (from 7.2% in BWD of <5% to 16.7% in BWD of >20%).     77  Table 3-3 Literature used for estimation of sample size  Study Outcome of Interest Rate of outcome with less than 10% discordance Absolute difference in proportion that we want to detect Rate in comparison group Number of twin pairs inquired per group   Wen 23 Fetal death 0.13 40% 0.26 764  Neonatal death 0.23 40% 0.46 377 Cheung 63 Congenital anomalies 0.05 50% 0.08 1471  Apgar score 5’> 7 0.07 50% 0.14 1022 Mazhar 5 CS  0.21 5% 0.26 1128  PIH 0.12 40% 0.17 839 M vs. F within SD twins Outcome of Interest Rate of outcome for male  Absolute difference in proportion that we want to detect Rate in comparison group Number of twin pairs inquired per group   Donaldson 33 Infant death 0.1 50% 0.15 686 Shinwell91  CS 0.69 20% 0.83 150  Preterm birth 0.77 20% 0.92 85  RDS 0.75 20% 0.90 100  GH 0.03 100% 0.06 749 Melamed31 Preterm birth 0.07 50% 0.12 1022  Preeclampsia 0.07 100% 0.14 300 M in SD in C twins      Melamed31 NICU 0.29 20% 0.38 431 Shinwell91 Pneumothorax 0.12 40% 0.16 839  RDS 0.77 10% 0.85 409 F in SD in C twins      Melamed31 NICU 0.31 20% 0.37 917 Shinwell91 RDS 0.75 10% 0.83 466  Infant mortality 0.18 50% 0.27 337 M vs. F from C twins      Shinwell91 RDS 0.77 10% 0.85 377  Infant mortality 0.22 50% 0.33 258 M: Male;  F: Female;  SD: Sex discordant;  C: Concordant;  MM: male-male;  FF: female-female;  CS: Cesarean section;  PIH: Pregnancy induced hypertension;  GH: Gestational hypertension;  NICU: Neonatal intensive care unit,  RDS: respiratory distress syndrome Analytical exclusion and inclusion criteria   78  Given the estimated existing number of samples for mono- and dichorionic twins in C&W hospital, we have 99% power to obtain the differences between mono- and dichorionic twin gestations stated above. Twin pregnancies with 20 weeks of gestation and higher who could be linked with mothers were included in the study. All chapters have similar exclusion criteria except Chapter 8 (on stillbirth), and Chapters 11 and 12 (both about maternal outcomes). We excluded cases with previous cesarean section from the analytical data in Chapter 11 and 12 to control for the impact of repeated cesarean section. For Chapter 8 where we analysed stillbirth, we did not exclude cases with single stillbirth for the sake of clarity and comprehensiveness surrounding the fate of twins with stillbirth.  For all other chapters, the excluded cases were as follows: cases with congenital anomalies, TTTs, <500 grams’ birth weight, one stillbirth, those who had a reduction procedure in multiple pregnancies greater than two (3 to 2 or 4 to 2) and papyrus placentas/fetuses. We chose to exclude the latter category of vanishing twins from this analysis because of association between sex discordance and vanishing twins.121  Each methods section presented in Chapter 5 to 12 describes the exact account of excluded cases for both province-data and hospital-data.   79  CHAPTER 4 A SYSTEMATIC REVIEW OF BWD AND PERINATAL OUTCOMES  4.1 INTRODUCTION Twin pregnancies are known to be associated with fetal and newborn complications.11,23,44,122  These complications include, but are not limited to, stillbirth, intrauterine growth restriction, neonatal mortality, congenital anomalies, and preterm birth.5 Risk factors for adverse fetal and perinatal outcomes are linked to BWD. BWD reflects a disparity in birth weight between the larger and smaller infants of a twin set. This phenomenon is well-documented and occurs in as many as 10-29% of twin pregnancies.15,44,123,124   BWD is believed to stem from a number of factors, which include TTTs, twin anemia polycythemia syndrome (TAPS),  unequal placental sharing, abnormalities of umbilical cord insertion in MC pregnancies13 and discordance for fetal anomalies.   In DC twins, placental pathology, genetic and sex discordance along with discordance for major fetal anomalies, likely are the main causes of BWD.14 Apart from the intra-uterine environment, biological variation between the twins could also result in BWD. The biological variation is associated with low degree of discordance. A larger degree of discordance, however, may indicate a pathologic process associated with adverse fetal and neonatal outcomes. The literature suggests that BWD of higher than 15% to 25% is more likely to represent a pathological process, resulting in adverse perinatal outcomes.125  To date, there have been conflicting results within the literature over the extent of BWD association with perinatal outcomes.11 Investigations, whether on a small15,16,123 or large126 sample size, with retrospective16,19 or prospective5 study designs, and using hospital-19,63,126,127 or  80  population-based23 studies, have found conflicting results on perinatal mortality and morbidity rates associated with BWD.   4.2 OBJECTIVES The present systematic review has two objectives:   To investigate the incidence of BWD specific to twin gestation   To study the association between BWD and adverse fetal and neonatal outcomes 4.3 METHODS  The methodology follows the MOOSE Statement128 and is explained under six categories: Search strategy, databases, study selection, data extraction, quality assessment, and statistical analysis. Search strategy A comprehensive list of Mesh terms was attained by three means. First, the definition of each concept was extracted from MEDLINE. Second, grey literature, conference proceedings, and reference lists of published articles were explored. Third, specialists in the field were consulted to identify the Mesh terms. These included “birth weight discordance”, “birth weight discordant”, “twin”, “multiple pregnancy”, “pregnancy outcome”, “fetal outcome”, and “neonatal outcome”. Appendix II shows the search strategy.  Boolean logic was used to combine the concepts and eliminate irrelevant articles. Filters were employed to limit the search to observational studies only. Non-English literature (e.g. French, Germany) was also included.  Databases The following databases were searched, from the earliest available date (mentioned in brackets) to July 2015: MEDLINE (1996), Cochrane Central Register of Controlled Trials (1991),  81  EMBASE (1980), CINAHL (1982), Database of abstracts of reviews of effects (1991), Web of Science (1990), and EBSCO (1946). Proceedings and conference abstracts were also explored through Paper First (1993), Pro Quest Digital Database (1993), and Proceeding First (1993) databases. The references of the retrieved articles were checked to find new articles.  Study selection Articles were included in the review if they were observational studies, studies performed on humans, twin-pregnancy studies on BWD as a risk factor associated with fetal and neonatal outcomes, and sufficient data were provided to analyze or report the risk ratio (RR). Articles were then stored in a reference manger, Mendeley, and duplicates were deleted.  Figure 4-1presents the inclusion and exclusion stages of articles. Promotion or exclusion of articles was performed manually. The first reviewer, author of the dissertation, screened all the selected articles. Five percent of the articles were randomly selected. They were then reviewed at title, abstract, and full paper screening levels by a second reviewer, a hired researcher (Dr. M. Molainejad). Kappa coefficient for agreement level between the reviewers was estimated to be 0.78. Differences of opinions were resolved by consensus.  Data extraction The following information was extracted from each full paper: study objective, number of subjects, study site(s) and settings, ethical approval, inclusion and exclusion criteria, and number of adverse outcomes for BWD and concordant twins; see Appendix III for the complete list. Adverse outcomes were recorded in 2×2 tables for ease of RR calculation. Kappa score was estimated to be 0.70 and disagreement was resolved by referring to the article. A narrative review of chorionicity information was added using snowballing technique which means  82  searching the references of studies that mentioned chorionicity in the reference list of selected papers. Quality assessment Quality assessment was undertaken using a modified version of the Downs and Black checklist,129 as introduced in Table 4-1. The tool contained 20 questions (20 Q’s) examining these items: clearance (5 Q’s), external validity (2 Q’s), and internal validity (13 Q’s on bias, exposure measurement, and confounding). The median quality score was 22, and the articles were categorized into two groups of low (below median) and high quality (above median).  Statistical analysis Meta-analyses were conducted using the generic inverse variance with a fixed- or random-effect model in Revman 5.2. The RR (95%CI), test statistics for the interaction, and statistical significance of the analysis were computed. For papers not reporting RR and 95%CI, the raw data (i.e., number of events and total number of samples in the exposed and unexposed groups) were used to estimate RR.  The heterogeneity across individual studies was quantified by the I2 test.130 Low, moderate, or high degrees of heterogeneity were approximated by I2 values of 25%, 50% and 75%, respectively. If the I2 value was larger than 50%, a random-effect model was used. Reasons for heterogeneity were investigated by eyeballing extreme RR and sensitivity analysis.   83  Subgroup analyses were conducted according to the quality score, study design, location, and publication year (based on use of ultrasound for BWD diagnosis). Studies with extreme RR or maximum weight were removed to examine their impact on the overall RR.  Inverted funnel plots were visually inspected to detect the potential for publication bias. The plots were produced for meta-analyses with more than 10 studies. An asymmetric funnel indicates an overestimation of the exposure effect.131 4.4 RESULTS Once the procedure described in Figure 4-1 was implemented, more than 2500 article titles were identified from the literature review, out of which 40 articles were finally selected. Seventy percent of these articles had a quality score > 22 with a mean of 21.86 ± 1.1 and a median of 22. Table 4-1 gives the quality score details. The 40 studies reviewed were published between 1986 and 2015. Eleven articles were published prior to 2000. Nineteen articles were from the USA and Canada, 11 articles were from Europe, and the remaining articles were from the Middle East/Asia. There were two matched case-control study designs,49,132 two prospective cohort studies,16,43 and 37 retrospective cohort designs.7,9,22–53 Only 13 studies reported maternal age, average maternal age was 29.7 years. Table 4-2 summarizes the characteristics of the articles and the incidence of BWD.     84    Figure 4-1 Flow chart of included and excluded studies for a systematic review of BWD and perinatal outcomes Studies used for creating meta-analysis (n=40)Records excluded Records identified through database searching risk factors on onset of MS  (n =4,597)  EMBASE (n=3,082)  MEDLINE(n=760)  EBSCO (n=155)  Cochrane Central Register of Controlled Trials (n=6)  Web of Science (n=592) Screening Included Eligibility Identification Additional records identified (n=50)  Paper First (1993) (n=8)  Proceeding First (1993) (n = 39) ProQuest (n=3) Records after duplicates removed,  Stage 1a: title screening (n = 2519) Stage 1: abstract screening (n=212)  Records excluded  (n =105) Duplicates (n=1445) Full-text articles not found (n=2) Stage 2: Full-text articles assessed for eligibility  (n =60) Full-text articles excluded, with reasons  (n =17) Full-text articles included (n=42) Stage 3: Fulfilled quality assessment,  (n =40) Two cross-sectional studies excluded  85  Table 4-1 Quality assessment of included articles by modified Dawn and Black tool Study Year Clarity External validity Internal validity Total Score     Bias Exposure  Confounding  Alam 2009 4 2 5 6 2 19 Amaru 2004 5 2 5 8 4 24 Appleton 2007 5 2 6 8 3 23 Audibert152 2002 5 2 5 7 3 22 Blickstein 2004 5 2 5 8 3 23 Branum 2003 5 2 5 8 3 23 Breathnach 2011 5 2 6 8 3 24 Canpolat 2006 5 2 5 7 3 22 Cheung 1995 5 2 5 8 3 23 Cohen 2002 5 2 5 8 3 23 Cooperstock 2002 5 2 5 8 3 23 Demissie 2002 5 2 5 8 3 23 Fakeye 1986 3 2 5 7 2 19 Fraser 1994 5 2 5 7 2 21 Graf153 1998 4 2 5 8 3 22 Harper 2012 4 2 5 7 3 21 Hollier 1999 5 2 5 8 3 23 Hsieh 1994 5 2 5 7 3 22 Jakobovits 1992 5 2 5 7 3 22 Kalish 2005 4 2 5 7 3 21 Kato 2006 5 2 5 7 3 22 Kilic 2006 5 2 5 7 3 22 Kontopoulos 2005 5 2 5 8 3 23 Lee 2007 4 2 5 7 3 21 Lewi 2008 5 2 5 7 3 22 Lopriore 2012 4 2 5 7 3 21 Mahony154 2009 4 2 5 7 3 21 Nawab 2008 4 2 5 7 3 21 Patterson 1990 4 2 4 8 3 21 Queiros 2010 4 2 5 7 3 21 Sanghi 2012 4 2 5 7 3 21 Severinski 2004 5 2 5 7 3 22 Smiljan 2004 5 2 4 7 3 21 Szymanski155 2009 5 2 5 7 3 22 Talbot 1997 5 2 5 7 3 22 Wen 2006 5 2 5 7 3 22 Vergani 2004 4 2 5 6 3 20 Victoria 2001 4 2 5 8 3 22 Wen 2005 5 2 5 8 3 23 Yalcin 1998 4 2 5 8 3 22 D’Antonio 2013 5 2 5 8 3 23 Median (Range) 5 (3-5) 2(2) 5 (4-6) 7 (6-8) 3 (2-4) 22 (19-24)   86  Table 4-2 Description of the 40 articles reviewed Author, year Country Twins Population- based Enrollment period BWD cut-off point %& Incidence % Maternal age Mean (SD)& Gestational age Weeks Chorionicity Cohort studies Alam, 2009 Brazil 302 No 1998-2004 ≥20 26.5 29.1 (6.2) ≥26  Yes Amaru, 2004^ USA 1,318 No 1992-2001 ≥20 16.0 - ≥24  Yes Appleton, 2007 Portugal 230 No 1989-2002 ≥20 23.0 29.7 (4.7) ≥34 No Audibert, 2002 France 346 No 1996-1999 ≥20 22.0 31.9 (4.4) ≥24 No Blickstein, 2004 USA 10,683 Yes 1995-1997 ≥25 8.2 29.2 (6.1) ≥28 No Branum, 2003 USA 128,168 Yes 1995-1997 ≥15 26.5 - 28-41 No Breathnach, 2011 Ireland 977 No 2007-2009 ≥15 31.0 33&& <22 Yes^^ Canpolat, 2006 Turkey 266 No 2000-2004 ≥20 20.0 - 33 No Cheung, 1995 Canada 122 No 1989-1992 ≥10 58.2 27.8 (4.5) 25-34 Yes^^^ Cohen, 2001 France 2121 No 1984-1998 ≥30 39.0 30.5&& ≥25 Yes^ Cooperstock, 2000 USA 9931 Yes 1978-1990 ≥30&&& - - ≥17 No Demissie, 2002 USA 297,155 Yes 1995-1997 ≥20&&& 8.1 - ≥20 No Fakeye, 1986 Nigeria 622 No 1982-1983 ≥20 9.0 - - No Fraser, 1994 USA 1145 Yes 1960-1986 ≥25 13.0 - ≥24 No Graf, 1998 USA 213 No - ≥20 - - - No Harper, 2012 USA 1,145 No - ≥20 8.0 - >24 Yes^ Hollier, 1999 USA 1,370 No 1988-1996 ≥15&&& 29.0 - 24 No Hsieh, 1994 Taiwan 279 No 1986-1991 ≥15 39.0 - ≥24 No Jakobovits, 1992 Hungary 329 No 1975-1992 ≥15 15.2 - 24 No Kalish, 2005 USA 176,985 Yes 1995-1998 ≥25 8.1 - 28-40 No Kato, 2006 Japan 25,955 Yes 1995-1999 ≥25 9.0 - 24 No Kilic, 2006 Turkey 136 No 2003-2005 ≥15 39.7 27.2 (5.6) Mean 31.8 Yes& Kontopoulos, 2005 USA 170,223 Yes 1995-1998 ≥20 17.0 - Mean 36.5 No Lee, 2007 Korea 251 No 2002-2004 ≥20 15.0 33.2 (0.43) Mean 36.5 Yes& Lewi, 2008 Belgium 202 No 2002-2007 ≥25 14.0 30.4 (4) 16 Only MCDA Mahony, 2009 Dublin 1166 No 1997-2006 ≥20 24.1 - - Yes Nawab, 2008 USA 1597 No 2001-2004 ≥20 24.7 - Mean 32.6 No Patterson, 1990 USA 194 No 1982-1985 ≥20 26.2 - 26 No Queiros, 2010 Portugal 934 No 1994-2008 ≥25 9.2 - - No Severinski, 2004 Croatia 351 No 1993-2001 ≥20 15.1 29.3 (4.9) >25 No Szymanski, 2009 Croatia 280 No 2003-2009 ≥20 19.0 - - No Talbot, 1997 USA 119 No 1988-1995 ≥20 35.0 - 24-35 No  87  Author, year Country Twins Population- based Enrollment period BWD cut-off point %& Incidence % Maternal age Mean (SD)& Gestational age Weeks Chorionicity Wen, 2006 USA 147,262 Yes 1995-1997 ≥25 8.6 - <32 No Vergani, 2004 USA 335 No 1990-2000 ≥20 - 32.1 (4.7) 20 Yes^^ Victoria, 2001 USA 382 No 1993-1995 ≥25 11.0 28.9 (4.7) 20 Yes Wen, 2005 Canada 59,034 Yes 1986-1997 ≥20 17.0 - >20 No Yalcin, 1998 Turkey 384 Yes 1993-1995 ≥20 32.0 25.8 (N/A) 24-40 No D’Antonio UK 1261 No 2000-2010 ≥25     Matched case control studies Lopriore, 2012 Netherlands 94 No 2002-2011 ≥25 - - Mean 33.2 Yes^^ Sanghi, 2012 India 42 No 2004-2008 ≥15 - - - No  &SD: Standard deviation  &&Median      &&& Range from 10, 10-15, 15-20, 20-30 and >30 or other range  ^ Data was presented as adjusted effect measure for chorionicity. Subgroup analysis for chorionicity was not possible.  ^^ Data was presented as a composite index of perinatal morbidity, thus data on separate outcome was not available.  ^^^No result was provided for perinatal outcomes of MC or DC twins &Chorionicity was partly recorded. Information on the role of chorionicity on the relationship between BWD and adverse outcomes was not presented. MCDA: Mono chorionic diamniotic   88  The incidence of BWD ranged from 8.0% to 58.2%. The weighted incidence rate of BWD was estimated as 12.69% (95%CI of 4.69-12.69). From a total of 1190261 twins studied, approximately 1% exhibited growth discordance ≥30%. About 43% were ≥25% discordant while about 45% were ≥20% discordant. Finally, 10.99% showed ≥15% growth discordance and 0.01% of twins had growth discordance of greater than 10%.   Table 4-3 contains a brief account of the findings of the current systematic review. From the 40 studies, seven provided data on stillbirth while 14 articles presented data on neonatal outcomes. The numbers of participants for each outcome ranged between 2458 to 348158. The estimated RR for fetal and neonatal adverse outcomes varied from 1.16-4.70 with a median RR of 2.41. The following sections elaborate on the results.   Table 4-3 Risk ratio and 95%CI of perinatal outcomes in relation to BWD  Outcome Studies Participants RR 95%CI Heterogeneity% Fetal Outcomes Stillbirth 7 154329 3.99 3.78-4.22 43 Neonatal Outcomes Early neonatal death 9 348158 3.61 1.84-7.07 100 Late neonatal death 2 2458 3.61 1.84-7.07 67 Congenital anomalies 9 3233 1.87 1.41-2.49 42 Fifth minute Apgar score < 7 6 3097 1.91 1.46-2.50 63 Length of stay in NICU > 2 days 4 3097 2.49 1.40-4.46 92 Newborn septicemia 7 4438 1.39 1.05-1.84 58 Hypoglycemia 3 2688 1.89 1.38-2.59 0 Neonatal anemia 5 3753 1.16 0.87-1.54 0 Intraventricular haemorrhage 7 8354 1.24 0.91-1.68 0 Respiratory distress syndrome 11 9537 1.38 1.22-1.57 77 Chorionicity 1 - - - - CI: confidence interval   89  Stillbirth Using a fixed-effect model, stillbirth was found to be almost four times more likely in BWD twins compared to concordant twins when cut-off point was considered 15% and above as seen in Table 4.2 (RR=3.99, 95%CI 3.78-4.22; see Figure 4-2).  Seven studies, consisting of 156490 twins, examined the association between BWD and stillbirth. The heterogeneity was found to be moderate (I2=43%). Only one126 study had a CI crossing the line of no effect.   Figure 4-2  Meta-analysis for association between BWD and stillbirth  Neonatal mortality Neonatal mortality, defined as the death rate during the first 28 days of life, was reported either as early (0-7 days) or late (8-28 days) death. Two meta-analyses were conducted for these distinct variables. Two articles reported late neonatal156,157 and nine studies compared early neonatal death between BWD and concordant twins. Two forest plots were drawn for these two outcomes (Figure 4-4  90  and Figure 4-4). The mortality risks for early and late neonatal periods were 2.66 (95%CI 2.41-2.94) and 3.61 (95%CI 1.84-7.07), respectively. Both these results were found to be significant.   Figure 4-3 Meta-analysis for association between BWD and early neonatal mortality   Figure 4-4 Meta-analysis for association between BWD and late neonatal mortality  A moderate heterogeneity value of I2=54% for late neonatal mortality was found to be insignificant (p=0.14). The heterogeneity measure for early neonatal mortality was high, estimated to be 67% (p=0.002). The value of I2 did not change when the random-effect method was employed. Subgroup analyses and excluding studies with higher weight did not reduce the heterogeneity.       91  Congenital anomalies Nine studies (n=3233) were included in a meta-analysis to obtain the association between BWD and congenital anomalies. A significant association was found between congenital anomalies and BWD, described by RR=1.87 (95%CI 1.41-2.49), I2 = 42%, and p=0.08; see Figure 4-5.    Figure 4-5 Meta-analysis for association between BWD and congenital anomalies  Apgar score Six studies (n=3097) reported the fifth minute Apgar score < 7, yielding RR of 1.89 (95%CI 1.45-2.47) with a high level of heterogeneity, as I2=57% and p=0.03; see Figure 4-6.   Figure 4-6 Meta-analysis for association between BWD and the fifth minute Apgar score <7    92  Neonatal Intensive Care Unit (NICU) stay of longer than two days Four studies (n=3567) were included to determine the association between BWD and NICU-length-of-stay longer than two days. The meta-analysis yielded an RR of 2.49 (95%CI 1.39-4.46) and I2 =92%; see Figure 4-7.  Figure 4-7 Meta-analysis for association between BWD and NICU stay >2 days  Newborn septicemia A fixed-effect model was run for seven studies (n=4438), resulting in RR=1.39 (95%CI 1.05-1.84); see Figure 4-8. This finding is compatible with an increase in the frequency of newborn septicaemia when exposed to BWD.   Figure 4-8 Meta-analysis for association between BWD and newborn septicemia     93  Neonatal hypoglycaemia From the three studies included in the meta-analysis of neonatal hypoglycaemia (n=2688), a significantly higher risk of hypoglycaemia was found for BWD twins in comparison with non-concordant twins (RR=1.89; 95%CI 1.38-2.59; I2=77%, p=0.01); see Figure 4-9.   Figure 4-9 Meta-analysis for association between BWD and newborn hypoglycaemia  Neonatal anemia Anemia was reported in five studies (n=3753), none of which showed a significant risk of anemia for growth discordant twins compared to their concordant counterparts. The effect sizes of the studies lay near the line of no effect; see Figure 4-10. The meta-analysis of neonatal anemia resulted in a non-significant RR of 1.16 (95%CI 0.87-1.54). The heterogeneity was found to be zero.  Figure 4-10 Meta-analysis for association between BWD and newborn anemia    94  Respiratory Distress Syndrome (RDS) More than 9000 twin pairs were included in the meta-analysis of RDS in eleven studies. An evidence of severe heterogeneity was found with I2=77%. The effect estimate was 1.38 (95%CI 1.22-1.57), presenting evidence of a higher risk for RDS in BWD twins compared with concordant ones. No evidence of publication bias was found; see Figure 4-11.   Figure 4-11 Meta-analysis for association between BWD and respiratory distress syndrome  4.5 DISCUSSION The present review investigated the incidence of BWD and its association with the fetal and neonatal outcomes of twin pregnancies.  The findings of the review will be discussed below.  Incidence BWD is a common phenomenon in twin pregnancies.8,15,16,158 The value representing the incidence of discordance depends partly on how BWD is defined and the threshold of what is  95  considered concordant. Studies selected for this review all provided the same definition as noted in Equation 1, Chapter 3.  The incidence of BWD was found to be wide, ranging between 8.0-58.2%. One explanation for this wide spectrum is the heterogeneity in socio-demographic characteristics of the population surveyed.11 Moreover, confounding factors with a direct impact on birth weight could obscure the actual incidence of BWD in twin pregnancies. Some of these factors are gestational age,159 pre-pregnancy weight, maternal weight gain during pregnancy,151 maternal age,68 gestational diabetes,160 use of invitro-fertilization (IVF),161 concurrent maternal diseases such as hypertension,162 genetic anomalies,163 and chorionicity.164 These, in turn, have an impact on adverse perinatal outcomes. It is therefore recommended that future studies adjust the confounding variables to obtain an accurate estimate of BWD incidence and its influence on perinatal outcomes.      Role of BWD on Adverse Perinatal Outcomes With a sample size of over one million twins, BWD was found to be associated with nine out of the twelve variables, while three variables were not significantly affected by BWD. Here we discuss the former variables.  Stillbirth Our data analysis suggested that stillbirth was about four times more likely among twins with growth discordance compared with concordant twins. While the effect estimate calculated for stillbirth was statistically significant, the result should be interpreted with caution for a number of reasons.   96  The magnitude of the effect may be biased in the meta-analysis, as the time interval between the fetal demise and birth is likely not be known. Hence, one fetus continues to grow whereas the other does not; thus the longer the interval, the greater the discordance.  This would apply not just in spontaneous losses, but also those cases where selective /multifetal reduction has been performed.  It is therefore possible that many of these dead infants were misclassified as discordant co-twins. This will remain a potential risk of bias in the respective studies, leading to an effect overestimation. Some publications studied attempted to minimize the issue by excluding twins with less than 500 grams weight, twin gestations in which 1 twins died in utero.165 Prospective studies using ultrasound assessment and/or fetal surveillance may be better suited to study the association between BWD and stillbirth, although intervention bias may prevent stillbirths, so other parameters such as perinatal morbidity or early neonatal demise might better capture the negative effects of BWD.  Stillbirth has been defined differently in various studies. In some studies, it is defined as death of fetus at gestational age of more than 20 weeks,166 while in others it is classified as demise of infant weight more than 500 grams or born at gestational age greater than 20 weeks.167 This discrepancy could misclassify some fetal demise, giving rise to an under-estimation of the stillbirth frequency.   Death of one twin in a monochorionic twin gestation, sharing the same placenta, might hasten the death of the second twin. Thus, chorionicity should be taken into account.    97  Registration of stillbirths in various community-based or hospital-based settings is not always accurate. Also, registration rules and practices may differ in various countries168 or even in the maternity wards of a specific country.5 As a result, many fetal deaths between 20 and 28 weeks may not be reported.158  Although rare, how monochorionic twins with reversed arterial perfusion sequence (TRAP) are classified in these datasets is of great importance. TRAP is a condition where among twins who share a common placenta, one twin is usually developmentally normal and the other twin has a serious condition (e.g. missing a heart or head) that prevents it from surviving. Inclusion of these cases may overestimate the association between BWD and stillbirth.  It is difficult to interpret neonatal death when we are comparing preterm and term neonatal death while gestational age has not been taken in to account.  The timing of fetal death is relevant when interpreting the effects of growth discrepancy. Dead fetus does not grow. Thus, the longer the dead fetus stays in utero, the more the growth discordance will be. Growth discordance associated with fetal death near term, though, might have a different cause.    Neonatal mortality Neonatal mortality is reported either as early (0-7 days) or late (8-28 days) death. The effect estimates for these were relatively high (Table 4-3). RR for early or late neonatal mortality was similar (both were 3.61).  Time frame of definition for these variables is different suggesting that either baby twins may die early in neonatal period, or that they don’t die till after 28 days. The variation between studies and the estimated RR(s) for neonatal mortality should be interpreted in the light of the following study limitations:  98  o Early neonatal deaths are easy to monitor. However, most retrospective hospital-based studies underestimate the late-neonatal death because the death might be recorded in a setting different from where the baby is born.  o Use of tertiary referral hospital datasets, with the exclusion of cases that remain in lower acuity community hospitals, increases the risk estimate of neonatal mortality.  o Neonatal death is overestimated if confounding variables such as gestational age,169 birth weight of each twin,75 fetal anomalies170 or TTTs171 are not adjusted for. Specifically, twin-pairs with TTTs are more likely to be classified under severe BWD (with ≥ 30% growth discordance)172 as cases of larger twins with cases of hydrops fetal are heavier than the anemic co-twins.  Congenital anomalies Congenital anomalies are present in as much as 40% of twin pairs with BWD.173 Monochorionic twins are more likely to have one malformed twin, leading to lethal consequences.174 Thus, zygosity is an important factor to explore when investigating the relationship between BWD and congenital anomalies. Zygosity was not accounted for in some of the studies included in the meta-analysis. This bias results in observed moderate heterogeneity of I2=42%. In any study on the association of BWD with adverse perinatal outcome, inclusion of major congenital fetal anomalies will inexorably affect any perceived discordance-related morbidity. This is due to the strong association between fetal structural anomalies and aberrant growth patterns. Future studies should therefore consider the presence of congenital anomalies as a confounding variable when estimating the role of BWD in relation to other perinatal adverse outcomes.  It is possible that datasets with unusually large proportions of anomalous fetuses of  99  twin pairs may reflect a biased population, as a prenatally identified anomalous twin is likely to be referred and delivered at a tertiary care facility. Apgar score The RR of the 5th minute Apgar score < 7 was significantly higher among BWD twins (RR=1.89; 95%CI 1.45-2.47). This shows the role of BWD. Some of the potential confounding variables influencing this influence are gestational age, congenital anomalies, infections, hypoxia, hypovolemia, and trauma.175 Additionally, the time lapse between the delivery of the first and second twins is a predictor of Apgar score of the second twin. This is more pronounced for a second twin with breech presentation.176 Mode of delivery may also influence the Apgar score.177. The reviewed publications studied the Apgar score of several categories of BWD, namely ≥10%, ≥15%, ≥20%, and ≥30%. This partly explains a high heterogeneity of I2=67% as estimated in the meta-analysis. A sub-analysis of various BWD thresholds was not performed due to an insufficient number of articles.  NICU Length-of-stay greater than two days Twins with BWD had a higher risk of staying in a NICU longer than two days (RR=2.94; 95%CI 1.39-4.46). NICU admission alone is only a proxy for adverse neonatal outcomes when faced with complications associated with twinning.178 For instance, low birth weight infants may be more prone to hospital acquired infections, increasing their length of stay. Another example would be IVF cases with more complications as these cases may stay longer in the hospital. Therefore, articles which reported the “admission” with no mention of length-of-stay were not included in the review.   100  The meta-analysis for NICU stay yielded I2=92%. Possible causes of such high heterogeneity are as follows:  o Different cut off points used to assess BWD, in that BWD ≥ 20% was considered in three studies while BWD ≥15% was used in the other article.  This can be the cause of heterogeneity for all other variables under study. o Unclear number of infants admitted, in that NICU admission took place if either or both twins were admitted to the NICU.  o Uncertainty about length of stay, in that the length of stay in the NICU could be recorded as the longer stay of each twin. Future studies should take into account the fact that smaller twins among those with BWD are more likely to stay longer in the NICU. o Local practices and admission criteria for the NICU. o Differing practices among obstetricians with regards to gestational age of delivery of twin pairs, both complicated and uncomplicated. o Different definitions and descriptions of NICU that may exist across different studies. o Anomalous twins are more likely to be admitted to a higher level of nursery care. Newborn septicemia Neonatal sepsis accounts for over 50% of neonatal deaths.179 A non-significant high effect size was estimated for newborn septicemia (RR=1.31, 95%CI 1.05-1.84). A moderately high level of heterogeneity (I2=58%) was found which can be explained as follows.  o In addition to pathogens, immaturity of a neonate’s immune system (due to genetic predisposition) can make neonates susceptible to sepsis.180 Polymorphism, a genetic condition associated with septicemia, is more common among monozygotic twins.181 Hence,  101  the zygosity of twins should be accounted for in any investigation on the association between BWD and sepsis.  o The clinical manifestations of neonatal septicemia are usually unclear, demanding a high index of suspicion. Babies with septicemia who had been active and sucking well would suddenly become lethargic and inactive, refusing to suckle. Therefore, the diagnosis of septicemia may be delayed, leading to underestimation of septicemia. o Clinical and laboratory criteria for diagnosis of septicemia are partly responsible for the lack of large studies on association between BWD and septicemia.  This is because the diagnosis of septicemia is based on isolation of microorganisms from the blood, cerebrospinal fluid, urine, pleural fluid or pus.  Taking these biological specimens from infants may be cumbersome, requiring special skills. Additionally, accurate diagnosis demands two or three tests, rendering it time-consuming and expensive. These limitations lead to underestimation of septicemia. Neonatal hypoglycemia In the early hours of postnatal life, hypoglycemia is fairly common in preterm twins. Hypoglycemia in BWD as a clinically important condition may result from an abnormal intrauterine environment, leading to discordant growth.89 From a twin set with BWD, the infant with small for gestational age (SGA) is at higher risk of hypoglycemia since the small size generally reflects a scarcity of intra-uterine nutrients.  These infants, if not diagnosed and treated immediately after birth, are at risk of adverse neuro-developmental conditions.182   A cross-sectional study found a prevalence of 15% and 8% among 75 preterm and 80 term twin infants, respectively.63 Gestational age is, therefore, an important variable confounding the relationship between BWD and neonatal hypoglycemia. One out of three articles reviewed did  102  not consider gestational age as a confounding variable.146 Nevertheless, analysis resulted in a significant CI (RR=1.89, 95%CI 1.38-2.59). Another potential confounder is the diagnosis, prevalence and clinical management of diabetes, both pre-existing and gestational.183  Newborn anemia Two factors can contribute to newborn anemia in growth discordant twins:  o TTTs and Twin Anemia-Polycythemia Sequence (TAPS), in which blood flows from one fetus to the other, and  o Premature separation of the placenta from the uterine wall before delivery, leading to hemorrhage of fetal blood.  o Abnormalities of placentation, such as vasa previa, which are more common in twins and may lead to fetal hemorrhage.184  Studies included in the meta-analysis did not specify the presence of either TTTs or placenta previa. This could seriously confound the findings.   Respiratory distress syndrome RDS is caused by developmental insufficiency of surfactant production in the lungs and is the major cause of morbidity and mortality in preterm deliveries.185 Twins are known to be at increased risk of RDS.   The meta-analysis indicated a high heterogeneity of I2=77%, resulting from confounding factors such as sex, gestational age and birth order.181  Furthermore, unclear definitions of RDS contributed to selection of cases with other respiratory problems. For example, some studies defined RDS as “the need for supplemental oxygen in the presence of typical radiographic  103  findings in the absence of other causes of respiratory distress, such as pneumonia, and bronchopulmonary dysplasia.63,142  Others, however, defined RDS as diagnosed by compatible chest x-ray findings and arterial blood gas results.89 Furthermore, the benefits of antenatal corticosteroids in prevention of RDS is well known, and the rate of administration/use may not be the same across all studies and populations.186  Chorionicity Chorionicity determination in twin gestation is an important issue in the risk assessment of twin pregnancies. Twins who are connected to a single placenta (monochorionic or MC) are more prone to adverse perinatal outcomes compared to those with one placenta each (dichorionic or DC).72  As shown in Table 4-2, only 11 studies out of 40 provided information on chorionicity. From these, some reported the adjusted estimate for chorionicity in combination with other confounding variables when analyzing the role of BWD relative to adverse outcomes, some did not present the raw data, and others investigated a composite index of adverse outcomes (see Table 4-2).  Furthermore, studies that took chorionicity into consideration showed inconsistent results in terms of its effect on perinatal outcomes. We could not identify enough studies to conduct a subgroup analysis of adverse perinatal outcomes according to chorionicity. This Chapter, therefore, presents a narrative review of importance of chorionicity in the relationship between BWD and adverse outcomes.  Table 2-2 in Chapter 2 summarizes the characteristics of studies that considered chorionicity when investigating the influence of BWD on perinatal outcomes or used a surrogate measure instead of chorionicity (e.g., analysis of unlike sex twins) but are not presented in Table 4-2. Studies that did not report chorionicity but noted the importance of chorionicity were assessed.  104  In addition, studies listed in the references of these studies were added for a comprehensive review. A twin study involving term and preterm twins found that the worst outcome in MC twins was mediated by the greater risk of severe prematurity among MC rather than DC twins.52  Moreover, BWD was significantly larger among MC compared to DC twins. These findings led the authors to conclude that mono-chorionicity was at the root of both extreme prematurity and inter-twin discordance. However, the rates of adverse perinatal outcomes such as fetal death (4/36 vs. 3/44, p=0.69) and NICU stay >30 days (11/36 vs. 7/44, p=0.2) among cases with severe weight discordance (i.e., >25%) were not significantly different between MC and DC twins. Consistent with this result, Dubé et al. have found that a BWD >25% is equally present in MC and in DC gestations (24/280 vs. 47/728 p=0.3).79  A prospective cohort study of 356 twin gestations accounted for chorionicity by ultrasound assessment of the dividing membrane, neonatal sex, and examination of the placenta at birth. Multivariate analysis showed that BWD was associated with adverse outcomes independent of chorionicity.80  Alam et al.74 studied the early neonatal morbidity and mortality in 151 twin pregnancies with growth discordance at the Multiple Pregnancy Unit of Sao Paulo University Hospital between 1998 and 2004.  Chorionicity was defined in 84.2% of cases (n=127) by histopathology, in 6.6% (n=10) with first-trimester ultrasound, in 2.6% (n=4) using the presence of distinct placental sites at ultrasound scan, and in 6.6% (n=10) based on discordant fetal sex of the cases. An assessment of both concordant and discordant groups using chorionicity showed a worse outcome in the MC concordant sub-group. In concordant twin pregnancies, MC cases compared to DC cases  105  presented lower gestational age (34.3±3.2 vs. 36.2±2.7, p=0.004) and weight at birth (2067±582 vs. 2334±501, p=0.002) and a longer period of hospital stay (5.5 vs. 3.0, p=0.002).  Another prospective cohort study of 1028 twin pairs in Ireland used series of ultrasound (from 2007 to 2009) to determine chorionicity through standard ultrasound criteria including placental number, identification of lambda or T sign, inter-twin membrane thickness, and fetal sex at the first visit. Subsequent correlation was sought with placental pathologic examination. Fortnightly growth scans were performed from 16 weeks of gestation until delivery for MC twin pairs and from 24 weeks until delivery in DC pregnancies.47 Perinatal mortality, individual morbidity, and composite perinatal morbidity were all seen to increase with BWD exceeding 18% for DC pairs (hazard ratio 2.2, 95% CI, 1.6–2.9) and 18% for MC twins without TTT syndrome (hazard ratio 2.6, 95% CI 1.6–4.3).  Appleton et al.16 conducted a retrospective cohort study of 230 twin gestations that ended at 34 weeks of gestation or later. They attempted to study the role of chorionicity but encountered a low number of MC twins (15 in discordant twins and 35 in concordant twins), preventing them from extended evaluation based on placentation. The comparison between the growth concordant (<20%; n=176) and discordant groups (n=54) did not indicate statistically significant differences in the parameters, which included Apgar scores, birth weights, admission to NICU, significant morbidity, and perinatal mortality.  Kato et al. recorded chorionicity in 22 (26.8%) twin pairs (8 MC, 14 DC) of their concordant group and in 30 (55.5%) twin pairs (20 MC, 10 DC) of their discordant group. There was no difference between concordant and discordant groups for chorionicity.139  106  Chorionicity data was available in Mahony et al’s. study in 1128 twin pairs (96.7%) and similar rates of discordance were recorded in monochorionic diamniotic (MCDA) and dichorionic diamniotic (DCDA) twins [54/292 (18.4%) vs. 149/836 (17.85); p = 0.88]. While the incidence of stillbirth was greater in MCDA twins (3.9% vs. 1.3%, p < 0.001), stillbirth incidence was similar in growth discordant MCDA and DCDA twins (3/54 5.5%) vs. 4/149 (2.7%), p = 0.57).154 Chorionicity was partly known in Vergani et al.’s cohort of twin gestations in 202 pregnancies (60%), including 69 MC and 133 DC gestations.80  This study used a composite index for adverse perinatal outcomes inclusive of neonatal death or major neonatal morbidities (e.g. infant respiratory distress syndrome, pneumonia, bronchopulmonary dysplasia, disseminated intravascular coagulation, necrotizing enterocolitis). Presenting composite data such as this increases the study power but does not serve our review. Victoria et al. illustrated that severely discordant (≥25% BWD) MC and DC twins had significantly worse perinatal mortality and morbidity than mildly discordant and concordant twins.52 Monochorionic twins were more likely to remain more than 10 days in the NICU (73 of 118 [40.4%] vs. 137 of 576 [23.7%]; OR 2.27; 95% CI 1.53, 3.22; P < .01), and more than 30 days in the NICU (25 of 178 [14%] vs. 37 of 576 [6.4%]; OR 2.38; 95% CI 1.34, 4.20; P < .01). D’Antonio et al’s. study included 2161 twin pregnancies (302 MC and 1859 DC). The total risk of perinatal loss (stillbirth and neonatal death) in twins with a BW discordance of ≥25% (60.9 per 1000 fetuses) was significantly greater than that in twins with a BW discordance of<25% (8.6 per 1000 fetuses), with a hazard ratio of 7.29 (95% CI, 4.37–12.00; P<0.0001).   107  Our narrative review of chorionicity data indicates the importance of chorionicity data in assessment of growth discordance and adverse fetal and neonatal outcomes. Studies used for the current review determined chorionicity using various methods including pathology reports, ultrasound screening during pregnancy, and fetal sex determination in ultrasound.  Future reviews should focus on the reliability of ultrasound versus pathological determination of chorionicity data.   4.6 SUMMARY In conclusion, the present systematic review screened more than 2500 articles containing approximately 1.2 million twins. The range of BWD in the literature reviewed varied between 8.0%-58.2%. This too-wide range makes it difficult to establish a threshold for the correlation of BWD with perinatal mortality and morbidity. Included studies did not always control for the impact of gestational age. Thus, it is not known whether or not BWD directly relates to adverse outcomes, or whether BWD predicts preterm birth which then begets complications. The impact of iatrogenic prematurity is also not known. The meta-analyses performed suggested that BWD increases the risk of stillbirth, neonatal mortality and several neonatal morbidities. This finding must be interpreted cautiously as the respective heterogeneity levels were found to be moderate to high. Chorionicity should be considered as an inseparable part of studies aiming to investigate perinatal mortality and morbidity.    108  CHAPTER 5 BWD AND PLACENTAL PATHOLOGY  5.1 INTRODUCTION Independent of gestational age at delivery, twins with significant growth discordance have poorer perinatal outcomes. 187,188120  The pathophysiology of BWD has been studied broadly. In MC twins, significant statistical differences are attributed to hemodynamic factors such as TTTs, 184 unequal placenta sharing,189 and placenta cord insertion.190 Apart from unequal placenta sharing and placenta cord insertion, in DC twins a pathological entity or variation in genetic makeup of twins is assumed.47 Placenta examinations, gross and microscopic, are useful to uncover the nature of intra-uterine impairment of twins’ growth.   5.2 OBJECTIVE We aimed to assess the pathological characteristics of placenta and cord in relation to BWD in twin gestations.  5.3 METHODS Data was collected from pathological records, retrospectively. Here, we briefly describe how the records were obtained. Placentas were collected and labeled as A (1 cord clamp) or B (2 cord clamps) according to birth order. Placentas were placed in plastic bags after delivery and kept at 4C until processed, usually within 24 hours of delivery. Placenta examination was carried out in the pathology department of the C&W hospital.  Pathologists who examined the placentas had access to the clinical information. The placentas were placed on a clean surface, adherent clots were removed, and the membranes and umbilical cords were excised before they were weighed. There was a systematic approach to attribute placental mass to each twin so that the total placental weight was recorded for each placenta. In dichorionic placentas with fused placentas, the proportion of placenta belonging to each twin was determined by measuring the length,  109  width, and thickness in each of the two placental disks. Measurement of placental thickness was carried out in three areas of the placental disk, and the mean thickness was then recorded.  Cord length and its distances from the placenta margin, from the membrane and from the other cord were also measured. Umbilical cord insertions into the disc of the placenta and more than 1 cm away from the marginal border were defined as (para) central, cord insertions within 1 cm of the disc edge were defined as marginal and cord insertions directly into the membranes were defined as velamentous. The number of vessels in the cord was recorded. A composite variable was created from cord properties inclusive of cord prolapse, number of cord vessels less than 3, and existence of cords knots or entanglements. The evaluation of placental chorionicity was performed by examination of the inter-twin membrane. Separated twin placentas were examined in the same way as those of singletons. Fused placentas can be MC or DC. The dividing membrane was examined to identify chorionicity.  The dividing membrane in a MC pregnancy is thin and translucent without any chorionic layer, while that of a DC placenta is thicker as it contains two chorionic layers between the amniotic sacs. The dividing membrane was then sampled as a membrane roll or in “T section” form. Identification of T form was considered confirmation of chorionicity.  Equal placental sharing was defined as 40% to 60% of the placenta attributed to each twin. We chose this range because preliminary data revealed that twins with 40/60 sharing and 50/50 sharing had similar gestational ages at delivery and degree of BWD.116 Unequal placental sharing was defined as 1 twin receiving blood from more than 60% of the placenta. In histological examination of the placenta, vascular-thrombotic lesions (infarction, chorangioma, subchorial fibrin deposition, and retro placental hematoma) were recorded.  110  Arteries were identified as vessels that are situated superficial in relation to the veins. A composite score was created for all of the pathological lesions for the purpose of analysis. Anastomosis between fetal vessels was recorded. Anastomosis was identified by the presence of an impaired vessel from one twin feeding an area drained by the co-twin.  Injection studies were performed in fresh specimen to identify unidirectional arteriovenous shunt(s) between donor and recipient.  Placentas were also assessed for maternal or fetal inflammatory response corresponding to chorionisits, chorioamnioitis, and chorionic villi inflammation. A composite variable was created using these items plus cases with inflammation or infection of the maternal or fetal side of the chorionic membrane.    Diagnosis of TTTs was made by the referring obstetrician and was designated on the pathology requisition. Placenta abruptio and invasive of trophoblast such as placenta accreta was also recorded.  Completed pathology reports for twins were printed from online pdf records or from hard copies of pathology records stored in hospital charts. Data was then abstracted from these records into an Excel database. Twin pregnancies were included in the data if mothers and babies were linked, babies had a calculated estimated gestational age at birth ≥20 weeks, and the mother did not have a termination procedure.  The pathology data were then linked to delivery outcome data such as gestational age and birth weight by PSBC. The link was possible using the PHN, maternal and baby ID and date of birth.   111  The final linked records were stored on Secure Research Environment on a Virtual Private Network and contained 1571 pairs (n=3142 individual twins) of twin pregnancy data.   For definition of placenta and cord composite variables and BWD, see Chapter 3. To identify the trend of change in growth discordance, twins were divided into four groups: 10% or less, 11-20%, 21-30% and more than 30%. Twins were then divided by three cut off points (10%, 20% and 30%) and the results were compared using 20% and 30% cut off points.  Fetal growth restriction was calculated using gestational age and BWD. Small for gestational age (SGA) was defined as birth weight less than the 10th percentile, whereas appropriate for gestational age (AGA) was defined as birth weight between the 10th and 90th percentile, according to a twin-adjusted gestational age nomogram.65 We excluded pathology reports of twins born with congenital anomalies, missing birth weight data and those born at less than 20 weeks of gestation.  Some other known confounders from the literature were excluded. These are cases with TTTs,132 and twins with one single stillbirth.191 In addition twins with smaller than 500 gram weights and those who underwent reduction procedures in utero were excluded as the growth pattern might be different in these cases.192 No exclusions were made because of pregnancy complications (e.g. hypertension, diabetes), and therefore selection bias is not expected to have occurred.  The frequency of placental and cord characteristics was compared between MC and DC twin pregnancies. The placenta and cord characteristics were then assessed as a factor in BWD and growth restriction. Frequency of occurrence of placenta/cord outcomes were compared between moderate (>20%) or severe (>30%) growth discordant twins and concordant twins. Comparisons were made between lighter twins of BWD pairs and the larger co-twins and concordant controls.  112  The relative frequencies of adverse events were also analyzed between twins with SGA compared with AGA/LGA twins. Analyses were then stratified by chorionicity. 5.4 RESULTS Of 6665 twin pairs born in BC, 1571 pairs of records (23.6%, registered at C&W hospital) had complete placenta pathological examination data available for analyses. After exclusion of cases noted in the methods section above, there were 1493 pairs of twins (n=2986 twins) in the analytical data. Of these, 768 twins (26%) were MC and 2218 twins (74%) were DC. Table 5-1 illustrates the clinical characteristics of the cohorts.  Table 5-1 Comparing clinical characteristics of monochorionic and dichorionic twins in the cohorts born in C&W hospital (1493 pairs, n=2986)    Overall N=2986 Monochorionic N=768  Dichorionic  N=2218 P value Birth weight, g, mean±SD 2356.58±596.41 2120±690.17 2356.89±626.73 0.01 Gestational age, Wks 34.56±3.36 33.85±3.69 34.78±3.22 0.01 BWD^  10% or less 11-20% 21-30% >30%  1656(55.4%) 834(28.9%) 317(10.6%) 179(5.9%)  446(58.1%) 164(21.3%) 91(11.8%) 67(8.8%)  1210(54.8%) 670(30.0%) 226(10.2%) 112(5.0%)  0.01 BWD >10% 1656(55.4%) 446(58.1%) 1210(54.5%) 0.76 BWD>20% 496(16.6%) 158(20.5%) 338(15.2%) 0.01 BWD>30% 194(6.6%) 75(9.76%) 119(5.3%) 0.01 Sex Male Female  1534(51.4%) 1452(48.6%)  404(52.7%) 362(47.1%)  1128(50.9%) 1090(49.1%0  0.20 Growth^  AGA LGA SGA  2133(72.7%) 629(21.5%) 170(5.8%)  521(74.2%) 120(17.1%) 61(8.7%)  1612(72.3%) 509(22.8%) 109(4.9%)  0.01 g: gram;  Wk: Week;  SD: standard deviation;  BWD: birth weight discordance      AGA: appropriate for gestational age;  LGA: large for gestational age,  SGA: small for gestational age; ^ANOVA test   113  Monochorionic twins were delivered at an earlier mean gestational age and were on average 236 grams lighter than their dichorionic counterparts. No significant difference was found between the MC and DC groups in terms of twins’ sex. Relationship between BWD and chorionicity  Investigating the trend of change in BWD amongst MC and DC cohorts, twins were compared in four groups: 10% or less, 11-20%, 21-30% and more than 30% growth discordance. Frequency of BWD≥30% was higher in MC twins compared with DC twins (8.8% vs. 5.0%, p=0.01). Similarly, BWD of 21-30% had a higher frequency in MC gestations compared with DC ones.  We then stratified BWD three times to ≥10%, ≥20% and ≥30%. BWD of ≥10% was found to be similar between MC and DC cohorts. Overall, 55.4% of the cohort were growth concordant (≥10% BWD), while 16.6% were moderately discordant (≥20% BWD). Severe discordance was found in 6.4% of twins (BWD≥30%). Analysis of stratified data for chorionicity showed that both moderate and severe discordant twin groups were statistically significantly different in terms of frequency in MC and DC cases.  Higher frequencies of BWD twins (≥20% or ≥30%) were found in MC cases (p =0.01, Table 5-1).  Chorionicity and pathology results Table 5-2 shows characteristics of placenta and cord, overall and stratified by chorionicity.  Compared to DC placentas, placenta anastomosis and unequal placenta sharing were significantly more common in MC placentas (p=0.01 for both comparisons). Average values for placenta weight, length, width and thickness were also higher in MC placentas compared to DC placentas. MC placentas had shorter cords than DC placentas (p=0.01). Cord length was longer in DC twins (p=0.01). Cord distances from the margin and from the other cord were longer in  114  DC placentas (p=0.01, p=0.05, respectively). Insertion of cord was found to be an insignificant variable in this class of comparisons.  Table 5-2 Comparing characteristics of placenta and cord by chorionicity for twins born in C&W hospital (1493 pairs, n=2986)   Overall N=2986 Monochorionic N=768  Dichorionic  N=2218 P  Placenta  Chorionic villi inflammation 264(9.0%) 54(7.7%) 210(9.4%) 0.09 Chorionitis 213(7.3%) 54(7.7%) 159(7.1%) 0.34 Anastomosis 263(9.0%) 229(32.6%) 34(1.5%) 0.01 Unequal placenta sharing 138(30.4%) 65(40.0%) 73(24.9%) 0.01 Composite of inflammation 388(13.2%) 82(11.7%) 306(13.7%) 0.09 Composite placenta lesions 54(13.2%) 18(2.6%) 36(1.6%) 0.07 Cord composite 8(0.3%) <5 (0.7%) <5 (0.2%) 0.29 Placenta other 36(1.2%) 7(1.0%) 29(1.3%) 0.34 Placenta weight, g, mean±SD 535.05±228.99 686.18±216.77 500.46±217.46 0.01 Placenta length, cm, mean±SD 22.07±5.17 23.53±4.24 21.73±5.31 0.01 Placenta width, cm, mean±SD 17.01±3.98 19.08±3.85 16.52±3.85 0.01 Placenta thickness, cm, mean±SD 2.13±0.46 2.24±0.54 2.09±0.44 0.01 Cord  Cord length, cm, mean±SD 27.44±12.67 25.77±12.42 27.99±12.71 0.01 Cord distance from margin, cm, mean±SD 4.61±2.88 4.40±2.36 4.68±2.26 0.01 Cord distance from membrane, cm, mean±SD 6.62±3.99 6.91±4.24 4.67±2.26 0.21 Cord distance from other cord, cm, mean±SD 14.15±8.32 13.28±5.30 14.50±9.24 0.05 Cord insertion type Marginal Central Velamentous  2024(69.0%) 481(16.4%) 125(4.3%)  472(76.4%) 114(18.4%) 32(5.2%)  1552(77.1%) 367(18.2%) 93(4.6%)  0.84 SD: Standard deviation    115  Relationship between BWD and pathology results  The relationship between BWD and pathology results was then analyzed (Table 5-3). Two thresholds of >20% and >30% were selected for further analysis as these levels were found to be statistically significantly different in the comparison between MC and DC (see Table 5-2).  Higher frequencies of unequal placenta sharing were found in twins with moderate growth discordance (BWD≥20%) as well as severe growth discordance (BWD≥30%). Twins with severe discordance had a higher frequency of velamentous cord insertion (VCI) compared with both central (32.7% vs. 19.5%, p=0.02) and marginal (12.1% vs. 5.4%, p=0.01) insertions. Twins with growth discordance (at both levels of moderate and severe growth discordance) had shorter cords (p=0.01) for both moderate and severe growth discordant twins). Lower frequencies of chorionitis and composite inflammation were found in growth discordance of ≥30% compared to growth concordant twins. Infection probably causes preterm delivery, and therefore may not allow BWD to occur.  Unadjusted and adjusted relationship between BWD and pathology findings  The variables of interest for regression analysis were anastomosis, unequal placenta sharing, cord length and cord insertion types. Twins with ≥30% BWD were less likely to develop anastomosis (0.57, 95%CI 0.33 to 0.98) compared with those with <30% BWD, after adjustment for chorionicity. Odds of unequal sharing of placentas were significantly higher in moderate and severe growth discordant twins (Table 5-4).  116  Table 5-3 Comparisons of placenta and cord characteristics in groups with and without ≥20% BWD or ≥30% growth discordance for twin gestations registered at C&W hospital (1466 pairs, n=2932)    Moderate growth discordance Severe growth discordance  less than 20% N=2438 20% and more N=494 P less than 30% N=6102 30% and more N=226 P Placenta Chorionic villi inflammation 221(9.1%) 43(8.7%) 0.43 251(9.2%) 13(6.7%) 0.15 Chorionitis 179(7.3%) 34(6.9%) 0.39 206(7.5%) 7(3.6%) 0.02 Anastomosis 209(8.6%) 54(10.9%) 0.06 245(8.9%) 18(9.3%) 0.48 Unequal placenta sharing 74(22.2%) 64(52.9%) 0.01 114(28.6%) 24(43.6%) 0.02 Composite of inflammation 325(13.3%) 63(12.7%) 0.38 372(13.6%) 16(8.2%) 0.04 Composite placenta lesions 49(2.0%) <5 (1.0%) 0.08 53(1.9%) <5 (2.2%) 0.11 Cord composite 7(0.3%) <5 (1.0%) 0.59 7(0.3%) <5 (2.2%) 0.42 Placentation 33(1.4%) <5 (1.0%) 0.12 34(1.2%) <5 (2.2%) 0.57 Average placenta weight, g 534.37±228.00 538.62±234.54 0.77 534.13±230.36 550.07±205.61 0.49 Average placenta length, cm 22.02±5.14 22.35±5.29 0.32 22.04±5.16 22.61±5.23 0.27 Average placenta width, cm 16.98±3.96 17.16±4.05 0.48 17.00±3.99 17.14±3.84 0.73 Average placenta thickness, cm 2.13±0.47 2.12±0.43 0.74 2.12±0.46 2.15±0.46 0.64 Cord       Cord length, cm 28.14±12.73 24.19±11.89 0.01 27.76±12.61 23.07±12.68 0.01 Cord distance from margin, cm 4.61±2.24 4.63±2.45 0.54 4.63±2.29 4.39±2.31 0.09 Cord distance from membrane, cm 6.54±3.84 6.90±4.62 0.81 6.63±3.93 6.25±4.69 0.22 Cord distance from the other cord, cm 14.23±8.85 13.81±5.60 0.30 14.29±8.53 12.54±5.10 0.46 Cord insertion type Marginal Central Velamentous  1688(77.6%) 388(17.8%) 99(4.6%)  336(73.8%) 93(20.4%) 26(5.7%)  0.21  1901(77.4%) 446(18.2%) 108(4.4%)  123(70.3%) 35(20.0%) 17(9.7%)  0.01 Velamentous vs. marginal 99(5.5%) 26(7.2%) 0.39 108(5.4%) 17(12.1%) 0.01 Velamentous vs. central 99(20.3%) 26(21.8%) 0.14 108(19.5%) 17(32.7%) 0.02 Marginal vs. central 1688(81.3%) 336(78.3%) 0.08 1901(81.0%) 123(77.8%)  0.19   117  Table 5-4 Unadjusted and adjusted odds of anastomosis and unequal placenta sharing in relation to BWD>20% and >30%.  BWD ≥20% BWD≥30%   Unadjusted Adjusted*   Unadjusted Adjusted*  Placenta Anastomosis 1.30(0.95-1.79) 0.92(0.64-1.31) 0.96(0.58-1.59) 0.57(0.33-0.98)* Unequal placenta sharing 4.26(2.70-6.70) 4.56(2.70-6.70)* 2.01(1.08-3.72) 2.00(1.07-3.76)* *Adjusted for chronicity   Linear regression analysis suggested that compared to the reference category (BWD below 30%), cord length was 3 cm shorter in twins with severe BWD (≥30%) while adjusted for chorionicity, sex discordance and gestational age (Table 5-5). The impact of chorionicity, however, was of great interest as the same model suggested that adjusted for BWD, sex discordance and gestational age; DC twins had shorter cords compared to MC by about 2 cm. Table 5-5 Regression analysis of cord length and type of cord insertion, beta, significance of  each term in the model, and 95% confidence interval. Dependent variable Independent variable Beta(SE) P value 95%CI Cord length BWD≥30%* Chorionicity  Sex discordance Gestational age -3.13±0.84 0.98±0.53 -1.20±0.46 0.684±0.07 0.060 0.001 0.010 0.001 -0.06 to -2.02 0.89 to 3.12 -2.10 to -0.29 -0.52 to 9.06 Marginal vs. central Velamentous vs. central  Velamentous vs. marginal  BWD≥30%* -0.05±0.25 0.86±0.17 2.23±0.26 0.830 0.001 0.001 -5.40 to 0.432 0.537 to 1.185 1.73 to 2.73 *Adjusted for chorionicity (Ref=MC), sex discordance and gestational age Adjusted for chorionicity and gestational age, ordinal regression analysis of cord insertion type suggested that BWD≥30% was related to type of cord insertion for two comparisons: “velamentous vs. central” and “velamentous vs. marginal” insertion. If the twin pair had severe growth discordance (≥30%), his/her ordered log-odds of having a VCI compared with marginal increased by 2.23. Similarly if a twin pair had severe growth discordance, his/her ordered log- 118  odds of having VCI decreased by 0.86 times of that of a twin with central cord insertion (Table 5-5). Sex discordance is a significant term in the model representing a relationship between cord length and BWD. The above analysis was repeated for BWD≥20% with no significant results (data are not shown). Lighter discordant versus heavier concordant twins A lighter twin was defined as a twin who was lighter than the co-twin within a growth discordant pair. The results for the lighter discordant twins (n=236) were compared to the combined group of heavier co-twins and twins with concordant birth weight (n=1214). Fetal weight to placenta (F/P) ratio was estimated for these two categories. Lighter twins have significantly lower F/P compared with the comparison group (Figure 5-1). Unequal placenta sharing was more frequent among lighter discordant twins compared with the comparison group (52.6% vs. 19.9%, p=0.01). Lighter discordant twins had heavier placentas (584.44±227.61 vs. 511.41±229.29, p=0.01), longer placenta length (23.18±5.01 vs. 21.68±5.03, p=0.01), wider placenta width (17.82±4.16 vs. 16.61±3.89, p=0.01) and shorter cords (27.48±12.47 vs. 25.25±12.03, p=0.01) compared to the combined heavier twins and twins with concordant birth weight.   119   Figure 5-1 A comparison of fetal to placenta weight ratio between lighter discordant twins (n=236) and combined group of heavier co-twins and twins with concordant birth weight (n=1214)  These results were then stratified by chorionicity to determine if these relationships were present in both MC and DC twins. Unequal placenta sharing was more frequent in lighter discordant twins of DC compared to the heavier DC cohort (Figure 5-2). Similarly, lighter discordant twins in the MC cohort had significantly higher frequencies of unequal placenta sharing when compared with their heavier DC counterparts.  In the DC cohort, a significant association was found between placenta weight and BWD. Compared with heavier twins, lighter twins had heavier placentas (549.30±218.57 vs. 482.52±221.02, p=0.01).  Similarly, lighter twins had wider and longer placentas compared to heavier twins (p=0.01 for both variables).   4.354.44.454.54.554.64.654.74.75Lighter twin Heavier twinFetal placenta ratio 120   Note: A lighter twin was defined as a twin who was lighter than the co-twin within a growth discordant pair.  Figure 5-2 Relative frequency of unequal placenta sharing in birth weight discordant twin pairs registered in C&W hospital (1466 pairs, n=2932)  In the MC cohort, cord length was found to be significantly shorter among lighter twins (24.90±11.64 vs. 25.63±12.03, p=0.01) compared with heavier twins. Type of cord insertion was not significantly different in any of the comparisons. Fetal growth restriction and pathology results Overall, the majority of twins were AGA (72.7%, Table 5-1). Stratified growth restriction by chorionicity showed a statistically significantly different growth restriction in the MC group compared with DC (p=0.01).  Overall, 5.8% of twins were SGA. The MC cohort had twice as many SGA twins compared with the DC cohort (8.7% vs. 4.9%). The difference between MC and DC was negligible in terms of frequencies of AGA, while LGA in MC gestations was 5% lower than in DC gestations (Table 5-1).  010203040506070MC_lighter MC_Heavier DC_Lighter DC_heavierUnequal placenta sharingMC_lighterMC_HeavierDC_LighterDC_heavier 121  A further analysis of the result was conducted to determine the association between fetal growth restriction and placental pathology, comparing SGA (n=170) and AGA (n=2133) cohorts. LGA twins were left out of this comparison (n=629). The SGA cohort had higher frequencies of unequal placenta sharing (48.7%) versus AGA twins (28.7%, p=0.01). Furthermore, the SGA cohort had lighter placenta (455.83±200.33 vs. 538.51±229.59, p=0.01) and smaller placenta dimensions (length and width, p=0.01 for both) compared to the AGA group. In addition, length of cord (22.46±11.21 vs. 27.75±12.69, p=0.01) and its distance from the margin of placenta (4.04±1.93 vs. 4.64±2.30, p=0.01), and the other cord (11.61±5.06 vs. 14.30±8.45, p=0.01) were statistically significantly shorter than the AGA cohort.   When further analysis was conducted stratifying the data according to chorionicity, a higher frequency of unequal placenta sharing was observed in SGA twins compared to AGA in the DC (p=0.02) cohort but not in MC twins (p=0.26). The placenta was lighter among SGA twins compared to AGA in the MC cohort (532.48±200.74 vs. 696.24±214.25, p=0.01). Similarly, placenta dimensions were smaller in SGA twins (p= 0.01 for both width and length of placenta) compared to AGA twins. A comparable pattern was observed for the DC cohort. SGA twins also had shorter cords compared to their counterparts in both the MC (21.38±11.15 vs. 26.17±12.46, p=0.01) and DC (23.06±11.24 vs. 28.25±12.73, p=0.01) cohorts. Cord insertion type, chorionitis, and chorionic inflammatory villi were not significantly different in any of the comparison groups, nor did the composite scores. Small numbers in these comparisons were of concern. 5.5 DISCUSSION The aim of this study was to assess the pathological characteristics of placenta and cord in relation to BWD in twin gestations. Our findings suggest that MC twins had a higher frequency of BWD≥30% than DC twins (9.5% vs. 5.0%, Table 5-1). The placenta origin of growth  122  discordance was due to presence of anastomosis, and unequal placenta sharing. Other factors significantly associated with growth discordance were inclusive of the size and weight of the placenta, cord length and cord insertion type.  Twin pregnancies with evidence of growth discordance have a higher risk of adverse outcomes.193–195 BWD has been attributed to chorionicity in many studies.72,83,196 Microscopic and macroscopic examination of the placenta is therefore important in the evaluation of growth discordance.197 Studies in singleton pregnancies have shown smaller, lighter placentas with morphological changes in villus structure in placentas of fetuses affected by intra-uterine growth impairment.198–200  The impaired fetal growth has been attributed to defective trophoblast invasion201 and placental infarcts 57 leading to impaired development of the utero-placenta circulation. The pathophysiological changes in the placenta and its relation with BWD have been studied in the literature.52,57  However, these studies are limited by relatively small numbers, inadequate control of confounding factors, and inclusion of twins with congenital anomalies. Moreover, in cases of multi-central studies, placenta examination was carried out in different pathology laboratories that applied different approaches in examination of placentas and cords. This could lead to variation between pathology results. Our data is based on a large cohort of twin data collected over a 10-year period. We excluded pathology reports of twins born with congenital anomalies, missing birth weight data and those born at less than 20 weeks of gestation.  We have also excluded cases with TTTs, twins with one single stillbirth, twins with smaller than 500 gram weights and those who underwent reduction procedures in utero. Standardized pathology examination techniques have been carried out in all  123  twin pregnancies in one hospital, hence lower variation in pathological reports is expected in our study.   Studies in the literature related to pathological findings of BWD twins are scarce. Several studies have focused on MC placentas only and suggested that growth discordance in twin pregnancies is attributed to TTTs 202,203 while others evaluated a single element, for instance placenta size/ placenta weight,204,205 placenta pathology or cord abnormalities in relation to BWD of twins.206,207 Our study provides a comprehensive picture of pathological changes of the placenta and cord in both MC and DC twins with or without BWD.  Anastomosis In our hospital-based cohort of 1466 twin pregnancies we have established associations between moderate and severe BWD and chorionicity. There were higher frequencies of placenta anastomosis among MC twins than DC ones as expected (Table 5-2).  Placenta anastomosis in MC is ascribed as the cause of BWD, while in DC twins discordant growth is linked to differences in placental mass or differences in placental parenchymal lesions.208  The number of cases of placenta anastomoses in our study was around 263 (9.0% of total population).  This observation is based on data where twins with TTTs were excluded. Our study limitation is a non-differentiation diagnosis between “arterial-to-arterial” anastomosis versus “vein- to- artery” anastomosis.    Unequal placenta sharing Unequal placenta sharing is considered an abnormality of placentation as it affects the available nutrient delivery, gas exchange and waste removal which may directly impact fetal growth.189 Apart from genetic potential, crowding in utero, and placenta insufficiency, unequal sharing of  124  placenta mass (or an unequal placental area) is an explanation for discordant fetal growth in DC.57 Unequal placenta sharing is also reported in MC twins. A retrospective study reported a prevalence of 154 placentas out of 395 MC twin placentas(39%).208  Our study provides data for both MC and DC cohorts where we showed a higher frequency of this condition in MC twins compared to DC.  The relationship with BWD and unequal placenta sharing was also studied in a prospective cohort of 522 MC/DC twin pregnancies (1997-2003) using Dye injection studies on fresh postpartum placentas.  Unequal placenta sharing was defined as 1 twin receiving blood from >60% of the placenta. The study revealed that 50% of the cases with BWD>20% had unequal placenta sharing compared with 10.4% of the cases without unequal placenta sharing (p=0.01).209  The phenomenon of increasing placenta territory discordance leading to discordant growth is thought to result from unequal splitting of the initial cell mass.210  Another small prospective study of 100 MC placentas investigated the relationship between placenta territory and BWD from pregnancies not complicated by TTTs. All samples were from pregnancies with two live born twins. Placental territory discordance was calculated by dividing the venous surface area of the larger placental part by that of the smaller. Unequal placental sharing was defined as a placental territory discordance of ≥1.5. Placental territory discordance increased with BWD (p=0.001).189 Our study showed that regardless of the level of BWD severity, a higher percentage of sharing of placenta is expected among twins with BWD compared to concordant twins.  The relationship between BWD and placenta sharing remained strong even after adjustment for chorionicity (OR 4.56 vs. 4.26 for BWD≥20% and 2.01 vs. 2.00 for BWD≥30%).  Higher odds for the threshold  125  level of ≥20% compared to ≥30% could be the result of various depths of anastomosis or type of anastomosis (artery-artery vs. artery-vein).  Sex discordance was found to be a significant predictor in the relationship between BWD and placenta characteristics. To the best of our knowledge, no study to date has investigated this relationship. These results will be discussed in the next Chapter, where we examine the role of sex discordance in relation to placenta characteristics. Placental lesions Placental histological examination is critical in the evaluation of intrauterine growth impairment. Singleton pregnancies are shown to have higher rates of morphological changes in villus structure in placentas of fetuses affected by intrauterine growth restriction.198,200 The abnormalities in the villus structure are due to defective trophoblastic development leading to impaired utero-placental circulation. An association between placenta infarcts and impaired utero-placental circulation of growth restricted fetuses is reported in singleton pregnancies.201   In twin pregnancies, such a relationship has also been previously studied 52,211 but publications on gross and histological placental examination of MC and DC twins are rare. Two retrospective studies evaluating placenta histology have been limited by a rather smaller sample size.52,57  A retrospective study of 147 twin pairs (99 DC and 48 MC) suggested that in DC twins a BWD >20% is attributable to a greater number of placental lesions in the lighter twin than in the heavier twin (p < 0.05).57 The other study included 388 DC and 89 MC twin pregnancies and found significantly more vascular thrombotic lesions in the placental domains of smaller twins in DC pairs.52  126  A prospective study of 668 twin pairs (21.1% MC and 78.9% DC) analysed a composite variable including evidence of infarction, retro-placental hemorrhage, chorangioma, subchorial fibrin, and abnormal villus maturation.  These histological abnormalities were more frequent in placentas of smaller twins of BWD pairs (p=0.02) and in placentas of SGA infants (p=0.01).  Such associations were observed in DC twins but not in MC ones.207 We did not find these differences in our study, potentially due to a smaller number of lesions in our study population which could be due to the retrospective nature of our data. The incidence of composite placenta lesions in our study was 54 (13.2%) with a slightly higher frequency in MC compared to DC twins (2.6% vs. 1.6%), although this difference was not significant.  Another reason for differences between our findings and the above-mentioned study could be our higher number of twins, specific differences in defining the composite variable, or measurement bias. Our composite placental pathological lesions included placental infarction, chorangioma, subchorial fibrin deposition, and retro-placental hematoma.  Cord length Information about cord length in twin pregnancies is scarce. Twins are reported to have, on average, a 7.9 cm shorter umbilical cord compared with singleton pregnancy.212 A higher incidence of categorised short umbilical cords (less than 35 cm in length) is also reported in twins compared to single pregnancies.38 In our research, the average length of umbilical cord was shorter in MC compared to DC twins and in growth discordant twins compared to concordant ones.  The relationship between BWD and length of cord has been investigated in the literature controlling for the impact of gestational age.213  We investigated such a relationship in our study.  127  Adjusted for gestational age, sex discordance, and chorionicity, the BWD was statistically significantly related to cord length suggesting that shorter cords by average of 3.13 cm are associated with severe growth discordance.  Additionally, our results suggest that shorter cords yield lighter twins. Such a finding has not been reported in the literature.  A small study of 24 MC and 200 DC placentas214 showed statistically significant differences between the length of the umbilical cords in MC twins and in DC ones. Unlike our study, this study included TTTs cases in the analysis. No information about BWD was reported in this study.  Our findings should be interpreted carefully as cord length is always affected by where it is clamped by the health care professional. It is easy to randomly cut the cord at different lengths. For instance, blood withdrawal is required to measure cord blood gases. In such cases, cord can be cut longer. Moreover, all cord segments might not be handed in for pathology examination. Although, the pathology reports did make a note of several pieces of cords examined macroscopically and the total length of card was recorded.  Velamentous cord insertion (VCI) An association between VCI and compromised fetal growth in singleton pregnancies is a longstanding discovery.215 In twin gestations, the effect of VCI is dependent on chorionicity. A high incidence of VCI in twins (21%) was reported to be significantly more common in MC than DC twin pairs (18% vs 6%; p=0.001).60 Moreover, the effect of VCI on BWD has been reported to be 13.5 times greater for MC twin gestations than for DC twins.  In our study, VCI was associated with BWD. Severely growth discordant twins had higher odds of getting their nutrients from a placenta with velamentous cord insertion than a placenta with a central cord.  This finding has been reported in previous studies with a smaller sample size of  128  MC twins. 216  A matched case-control study of 47 twin pairs compared the placental characteristics of MC placentas from pregnancies with BWD >25% with a control group of MC placentas without BWD, matched for gestational age at birth.  The rate of VCI was significantly different in BWD twins compared with concordant ones (30% vs. 16%, p=0.036).132 In our study we adjusted the impact of gestational age and found similar results. A prospective study of 319 twins also found a higher frequency of VCI in BWD of ≥20% compared with growth concordant twins (22% vs. 8%, p=0.001). A total of 369 MC placentas were analyzed in another study of MC placentas, and the frequency of VCI was found to be 36% in BWD twins compared to 21% in normal MC group (p=0.06).217 A higher incidence of VCI is reported in twins with TTTs than in non-TTTs placentas. In view of this finding, an etiologic role for VCI in the impaired growth of fetus is linked with TTTs. The authors proposed that  TTTs could result from hemodynamic instability due to reduced blood flow to the donor twin with a VCI, hence growth impairment in a donor twin.218  In our study, we have excluded twins with TTTs and still found a relationship between VCI and BWD despite adjustment for chorionicity.  This conclusion is in the light of the fact that we were unable to show any association between VCI and placenta sharing, or other placenta lesions. Thus, the reason(s) for association between BWD and VCI remains unknown.  Marginal cord insertion  Our findings suggest that it was more likely to show growth discordance at birth if the cord is inserted into the placenta marginally rather than centrally.  This finding is in agreement with the result of a study of 60 MC twins in that when one placental cord insertion site was central and  129  the other was marginal, 33% of twin pairs exhibited a BWD of at least 20%.219  The study did not separate the marginal and velamentous cord insertion due to low numbers.  A prospective larger study of cord insertion type and growth discordance was also conducted on a large population of 806 twin pairs (165 MC, 651 DC). Monochorionic twins had higher rates of marginal (p=0.0068) placental cord insertion.  A significant association between non-central (combined marginal and velamentous) placental cord insertion sites and growth abnormalities was reported in MC twins but not in DC twins.120 Our study finding is in agreement with this study because our ordinal regression model showed higher odds of BWD in marginal versus velamentous cord insertions. These odds were adjusted for chorionicity, sex discordance and gestational age. 5.6 SUMMARY Our study focused on the relationship between the type of the twin placenta and BWD of the twins in a large cohort study from one major hospital in BC. The large sample size, similar chorionicity verification in one laboratory, and ability to control for confounding variables such as gestational age and sex discordance enabled us to control for several major confounders.   First, MC twins had a higher frequency of BWD≥30% than DC twins, suggesting that the single placenta is less efficient than the DC placenta to nurture twins. In other words, there is more competition for nutrition and oxygen between pairs among MC twins compared with DC twins.   Secondly, we attempted to assess the placental origin of aberrant growth among twins, mainly in correlating the presence of anastomosis, unequal placenta sharing, and pathological lesions as well as investigating the size of the placenta, inflammatory reactions, cord length and cord  130  insertion type. From these variables, anastomosis, unequal placenta sharing, cord size and cord insertion type were found to be the key elements that impacted growth aberration.  We hypothesised that MC placentas compensate for lack of nutritional flow by infiltrating to other placenta surfaces, mainly due to the following findings: 1) Higher rates of severe growth discordance were negatively associated with higher frequencies of anastomosis. 2) Placentas in MC twins were more likely to have shared arteries/veins.  Compensation for lack of vascular sufficiency would mean a fused placenta or sharing more portions of the placenta.  This translates to higher rates of unequal placenta sharing among growth discordant twins irrespective of chorionicity, as suggested by the result of multivariable regression models.   Other than anastomosis and unequal placenta sharing, the shorter cord length and non-central cord insertion type were associated with BWD.     131  CHAPTER 6 FETAL SEX AND PLACENTAL PATHOLOGY  6.1 INTRODUCTION Adverse perinatal outcomes and fetal growth have been found to be associated with fetal sex.32  The placenta, as an important part of the fetus’s nutritional route, plays a central role in mediating growth and development of fetuses. Sex differences in growth discordance and fetal/neonatal morbidities and mortality are likely to be mediated by sex-specific placenta functions. None of the findings in the literature has led clinicians to pay attention to sex during perinatal screening, mostly because the mechanisms that confer these differences between the sexes are unknown. It is also possible that the association between adverse perinatal outcomes and fetal sex is not that strong or that other factors, rather than fetal sex, such as chorionicity play a stronger role. 6.2 OBJECTIVE Our study aimed to investigate the pathological characteristics of the placenta associated with fetal sex.  6.3 METHODS In brief, a retrospective cohort of twin pregnancies born in British Columbia Women’s Hospital for a period of a decade was studied. Data were abstracted from hospital charts in the pathology department. For definitions of variables related to placenta and cord variables, see Chapters 3 and 5. Twin pregnancies were included in the data if mothers and babies were linked, babies had a calculated estimated gestational age at birth ≥20 weeks, and the mother did not have a termination procedure. We excluded cases with congenital anomalies, TTTs, <500 grams’ birth  132  weight, one stillbirth, those who had a reduction procedure in a multiple pregnancy greater than two (3 to 2 or 4 to 2), and papyrus placentas. We chose to exclude the latter category of vanishing twins from this analysis because of an association between sex discordance and vanishing twins.121  No exclusions were made because of pregnancy complications, and therefore selection bias is not expected to have occurred. Information about placenta examination, linkage of data and ethical approval is provided in Chapters 3 and 5. Fetal to placenta (F/P) ratio was calculated by dividing birth weight by placenta weight in grams. The frequency of placental and cord characteristics was compared between MC and DC twin pregnancies. Incidences of placenta and cord characteristics were compared between male and female. Bivariate analysis was used to determine significant variables that were to be included in the regression analysis. We then investigated the relationship between sex pairing and pathological findings. Sex pairing was a composite of the twin’s sex and with what type of twin pair he or she was identified: male from a male-male pair, male from a male-female pair, female from a female-male pair, and female from a female-female pair. Infants from a female-female pairing were considered as the referent group. We were interested to account for a direct association between sex and pathological findings; hence we first chose a population average analysis. Second, we estimated the total association of fetal sex pairing on pathological outcomes, therefore presenting the comparisons employing a subject-specific approach. For more information about definition of variables, see section 5.3. 6.4 RESULTS The result of placenta examination was available for 1571 twin pairs born in C&W hospital. There were 1493 pairs (n=2986) in the analytical data.   133  From 2986 cases included in the study, 26% (n=768) had MC and 74% (n=2218) had DC placentas. Clinical characteristics of the cohort are shown in Table 6-1. The majority of twins were male (51.4%). About 2/3 of twins were sex concordant (67.5%). Of the sex concordant twins, similar percentages were male-male (MM: 35.1%) and female-female (FF: 32.3%). Of the sex discordant twins, the male-female (MF: 32.5%) group constituted about 1/3 of the whole population, and about 33% of infants were in the female-female group; this last group was considered the reference group for further analysis. MC twins were on average 341 grams lighter than their DC counterparts and were delivered at an earlier mean gestational age. The mode of delivery was more frequently cesarean section than vaginal delivery (62.2% vs. 37.8%, respectively). DC twins a had higher frequency of cesarean section than MC twins (p=0.01) Table 6-1 Comparing clinical characteristics of monochorionic and dichorionic twins born in C&W hospital (1493 pairs, n=2986)   Overall Monochorionic N=768  Dichorionic  N=2218 P Sex Male Female  1534(51.4%) 1451(48.6%)  404(52.7%) 362(47.1%)  1128(50.9%) 1090(49.1%0  0.20 Sex discordance  Concordant Discordant  1930(67.5%) 930(32.5%)  748(97.2%) 20(2.6%)  1252(56.5%) 966(43.5%)  0.01 Sex composition Male-Male Female-Female Male-Female  1005(35.1%) 925(32.3%) 930(32.5%)  392(51.0%) 356(46.4%) 20(2.6%)  649(29.3%) 603(27.2%) 966(43.5%)  0.01 Gestational age mean±SD 34.25±3.68 33.02±4.30 34.69±3.32 0.01 Birth weight mean±SD 2245.22±696.45 1991.21±761.95 2332.03±646.68 0.01 Mode of delivery Vaginal Cesarean section  1082(37.8%) 1778(62.2%)  324(42.2%) 444(57.8%)  758(36.2%) 1334(63.8%)  0.01 MC: Monochorionic; DC: Dichorionic; SD: standard deviation  134  Relationship between pathology results and fetal sex: A population average approach Pathology outcomes of the study samples in response to sex are summarized in Table 6-2. The sample was composed of 1534 males (51.4%) and 1452 females (48.6%). Two cases were identified with unknown sex and were not included in the analysis.  Table 6-2 Characteristics of placenta and cord, overall and stratified by sex for twins born in C&W hospital (1493 pairs, n=2986)   Overall Male N=1534 Female N=1452 P  Placenta Chorionic villi inflammation 264(9.2%) 149(10.1%) 115(8.3%) 0.09 Chorionitis 208(7.3%) 121(8.2%) 87(6.3%) 0.04 Anastomosis 252(8.8%) 155(10.6%) 96(6.9%) 0.01 Unequal placenta sharing* 142(32.6%) 84(37.2%) 58(27.6%) 0.04 Composite of inflammation 378(13.2%) 214(14.6%) 164(11.8%) 0.03 Composite placenta lesions 50(1.7%) 33(2.2%) 17(1.2%) 0.05 Cord composite 9(0.3%) 6(0.4%) <5 (0.3%) 0.36 Placenta others^ 36(1.3%) 20(1.4%) 16(1.2%) 0.74 Placenta weight, g, mean±SD 532.62±228.97 536.57±228.44 528.56±229.68 0.46 Placenta length, cm, mean±SD 22.02±5.14 22.16±5.21 21.87±5.06 0.23 Placenta width, cm, mean±SD 532.62±228.97 17.05±16.94 16.94±3.93 0.55 Placenta thickness, cm, mean±SD 2.12±0.47 2.11±0.51 2.13±0.41 0.37 Cord Cord length, cm*, mean±SD 27.33±12.70 28.03±13.14 26.57±12.21 0.02 Cord distance from margin, cm, mean±SD 4.60±2.27 4.58±2.23 4.63±2.31 0.59 Cord distance from membrane, cm, mean±SD 6.52±5.14 6.63±3.96 6.40±4.01 0.41 Cord distance from other cord, cm, mean±SD 14.14±8.43 6.63±3.95 6.40±4.01 0.24 Cord insertion type Marginal Central Velamentous  1981(69.3%) 469(16.4%) 118(4.1%)  1029(78.0%) 228(17.3%) 62(4.7%)  950(76.2%) 241(19.3%) 56(4.5%)  0.41 *Mann Whitney test; SD: standard deviation ^Refers to Placenta previa, placenta abruptio, placenta accreta, and other types of placenta separations that causes hemorrhage  135   Higher frequencies of chorionitis, anastomosis, unequal placenta sharing, inflammation and placenta lesions were found in males compared with female twins. Cord length was about 1.5 cm longer in males compared with female twins. These variables were then further analysed by regression analysis.  For the variable of cord length, risk ratio and for the others odds ratios were estimated (Table 6-3).  Table 6-3 Regression analyses of pathology findings, comparing male and female twins born in C&W hospital (1493 pairs, n=2986)   Unadjusted (95%CI) Adjusted (95%CI) Chorionitis 1.34(1.01-1.79) 1.38(1.04-1.84)* Anastomosis 1.59(1.22-2.07) 1.63(1.22-2.19)** Unequal placenta sharing 1.5(1.03-2.33) 1.72(1.11-2.64)* Composite of inflammation 1.27(1.02-1.58) 1.30(1.05-1.62)* Composite placenta lesions 1.86(1.03-3.35) 1.83(1.02-3.31)* Cord length 1.46 (0.48-2.44) 1.35(0.39-2.30)*** Ref: Female; *Adjusted for chorionicity; BMI>30% and gestational age; ** Adjusted for chorionicity and gestational age; ***Adjusted for BWD>30% and gestational age; &Two individual twins with unknown sex were excluded  Compared with females, the odds of chorionitis, anastomosis, unequal sharing of placenta, placental inflammation and lesions were higher in male twins, adjusted for chorionicity, BWD≥30% and gestational age.   Relationship between pathology results and fetal sex: A subject-specific approach We were interested in estimating the total association of fetal sex pairing on pathology adverse outcomes. Of the eight categorical outcomes under study, unequally shared placenta and anastomosis ns placental inflammation were the most common (Table 6-4).  Seven variables  136  were found to be significantly differently distributed between the sex pairing groups, including anastomosis, unequal placenta sharing, composite of inflammation, placenta weight, length and width. Cord length was also significantly different between sex pairs of different types. However, except for anastomosis and unequal placenta sharing, the difference between categories was clinically small. Thus no further analysis was performed.  137  Table 6-4 Association between sex pairing and placental /cord pathology findings in twins born at C&W hospital (1493 pairs, n=2986)   F from FF n=925 M from MM n=1005 M from MF n=483 F from MF n=483 P Placenta Chorionic villi inflammation 74(8.0%) 112(11.1%) 37(8.0%) 41(8.8%) 0.07 Chorionitis 53(5.7%) 89(8.9%) 32(6.9%) 34(7.3%) 0.07 Anastomosis 88(9.5%) 144(14.3%) 11(2.4%) 8(1.7%) 0.01 Unequal placenta sharing 47(30.1%) 72(42.1%) 12(21.8%) 11(20.4%) 0.01 Composite of inflammation 104(11.2%) 160(15.9%) 54(11.6%) 60(12.9%) 0.01 Composite placenta lesions 10(1.1%) 23(2.3%) 10(2.2%) 7(1.5%) 0.19 Cord composite <5 (0.5%) <5 (0.5%) <5 (1.0%) 0(0.0%) - Placenta others^ 11(1.2%) 17(1.7%) <5 (1.0%) <5 (1.0%) 0.38 Placenta weight, g, mean±SD 557.17±231.99 558.17±232.75 495.46±214.42 475.52±215.88 0.01 Placenta length, cm, mean±SD 22.17±4.91 22.46±5.24 21.55±5.11 21.30±5.29 0.01 Placenta width, cm, mean±SD 17.44±3.93 17.47±4.20 16.36±3.79 16.99±4.02 0.01 Placenta thickness, cm, mean±SD 2.16±0.41 2.11±0.48 2.11±0.54 2.08±0.42 0.10 Cord Cord length, cm*, mean±SD 25.97±11.98 28.08±13.22 27.92±12.87 27.78±12.58 0.01 Cord distance from margin, cm, mean±SD 4.62±2.39 4.53±2.41 4.67±2.22 4.64±2.15 0.75 Cord distance from membrane, cm, mean±SD 6.45±4.08 6.62±3.99 6.66±3.89 6.30±3.83 0.85 Cord distance from other cord, cm, mean±SD 13.35±5.71 14.75±11.52 13.68±5.45 14.73±8.43 0.16 Cord insertion type Marginal Central Velamentous  612(75.0%) 162(19.9%) 42(5.1%)  724(79.5%) 143(15.7%) 44(4.8%)  305(74.8%) 85(20.8%) 18(4.4%)  338(78.4%) 79(18.3%) 14(3.2%)  0.13 ^Refers to Placenta previa, placenta abruptio, placenta accreta, and other types of placenta separations that causes hemorrhage SD: Standard deviation; M: male; F: female; FF: female-female; MM: male-male; MF: male-female  138  Table 6-5 Logistic regression analysis of anastomosis in twins born at C&W hospital (1493 pairs, n=2986)   Anastomosis Beta ± SE P Odds (95%CI) -2 Log likelihood Model 1 F from FF vs. M from MM M from MF F from MF Ref. 0.46±0.14 -1.47±0.33 -1.79±0.37  0.001 0.001 0.001  1.59(1.20-2.11) 0.23(0.12-0.44) 0.17(0.08-0.35)  1592.14 Model 2 F from FF vs. M from MM M from MF F from MF Chorionicity Ref. 0.54±0.16 0.37±0.37 0.21±0.42 3.26±0.23  0.001 0.318 0.616 0.001  1.71(1.26-2.33) 1.45(0.70-3.01) 1.24(0.54-2.83) 25.96(16.53-40.77)  1235.71 M: male; F: female;  FF: female-female;  FF: female-female;  MM: male-male;  MF: male-female;  CI: confidence interval;  SE: standard error Table 6-5 shows a logistic regression analysis for categorical variable of anastomosis. Model 1 represents an unadjusted association between sex pairing and anastomosis while model 2 is an adjusted model for chorionicity.  The second model is a better fit due to lower -2log likelihood. Compared to females from the FF group (the reference category), twins of either sex from mixed-sex pairs were less likely to have anastomosis. Males from MM pairs had a statistically significant increase in their odds of anastomosis compared with females from FF pairs (1.71; 95%CI 1.26-2.33).  The odds of unequal placental sharing were highest in males from male-male pairs (1.75, 95% 1.14-2.69) compared with the females of concordant pairs (FF, data is not shown). After adjustment for chorionicity, the odds remained statistically significantly high (1.74, 95%CI 1.13-2.69) compared to the reference category. Similarly, the adjusted (for chorionicity and  139  gestational age) odds of composite inflammation were higher in males with male-male status compared to females of female-female pairs (1.52. 95%CI 1.18-1.94). Table 6-6 Linear regression analyses of placenta weight and sizes in twins born at C&W hospital (1493 pairs, n=2986)   Model 1 Sex pairing Model 2 Sex pairing and chronicity Model 3 Sex pairing, chronicity and gestational age Placenta weight Beta± SE 95%CI R2  -15.35±5.11 -25.37 to -5.33 0.071  -12.37±4.87 -21.94 to -2.81 0.309  -9.45±4.46 -18.19 to -0.70 0.495 Placenta length Beta± SE 95%CI R2  -0.33±0.12 -0.55 to -0.10 0.068  -0.13±0.12 -0.36 to 0.108 0.147  -0.07±0.12 -0.29 to 0.16 0.302 Placenta width Beta± SE 95%CI R2  -0.02±0.01 -0.04 to -0.00 0.055  -0.01±0.01 -0.03 to 0.01 0.109  -0.07±0.01 -0.03 to 0.01 0.237 Ref: Female from FF,  Other groups: female-female; female-female; male-male; male-female;  CI: confidence interval;  SE: standard error;  R2: R squared  Linear regression analyses were used to analyse the association between sex pairing and placenta weight, length, and width. Three models for each variable are presented in Table 6-6.  The first model is unadjusted while the 2nd and 3rd models are adjusted step by step. Although the role of sex pairing on placenta size is reduced by introducing co-variates at each step of the analysis (from model 1 to model 3), placenta weight is moderately correlated with sex (Beta ± SE =           -9.45±4.46). Model three is the best model as R squared shows that the model explains about 50% of variation.   From one level of sex pairing to the other, placenta weight decreases, on average, by about 10 grams. In other words, a male from either a MM pair or a mixed-sex pair has a statistically significant reduction in their placental weight compared with infants from a FF  140  twin pair.  Similar to placenta weight, placenta sizes (length and width) on average were shorter among other categories compared with females of FF pairs, the reference category. These values are not clinically significant. Figure 6-1 shows the average of fetal to placenta weight ratio (F/P) in different categories of sex pairing. The average F/P ratio was 4.88±2.46. Comparing all groups with each other, an ANOVA test was significantly different (F=18.02, p=0.001). Females from female-female twin sets had the lowest F/P ratio while males from sex discordant twin pairs had the highest ratio. A female sharing the uterus with a male mate had a higher F/P ratio compared with the reference category, females from female-female pairs. Linear regression analysis did not reveal any statistically significant result for the relationship between sex pairing and F/P ratio.  Figure 6-1 Fetal to placenta weight ratio in sex pairing categories in twins born at C&W hospital (1493 pairs, n=2986) M: male; F: female; FF: female-female; FF: female-female; MM: male-male; MF: male-female  0123456Fetal placenta ratioF from FFM from MMM from MFF from MF 141  6.5 DISCUSSION This analysis of a retrospective cohort of pathological data revealed an association between twin sex and pathology characteristics of placenta and cord. A paucity of data exists in the literature regarding the relationship between sex and pathology of the placenta and cord. Previous studies of pathology findings tend to focus on factors other than infant sex, such as BWD.220 If sex was noted as a secondary risk factor of interest, then simple analytical differences between male and female in the placenta’s adverse outcomes were investigated rather than the risk of adverse pathological events by sex pairing.221 Given that the intra-uterine environment shared by a set of twins could impact the outcomes of interest in this study, we decided to utilise two analytical approaches to investigate the role of sex in relation to pathological outcomes. We first analysed the data using a population-average approach where a selected twin from the population is compared with another twin who is also selected from the population. We then compared each twin with his or her mate, illustrating the use of a subject-specific-model rather than a population-based approach. Hence, sex pairing comparisons were adopted for analysis.  Based on the population average approach, we found that males had higher incidences of chorionitis, anastomosis, unequal sharing of placenta, placental inflammation and placental lesions than females. Length of cords was slightly higher for males than females. To the best of our knowledge, there is no study to date that replicates these findings in such a comprehensive manner.   142  Vascular Anastomosis Evidence suggests that sex specific adaptation of the placenta may be central to the differences in fetal growth and survival.222  Studies consistently report that adverse fetal and neonatal morbidity relates to sex specific differences.223,224  However, sex specific differences of placental and cord pathology is rarely noted in the literature. Our study showed the importance of sex-specific findings related to the placenta and cord. Using the subject-specific approach, we found that males whose co-twin also was a male were at an increased risk for anastomosis compared with females from female-female twin pairs. Vascular anastomosis is characterized as an artery or vein captured by one twin where its partner artery or vein is captured by the second twin.  Differential vascular capture may result in discordant twin outcomes. A higher frequency of anastomosis in males compared with females was found in our analysis. Vascular anastomosis or sharing is almost foreseeable in MC twinning.225 When we controlled for the effect of chorionicity, the adjusted odds for anastomosis was still 63% higher in males than females (Table 6-3). Similarly, chorionicity adjustment for analysis of the relationship between anastomosis and sex pairing showed 70% higher odds of anastomosis in males from male-male twin sets compared with females from female-female twin sets (Table 6-5).  Moreover, the presence of a male in the uterus led to increased odds of vascular anastomosis in females who share the uterus with males (24% increase in odds, 95%CI 0.54-2.83). This finding suggests that both single and double males in the uterus are associated with higher frequencies of anastomosis. This finding, however, should be interpreted carefully as we had a few cases of discordant gender among MC twins. These cases have been reported in case studies before and could be in the form of 46XY and 46XO (due to genetic mitotic disjunction) or cases of mosaicism.226 It is  143  also possible that these cases of discordant gender among MC twins are misclassified in the laboratory.  Fetal sex can be diagnosed in utero by routine ultrasound examination, and based on our findings should be an integral part of prenatal investigations in twins. However, sex is generally ignored when studying obstetric complications of pregnancy due to two main reasons: 1) lack of evidence and 2) unknown mechanisms that confer pathological differences between the sexes. A MC twin placenta is designed to feed one fetus; hence, attempts to cater for the needs of twin fetuses are often suboptimal leading to growth discordance. DC placentas are designed to function separately, but vascular anastomoses are found in pathological examination. This could be due to a compensation mechanism adopted by the placenta for the lack of proper placental size to nourish competing twins. The twin fetal circulations are, therefore, rarely detached and several inter-twin vascular transportations of various kinds may be present, irrespective of chorionicity.  In these circumstances, knowledge of fetal sex and sex pairing can assist clinicians to predict the adverse pathological findings and thus be prepared for clinical intervention or closer monitoring of twins. Our findings suggest that male fetuses deserve more attention, as higher frequencies of anastomosis are expected for them. Similarly, chorionitis, unequal placenta sharing, composite inflammation, and placental lesions are found with higher risk in gestations where at least one of the co-twin is male. Unequal placental sharing Unequal sharing of the placenta is often found in MC placentas, which in turn results in growth discordance.203,227 Whether or not the unequal placenta sharing is more common in males because of genetic predisposition, or the discharged male hormones causes the unequal pattern of  144  parenchymal sharing, remains unknown. Further prospective studies are needed to answer this important question by investigating male hormones and their association with placental unequal sharing among monozygotic twins versus dizygotic twins.  Placental inflammation In singleton pregnancies, intra-uterine infection/inflammation of male fetuses has been attributed to higher mortality of male newborns. A histological examination of placentas and umbilical cords of 446 infants born at 23 to 32 weeks, cultured for aerobic and anaerobic bacteria, showed that male infants were significantly more likely to have placental membrane bacterial infection than female infants.228 Another prospective study of 437 singleton pregnancies delivered between 22 to 32 weeks of gestation investigated the histopathological placental findings. Placental histology showed no association between fetal sex and lesions of acute inflammation (p=0.08).229 Chronic placental inflammation, however, was significantly more noted in male than female fetuses. The authors hypothesised that the chorionic placental inflammation is observed more frequently in males because of some kind of immunological response. In other words, the mother’s body consider the male as a foreign body. This causes a maternal immune response against the male trophoblast, a situation that is also found in transplant rejection.230 We hypothesis that a similar mechanism is more likely to apply to twin pregnancies, as according to our results males of both sex concordant and sex discordant twin pairs had higher frequencies of inflammation compared with females of female-female twin pairs.  No twin study to date has investigated the relationship between sex pairing and placental inflammation. This finding is interesting to be pursued further by basic scientists.    145  Placenta weight Placental weight is a measure which reflects several aspects of placental development, counting the expanding growth of the chorionic disc, increased number of chorionic villi, and increased surface area for vascular nutrient interchange between maternal and fetal sides of the placenta. Therefore, the growth of the chorionic disc is the main determinant of its transfer capacity and an indicator of potential biological fetal growth.231  Placenta weight was not statistically significantly different when the average placenta weight in the male population was compared with that of the female. However, the association between sex and placenta weight was revealed using the analytical subject-specific approach. The adjusted relationship between sex and average placental weight showed a statistically significant but clinically insignificant reduction of 5.9 grams in placenta weight when each group (M from MM, M from MF and F from MF) were compared with females of female-female pairs.  We then calculated the fetal/placenta weight (F/P) ratio which is used as an index of placental nutrient efficiency and has been associated with adverse placental and perinatal outcomes.232 Studies on singleton pregnancies have shown an association between F/P weight ratio and fetal sex. In a retrospective large Japanese study of 53650 singleton pregnancies, F/P was smaller in female than male infants irrespective of parity and gestational age.233 Our finding is in agreement with this study. These data may be useful for further clarifying the fetal-placental pathophysiology in fetal growth restriction and its related conditions. 6.6 SUMAMRY In summary, ultrasound screening of twin gestation is mostly focused on the anatomical features of fetuses, on amniotic fluid volumes, and on umbilical vascular flow rather than sex. Clinicians  146  look for complications such as TTTs, growth discordance, fetal death of one twin, and the differences between superficial and deep vascular connections, rather than sex.  Fetal sex is an important predictor in placental findings including anastomosis, unequal placenta sharing, placental lesions and placenta inflammation. These signs and symptoms should be thoroughly screened, especially if the fetus is male and is from a male-male pair. Although these findings are clinically important, they might not lead to change in practice until more evidence is provided by basic sciences as what are the mechanisms behind these results.     147  CHAPTER 7 OPTIMAL THRESHOLD FOR BIRTH WEIGHT DISCORDANCE, DOES KNOWLEDGE OF CHORIONICITY MATTER? 7.1 INTRODUCTION The perinatal mortality of twin pregnancies is increased compared with singletons.1,2 BWD is found to be a predictor of mortality and morbidity in twin gestation.3–5  However, the literature lacks consensus on what constitutes discordant growth. Proposed thresholds for clinically meaningful BWD rely on determination of that level of discordance above which perinatal mortality increases. This level of discordance ranges from 18 to 30%.6–11 The wide variation observed in reported “significant” thresholds for discordant growth may be explained by a number of factors. 1) Some studies did not exclude cases of major congenital anomalies.18 Given the strong association between abnormal major fetal anomalies and aberrant growth patterns, their inclusion can bias the study. 2) Some other studies did not acknowledge the critical impact of mono-chorionicity,149 which is associated with TTTs and related perinatal adverse outcomes. Not excluding cases of TTTs, therefore, creates a challenge in the determination of an optimal BWD threshold. 3) The clinical importance of BWD as an independent predictor of adverse perinatal outcomes depends on the optimal balance between the sensitivity and specificity of BWD threshold associated with perinatal loss. Such a balance can be compromised if the measured adverse outcomes or predictors are not defined clearly. For example, stillbirth may be defined as death of the fetus in utero prior to birth or fetal demise before and during labour. 4) Prediction of birth weight is less accurate in twins.7 The “actual BWD measured” after birth, rather than the “estimated inter-twin birth weight disparity” measured during ultrasound screening, is more accurate in determining the threshold level associated with the adverse perinatal outcomes. For this reason, we chose to analyse the actual birth weight data to find the  148  optimum threshold level beyond which perinatal loss is to be expected. Studies that employ “estimated fetal birth weight” (EFW) rather than the “actual birth weight” for research purposes to determine growth discordance lack precision. Ultrasound screening is not reliable in providing an accurate estimate of BWD due to higher random errors in ultrasound readings of twin gestations compared with singletons.12 5) Estimated BWD thresholds in the literature are based on simple linear regression analysis. Employment of receivers operating characteristic (ROC) curve analyses and survival analysis best serve this type of analysis.  We assessed “actual fetal birth weight” of twins born with no congenital anomalies or TTTs in which we tested the predictive ability of BWD by means of ROC curve analyses and survival analyses. The impact of chorionicity on reliability of BWD in detecting perinatal loss was estimated. The current study aimed to establish the optimal threshold of BWD for prediction of perinatal mortality and morbidity.   7.2 OBJECTIVE  The study objectives were 1) To establish the optimal threshold of BWD for prediction of stillbirth, perinatal mortality and morbidity in twins born in hospital cohort adjusted for chorionicity. 2) To repeat the analysis in a larger twin cohort without chorionicity information and compare the cut-off points between the two populations. 7.3 METHODS For definitions of BWD and chorionicity, see Chapter 3. We excluded cases with TTTs, congenital anomalies, newborns <500 grams in weight, those with reduction procedures, papyrus twins and placentas, as well as twin pairs with one stillbirth.  149  The predictive performance of BWD from 20 weeks’ gestation for perinatal loss was assessed using ROC analysis. After establishing an optimal cut-off for BWD, the prospective risk of perinatal loss above and below this cut-off was calculated. Positive likelihood ratio was used to estimate test accuracy. LR+ values >10 were considered to provide a ““conclusive”241 increase in the likelihood of disease while scores 5-10 and <5  were categorized as “moderate” and “marginal” increases in likelihood of disease, respectively.117 Survival trends from 20 weeks’ gestation until delivery according to different degrees of BWD           were assessed using time-to-event analysis (Kaplan–Meier), in which duration of gestation was used as the time scale and stillbirth/perinatal mortality as the event. Survival data were plotted as cumulative percentage without event and compared by the log-rank test. Finally, cox proportional hazard analysis was carried out in order to take confounders such as chorionicity into account for the estimation of survival rate among those with and without growth discordance in studying different levels of BWD.   7.4 RESULTS Determining the optimal threshold for BWD: Analysis with chorionicity information A total of 1493 twin pregnancies (n=2986 individual twins) were included in the analysis. The cohort included 384 MC (n=768) and 1109 DC (n=1109) twin pairs.  ROC analysis BWD was found to be predictive of stillbirth, perinatal mortality and morbidity with areas under the ROC curves of 0.70 (95%CI 0.60-0. 80), 0.64(95%CI 0.57-0.70), and 0.54(0.51-0.56), respectively (Table 7-1). Stratifying the data for chorionicity did not change the significance of the tests.   150  Table 7-1 Distribution of perinatal loss, area under curve and 95% CI of twins born at C&W hospital (1493 pairs, n=2986)  Outcome Count (%) Area under curve (SE) 95%CI Stillbirth  MC DC 51(1.8%) 37(4.4%) 17(0.8%) 0.70 (0.05) 0.68(0.05) 0.68(0.09) 0.60-0.80 0.57-0.79 0.50-0.87 Perinatal mortality MC DC 118(2.4%) 67(8.7%) 51(2.4%) 0.64(0. 03) 0.66(0.04) 0.57(0.05) 0.57-0.70 0.58-0.75 0.48-0.67 Perinatal morbidity MC DC 833(27.9%) 255(35.0%) 578(25.6%) 0.54(0.01) 0.59(0.02) 0.52(0.01) 0.51-0.56 0.54-0.63 0.49-0.55 SE: standard error;  CI: Confidence interval;  MC: Monochorionic;  DC: Dichorionic The predictability of BWD of ≥30% and ≥35% was better than any other thresholds (Table 7-2). For stillbirth, an inter-twin discordance of ≥30% and ≥35% had the best combination of sensitivity and specificity.  A BWD threshold of ≥35% was 35% sensitive and 96% specific.  A BWD of 35% was also found to have LR+ of 8.33, indicating a moderate increase in the likelihood of stillbirth, while a threshold of 30% was a marginal indicator of stillbirth (LR+ of 5.86). BWD of ≥35% was similarly a moderate predictor of perinatal mortality (sensitivity: 25%, specificity: 96% and LR+ of 6.24) (Table 7-3).The positive likelihood ratio for perinatal morbidity was low for all threshold levels (Table 7-4).      151  Table 7-2  Unadjusted diagnostic indices of selected thresholds of BWD for overall stillbirth and according to chorionicity in twins born in C&W hospital (1493 pairs, n=2986)  Stillbirth among MC twins  BWD Cut-off Sensitivity% Specificity % PPV% NPV% LR+ LR- 5% 0.85(0.68-0.94) 0.32(0.28-0.35) 0.05 0.98 1.25 0.46 10% 0.74(0.55-0.86) 0.51(0.47-0.54_ 0.06 0.98 1.49 0.52 15% 0.73(0.55-0.86) 0.65(0.62-0.69) 0.08 0.91 2.11 0.41 20% 0.56(0.38-0.72) 0.75(0.71-0.78) 0.09 0.97 2.21 1.60 25% 0.47(0.30-0.65) 0.82(0.79-0.85) 0.11 0.97 2.68 0.64 30% 0.38(0.23-0.56) 0.86(0.84-0.89) 0.12 0.97 2.92 0.71 35% 0.32(0.18-0.51) 0.92(0.89-0.94) 0.16 0.97 4.09 0.73  Stillbirth among DC twins  BWD Cut-off Sensitivity% Specificity % PPV% NPV% LR+ LR- 5% 0.76(0.50-0.92) 0.32(0.30-0.33) 0.00 0.99 1.12 0.74 10% 0.59(0.33-0.80) 0.55(0.53-0.57) 0.01 0.99 1.31 0.74 15% 0.59(0.33-0.81) 0.73)0.71-0.75) 0.02 0.98 2.20 0.56 20% 0.59(0.33-0.81) 0.85(0.83-0.86) 0.03 0.99 3.84 0.49 25% 0.59(0.33-0.81) 0.92(0.91-0.93) 0.06 0.99 7.53 0.45 30% 0.59(0.33-0.81) 0.94(0.93-0.95) 0.08 0.99 10.17 0.44 35% 0.41(0.19-0.67) 0.97(0.96-0.98) 0.10 0.99 14.01 0.61  Stillbirth overall  BWD Cut-off Sensitivity% Specificity % PPV% NPV% LR+ LR- 5% 0.82(0.69-0.91) 0.32(0.30-0.34) 0.02 0.99 1.20 0.56 10% 0.68(0.54-0.80) 0.54(0.52-0.56) 0.03 0.97 1.49 0.58 15% 0.67(0.54-0.80) 0.71(0.69-0.73) 0.04 0.99 2.37 0.44 20% 0.57(0.42-0.70) 0.82(0.81-0.83) 0.05 0.99 3.17 0.53 25% 0.51(0.37-0.65) 0.89(0.88-0.91) 0.08 0.99 4.92 0.55 30% 0.45(0.31-0.60) 0.92(0.91-0.93) 0.09 0.99 5.86 0.59 35% 0.35(0.23-0.50) 0.96(0.95-0.96) 0.13 0.99 8.33 0.67 MC: Monochorionic;  DC: Dichorionic;  PPV: positive predictive value;  NPV: negative predictive value;  LR+: Positive likelihood ratio;  LR-: Negative likelihood ratio.   152  Table 7-3 Unadjusted diagnostic indices of selected thresholds of BWD for overall perinatal mortality and according to chorionicity in twins born in C&W hospital (1493 pairs, n=2986)   Perinatal mortality among MC twins  BWD Cut-off Sensitivity% Specificity % PPV% NPV% LR+ LR- 5% 0.78(0.65-0.86) 0.32(0.29-0.36) 0.01 0.94 1.14 0.70 10% 0.70(0.58-0.80) 0.51(0.48-0.55) 0.12 0.94 1.45 0.58 15% 0.70(0.58-0.80) 0.67(0.63-0.70) 0.17 0.95 2.10 0.45 20% 0.54(0.41-0.66) 0.76(0.73-0.79) 0.18 0.94 2.23 0.61 25% 0.43(0.31-0.56) 0.83(0.80-0.86) 0.20 0.94 2.62 0.68 30% 0.36(0.25-0.49) 0.88(0.85-0.90) 0.22 0.93 2.95 0.07 35% 0.30(0.20-0.42) 0.93(0.91-0.95) 0.29 0.93 4.27 0.75  Perinatal mortality among DC twins  BWD Cut-off Sensitivity% Specificity % PPV% NPV% LR+ LR- 5% 0.73(0.58-0.84) 0.32(0.30-0.34) 0.03 0.98 1.06 0.86 10% 0.59(0.44-0.72) 0.55(0.53-0.58) 0.03 0.98 1.32 0.74 15% 0.45(0.31-0.59) 0.73(0.71-0.75) 0.04 0.98 1.69 0.75 20% 0.24(0.13-0.38) 0.85(0.83-0.86) 0.04 0.96 1.52 0.90 25% 0.24(0.13-0.38) 0.92(0.91-0.93) 0.07 0.98 3.00 0.83 30% 0.24(0.13-0.38) 0.94(0.93-0.95) 0.09 0.98 4.07 0.81 35% 0.18(0.09-0.31) 0.97(0.96-0.98) 0.13 0.98 6.10 0.85  Perinatal mortality Overall  BWD Cut-off Sensitivity% Specificity % PPV% NPV% LR+ LR- 5% 0.75(0.66-0.83) 0.32(0.30-0.34) 0.05 0.96 1.11 0.77 10% 0.65(0.56-0.74) 0.54(0.53-0.56) 0.06 0.97 1.43 0.64 15% 0.59(0.49-0.68) 0.72(0.70-0.73) 0.08 0.98 2.09 0.57 20% 0.41(0.32-0.50) 0.82(0.81-0.840 0.09 0.97 2.30 0.72 25% 0.35(0.26-0.44) 0.89(0.89-0.91) 0.13 0.97 3.45 0.73 30% 0.31(0.23-0.40) 0.93(0.92-0.940 0.15 0.97 4.12 0.75 35% 0.25(0.17-0.34) 0.96(0.95-0.97) 0.21 0.97 6.24 0.79 MC: Monochorionic;  DC: Dichorionic;  PPV: positive predictive value;  NPV: negative predictive value;  LR+: Positive likelihood ratio;  LR-: Negative likelihood ratio   153  Table 7-4 Unadjusted diagnostic indices of selected thresholds of BWD for overall perinatal morbidity and according to chorionicity in twins born in C&W hospital (1493 pairs, n=2986)  Perinatal morbidity MC  BWD Cut-off Sensitivity% Specificity % PPV% NPV% LR+ LR- 5% 0.68(0.63-0.71) 0.41(0.34-0.49) 0.79 0.27 1.16 0.76 10% 0.49(0.44-0.53) 0.61(0.53-0.68) 0.81 0.26 1.26 0.83 15% 0.38(0.34-0.42) 0.81(0.74-0.87) 0.87 0.27 2.09 0.75 20% 0.27(0.23-0.31) 0.87(0.81-0.92) 0.88 0.26 2.24 0.82 25% 0.19(0.15-0.22) 0.93(0.88-0.96) 0.90 0.25 2.85 0.86 30% 0.13(0.10-0.16) 0.95(0.90-0.97) 0.90 0.24 2.78 0.90 35% 0.07(0.05-0.10) 0.98(0.95-0.99) 0.95 0.23 6.44 0.93  Perinatal morbidity DC  BWD Cut-off Sensitivity% Specificity % PPV% NPV% LR+ LR- 5% 0.69(0.67-0.76) 0.32(0.28-0.36) 0.75 0.26 1.02 0.93 10% 0.45(0.43-0.48) 0.56(0.51-0.60) 0.75 0.26 1.04 0.96 15% 0.27(0.25-0.30) 0.76(0.72-0.79) 0.75 0.28 1.16 0.94 20% 0.16(0.14-0.18) 0.86(0.83-0.89) 0.78 0.26 1.23 0.96 25% 0.08(0.07-0.10) 0.94(0.91-0.95) 0.81 0.26 1.51 0.96 30% 0.06(0.05-0.08) 0.96(0.94-0.97) 0.84 0.26 1.87 0.96 35% 0.03(0.02-0.04) 0.98(0.96-0.96) 0.84 0.25 1.80 0.98  Perinatal morbidity Overall  BWD Cut-off Sensitivity% Specificity % PPV% NPV% LR+ LR- 5% 0.69(0.67-0.70) 0.34(0.31-0.38) 0.76 0.27 1.05 0.89 10% 0.46(0.44-0.48) 0.57(0.53-0.60) 0.76 0.26 1.09 0.93 15% 0.30(0.28-0.32) 0.75(0.72-0.78) 0.79 0.26 1.24 0.92 20% 0.19(0.17-0.20) 0.86(0.84-0.89) 0.81 0.26 1.46 0.92 25% 0.11(0.10-0.12) 0.93(0.91-0.95) 0.85 0.11 1.88 0.94 30% 0.08(0.07-0.09) 0.96(0.94-0.97) 0.86 0.25 2.17 0.95 35% 0.04(0.03-0.05) 0.98(0.96-0.99) 0.88 0.25 2.59 0.97 MC: Monochorionic; DC: Dichorionic;  PPV: positive predictive value;  NPV: negative predictive value;  LR+: Positive likelihood ratio;  LR-: Negative likelihood ratio   154  We stratified the data according to chorionicity (Table 7-2 and 7-3). In MC twins, BWD≥35% was found to be 32% sensitive and 92% specific with marginal likelihood of stillbirth detection (LR+ of 4.09). In DC twins, BWD≥30% and ≥35% had higher positive likelihood ratios indicating conclusive evidence of stillbirth (10.17 and 14.01, respectively).   Perinatal mortality detection using BWD of various thresholds was less conclusive when stratified for chorionicity (Table 7-3). In MC twins, BWD≥35% had the marginal likelihood of spotting perinatal mortality (LR+ of 4.27). In DC twins, a BWD of ≥35% had a moderate likelihood of perinatal mortality detection (LR+ of 6.10). We therefore chose cut-off points of ≥30% and ≥35% BWD for further analysis as these two thresholds were statistically significantly expressive in identifying perinatal loss. The positive likelihood ratio for perinatal morbidity was only moderately accurate for MC twins at ≥35% BWD threshold level (LR+ =6.44) (Table 7-4).  Figure 7-1 and Figure 7-2 summarize the results of LR+ for two threshold levels of ≥30% and ≥35% BWD for stillbirth, perinatal mortality and morbidity.           155   Figure 7-1 Positive likelihood ratio for ≥30% BWD threshold level in overall, MC and DC twins born among twins born in C&W hospital            Figure 7-2 Positive likelihood ratio for ≥35% BWD threshold level in overall, MC and DC twins born among twins born in C&W hospital   024681012Stilllbirth Perinatal mortality Perinatal morbidity≥30% BWD Threshold LevelAll MC DCConclusive0246810121416Stilllbirth Perinatal mortality Perinatal morbidity≥35% BWD Threshold LevelAll MC DCMarginal Moderate Marginal Moderate Conclusive  156  Survival analysis Figure 7-3 shows two Kaplan-Meier analyses of perinatal loss for MC and DC twins. Log rank test suggested a statistically significant difference of survival trend between those who had BWD ≥30% compared to those with lesser degree of discordance (p=0.01) in both strata (MC and DC).      Figure 7-3 Kaplan-Meier analyses of perinatal loss for MC and DC twins born in C&W hospital (1493 pairs, n=2986)  The result of multivariate cox proportional hazard analysis suggested that total risk of perinatal mortality (stillbirth and neonatal death) in twins with a BWD of ≥30% was statistically significantly greater than that of growth concordant twins, with a hazard ratio of 4.57(95%CI 3.06-6.82) after adjusting for chorionicity. The total risk of perinatal loss in twins with a BWD of ≥35% was 6.26 (95%CI 4.06-9.64), controlling for chorionicity. The hazard ratio of perinatal mortality for BWD ≥30% was 4.16(95%CI 2.41-7.17) after adjustment for chorionicity and cesarean section.  The cesarean section term in this model was significant (HR: 3.86, 95%CI 2.38-5.06). Cox proportional hazard ratios of perinatal morbidity were 1.76(95%CI 1.52-2.05)   157  and 2.42 (95%CI 1.98-2.95) for BWD of ≥30% and ≥35%, respectively.   Although the log rank test was significant in both DC and MC twins when survival of twins was investigated, the average survival for MC twins was 34.27 weeks of gestation (95%CI 33.97-34.57) while that of DC twins was 35.36(95%CI 35.18-35.54). Determining the optimal threshold for BWD: Analysis without chorionicity information A total of 6328 twin gestations (n=12656 individual twins) were included in the study. We excluded twins with TTT, congenital anomalies, single stillbirth, and birth weight <500 grams. No chorionicity information was available for this cohort. Perinatal mortality from 20 weeks of gestation was 1.5% (193/12656 twins) including 68 (0.5%) stillbirths where both twins were stillbirths.  BWD was a significant predictor of stillbirth (area under curve=0.74, 95%CI 0.70-0.78) and perinatal mortality (area under the ROC curves of 0.66, 95%CI 0.63-0.70).  ROC analysis ROC curve analysis of twins born in BC yielded similar results to those born in C&W Hospital. The predictability of BWD of ≥30% and 35% was more optimal than any other cut-off points (Table 7-5 and Table 7-6). For stillbirth, an inter-twin discordance of ≥30% and ≥35% had the optimal combination of sensitivity and specificity, although with lower rates. A BWD threshold of ≥35% was only 18% sensitive but 98% specific.  BWD of ≥30% was 24% sensitive and 97% specific with a LR+ of 6.79 indicating a moderate increase in the likelihood of stillbirth while a cut-off of 35% was a convincing indicator of stillbirth (LR+ of 10.38). BWD of ≥35% was a moderate predictive of perinatal mortality (sensitivity: 12%, specificity: 98% and LR+ of 7.32). The cut-off point of ≥30% and ≥35% BWD were chosen for survival analysis as these were two   158  levels of difference that identified more perinatal loss. Figure 7-4 summarizes the LR+ for perinatal outcomes at two BWD threshold levels of ≥30% and ≥35%.   Table 7-5 Unadjusted diagnostic indices of selected thresholds of BWD for stillbirth, perinatal mortality and morbidity for twins born in BC (6328 pairs, n=12656)  Stillbirth  BWD Cut-off Sensitivity% Specificity % PPV% NPV% LR+ LR- 5% 0.79(0.68-0.88) 0.33(0.32-0.34) 0.00 0.99 1.18 0.63 10% 0.62(0.49-0.73) 0.57(0.56-0.58) 0.00 0.99 1.45 0.67 15% 0.53(0.41-0.65) 0.76(0.78-0.76) 0.01 0.99 2.17 0.62 20% 0.38(0.27-0.51) 0.86(0.86-0.87) 0.01 0.98 2.79 0.72 25% 0.29(0.19-0.42) 0.93(0.93-0.93) 0.02 0.99 4.22 0.76 30% 0.24(0.14-0.36) 0.97(0.96-0.97) 0.04 0.99 6.79 0.79 35% 0.18(0.09-0.29) 0.98(0.98-0.99) 0.05 0.99 10.38 0.84  Perinatal mortality  BWD Cut-off Sensitivity (%) Specificity (%) PPV(%) NPV(%) LR+ LR- 5% 0.73(0.66-0.79) 0.33(0.32-0.34) 0.02 0.98 1.09 0.82 10% 0.53(0.46-0.60) 0.57(0.56-0.58) 0.02 0.98 1.25 0.81 15% 0.38(0.31-0.45) 0.75(0.74-0.76) 0.02 0.98 1.55 0.82 20% 0.24(0.18-0.31) 0.86(0.85-0.86) 0.02 0.98 1.74 0.88 25% 0.19(0.14-0.25) 0.93(0.92-0.93) 0.04 0.98 2.69 0.87 30% 0.14(0.09-0.19) 0.95(0.96-0.97) 0.05 0.98 4.10 0.89 35% 0.12(0.07-0.18) 0.98(0.98-0.99) 0.11 0.98 7.32 0.89  Perinatal morbidity  BWD Cut-off Sensitivity (%) Specificity (%) PPV(%) NPV(%) LR+ LR- 5% 0.70(0.69-0.68) 0.33(0.32-0.34) 0.26 0.25 1.06 0.86 10% 0.49(0.47-0.51) 0.59(0.58-0.60) 0.28 0.71 0.77 0.22 15% 0.31(0.30-0.33) 0.77(0.77-0.78) 0.32 0.77 1.44 0.87 20% 0.20(0.18-0.21) 0.88(0.87-0.88) 0.36 0.76 1.73 0.90 25% 0.12(0.11-0.13) 0.94(0.94-0.95) 0.42 0.76 2.24 0.92 30% 0.07(0.06-0.08) 0.98(0.96-0.98) 0.49 0.75 2.96 0.95 35% 0.04(0.03-0.05) 0.99(0.98-0.99) 0.64 0.75 5.46 0.96 PPV: positive predictive value;  NPV: negative predictive value;  LR+: Positive likelihood ratio;  LR-: Negative likelihood ratio.   159   Figure 7-4 Positive likelihood ratio for ≥30% and ≥35% BWD threshold level among twins born in BC  Survival analysis There were 226 individual twins with ≥35% BWD of whom 23 died during the perinatal period. Figure 7-5 shows cumulative survival functions of perinatal loss comparing twins with and without ≥35% BWD and ≥30% BWD. In both figures (left and right), twins with growth discordance were lost faster compared with those who were growth concordant. We did not have chorionicity data for stratification for this data but adjusted for sex-discordance and gestational age.   024681012Stilllbirth Perinatal mortality Perinatal morbidity≥30%  and ≥35% BWD Threshold Levels30% 35%ConclusiveMarginal Moderate   160    Figure 7-5 Cumulative survival function of perinatal loss for twins born in BC after adjustment for gestational age and sex discordance (6328 pairs, n=12656) BWD_30: BWD threshold 30%; BWD_35: BWD threshold 35%  Adjusted multivariate cox proportional hazard analysis suggested that the total risk of perinatal mortality (stillbirth and neonatal death) in twins with a BWD of ≥35% was statistically significantly greater than that of growth concordant twins, with a hazard ratio of 8.09(95%CI 5.23-12.52), after adjusting for maternal age. The total risk of perinatal loss in twins with a BWD of ≥30% was 4.64 (95%CI 3.09-6.96), controlling for maternal age. The hazard ratio of perinatal morbidity for BWD ≥30% was 2.54(95%CI 2.22-2.91) after adjustment for sex discordance and cesarean section.      161  Table 7-6 Comparing survival analysis for perinatal mortality in two cohorts with and without chorionicity adjustment   ≥30% BWD HR (95%CI) ≥35% BWD HR (95%CI) Twins born in C&W hospital without chorionicity adjustment (n=2986) Twins born in C&W  Hospital with chorionicity adjustment (n=2986) 5.65(95%CI 3.81-8.39) 4.57(95%CI 3.06-6.82) 8.21(95%CI 5.37-12.54) 6.26 (95%CI 4.06-9.64) Twins born in BC without chorionicity adjustment (n=12656) 4.64 (95%CI 3.09-6.96) 8.09(95%CI 5.23-12.52) CI: Confidence interval;  HR: hazard ratio Comparing two cohorts, with and without chorionicity information Table 7-6 compares the result of survival analysis for two cohorts with (n=2986) and without (n=12656) chorionicity adjustment for two threshold levels of growth discordance (≥30% and ≥35%). At a threshold level of ≥30%, similar hazard risks were found for both cohorts (4.57 vs. 4.64), although the unadjusted hazard risk for ≥30% threshold level was smaller in C&W databased compared to BC database.  The risk of perinatal loss is 8 times higher in severely growth discordance twins (≥35%BWD) than those with less growth discordance in the cohort that was not adjusted for chorionicity (BC population). A similar, but lower, figure for the other cohort, where chorionicity adjustment was possible, was found (Hazard Risk of 6.26), suggesting that the risk of perinatal loss is more than 6 times higher in severely growth discordant twins compared with the reference category. Although these risks (8.09 and 6.26) are not directly comparable, this analysis suggests that adjustment for chorionicity reduces the hazard risk of perinatal loss. Nevertheless, BWD remains a strong predictor at threshold level of greater or lower than 30% and 35%, irrespective of chorionicity.  We have selected the threshold level of   162  ≥30% for analyses of adverse outcomes in the remaining chapters, as this is the level beyond which perinatal loss increases dramatically by a hazard ratio of more than 4 times compared to concordant growth (below 30%).   7.5 DISCUSSION Growth discordance among twins relates to physiological conditions involving the fetus or the placenta. A high risk of perinatal loss is associated with severe BWD. Therefore, screening for BWD and its severity during ultrasound examination of twin pregnancy is a routine practice. Despite the universal screening for BWD, there is no agreement as to the optimum cut-off of BWD able to predict perinatal loss. The absence of consensus in the literature is due to lack of chorionicity data,75,126 and inclusion of cases with compromised birth weight (e.g. single fetal demise or reduction procedures). Moreover, these studies suffer from inadequate sample size, hence a smaller number of perinatal losses.  We studied two large cohorts of twin pregnancies: one with and one without chorionicity information. Our retrospective large cohorts of twins excluded cases that would influence the association between growth discordance and perinatal loss such as single fetal demise, chromosomal and structural malformations known to exhibit aberrant fetal growth, and cases with TTTs.  For our cohort with chorionicity information, we were able to additionally exclude cases with reduction procedures and twins with papyrus fetus or placentas, making our sample more robust for analysis.    163  Our study provides several unique contributions to the literature.   First, congenital anomalies and TTTs can cause as much as 40% discordance in birth weight, presenting important potential confounders.18–20 Major congenital anomalies affect the rates of mortality and may also lead to morbidity in the non-affected co-twins. We eliminated the impact of these confounders by excluding them from the analysis. Second, we excluded cases with one single stillbirth. There is a relationship between perinatal loss and BWD as loss of one twin during pregnancy will make BWD a highly unreliable measure due to fetal maceration. Moreover, twins with papyrus twins were also excluded from our smaller cohort using pathology reports, making it a fit sample to analyse the relationship between BWD and perinatal loss. Exclusion of twins with less than 500 grams’ weight would further reduce the probability of finding an incorrect association between BWD and perinatal loss. Third, we were able to analyse two cohorts: 1) A rather large cohort of twin gestations with chorionicity information, from a hospital based study. 2) A province-wide cohort of twin gestations without chorionicity data but with 4.4 times the size of the first cohort (12656/2860). Comparing the results between these two cohorts would clarify whether or not other studies without chorionicity can rely on their analysis. In addition, the larger sample of data will increase the study power, reduce the confidence intervals and variations around the calculated estimates, and make the extrapolation of data to the province more applicable.  Employing ROC curve analysis, we identified growth discordance of ≥30% as the optimal threshold for identifying perinatal loss. In DC twins, to identify cases of stillbirth only, a threshold level of ≥30% was conclusive. Furthermore, moving from 10% to 30% BWD threshold   164  level of, sensitivity doesn’t change for stillbirth (sensitivity figures remains unchanged as 0.59, See Tble 7-2) till 30% BWD threshold level. In other words, up until 30% threshold level, increasing thresholds only increases specificity. This was not the case for MC twins. There was no ideal threshold level for MC when analysing stillbirth. Similarly, a threshold level of ≥35% was moderately predictable of perinatal mortality in DC twins. In MC twins, threshold levels of ≥35% and ≥30% were only marginally able to detect an increase in the likelihood of stillbirth or perinatal loss.  This is probably due to the fact that we have excluded cases with TTTs, a phenomenon unique to MC twin gestations.  Survival analysis enabled us to estimate the adjusted hazard ratio of survival in two study cohorts where we could adjust for the impact of chorionicity (in the C&W dataset) and sex-discordance (in the province-based dataset). The findings suggested that MC twins tend to have shorter gestational age in a relative comparison with DC twins (34.27 weeks vs. 35.36 weeks). Although, a one-week difference might not be clinically important, the shorter gestational age in MC twins may relate to earlier termination of pregnancy by cesarean section.  Cox proportional hazard analysis confirmed this observation, as entering cesarean section in the analytical model of perinatal loss and BWD was found to be statistically significant. Previous studies have found that a BWD of 18% or higher is an independent predictor of neonatal mortality and morbidity.4,7,18,21,22  Our findings are consistent with those of a small study of 122 live born twin sets that found a greater than 30% birth weight difference was associated with a significant increase in rates of infant mortality (25%).23 The study subjects were live born twins delivered between 25 and 34 weeks of gestation and excluded cases of congenital anomaly. Another study on 335 twin gestations included preterm infants (>37 weeks’   165  gestation) and found BWD to be a significant predictor of adverse neonatal outcomes.24  Adverse outcome was defined as neonatal death or major neonatal morbidities. A BWD of 30% or greater was found to minimize the false-positive rate and maximize the positive likelihood ratio so that a discordance was present 3.6 times more frequently among twins with adverse outcomes than those in the comparison group with good outcomes.   Studies that noted an excess of adverse neonatal outcomes in twins born with discordance of other than 30% included cases of congenital anomalies in the study analysis,11 did not separate stillbirth from neonatal mortality,24 employed ROC curves plotted based on simple logistic regression rather than multivariate analysis,24–27 deviated from a normal population by exclusively studying a subgroup of twins such as preterm twins or monochorionic twins,23,24,28 and were hospital based.6,8,29 Studies conducted in hospitals or referral centres over-represent high risk pregnancies, monochorionic pregnancies, and cases of TTTs, some of which require laser coagulation for the purpose of placental anastomosis, a procedure done in referral centers only. It is challenging to extrapolate the conclusions of such studies to a population of non-malformed, non-TTTs twins of a wide range of gestational age.  We found that BWDs of ≥30% and ≥35% are optimal predictors of stillbirth and perinatal mortality in our two cohorts (hospital-based and population-based). LR+ is a measure of the reliability and strength of a clinical test to predict a condition of interest.31 Our results suggest that LR+ for a BWD of ≥35% is a conclusive predictor for stillbirth while ≥30% is a moderate predictor. Our adjusted survival analysis found similar hazard risks for both cohorts (C&W and BC datasets) at a threshold level of ≥30%. We therefore, choose this threshold level to analyse the study objectives, presented in future chapters.   166  The SOGC guidelines currently use an estimated fetal weight discordance of >20% as a starting point to monitor twin gestation more closely.249 The authors of the guidelines took the more conservative of the thresholds in the literature given the current limits of ultrasound EFW at that time.  Our study results, using the preferable measure of the actual BWD, suggest that higher threshold levels of (30% or 35% and higher) are associated with perinatal mortality. Our result may assist clinician to choose a different time point to start the routine ultrasound screening.  Since, 30% BWD is associated with higher hazard of perinatal loss, it logically makes sense to start the screening earlier. The practical clinical choice of the threshold used for ultrasound screening of twin pregnancies, however, will depend on many factors, including the amount of local clinical resources.  In any case, the timeline chosen by clinician to start ultrasound screening should aim at improved predictability, lower false positive rates, and subsequently lower patient anxiety and resource utilization. Research-oriented BWD or threshold derived from birth data is based on the actual birth weight, rather than EFW. Clinicians, however, have only access to EFW via ultrasound screening. The question remains how the result of this study can benefit clinicians? Relying on growth discordance estimations from our research, clinicians may prevent perinatal loss under certain assumptions, only. The first assumption is that EFW is a reliable estimate, close to actual birth weight, in determining the fetal birthweight. Secondly, severe discordance of birth weight is easier to detect in ultrasound compared to mild or moderate growth discordance. Thus, to predict perinatal loss, if the optimized threshold levels for BWD are at the higher end (≥30% or ≥35%), ultrasound findings can be relied upon. This premise is not true if the BWD thresholds are found to be at the lower end (≥20%, ≥25%). The current study suggests that severe growth discordances of ≥30% is associated with perinatal loss, threshold that is easily identifiable by   167  ultrasound screening even via EFW. Further studies are necessary to compare FEW with actual birth weight data in determining the optimum threshold for BWD, pointing at perinatal morbidities.  7.6 SUMMARY In brief, we found that a BWD cut-off limit of 30% and higher has optimal accuracy for detecting perinatal mortality. Clinical decisions over the frequency of ultrasound screening of twin gestations can be accordingly adjusted. In other words, serial ultrasound monitoring of twin gestations can be started before a BWD of larger than 30% is discovered in ultrasound. These results hold promise until further studies are conducted by serial ultrasound screening collecting chorionicity information and comparing FEW and actual birth weight data.     168  CHAPTER 8 STILLBIRTH AND BIRTH WEIGHT DISCORDANCE  8.1 INTRODUCTION Birth weight discordance, which occurs when there is a disparity in birth weight between the larger and smaller infants of a twin set,11 is a common phenomenon in twin gestations. The mechanism of BWD has been attributed to unequal placental sharing,116 abnormalities of umbilical cord insertion,13 placental pathology,14 and TTTs.7 BWD occurs in 10-29% of twin pregnancies8–10 and it has been associated with a number of perinatal outcomes, such as stillbirth,24,250 respiratory distress syndrome, necrotizing enterocolitis, IVH, and sepsis.143 The most frequent reported outcome of BWD in the literature is fetal death.142,251–254 To date, some studies have reported increased fetal death rates among BWD twins as compared to weight concordant twins,139,194,242,252 while others have not.5,20  A possible explanation for the lack of agreement among studies is methodological limitations.  For example, no study to date has assessed potential confounding factors related to obstetric history (stillbirth, congenital anomalies, low birth weight) in the association between BWD and stillbirth. Other methodological problems include an unclear definition of stillbirth,167 small sample size,15,44 and lack of correction for the correlation in stillbirth between twins.143 To overcome these shortcomings, we carried out a study to determine how BWD affects stillbirth rate in association with other risk factors, adjusting for demographic, behavioural, fetal, clinical and obstetrical confounding factors.  Our study also determined the effect of fetal growth, parity, sex-discordance, gestational age, twin size and chorionicity on these associations.    169  8.2 OBJECTIVES To determine the stillbirth risk differences between twins in various BWD strata based on fetal growth, parity, sex discordance, gestational age, twin size and chorionicity. 8.3 METHODS   Briefly, we analysed data from a province-based retrospective cohort study using the PSBC data set, containing matched records of almost all twins born from 1 April 2000 to 30 December 2010.  We excluded the twin records with missing birth weight, gestational age, and infant sex. Records with TTTs, congenital anomalies, and those with less than 500 grams’ weight were also excluded. For our hospital cohort, using pathology data, we were able to identify twins with reduction procedures. These were also excluded from our hospital cohort data. After excluding these cases, our analytical data of the population-based data set was 12814 individual twins. The number of twins in this analytical dataset is higher than in other chapters because twins with single stillbirths were not excluded. Part of this cohort included 2992 twins (following exclusion of cases as described above) who were registered at the British Columbia Women’s Hospital. Pathology reports were available for this cohort, enabling us to analyse the relationship between growth discordance and stillbirth in the light of chorionicity data.  In these databases, twin sets were labeled with a unique identifier linking the members with one another and with their mothers. The method for compilation of this data file and linkage process with the provincial data set is described in Chapter 3.   Twins with a BWD of <30% were chosen as growth concordant and the reference category (see Chapter 7 for why this threshold was identified as an optimal one).  The outcome of interest was stillbirth, defined as the complete expulsion or extraction from the maternal body after at least 20   170  weeks of gestation or after attaining a weight of at least 500 grams of a fetus with no breathing, beating heart, pulsation of the umbilical cord or unmistakable movement of voluntary muscle.4   Variables examined in multivariate analyses included   Socio-demographic characteristics of the mother including maternal age, maternal education (<12 years/ ≥12 years), parity (nulliparous, multiparous),   Maternal behavior consisting of maternal smoking (never smoker, former smoker, current smoker), and number of antenatal visits (<7, 8-10,>11).   Fetal characteristics constituted fetal sex (male/female), fetal sex concordance (concordant, discordant), and fetal growth (small for gestational age [SGA], appropriate for gestational age [AGA], large for gestational age [LGA]).   Clinical confounding factors included maternal diabetes (yes/no), PIH (yes/no), preeclampsia (yes/no), in vitro fertilization (IVF) (yes/no), maternal weight gain during current pregnancy.  Obstetric history was inclusive of history of stillbirth (yes/no), previous baby with low birth weight (yes/no), previous pregnancy with congenital anomaly (yes/no), and pre-pregnancy weight (<70 kg/ ≥70 kg). The statistical analysis, briefly, included estimating distribution of birth characteristics based on BWD and GEE modeling for bivariate and multivariate analysis, accounting for non-independent occurrence of stillbirth among twin pairs and the strength of association between BWD and stillbirth. The GEE method allows data for twin births to be treated as cluster measures of the same birth. Clustering of twins within mothers is often overlooked which can lead to over-estimation of effects if neglected.255     171  For the multivariate model of stillbirth, data were adjusted for covariates based on variables available in the dataset, clinical and epidemiological evidence of confounding effects in the literature, and results of the bivariate analysis. Crude and adjusted odds ratios (ORs) and 95% confidence intervals (95%CIs) of stillbirth for BWD were then calculated for the study population and for subgroups classified by fetal growth, parity, gestational age, sex discordance, individual fetal size and chorionicity.  To control for intrauterine fetal growth, infant birth weight for gestational age was classified as follows: SGA (<10th percentile), AGA (10-89th percentile) and LGA (90th percentile and more) based on twin growth curves derived from a clinical twin population with reliable sex-specific gestational age estimates.65 Analysing our province-based cohort, sex-discordant stratification was performed as a proxy to evaluate confounding by chorionicity. Sex-discordant analysis has been used as a substitute for chorionicity.20,256 Because of the higher stillbirth rate reported for lighter twins,257 we fitted separate models for heavier and lighter twins to adjust for fetal growth. Finally, the relationship between BWD and stillbirth was also investigated for subcategories of gestational age (<32 weeks and ≥32 weeks) and parity (multiparous and nulliparous). 8.4 RESULTS A sample of 12814 twins was analysed.  Rates of missing data were <1% for all the variables except for maternal education (85%), in vitro fertilization (75%), and maternal smoking (67%). During the 10-year study period, the incidence of twins with BWD ≥30% was 4.2%. Mean intra-pair birth weight difference was 11.18±9.75 SD.     172  The overall rate of stillbirth was 1.1% (n=146), of which 0.2% died after onset of labour (n=21), 0.7 % (n=90) died prior to onset of labour and 0.3% (n=35) were recorded as unknown timing. Fetal death rates were 0.7% and 11.0% in infants with a BWD of <30% and ≥30%, respectively.  The distribution of birth characteristics based on the percentage of BWD levels is listed in Table 8-1. The associations between different BWD levels and the following variables were found to be significant: maternal age, parity, gestational age, maternal education, PIH, preeclampsia, intra-uterine growth restriction, in vitro fertilization, pre-pregnancy weight, and antenatal visits of less than eight.    The crude odds ratio for stillbirth was significantly higher for twins born with a BWD of ≥30% compared with non-discordant twins (17.26; 95%CI 12.25-24.32; Table 8-2). The BWD-related risk of stillbirth decreased from 17.26 to 13.69 (95%CI 7.32-25.62), but remained statistically significant, after adjusting for confounding factors including maternal behaviour, socio-demographic characteristics and clinical confounders (see section 8.3 for details).             173  Table 8-1 Distribution of birth characteristics according to the percentage BWD levels in twins born in BC (n=6407) Variable BWD<30% (N=6138) BWD≥30% (N=269) P  Maternal age, y <24 25-34 ≥35  701(11.4%) 3494(56.9%) 1943(31.7%)  40(14.9%) 132(49.1%) 97(36.1%)  0.01 Parity  Primiparous  3018(49.2%)  171(63.4%)  0.01 Gestational age, weeks  <32 32-36   ≥37  575(9.4%) 3122(50.9%) 2440(39.8%)  69(25.7%) 145(53.9%) 55(20.4%)  0.01 Maternal smoking Current smoker Former smoker Never smoker  575(9.4%) 419(6.8%) 1015(16.5%)  28(10.4%) 19(7.1%) 37(13.9%)  0.41 Maternal Education  ≤12  652(22.3%)  28(12.6%)  0.01 Maternal diabetes Yes  1326(10.8%)  224(8.2%)  0.06 Pregnancy induced hypertension Yes  380(6.2%)  29(10.8%)  0.01 Preeclampsia Yes  140(2.3%)  13(4.8%)  0.01 IUGR Yes  675(11.0%)  132(49.3%)  0.01 In vitro fertilization Yes  506(32.6%)  28(41.0%)  0.04 History of stillbirth Yes  69(1.1%)  <5 (1.9%)  0.65 Previous baby with LBW Yes  134(2.2%)  <5 (1.9%)  0.13 History of congenital anomaly Yes  69(1.1%)  <5 (1.9%)  - Pre-pregnancy weight <70 kg  3045(49.6%)  130(49.3%)  0.63 Pregnancy weight gain <12 kg  626(10.2%)  42(15.6%)  0.01 Antenatal visits <8  1854(34.2%)  111(46.4%)  0.01    174  Table 8-2 Risk of stillbirth associated with BWD, among twins and in respect to fetal growth among twins born in BC (6407 pairs, n=12814) Birth weight discordance Number of stillbirths (%) Crude OR (95%CI) Adjusted OR (95%CI) Quasi Likelihood Criterion (QIC)  All twins     <30%(n=12276 ) 87(0.7%) Ref Ref   471.34 ≥30% (n= 538) 59(10.9%) 17.26(12.25-24.32)  13.69(7.32-25.62)  SGA <30%(n=372) 15(4.0%) Ref Ref  229.78 ≥30% (n= 196) 44(22.4%) 6.89(3.72-12.76)  5.50(2.97-10.18)  AGA <30%(n=9141 ) 30(0.3%) Ref Ref  370.04  ≥30% (n=220 ) 8(3.6%) 11.46(5.19-25.29)  3.59(1.52-8.49)  LGA <30%(n=2763) 42(1.5%) Ref Ref  227.80  ≥30% (n= 220) 7(3.2%)  3.94(1.73-8.97)  5.18(1.68-15.97) OR: odds ratio;  CI: Confidence interval  Adjusted for maternal behaviour, socio-demographic characteristics and clinical confounders (see section 8.3 for details) The association between BWD and stillbirth after stratifying by fetal growth, parity, sex discordance, gestational age, and individual fetal size is shown in Tables 8-3 to 8-6. When the data was categorised to three groups of SGA, AGA and LGA, the growth discordant group persistently showed higher adjusted odds of stillbirth compared to the reference category (Table 8-2). Relatively, the adjusted risk of stillbirth was strong for SGA twins (5.50, 95%CI 2.97-10.18), and LGA (5.18, 95%CI 1.68-15.97) twins, while it was moderately strong for AGA (3.59, 95%CI 1.52-8.49). The crude and adjusted odds of stillbirth were much stronger in multiparous gestations of discordant twins than in the reference group (19.98 vs. 9.81, respectively, Table 8-3). In addition, the adjusted odds were relatively greater in primiparous gestations than the corresponding figures in multiparous gestations (14.70 vs. 9.81).    175  Table 8-3 Risk of stillbirth associated with BWD, among twins and subcategories of parity among twins born in BC (6407 pairs, n=12814) Birth weight discordance Number (%) Crude OR (95%CI) Adjusted OR(95%CI) Multiparous    <30%(n=6239) 39(0.6%) Ref Ref ≥30% (n=197) 22(11.2%) 19.98(11.60-34.42) 9.81(4.93-19.49) Primiparous    <30%(n=6037) 48(0.8%) Ref Ref ≥30% (n=341) 37(10.9%) 15.19(9.74-23.67) 14.70(8.67-24.94) OR: odds ratio;  CI: Confidence interval  Stratification by sex concordance showed higher crude odds of BWD-related stillbirth in growth discordant twins than in the reference category (28.87, 95%CI 15.54-53.64; Table 8-4).  After adjusting for potential confounders, the odds increased and remained statistically significant (30.43, 95%CI 13.98-66.22). Sex-discordant twin pairs were evaluated as a surrogate measure, as chorionicity data were not available for this cohort. Similar to concordant twins, the risk of stillbirth was high among sex discordant twins with high growth discordance (crude: 13.72, 95%CI 9.04-20.82; adjusted: 8.62, 95%CI 5.26-14.16) compared with the reference category.  The adjusted odds of stillbirth were 4.48 times higher for preterm twin pairs with growth discordance (95%CI 1.78-11.29) compared to non-discordant twins of the same gestation (Table 8-5).   176  Table 8-4 Risk of stillbirth associated with BWD, among twins and subcategories of sex concordance among twins born in BC (6407 pairs, n=12814) Birth weight discordance Number (%) Crude OR (95%CI) Adjusted OR(95%CI) Sex concordant    <30%(n=4210) 21(0.5%) Ref Ref ≥30% (n=174) 22(12.6%) 28.87(15.54-53.64) 30.43(13.98-66.22) Sex discordant    <30%(n=8066) 66(1.1%) Ref Ref ≥30% (n=364) 37(10.2%) 13.72(9.04-20.82) 8.62(5.26-14.16) Note: One twin had an unknown sex, thus the twin pair was considered missing for concordance;  OR: odds ratio;  CI: Confidence interval  A similar pattern was observed for twins who were born dead at 32 weeks of gestation or after (9.50, 95%CI 3.75-24.05).   Table 8-5 Risk of stillbirth associated with BWD, among twins and subcategories of preterm birth among twins born in BC (6407 pairs, n=12814) Birth weight discordance Number (%) Crude OR (95%CI) Adjusted OR(95%CI) <32 weeks    <30%(n=840) 68(8.1%) Ref Ref ≥30% (n=110) 30(27.3%) 4.26(2.62-6.93) 4.48(1.78-11.29) ≥32 weeks    <30%(n=11436) 19(0.2%) Ref Ref ≥30% (n=428) 29(6.850 43.67(24.28-78.55) 9.50(3.75-24.05) OR: Odds ratio; CI: Confidence interval; Adjusted for maternal behaviour, socio-demographic characteristics and clinical confounders (see section 8.3 for details)    177  Lighter individual twins with ≥30%BWD had greater adjusted odds of stillbirth than in the reference group (7.09, 95%CI 3.61-13.96, Table 8-6).  Table 8-6 Risk of stillbirth associated with BWD, among twins and subcategories of individual twin sizes among twins born in BC (6407 pairs, n=12814) Birth weight discordance Number (%) Crude OR (95%CI) Adjusted OR(95%CI) Lighter twins    <30%(n=6353) 42(0.7%) Ref Ref ≥30% (n=244) 11(4.5%) 13.49(9.20-19.78) 7.09(3.61-13.96) Heavier twins    <30%(n=6224) 69(1.10%) Ref Ref ≥30% (n=373) 49(13.1%) 7.09(3.61-13.95) 5.83(1.70-19.98) OR: Odds ratio;  CI: confidence interval Lighter refers to lighter twin compared to the co-twin in birth weight (grams) Adjusted for maternal behaviour, socio-demographic characteristics and clinical confounders (see section 8.3 for details) Subgroup analysis with chorionicity information  The hospital cohort of 1501 twin pregnancies (n=3002 individual twins) with chorionicity information was analysed. The rate of stillbirth from 20 weeks’ gestation was 1.0% (31/3002). There were 19 (0.7%) stillbirths within the growth concordant twins and 12 (5.4%) stillbirths within the growth discordant twins of ≥30% BWD. Our study found that an increase in BWD was associated with an increased risk of stillbirth. A dose effect of BWD on stillbirth is shown in Figure 8-1.     178   Figure 8-1 Frequency of stillbirth (%)  in BWD categories with four layers for twins born in C&W hospital (1501 pairs, n=3002; p=0.01) Table 8-7 Risk of stillbirth associated with BWD below or equal and above 30% for twins born in C&W Hospital (1501 pairs, n=3002) Birth weight discordant Number (%) Crude OR(95% CI) Adjusted OR(95%CI) <30%(n=2781) 19(0.7%) Ref Ref ≥30%(n=221) 12(5.4%) 8.34(3.99-17.42) 5.21(2.17-12.47) Note: Adjusted for gestational age, chorionicity, parity and maternal age;  OR: Odds ratio;  CI: confidence interval The crude odds of stillbirth in twins with a BWD of ≥30% was significantly greater than that of twins with a BWD of <30% (8.34; 95%CI 3.99-17.42;Table 8-7). In other words, the odds of stillbirth were about 8 times higher in twins with BWD≥30% than the reference category. Adjusted odds of stillbirth (for gestational age, parity, chorionicity and maternal age) was lower than the crude rate but still significant (5.21, 95%CI 2.17-12.47).  024681012>10% 11-20% 21-30% >30%Stillbirth  179  Time of stillbirth was recorded as unknown in 11 of cases while 19 were recorded as “prior to onset of labour” and 1 death occurred after onset of labour. We excluded cases with unknown timing and repeated the analysis. The odds of stillbirth in twins with severe growth discordance was much higher than the earlier estimation (11.89, 95%CI 4.64-30.49) but with a wider confidence interval due to smaller numbers of stillbirths (0.6%, 19/2985). When data were stratified based on chorionicity, MC twins with a BWD≥30% had a high risk of stillbirth compared with the reference category before (16.52; 95%CI 5.53-49.36) and after (15.32; 95%CI 5.43-43.21) adjustment for confounding variables. The odds for monochorionic twins were more than 4 times higher than the dichorionic twins (2.19, 95%CI 1.10-13.25). Table 8-8 Unadjusted and adjusted odds (95%CI) for stillbirth and BWD≥30% in two cohorts with and without chorionicity information  Unadjusted OR (95%CI) Adjusted OR (95%CI) Twins born in C&W with chorionicity (1501 pairs, n=3002) 8.34(3.99-17.42) 5.21(2.17-12.47) Twins born in BC without chorionicity (6407 pairs, n=12814) 17.26(12.25-24.32) 13.69 (7.32-25.62) OR: odds ratio;  CI: Confidence interval Adjusted OR for C&W database: chorionicity, gestational age, parity and maternal age Adjusted OR for BC database: sex discordance, maternal behaviour, socio-demographic characteristics and clinical confounders (see section 8.3 for details)  Table 8-8 summarizes the odds of stillbirth in twins with and without chorionicity information. The odds of stillbirth are higher for growth discordant twins of ≥30% BWD compared with those with lower than 30% growth discordant twins, after adjustment for chorionicity or sex discordance. The higher odds of stillbirth in the BC population relates to higher number of   180  stillbirths in the province-based population compared to a single hospital population (146 vs. 3\1). The rate of stillbirth in the BC population was comparable with that of C&W hospital (1.1% vs. 1.0 %, respectively).    8.5 DISCUSSION Growth discordance among twins has become an increasingly important health concern due to the increased rate of twin births in recent years. The present study found an association between stillbirth and growth discordant twin pairs (≥30%) after adjustment for maternal medical history, obstetric history, pre-pregnancy weight, weight gain during pregnancy, number of antenatal visits, and maternal behaviour.  Similar to our findings, several studies suggest that stillbirth is not observed until discordance reaches levels of 30% to 40%.63,76,88 Twin pairs with growth concordance had a low frequency of stillbirth in both our hospital-based and population-based datasets (0.7% and 0.7%). This could be due to the fact that growth discordance of less than 30% represents a normal variation between twin pairs, whereas growth discordance of 30% or higher indicates some factors which hamper normal growth and lead to fetal demise. We have identified some of these factors relating to placenta pathology, as discussed in Chapter 5. Distinguishing normal variation from pathological growth discordance is clinically important leading to a frequent practice of increased prenatal surveillance and early labour induction in these pregnancies. In the literature, the association between BWD with increased stillbirth is established.24 However, what is not clear is the extent to which fetal growth, parity, sex discordance, gestational age, chorionicity or individual fetal size may influence the stillbirth in the presence of potential confounding factors.80,242  To the best of our knowledge, our study is the first   181  population-based study that had access to a wealth of relevant confounding factors to adjust the multivariable models with.  Although this population-based data set lacked chorionicity data, we were able to (1) show that sex discordance is a suitable proxy for chorionicity, and (2) analyse the impact of chorionicity in a smaller cohort (C&W hospital dataset). The findings from both analyses suggest that ≥30% BWD is a good predictor of stillbirth.  Our stratified analysis, the results of which are discussed below, further helps to address important gaps in the literature.  Our subsample analysis of twins with chorionicity data showed persistently higher odds of stillbirth among twins with a BWD ≥30% even after adjusting for known confounding variables including parity, gestational age and maternal age. Monochorionic twins with a BWD≥30% had a high risk of stillbirth when compared with the reference category after adjustment for confounding variables (15.32; 95%CI 5.43-43.21). Although estimated wide confidence interval limits the strength of the interpretation, higher risk of stillbirth in MC twins can relate to placenta characteristics, discussed in Chapter 5.  Fetal growth An increased frequency of fetal death in SGA twins is reported at growth discordance levels of 18%,195 20%, and 30%.137  Our analysis showed that the odds of stillbirth were invariably greater in growth discordant twins (≥30%) in three subcategories of SGA, AGA and LGA compared with non-discordant twins. The odds of stillbirth were much stronger in LGA and SGA twins than AGA twins (5.50 and 5.18 vs. 3.59, respectively). The role of fetal growth in BWD-related stillbirth is subject to disagreement as some investigators have determined the fetal growth as an isolated finding rather than a perinatal risk factor for stillbirth.146,258 Others have drawn special   182  attention to SGA twins. Yinon et al 259 reported a higher stillbirth rate in the SGA-discordant group than in the AGA-discordant group (19.5% vs. 6%, P =0.04). Blickstein et al.8 also found a significantly higher stillbirth rate among those twin pairs in which their smaller twin was SGA. Our study suggests that BWD alone is predictive of stillbirth irrespective of fetal growth. Maternal parity has been shown frequently to be the most consistent key factor that influences twin growth.260–262  Similar to the finding of Wen et al.,143 our results suggest that primiparous mothers have a higher odds of stillbirth compared with multiparous mothers (14.70 vs. 9.81). These findings have important clinical implications. For instance, when an apparent discordant growth is detected in a primiparous mother by ultrasound screening during prenatal visits, closer monitoring or termination of pregnancy (via early induction or elective cesarean section) may be needed for her.  The mechanism by which the primiparous uterus yields stillbirth is not clear. One potential explanation is that primiparous mothers may be more prone to preeclampsia and hypertension which leads to a constricted vascular system, thus less oxygenation of the fetus.   Sex-discordance Conflicting data have been reported in the literature as to whether the fetal sex distribution of twin pairs influences fetal mortality.148  A population-based study from Sweden reported the stillbirth rate for sex-concordant pairs to be more than twice that of sex-discordant twin pairs.263 We stratified the data by sex concordance. The results suggested higher increased risks of stillbirth in sex-concordant twins compared with sex-discordant twins (30.43 vs. 8.26).  Sex-discordance is used as a substitute for chorionicity in our population-based database analysis as well as in the literature.264 Given sex-discordance as a proxy for chorionicity, it can be speculated that a difference in chorionicity was likely to change the risk of stillbirth more than   183  expected in comparison with estimates of stillbirth among sex-concordant twins.  This finding was supported by our hospital-based data analysis, as the stratified analysis by chorionicity showed different odds in MC and DC pregnancies (15.32 vs. 2.19). Gestational age Apart from chorionicity, parity and fetal growth, the findings of our large population-based study indicated that BWD differs significantly in various gestational periods. The BWD-related risk of stillbirth was higher in infants with gestational age ≥32 weeks compared with gestations that lasted less than 32 weeks. This finding is consistent with results reported by some investigators.5,140,143,265 This suggests that BWD remains an important risk factor even after adjusting for preterm birth. More ultrasound screening after 32 weeks of gestation is therefore recommended.  Individual fetal size Our study revealed that lighter twins of highly discordant pairs were subject to high rates of stillbirth. Similar to our findings, Branum et al. found that when BWD was ≥30%, a higher risk of stillbirth was observed among lighter infants.137 However, the risk of stillbirth was reduced in heavier twins.  Chen et al. also found that the smaller twin of a pair was at greater risk of stillbirth, especially as the growth discrepancy increased.266 Chorionicity Chorionicity has a significant influence on stillbirth.254  The odds of stillbirth among MC twin pregnancies were seven times larger than that of the DC twins. This result is consistent with the literature203,254,267 leading to recommendations for earlier delivery of pregnancy in MC gestations.     184  The obstetrics management in MC twins with BWD is not well established, as it depends on clinician judgment to choose between a more conservative approach, such as close monitoring via serial ultrasounds, or a more invasive approach, such as fetoscopic laser coagulation of the placental vascular anastomoses and selective feticide.  Choosing one type of management over the other could lead to stillbirth or save the twins. This calls for further research in the field of intervention studies. Nevertheless, our study reveals that a BWD of equal to or greater than 30% is a fine determinant of stillbirth even after adjustment for chorionicity. It then seems logical to seek the answers in the characteristics of the placenta and cord of MC twins with or without BWD. This discussion can be found in Chapter 5. Strength and limitation The inability to adjust for chorionicity is one of the consequences of a data deficiency in the PSBC database. The subgroup analysis of sex-discordant twins has been used in the literature as a proxy.20 We used such a proxy measure to address the lack of information on chorionicity when analysing the whole database.  To adjust for the impact of chorionicity, however, we collected data from one hospital in BC which had a higher rate of deliveries compared to other hospitals. We then analysed a group of 2986 twins for whom pathology data were linked with the outcome data. The similar predictive ability of growth discordance on stillbirth from both of these analyses suggests that sex discordance can be used as a proxy for chorionicity in studies where it is not available. Access to chorionicity is therefore important. However, we have shown that lack of access to chorionicity data, although limiting, should not stop researchers from using the proxy measure of sex discordance in analysis of the association of growth discordance and fetal demise.   185  Another limitation was incomplete data on assisted reproductive therapy, which was recorded only after April 2008, covering 25% of twins. We also lacked data on several maternal variables such as race and household income. Furthermore, data on maternal education, maternal smoking, drug use, and alcohol use was limited. Furthermore, data on sensitive information such as smoking should be carefully examined as we are not sure of its accuracy. Lack of knowledge regarding true rate of smoking among pregnant mothers could have public health implications.  NICU II and III admission could have counted some twins twice if they have been transferred from one to the other. We choose not to add these two levels of admissions as the severity of condition can be highlighted among those who at some point in time where admitted to NICU III vs. II. The strength of our study is five-fold. First, our large retrospective cohorts used PSBC data which were routinely validated and contained information on approximately 99% of births in the province of British Columbia.  Information in the registry is compiled from standardized forms completed by clinicians. The validity of the data is insured by quality control measures including built-in warnings in the data entry software and periodic checking of the data.4 Second, our data analysis takes the inter-twin relationship into consideration by conducting GEE modeling. Third, to our best of knowledge, this is the first population-based study in British Columbia on the role of BWD in stillbirth. In the literature related to BWD, all studies to date are hospital-based except for the study by Wen et al.,143 which excluded Ontario data and reported inconsistent data quality among all Canadian provinces.268  The current hospital-based study confirmed the results obtained from the population-based dataset, taking the analytical strategy one step further to control for chorionicity. Fourth, no study to date has comprehensively assessed potential confounding factors such as maternal socio-demographics, maternal behaviour, fetal   186  characteristics, clinical confounding factors, and obstetrics history. Fifth, we examined the role of BWD in relation to stillbirth within the subcategories of fetal growth, parity, sex discordance, gestational age, twin size, and chorionicity. The latter was done in a smaller population of twins born in C&W Hospital. Finally, we were able to exclude cases with impact on stillbirth or growth discordance. No study to date has excluded cases of papyrus fetuses and twins who underwent selected reduction from a hospital-based dataset. Papyrus fetuses are known as vanishing twins and are frequently the result of in-vitro-fertilization but can also represent natural selection for fetal demise. In either case, vanishing twins are found to be associated with adverse perinatal outcomes.269  We also excluded cases with congenital anomalies. Anomalies of one fetus in twin gestation are known to cause discordant growth.170 Some studies have regarded congenital anomalies as a confounding factor and therefore adjusted143 or excluded270 them from analysis of growth discordant twins. Only 1.4% of our twin population had congenital malformation. Furthermore, congenital anomalies were not related to BWD (p=0.201) and none of the stillborn twins were congenitally abnormal. Nevertheless, we excluded subjects with congenital anomalies in our analysis. Furthermore, TTTs has been almost invariably known to be associated with stillbirth in MC twin gestations due to placental anastomoses.119  In our study, cases coded with TTTs were also excluded from analysis.  Discussion of anastomosis and its impact on BWD is given in Chapter 5. A major limitation of our study is that we cannot eliminate the possibility of growth discordance being a result of an early in utero stillbirth. The only way to avoid this bias it to conduct a prospective study where using close ultrasound monitoring, the time of death is recorded. We have tried to control this bias by excluding stillbirth of less than 500 grams which would   187  eliminate some of the effect. We have also investigated the association between BWD and neonatal mortality and morbidity both of which are free from the above mentioned bias.   8.6 SUMMARY Our population-based GEE analysis of a comprehensive twin data set from BC (Canada) showed a substantially increased fetal mortality associated with a BWD ≥30%, indicating that this value is clinically significant. Repeated analysis in a smaller hospital-based population confirmed the same results. The association between BWD and stillbirth was found to be dependent on the following factors: fetal growth, parity, gestational age, twin size, sex discordance and chorionicity. Clinical decisions relating to levels of fetal surveillance, antenatal corticosteroid therapy, and timing of delivery should be driven by these factors.       188  CHAPTER 9 BWD AND ADVERSE PERINATAL OUTCOMES 9.1 INTRODUCTION Studies from Canada,1 the United States,271 Australia,272 Norway,273 England,274 and Korea275 have shown an increase of over 50% in the twinning rate in the last decade. The increased rate of twinning across the globe has affected the clinical perspective pertaining to specific perinatal complications of twin pregnancies. Associated complications such as increased perinatal mortality and morbidities are established in twin pregnancies compared with singleton gestations.276–279 The addition of discordant fetal growth exacerbates the potential for future complications. The literature is rich with studies retrospective and prospective,16,18,19 hospital-based15,63,87 and population-based studies,23 small 15–17,45and large20 producing widely disputed rates of perinatal mortality and morbidities. Some studies used a BWD of 20-30% to define growth discordance associated with higher rates of adverse perinatal outcomes,132 while other studies suggest that growth discordance is common and the majority of twins do well regardless of their weight difference.15,20,63  These studies adjusted variables most commonly thought to be confounders of the relationship between BWD, as an independent variable, and perinatal outcomes as dependent variables. These confounding variables are gestational age,63 intra-uterine growth restriction (IUGR),19 twin size,20 infant sex,62,239 birth order,20 sex pairing,280 maternal age,281 chorionicity,16,52,64,74,76 prenatal care,5 preterm birth,63 and pregnancy due to assisted reproductive technology.282,283 However, a few studies to date have adjusted for confounding factors related to maternal complications at pregnancy (e.g. diabetes, PIH and preeclampsia), prior obstetric history and maternal weight gain during pregnancy.  While the majority of studies have focused on presenting the association between BWD and perinatal morbidities such as respiratory distress syndrome, congenital anomalies,   189  hyperbilirubinemia, hypoglycemia, and anemia, a few studies have investigated the outcomes related to management of newborn viability at birth, and neonatal respiratory support in infancy. Our study primarily aimed at investigating the association between BWD and perinatal mortality. The secondary objective of this study was to assess the difference in short-term neonatal outcomes across discordant twins at birth, and in the first 28 days of life. This information will convey accurate prognostic information to parents and caregivers as the twinning rate increases in British Columbia and as the population of growth discordant twins admitted to the NICU during infancy increases throughout the province.     9.2 OBJECTIVE Our purpose was to examine the association between discordant growth and perinatal mortality and morbidities among twins with careful adjustment for maternal confounding factors and chorionicity.   9.3 METHODS In brief, we retrospectively evaluated data collected from the clinical records of all women with twin pregnancies who delivered anywhere in BC and were registered by PSBC. Thirteen thousand three hundred and thirty twins born between 2000 and 2010 were included in the study. Cases with major structural congenital anomalies, TTTs, those with a weight less than 500 grams and single stillbirths were excluded.  After excluding these cases, 12,656 twins remained in the dataset for analyses. We adopted the definition of growth discordance for twins as larger than 30% BWD. This threshold was assumed based on our receiver operating characteristic analysis as the optimal cut-off point associated with perinatal adverse outcomes (see Chapter 7).   190  Outcome variables related to perinatal mortality were early and late neonatal mortality. Early neonatal death was defined as the death of the baby within 7 days of birth, whereas late neonatal death was described as babies’ death within 28 days of birth.  Both of these variables included inpatient death and death identified through linkage with Vital Statistics.  Perinatal morbidity was divided into short-term and long-term morbidities. Short term morbidity was inclusive of 1st and 5th minute Apgar scores of less than 7; use of antibiotics, surfactant, steroids for lung maturation, or resuscitation drugs; whether or not the newborn underwent chest compression; and intermittent positive pressure ventilation [IPPV_ETT]. Diagnostic tests comparing growth discordant and concordant twin pairs were streptococcus culture test, cord arterial gases base and cord arterial gases PH <7.0 using blood from cord. The number of days spent under a ventilator and the number of days receiving continuous positive airway pressure (CPAP) were also analysed. Long term morbidity included newborn septicaemia, hypoglycemia, neonatal diabetes, anemia, apnea, IVH, pneumonia, respiratory distress syndrome (RDS), retinopathy and neonatal intensive care unit (NICU) admission. Admission in either NICU II or III signified the severity of neonate condition.5,284 A composite score was calculated using five of the above-mentioned variables: 5th min Apgar score< 7, cord gases < 7.0, NICU admission (level 3) > 2d, NICU admission (level 2) > 2 days, IVH and ventilator use. Confounding variables adjusted for in the analysis is similar to that of pervious chapter.  The difference in outcome between discordant and concordant twin pairs was determined employing sample t-tests or Mann-Whitney tests for quantitative variables (e.g. gestational age) and Chi-square tests (e.g. diabetes) for qualitative variables.  A p-value of <0.05 was considered significant. To account for the dependent manifestation of outcomes among twin pairs, the   191  strength of association between BWD and perinatal outcomes was estimated using GEE modeling. The parsimonious model was selected using the least Quasi-Likelihood under Independent Model Criterion (QIC) estimate.  Subgroup analysis of chorionicity Since PSBC did not include the chorionicity data, a subsample of pathology reports from BC Women Hospital was collected and linked with the outcome data (n= 3182). We excluded cases with TTTs, congenital anomalies, one stillbirth, birth weight less than 500 grams, and those who had reduction procedures. The final analytical dataset included 2986 twins. Chorionicity was then analysed in relation to perinatal mortality/morbidities and BWD.  9.4 RESULTS The study population comprised 12656 twins, of whom 452 (3.57%) were discordant by the percentage definition of ≥30%.  The growth concordant and discordant twins were similar with regards to maternal demographics including maternal age, education, smoking habits, BMI and weight gain during pregnancy. There were no significant differences between concordant and discordant twin pairs in the odds of prior congenital anomalies, low birth weight, stillbirth, and neonatal death.  There were, however, statistically significant differences between growth concordant and discordant twins for the following variables: gestational age, parity, number of antenatal visits, PIH, preeclampsia, and gestational diabetes (Table 9-1).      192  Table 9-1  Characteristics of mothers who delivered in BC according to BWD (n=6328).  <30% (N=6102) ≥30% (N=226) P  Gestational age, Wks, mean±SD 35.28±2.95 33.51±3.81 0.01 Maternal age, years, mean±SD 31.64±5.64 31.76±6.48 0.64 Education, years, mean±SD 14.62±2.85 14.67±2.66 0.93 BMI, kg/m2, mean±SD 24.35±5.03 24.46±5.92 0.69 Pregnancy weight gain, kg, mean±SD 18.60±7.46 17.85±7.47 0.12 Number of antenatal visits, mean±SD 8.96±3.59 8.15±3.80 0.01 Primigravida, n (%) 2998(49.1%) 143(63.5%) 0.01 Smoking, n (%)  Current  Former  Non-smoker  573(9.39%) 415(6.8%) 1010(16.5%)  155(68.5%) 27(11.9%) 15(6.63%)  0.08 Diabetes, n (%) 660(10.8%) 17(7.5%) 0.02 Pregnancy induced hypertension, n (%) 380(6.2%) 26(11.5%) 0.01 Preeclampsia, n (%) 140(2.3%) 24(10.6%) 0.01 Prior congenital anomaly, n (%) 67(1.1%) <5 (2.2%) 0.65 Prior low birth weight baby, n (%) 131(2.1%) <5 (2.2%) 0.24 Prior stillbirth, n (%) 57(0.9%) <5 (2.2%) 0.92 Prior neonatal death, n (%) 29(0.47%) 0(0.0%) - Perinatal outcomes are shown in Table 9-2. Perinatal mortality was more frequent among BWD twins compared with concordant twins. Growth discordant twins had higher frequencies of 1st minute and 5th minutes Apgar scores <7, antibiotic and steroid prescriptions, IPPV-ETT use and a longer average of ventilator days compared with growth concordant twins.  The two groups were similar in terms of surfactant use, chest compression, and use of resuscitation drugs. Comparable distributions of cord arterial base gases and cord arterial PH were also found between the two groups. Frequencies of Streptococcus positive tests were also alike in BWD and concordant twins.   193  Table 9-2 Overall adverse perinatal outcomes in relation to BWD in twins born in BC (6328 pairs, n=12656).  All infants  <30% (N=12204) ≥30% (N=452) P  Birth weight, g, mean ±SD 2446.71±598.51 2463.64±582.29 1984±812.89 0.01 Early Neonatal Death  171(1.4%) 146(1.2%) 25(5.5%) 0.01 Late Neonatal Death  193(1.5%) 166(1.4%) 27(6.0%) 0.01 Short term morbidity Apgar score of 1 minutes <7  2355(18.6%) 2227(18.3%) 128(28.4%) 0.01 Apgar score of 5 minutes <7  491(3.9%) 446(3.7%) 45(10.0%0 0.01 On Antibiotics 654(17.8%) 614(17.4%) 40(26.7%) 0.01 Strep test positive 452(3.6%) 438(24.6%) 14(26.9%) 0.70 Surfactant 192(5.2%) 180(5.1%) 12(8.0%) 0.11 Suction of trachea 51(0.4%) 48(0.4%) <5 (1.1%) 0.37 Steroids for lung maturation  1430(11.3%) 1336(10.9%) 94(20.8%) 0.01 Chest compression 30(0.2%) 28(0.2%) 5 (1.1%) 0.36 IPPV_ETT 403(3.2%) 377(3.1%) 26(5.8%) 0.01 Resuscitation Drug  196(1.5%) 187(1.5%) 9(2.0%) 0.44 Cord arterial gases base mean ±SD -3.25±3.41 -3.25±3.42 -3.49±3.22 0.23 Cord arterial gases PH mean ±SD 7.27±0.06 7.27±0.07 7.26±0.07 0.13 CPAP days mean ±SD 2.19±7.87 2.00±7.53 2.36±9.02 0.13 Ventilator days mean ±SD 0.57±4.74 0.53±4.61 1.59±7.42 0.01 Long term morbidities Newborn septicaemia 318(2.5%) 303(2.5%) 15(3.3%) 0.27 Hypoglycaemia 30(0.3%) 33(0.3%) <5 (1.1%) 0.12 Neonatal Diabetes 34(0.3%) 32(0.3%) <5(0.4%) 0.47 Anemia 297(2.3%) 282(2.3%) 15(3.3%) 0.16 Apnea 1242(9.8%) 1182(9.7%) 60(13.3%) 0.01 Intra ventricular haemorrhage 109(0.9%) 104(0.9%) 5(1.1%) 0.56 Pneumonia 100(0.8%) 93(0.8%) 7(1.5%) 0.06 Respiratory distress syndrome 1001(7.9%) 948(7.8%) 53(11.7%) 0.01 Retinopathy 91(0.7%) 82(0.7%) 9(2.0%) 0.01 NICU II stay >2 days  2302(18.2%) 1024(32.2%) 150(76.5%) 0.01 NICU III stay >2 days  811(6.4%) 737(46.3%) 74(64.9%) 0.01 Composite perinatal morbidity^ 3167(25.0%) 2942(24.1%) 225(49.8%) 0.01 ^ Inclusive of RDS, hypoxic ischemic encephalopathy, periventricular leukomalacia, necrotizing enterocolitis, sepsis   194  For the long term perinatal morbidities under study, apnea, RDS, retinopathy and stay longer than 2 days in NICU II and III were found to be higher in frequency among growth discordant twins (≥30%) in comparison to the growth concordant twins.  GEE analysis Multivariate logistic GEE analysis was used for dichotomous outcomes to assess whether discordance remained an independent risk factor for the perinatal outcomes after adjusting for maternal outcomes. After adjustment, growth discordance was still associated with increased odds of early (6.11, 95%CI 3.84-9.70) and late (2.83, 95%CI 1.47-5.43) neonatal mortality.  Higher odds of 5th minute Apgar scores <7, and NICU II /III admissions >2days were found in growth discordant twins (Table 9-3).The odds of composite perinatal morbidity was 2.32 (95%CI 1.73-2.87).               195  Table 9-3 Odds of perinatal mortality and morbidity associated with BWD in twins born in BC (6328 pairs, n=12656)  OR (95%CI) AOR^ (95%CI) Early Neonatal Death  6.11(3.84-9.70) 3.14(1.62-6.09) Late Neonatal Death  5.78(3.71-9.00) 2.83(1.47-5.43) Apgar score of 1 minutes <7  1.77(1.42-2.21) 1.16(0.90-1.47) Apgar score of 5 minutes <7  3.23(2.30-4.54) 1.64(1.11-2.43) On Antibiotics 1.54(1.02-2.32) 0.82(0.43-1.53) Steroids for lung maturation  2.15(1.67-2.71) 0.98(0.68-1.40) IPPV-ETT 2.39(1.58-3.61) 0.94(0.50-1.75) CPAP days  1.48(1.03-2.13) 0.99(0.69-1.43) Ventilator days 3.42(1.60-7.28) 1.31(0.65-2.64) Newborn septicaemia 1.52(0.89-2.58) 1.03(0.58-1.81) Anemia 1.67(0.98-2.85) 0.99(0.55-1.78) Apnea 1.39(1.03-1.87) 0.94(0.67-1.31) Intra ventricular haemorrhage 1.49(0.60-3.71) 0.49(0.16-1.46) Pneumonia 2.34(1.07-5.11) 1.76(0.80-3.88) Respiratory distress syndrome 1.69(1.24-2.30) 1.05(0.73-1.49) Retinopathy 3.42(1.69-6.89) 1.73(0.79-3.79) NICU II stay >2 days  1.54(1.07-2.23) 1.41(0.97-2.04) NICU III stay >2 days  2.90(1.83-4.59) 2.91(1.76-4.83) Composite perinatal morbidity 4.01(3.38-4.45 2.32(1.73-2.87) Adjusted for maternal behaviour, socio-demographic characteristics and clinical confounders (see section 8.3 for details)  Sex discordant GEE analysis Since we did not have access to chorionicity for the PSBC dataset; we stratified the data according to sex discordance as a proxy (Table 9-4). Multivariate GEE analysis of adverse perinatal outcomes in dislike-sex twins found higher odds of perinatal mortality for BWD twins.   196  From the perinatal morbidity outcomes, we found higher odds of composite perinatal morbidity (2.57, 95%CI 1.59-4.16), and NICU III stay of >2 days compared to the reference category (6.85; 95%CI 1.59-4.16). Other variables (perinatal mortality, Apgar score of 5 minutes <7, septicemia, anemia, apnea, IVH, pneumonia, RDS, retinopathy, NICU admission) ware not found to be associated with BWD. Table 9-4 Odds (95%CI) of perinatal outcomes associated with BWD in respect to sex discordance in twins born in BC (6328 pairs, n=12656)  Sex-concordant (n=8322) Sex-discordant (n=4334)  COR (95%CI) AOR^ (95%CI) COR (95%CI) AOR^ (95%CI) Early Neonatal Death   5.93(3.51-10.04) 2.83(1.35-5.93) 6.44(2.41-17.22) 4.63(1.04-20.51) Late Neonatal Death  5.42(3.26-9.01) 2.34(1.12-4.87) 6.89(2.78-17.04) 5.40(1.28-22.77) Apgar score of 5 minutes <7  3.41(2.31-5.03) 1.63(1.04-2.55) 2.69(1.32-5.45) 1.62(0.71-3.70) Newborn septicaemia 1.26(0.63-2.48) 0.78(0.38-1.61) 2.17(0.93-5.08) 1.63(0.64-4.17) Anemia 1.63(0.87-3.04) 0.93(0.47-1.84) 1.76(0.63-4.91) 1.16(0.37-3.62) Apnea 1.36(0.95-1.94) 0.89(0.60-1.33) 1.44(0.83-2.50) 1.06(0.57-1.96) Intra ventricular haemorrhage 1.55(0.56-4.30) 0.51(0.16-1.66) 1.24(0.16-9.31) 0.40(0.02-6.43) Pneumonia 1.87(0.67-5.20) 1.41(0.50-3.91) 3.49(1.04-11.74) 2.67(0.76-9.37) Respiratory distress syndrome 1.82(1.27-2.60) 1.82(1.27-2.60) 1.37(0.75-2.53) 1.37(0.75-2.51) Retinopathy 1.88(0.77-4.57) 3.64(1.63-8.12) 1.23(0.21-6.99) 2.75(0.64-11.88) NICU II stay >2 days  1.47(0.94-2.29) 1.29(0.82-2.02) 1.71(0.89-3.29) 1.74(0.91-3.34) NICU III stay >2 days  2.32(1.38-3.90) 2.15(1.17-3.95) 6.25(2.11-18.44) 6.85(2.04-23.01) Composite perinatal morbidity 4.25(3.47-5.21) 2.09(1.56-2.81) 3.41(2.50-4.65) 2.57(1.59-4.16) ^ Adjusted for maternal behaviour, socio-demographic characteristics and clinical confounders (see section 8.3 for details) Among sex-concordant twins, perinatal death, 5th minute Apgar score of <7, RDS, retinopathy and NICU III admission longer than 2 days and composite perinatal morbidity (2.09, 95%CI 1.56-2.81) has higher odds of being growth discordant compared with the reference group.      197  Subsample analysis Subsample analysis of 2986 twins born in C&W hospital was done next. Distribution of data in twins with and without growth discordance is shown in Table 9-5. Similar findings to that of the larger population, born in BC, was found. Additionally, frequencies of anemia and newborn septicemia were significantly higher in growth discordant twins compared to the reference category.  GEE analysis for the subsample of 2986 twins born in C&W hospital confirmed higher odds of early and late neonatal mortality for a BWD of greater than 30%, even after adjustment for chorionicity (Table 9-6).  Table 9-5 Comparing perinatal outcomes in twins born in C&W hospital (1493 pairs, n=2986)  All infants  <30% (N=2766) ≥30% (N=220) P  Birth weight, mean ±SD, g 2285.38±664.92 2324.58±635.04 1793.16±819.21 0.01 Placenta Monochorionic Dichorionic  728(24.4%) 2258(75.6%)  644(23.3%) 2122(76.7%)  84(38.2%) 136(61.8%)  0.01 Short term morbidity Early Neonatal Death  70(2.3%) 52(1.9%) 18(8.2%) 0.01 Late Neonatal Death  82(2.7%) 63(2.3%) 19(8.6%) 0.01 Apgar score of 1 minutes <7  169(5.7%) 562(20.3%)  75(34.1%) 0.01 Apgar score of 5 minutes <7  637(21.3%) 138(5.0%) 31(14.2%) 0.01 On Antibiotics 2224(7.5%) 201(7.3%) 23(10.5%) 0.08 Strep test positive 82(2.7%) 73(20.7%) 9(47.4%) 0.01 Surfactant 74(2.5%) 65(2.3%) 9(4.1%) 0.11 Steroids for lung maturation  594(19.9%) 516(18.7%) 78(35.5%) 0.01 IPPV_ETT 191(6.4%) 160(5.8%) 31(14.1%) 0.01 Resuscitation Drug  34(1.1%) 30(1.1%) <5 (2.2%) 0.32 Cord arterial gases base mean ±SD -4.00±3.00 -4.01±2.94 -3.86±3.61 0.55 Cord arterial gases PH mean ±SD 7.26±0.67 7.26±0.06 7.26±0.05 0.39 CPAP days mean ±SD 2.19±8.87 2.00±8.53 4.36±12.02 0.03 Ventilator days mean ±SD 1.57±8.5 1.37±7.92 4.06±14.02 0.01   198   All infants  <30% (N=2766) ≥30% (N=220) P  Long term morbidity Newborn septicaemia 158(5.3%) 137(5.0%) 21(9.5%) 0.01 Anemia 145(4.9%) 122(4.4%) 23(10.5%) 0.01 Apnea 452(15.1%) 400(14.5%) 52(23.6%) 0.01 Intra ventricular haemorrhage 77(2.6%) 70(2.5%) 7(3.2%) 0.55 Pneumonia 39(1.3%) 35(1.3%) <5 (2.2%) 0.48 Respiratory distress syndrome 416(13.9%) 355(12.8%) 61(27.7%) 0.01 Retinopathy 53(1.8%) 44(1.6%) 9(4.1%) 0.01 NICU II stay >2 days  674(22.6%) 601(33.3%) 73(41.7%) 0.02 NICU III stay >2 days  400(13.4%) 340(22.9%) 60(37.3%) 0.01 Composite perinatal morbidity 892(26.6%) 691(25.0%) 126(57.3%) 0.01  The odds of perinatal mortality were 2.71% (95%CI 1.43-5.10) in growth discordant twins compared with concordant ones after adjusting for chorionicity (Table 9-6). The odds of 5th minute Apgar score <7 was 2.01(1.24-3.26). Composite perinatal morbidity was found to be higher in growth discordant compared with growth concordant twins (57.3% vs. 25.0%, p=0.0, Table 9-4). Composite perinatal morbidity was significantly associated with BWD (2.23, 95%CI 1.73-2.87) after adjustment for gestational age and chorionicity. No other variable was found to be associated with BWD in the subgroup analysis. Adjusted perinatal mortality and morbidity outcomes stratified by chorionicity are shown in Figure 9-1. MC twins have higher frequencies of composite morbidity compared to DC twins. Table 9-6 Unadjusted and adjusted OR (95%CI) of perinatal morbidity in twins born in C&W hospital (1493 pairs, n=2986)  COR (95%CI) AOR* (95%CI) Early Neonatal Death  4.65(2.67-8.09) 3.15(1.66-5.94) Late Neonatal Death  4.05(2.38-6.90) 2.71(1.43-5.10)   199   COR (95%CI) AOR* (95%CI) Apgar score of 1 minutes <7  2.02(1.51-2.72) 1.36(0.97-1.90) Apgar score of 5 minutes <7  3.13(2.06-4.75) 2.01(1.24-3.26) Newborn septicaemia 2.02(1.25-3.27) 1.28(0.74-2.21) Anemia 2.53(1.58-4.04) 1.39(0.78-2.47) Apnea 1.83(1.31-2.54) 0.93(0.61-1.43) Intra ventricular haemorrhage 1.58(0.90-2.74) 1.55(0.89-2.71) Pneumonia 1.44(0.50-4.10) 0.91(0.31-2.61) Respiratory distress syndrome 2.60(1.90-3.57) 1.38(0.87-2.19) Retinopathy 2.63(1.27-5.47) 1.68(0.71-3.95) NICU II stay >2 days  1.43(1.04-1.96) 1.13(0.80-1.59) NICU III stay >2 days  2.00(1.42-2.82) 1.39(0.90-2.15) Composite perinatal morbidity 4.03(3.43-4.72) 2.23(1.73-2.87) *Adjusted for gestational age and chorionicity    Figure 9-1  Perinatal mortality and morbidity outcomes stratified by chorionicity in twins born in C&W hospital (1493 pairs, n=2986) 0102030405060708090Early neonataldeathLate neonataldeath5th min. Apgar<7NICU_III >2daysTotal perinatalmorbidityMCDC  200  Table 9-7 shows an overview of adjusted odds of perinatal mortality and morbidities in both the BC population and twins born in C&W hospital. Perinatal mortality, 5th minute Apgar score <7 and composite perinatal morbidity were found to be similarly associated with BWD in both cohorts.  Table 9-7 Adjusted OR (95%CI) of perinatal outcomes in twins registered at BC (6328 pairs, n=12656) and C&W hospital (1493 pairs, n=2986)  BC (n=12656) AOR^ (95%CI) C&W (n=2986) AOR^^ (95%CI) Early neonatal death  3.14(1.62-6.09) 3.15(1.66-5.94) Late neonatal death  2.83(1.47-5.43) 2.71(1.43-5.10) Apgar score of 1 minutes <7  1.16(0.90-1.47) 1.36(0.97-1.90) Apgar score of 5 minutes <7  1.64(1.11-2.43) 2.01(1.24-3.26) Newborn septicaemia 1.03(0.58-1.81) 1.28(0.74-2.21) Anemia 0.99(0.55-1.78) 1.39(0.78-2.47) Apnea 0.94(0.67-1.31) 0.93(0.61-1.43) Intra ventricular haemorrhage 0.49(0.16-1.46) 1.55(0.89-2.71) Pneumonia 1.76(0.80-3.88) 0.91(0.31-2.61) Respiratory distress syndrome 1.05(0.73-1.49) 1.38(0.87-2.19) Retinopathy 1.73(0.79-3.79) 1.68(0.71-3.95) NICU II stay >2 days  1.41(0.97-2.04) 1.13(0.80-1.59) NICU III stay >2 days  2.91(1.76-4.83) 1.39(0.90-2.15) Composite perinatal morbidity 2.32(1.53-3.41) 2.31(1.05-2.43) ^ Adjusted for maternal behaviour, socio-demographic characteristics and clinical confounders (see section 8.3 for details); ^^Adjusted for chorionicity and gestational age  9.5 DISCUSSION The association between BWD and adverse perinatal outcomes has been established in the literature.75,153,285–288 However, the literature is inconsistent in terms of adjusting the predictive   201  models for chorionicity. Our finding is consistent with the literature in that we found higher odds of perinatal mortality and morbidity in BWD twins compared with growth concordant twins. These findings were observed in both cohorts, with and without chorionicity data. We also found increased odds of a 5th minute Apgar score <7, and long stays in NICU in growth discordant twins compared with the reference category. A hospital-based study of 150 twin gestations found an increased admission to the NICU in BWD twins compared to concordant twins (14.3% versus 3.5%, p=0.032 for larger infants and 20.0% versus 4.3%, p-0.007 for smaller infants).87 A study on 253 pairs of twins who were delivered in a hospital in Pakistan found that 12% of growth discordant and 7% of concordant infants required NICU admission. Similarly, mean growth discordance for 346 twins born in a hospital in the USA was independently associated with NICU admission.76  Previous studies have demonstrated that the risk associated with discordant twins is related to the increased incidence of prematurity and the growth restriction of the twin pairs. To reduce mortality, resuscitation interventions and respiratory supports are used to manage complications associated with prematurity complications. The literature lacks the frequency with which twins with BWD undergo these interventions both at birth (e.g. resuscitation medication, cord blood testing for gases) and after birth (e.g. respiratory support, described here as the need for a ventilator, CPAP, or oxygen during hospitalization).  Our findings suggest that at birth, the incidence of resuscitation medication used and the result of cord arterial gases testing are similar between growth discordant and concordant twins. However, there is a higher frequency of IPPV-ETT use among growth discordant twins compared to those with less than 30% growth discordance. With regards to infection, we found that BWD infants have a higher rate of antibiotic prescription (26.7% vs. 17.4%) compared to growth concordant infants even though   202  the strep test results similar between growth discordant and concordant twins (26.9% vs. 24.6%, p=0.70, Table 9-2).  Our findings demonstrate that the BWD twins required early respiratory support (more ventilator days than growth concordant twins); this finding is consistent with those of prior studies that identified the same difference.157,289 Information on adverse neonatal outcomes of growth discordant opposite-sex twins within the early hospitalization period is scarce. Our stratified analysis of sex-discordant twins was conducted as a substitute for chorionicity. This type of analysis has been adopted by other studies where no access to chorionicity data was possible.256,290  The adjusted odds (for sex discordance and other confounders) of perinatal mortality and morbidity were higher in BWD twins compared to concordant twins. These findings are also consistent with the literature.16,46,63,146,148,154  If we assume that differences in chorionicity confound the association between BWD and adverse perinatal outcomes, we would expect similar or lower odds among sex-discordant twin pairs who have dichorionic placentas.  Two other studies that used sex discordance as a surrogate variable for chorionicity found that BWD and neonatal mortality were associated in both same- and opposite- sex twins.63,134 When we stratified the data according to sex discordance, the odds of early and late neonatal mortality in opposite-sex twins with growth discordance were higher than the corresponding odds in the sex-concordant data (Table 9-4). This result is in contrast with a large study of 4091 unlike-sexed and 10875 like-sexed twins with gestational age of 28 weeks and longer. This inconsistency might be due to several factors. Pregnancy loss, in the publication, was twice as high in like-sexed compared with unlike-sexed pairs, and only in like-sexed pairs was pregnancy loss strongly correlated to BWD. This study only included twins with a birth weight of 500   203  grams or more.  Furthermore, twins with TTTs and congenital anomalies or single stillbirths were not excluded.291 Regarding perinatal morbidity, the findings from our subsample data analysis, taking chorionicity into consideration, were consistent with the results from the literature. A prospective multicentre study of 1028 twin pairs over a period of two years used a composite measure of perinatal morbidity inclusive of RDS, hypoxic ischemic encephalopathy, periventricular leukomalacia, necrotizing enterocolitis and sepsis. Adjusting for gestational age at delivery, perinatal mortality and composite perinatal morbidity were associated with BWD in both MC and DC twins.248 Another report from the same study found that a composite measure of adverse perinatal outcomes, which included any of the morbidity measures described above or perinatal death, was more frequent in monochorionic twins.292 Our adjusted odds suggested that composite perinatal morbidity was significantly higher in BWD twins. We have included five variables to form the composite for our analysis: 5th minute Apgar score, NICU admission, IVH, ventilator use and cord blood gas <7.  Even though this composite does not include important variables such as necrotising enterocolitis, we were able to show the significant association between BWD and perinatal morbidity. The literature lacks a validated composite index from perinatal morbidity that would be easily available in most databases. Information obtained from our study is important for caretakers of neonates because the short-term outcome differences of discordant twins draws nurses’ and parents’ attention to the extent of care these babies need. As the rate of twinning, prematurity and associated complications increases, this knowledge is necessary for both healthcare professionals and parents to better comprehend and anticipate outcomes in this group of twins who are “high risk”. Future research   204  to assess the role of chorionicity is beneficial to further clarify the most specific group of twins in need of NICU services.   It is worth noting that coding for prenatal NICU admission, often is related to maternal admission and is done differently in various institutions. This might lead to differences in effect measures between BC population and C&W hospital. 9.6 SUMMARY  Twin pregnancies with BWD are at high risk of prenatal mortality, longer stays in the NICU and lower Apgar score, and composite perinatal morbidity, irrespective of chorionicity. With increased twin pregnancy, prematurity, and higher utilization of NICU resources at both delivery time and during infancy, this knowledge is necessary for both health care providers and parents of twins to anticipate outcomes in this highly specific population.   205  CHAPTER 10 FETAL SEX AND ADVERSE PERINATAL OUTCOMES  10.1 INTRODUCTION The rate of twinning is increasing in Canada. The twin birth rate increased from 2.5 per 100 births in 2000/2001 to 3.1 per 100 in 2007/2008.105 This may be explained by the increased use of assisted reproductive technology among the aging maternal demographic.2, 3  In British Columbia, the average age of women having their first child increased from 21.5 in 2000 to 25.0 in 2008.293 It is hypothesized that older women produce more follicle stimulating hormone. Thus, compared to younger population of women, older women are likely to ovulate more than one egg, hence increasing the risk of twining, independent of maternal age.294     Twin pregnancies are known to have increased perinatal complications.5  These adverse perinatal outcomes are attributed to BWD,6,7 prematurity,298 placenta/cord complications,102 method of delivery,299 and sex discordance.244 It is hypothesized that the adverse perinatal outcomes in twin gestations could be the result of the mere presence of a male fetus in the uterus along with a female that has a profound effect on the perinatal morbidity and mortality of the female fetus.25  Animal studies12, 13 have hypothesized that androgens are transferred between fetuses across membranes, and have shown that female fetuses in discordant twin pairs have higher concentrations of serum testosterone than female fetuses of female-female pair.  In addition, androgen is found to have an inhibitory effect on lung development.27  In theory, this could explain the increased risk of respiratory morbidity (e.g. RDS) found among male-female twin pairs compared to female-female twin pairs (1.3; 95%CI 1.1–1.7).31    206  Based on the same principle, previous studies have attempted to assess the association between fetal sex and other adverse perinatal outcomes. However, the findings from these studies remain indecisive. Male fetal death among male-male twins was found to be higher than300 or identical to 91 male mortality among male-female twins. Other studies have found contrary results. For instance, higher fetal death rates have been observed for males within male-female twins compared with male-male with33 or without91 controlling for gestational age.  Similarly, the incidence of neonatal mortality was compared between male versus female within sex discordant twin pairs in low birth weight infants of 24-34 weeks twin gestation (n=2448).91  After adjustment for gestational age, male sex was associated with increased odds for neonatal mortality (1.36, 95%CI 1.05-1.77) in comparison with females.  Another retrospective study of dizygotic twin pregnancies detected high odds of prematurity (1.7, 95%CI 1.2-2.6) for male-male twin pairs compared to female-female pairs.  Klein et al. study (n=125)301 also found that the risk of preterm birth before 35 weeks’ gestation was highest in the male-male (20.2%), lowest in the female-female (15.7%) and intermediate in male-female (17.9%) twins (p<0.001).   Shinwell et al’s91 study (n=2448) on very low birth weight preterm twins showed that the male mortality rate was identical in sex-discordant and sex-concordant (male-male) twin pairs (21.6%).  Similar rates were observed for retinopathy of prematurity (7.1% vs. 7.9%, p>0.05) and periventricular intraventricular haemorrhage (14.4% vs. 14.6%, p>0.05) for males in sex-discordant twin sets compared with males in sex-concordant twins. However, discordant twins had lower risks for pneumothorax (9.7% vs. 11.7%, p<0.05), periventricular leukomalacia (6.8% vs. 9.8%, p<0.01), respiratory distress syndrome (73.6% vs. 76.6%, p=0.007) and   207  bronchopulmonary dysplasia (11.9% vs. 13.7%, p=0. 04) due to the inhibitory effect of androgen on lung development.31 One major limitation of some of these studies is the lack of a large population-based sample. Another important limitation is inadequate adjustment for confounding factors such as maternal characteristics, medical complications during pregnancy and obstetric history. But most importantly, these studies often analyzed data using an inappropriate regression analysis that can be misleading. Twins provide naturally matched pairs or clusters with within-twin-pair and between-twin-pair effects. This type of data needs a specialized standard regression models that reflect the paired structure of the data, which induces correlation between twins.302 Generalized estimating equations procedures that accommodate correlated clustered data were used in our data analysis to build up such models. Moreover, we used a large population-based province-wide sample of twins to examine the relationship between fetal sex pairing in twin pregnancy and fifteen adverse perinatal outcomes. Melamed et al.300 investigated the role of fetal sex relative to perinatal morbidities but did not present perinatal mortality and did not use marginal methods to account for the clustering effect of twinning. A recent publication that has used GEE modeling to obtain the effect of fetal sex combination on perinatal outcome did not look into a wide range of perinatal morbidities, was not population-based, and did not control for maternal characteristics, gestational age and birth order.81 We examined the association between fetal sex pairing in twin pregnancies and perinatal outcomes using a population-based and province-wide (from C&W hospital), containing chorionicity information, sample of 10 years of registered twin pregnancies. We used GEE modeling and adjusted for a wide range of confounding variables including chorionicity.    208  10.2 OBJECTIVE  To examine the association between fetal sex and adverse perinatal outcomes in twin pregnancies with or without chorionicity.   10.3 METHODS In brief, we accessed a population-based retrospective cohort study of 6328 twin gestations (12656 twins) in BC, Canada for the period of 2000-2010 using the PSBC data. Variables considered for this analysis are described in Chapters 3 and 9. Demographic and clinical characteristics of mothers and adverse perinatal outcomes were summarized according to the following sex pairings: males from male-male twins, females from female-female twins, males from male-female twins and females from male-female twins. We have excluded cases with unknown sex, TTTs, congenital anomalies, birth weight less than 500 grams, and those whose co-twin died in utero during pregnancy. General estimating equation models for binary outcomes were used for multivariate analysis accounting for the non-independent occurrence of outcomes among twin pairs and the strength of association between sex discordance and the outcome of interest. The method allows data for twin births to be treated as cluster measures of the same birth. Clustering of twins within mothers is often overlooked, which can lead to over-estimation of effects if neglected.255   For the multivariate model of outcomes, data were adjusted for covariates based on variables available in the PSBC dataset, clinical and epidemiological evidence of confounding effects in the literature, and results of the bivariate analysis. These confounding variables are listed in method section and section 8.3 including variables such as maternal behaviour (e.g. number of   209  antenatal visits, smoking habits), socio-demographic characteristics and clinical confounders. In sequential models, one potential confounding variable was tested at a time.  Retaining the most statistically significant variable (p value of less than 0.05), the remaining variables were entered separately in sequential models, and then the process repeated until all potential variables had been tested. Adjusted odds ratios for the final model (and 95% CI) were then reported. The models with the least Quasi-Likelihood under Independent Model Criterion (QIC) estimate was chosen as the best model. All analyses were repeated after restricting the study population to sex-discordant twins, in order to gauge confounding by chorionicity using placenta pathology data available from C&W hospital.  10.4 RESULTS A total of 13330 newborn twins were born in British Columbia from 2000 through 2010, of which 14 twins were listed with unknown sex and were deleted, leaving 13316 records for analysis.  We further excluded cases with TTTs, congenital anomalies, birth weight less than 500 grams, and those whose co-twin died in utero during pregnancy. There were 12642 cases of twins (n=6321 pairs) remaining in the analytical dataset. The sample included 6371 males and 6281 females. More than 2/3 of the infants were sex-concordant (n=8322). Approximately equal numbers of infants were from male-male (n=4202), female-female (n=4106) and male-female (n=4434) gestations.  Table 10-1 summarizes maternal characteristics and gestational age. Gestational age was significantly different between the three groups of male-male, female-female, and male-female twin pairs (p=0.04). The groups were similar in terms of baseline weight gain during pregnancy, and number of antenatal care visits. ANOVA tests showed significant differences between   210  groups for the following variables: maternal age, maternal school years, BMI and parity.  Maternal complications during pregnancy were similar between groups. History of stillbirth and low birth weight were statistically significantly different between the three sex groups. The sex-discordant group had the highest rate of in vitro fertilization (p=0.01). Table 10-1 Maternal characteristics of twin gestations born in BC according to sex combinations (6321 pairs, n=12642)  Female-Female (4106 pairs) Female-Male (4334 pairs) Male- Male  (4202 pairs) ANOVA Gestational age, Wks, Mean±SD 35.15±3.02 35.30±3.00 35.15±3.14 0.04 Maternal age, years, Mean±SD 31.25±5.68 32.35±5.67 31.26±5.72 0.01  Education, years, Mean±SD 14.79±2.87 14.75±2.81 14.32±2.85 0.01 BMI, kg/m2, Mean±SD 24.26±4.99 24.60±5.25 24.20±4.92 0.01 Pregnancy weight gain, kg, Mean±SD 18.54±7.87 18.49±7.05 18.71±7.44 0.54 Number of antenatal visits, Mean±SD 8.89±3.72 8.89±3.44 9.00±3.63 0.32 Primigravida, n (%) 2076(50.6%) 2236(51.6%) 1960(46.6%) 0.01 Smoking, n (%)  Current  Former  Non-smoker  368(27.1%) 286(21.1%) 703(51.8%)  404(27.8%) 292(20.1%) 758(52.1%)  426(32.2%) 280(21.1%) 618(46.7%)  0.02 Diabetes, n (%) 412(10.0%) 492(11.4%) 450(10.7%) 0.15 Pregnancy induced hypertension, n (%) 266(6.5%) 274(6.3%) 266(6.3%) 0.94 Preeclampsia, n (%) 108(2.6%) 110(2.5%) 84(2.0%) 0.12 Prior congenital anomaly, n (%) 38(0.9%) 52(1.2%) 50(1.2%) 0.39 Prior low birth weight baby, n (%) 82(2.0%) 66(1.5%) 120(2.9%) 0.01 Prior macrocosmic baby, n (%) 160(3.9%) 188(4.3%) 166(4.0%) 0.53 Prior stillbirth, n (%) 18(0.4%) 60(1.4%) 40(1.0%) 0.01 Prior neonatal death, n (%) 22(0.5%) 20(0.5%) 16(0.4%) 0.58 IVF, n (%) 249(25.2%) 492(43.5%) 314(28.8%) 0.01 IVF: In vitro fertilization;  BMI: Body mass index Fetal weight was highest in male infants belonging to male-female twin pairs. Figure 10-1 shows the birth weight of twins according to their sex-pairing.    211   Figure 10-1 Birth weight of infants according to sex pairing F=female, MM=male-male, FF=female-female, MF=male-female .Table 10-2 shows the frequency of adverse perinatal outcomes and mean± standard deviation (SD) of birth weight in four sex pairing groups: females from female-female gestations (n=4106), females from male-female gestations (n=2165), males from male-male gestations (n=2166), and males from male-female gestations (n=2165).  From the categorical perinatal outcomes under study, prolonged stay in NICU III > 2 days (47.5%), prolonged stay in NICU II > 2 days (18.2%), positive streptococcus test (24.7%) and 1st minute Apgar score of <7 (18.6%) were the most common, while stillbirth (0.5%) and chest compression (0.2%) were the least frequent outcomes.  The average number of days that infants were under CPAP was associated with sex pairing (p=0.01). Chi-square test showed significant differences between the four sex-pairing groups for 230023502400245025002550F from FF M from MM M from MF F from MFBirth weight  212  perinatal mortality (stillbirth, early and late neonatal mortality, Table 10-2) and Apgar scores (1st and 5th minutes). Males from the male-male group had the highest frequency of IPPV resuscitation intervention compared with the other groups (p=0.01). The highest frequency with which surfactant was prescribed belonged to females from female-female twin pairs (6.8%) compared with the other groups.  Moreover, anemia, apnea and composite perinatal morbidity were statistically significantly associated with sex pairing. The incidence of apnea was highest among female from female-female twin pairs compared with other categories. Composite perinatal morbidity was highest in frequency among males of male-male twin pairs (26.7%) .  213  Table 10-2 Association between twin sex pairing and perinatal outcomes in twins born in BC (6328 pairs, n=12656)    All infants     Female from a FF Gestation n=4106 Female from a MF Gestation n=2165 Male from a MM Gestation n=2166 Male from a MF Gestation n=2165 P  Stillbirth 68(0.5%) 30(0.7%) 6(0.3%) 20(0.5%) 8(0.4%) 0.05 Early Neonatal Death 171(1.4%) 60(1.5%) 18(0.8%) 66(1.6%) 23(1.1%) 0.05 Late Neonatal Death 193(1.5%) 65(1.6%) 21(1.0%) 78(1.9%) 25(1.2%) 0.02 Short term morbidity Apgar score of 1 minutes <7 2355(18.6%) 803(19.6%) 402(18.6%) 794(19.0%) 350(16.2%) 0.01 Apgar score of 5 minutes <7 491(3.9%) 172(4.2%) 62(2.9%) 180(4.3%) 73(3.4%) 0.01 On Antibiotics 65(17.8%) 208(18.3%) 102(16.1%) 230(18.3%) 114(17.9%) 0.63 Strep test positive 452(24.7%) 132(23.0%) 72(23.3%) 176(27.5%) 72(23.3%) 0.24 Surfactant 192(5.2%) 77(6.8%) 17(2.7%) 74(5.9%) 24(3.8%) 0.01 Suction of trachea 51(0.4%) 15(0.4%) 12(0.6%) 14(0.3%) 9(0.4%) 0.58 Steroids for lung maturation 1430(11.3%) 466(11.3%) 231(10.7%) 502(11.9%) 231(10.7%) 0.32 Chest compression 30(0.2%) 13(0.3%) <5 (0.2%) 8(0.2%) 6(0.3%) 0.47 IPPV_ETT 403(3.2%) 132(3.2%) 50(2.3%) 165(3.9%) 56(2.6%) 0.01 Resuscitation Drug 196(1.5%) 58(1.4%) 35(1.6%) 76(1.8%) 27(1.2%) 0.29 Cord arterial gases base mean ±SD -3.25±3.40 -3.13±3.39 -3.41±3.55 -3.20±3.36 -3.41±3.35 0.04 Cord arterial gases PH mean ±SD 7.27±0.07 7.27±0.07 7.27±0.07 7.27±0.07 7.26±0.07 0.01 CPAP days* mean ±SD 2.11±2.72 2.01±2.49 3.05±1.99 2.06±2.15 3.17±3.44 0.01 Ventilator days* mean ±SD 0.56±4.73 0.61±5.47 0.43±4.14 0.63±4.47 0.48±4.27 0.30   214     All infants     Female from a FF Gestation n=4106 Female from a MF Gestation n=2165 Male from a MM Gestation n=2166 Male from a MF Gestation n=2165 P  Long term morbidities Newborn septicaemia 318(2.5%) 109(2.7%) 48(2.2%) 110(2.6%) 50(2.3%) 0.64 Hypoglycaemia 36(0.3%) 12(0.3%) <5 (0.2%) 15(0.4%) <5 (0.2%) 0.62 Neonatal Diabetes 34(0.3%) 8(0.2%) 8(0.4%) 16(0.4%) <5 (0.2%) 0.10 Anemia 297(2.3%) 114(2.8%) 47(2.2%) 103(2.5%) 33(1.5%) 0.01 Apnea 1242(9.8%) 445(10.8%) 181(8.4%) 422(10.0%) 194(9.0%) 0.01 Intra ventricular haemorrhage 109(0.9%) 33(0.8%) 11(0.5%) 46(1.1%) 19(0.9%) 0.11 Pneumonia 100(0.8%) 30(0.7%) 16(0.7%) 38(0.9%) 16(0.7%) 0.79 Respiratory distress syndrome 1001(7.9%) 335(8.2%) 159(7.3%) 351(8.4%) 156(7.2%) 0.27 Retinopathy 91(0.7%) 28(0.7%) 16(0.7%) 35(0.8%) 12(0.6%) 0.64 NICU II stay >2 days  2302(18.2%) 764(69.3%) 373(67.6%) 778(67.9%) 385(67.5%) 0.83 NICU III stay >2 days  811(47.5%) 280(47.5%) 126(46.7%) 273(48.4%) 132(46.8%) 0.95 Composite perinatal morbidity^  3167(25.0%) 1048(25.5%) 477(22.0%) 1123(26.7%) 513(23.7%) 0.01 ^ includes 5th min Apgar score< 7, cord gases < 7.0, NICU admission (level 3) > 2d, NICU admission (level 2) > 2 days, IVH and ventilator use;  IPPV_ETT: intermittent positive pressure ventilation;  NICU: Neonatal intensive care unit; CPAP: Continuous positive airway pressure; *Kruskal Wallis Test;  215  Regression modeling The results of the GEE analysis are shown in Table 10-3. Considering females from female-female twin pairs as the reference category, females from discordant twin pairs (0.37, 95%CI 0.15-0.89) and males from concordant twin pairs (0.42, 95%CI 0.20-0.86) were less likely to die in utero. The odds of early neonatal death were 7