UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Risk factors for drug resistant TB in British Columbia Moniruzzaman, Akm 2004

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

Item Metadata

Download

Media
831-ubc_2004-0566.pdf [ 4.24MB ]
Metadata
JSON: 831-1.0091623.json
JSON-LD: 831-1.0091623-ld.json
RDF/XML (Pretty): 831-1.0091623-rdf.xml
RDF/JSON: 831-1.0091623-rdf.json
Turtle: 831-1.0091623-turtle.txt
N-Triples: 831-1.0091623-rdf-ntriples.txt
Original Record: 831-1.0091623-source.json
Full Text
831-1.0091623-fulltext.txt
Citation
831-1.0091623.ris

Full Text

RISK FACTORS FOR DRUG RESISTANT TB IN BRITISH COLUMBIA by AKM MONIRUZZAMAN MBBS, MBA Health Care and Epidemiology A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF Master' of Science in FACULTY OF GRADUATE STUDIES (Health Care and Epidemiology) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA SEPTEMBER 2004 Copyright Akm Moniruzzaman THE UNIVERSITY OF BRITISH COLUMBIA FACULTY OF G R A D U A T E STUDIES Library Authorization In presenting this thesis in partial fulfillment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Name of Author ( p l e a s e print) Date (dd/mm/yyyy) Title of Thesis: TW T I S M co Luhi VIA Degree: MS, Year: Department of 14 £ A L T~ti £ . A £ & /L BP£D S f^A ID L 0 <r~y The University of British Columbia Vancouver, B C Canada ABSTRACT Objectives: A) To investigate the epidemiological risk factors associated with resistance to anti-tuberculosis drugs in British Columbia B) To determine i f there are differences in risk factor characteristics among different categories of drug resistant (mono-drug, poly and multi-drug resistance) TB cases. Methods: The study was carried out in two phases. The first phase of the study was retrospective in design. The data examined in this study was collected by the division of TB control of British Columbia and was obtained from retrospective review of charts, prescriptions, files of the TB subjects or other clients who attended the provincial (BC) TB service. The second phase involved a matched case control study design (using a sample from the total study population). Cases were all resistant TB subjects while controls were selected randomly from the pool of sensitive subjects matched on age group and sex. Results: A total of 3041 TB subjects were eligible for this study over a period of 12 years (1990-2001). Of those eligible, there were 295 (10%) resistant cases and 2746 (90%>) sensitive cases. The significant risk factors for drug resistance found in this study were immigration status, ethnicity, birth country, last country of residence, a history of previous TB, place of diagnosis and case category. With respect to immigration, foreign born people are four times more likely to present with drug resistance than the Canadian-born people. Among the ethnic groups, Chinese, Southeast Asian and the Rest of the Asian are two to three times more likely to present with drug resistance than Caucasians. According to birth country, people born in China, Southeast Asia, and other parts' of Asia were four times more likely to present with drug resistance than Canadian-born people. Subjects with a history of previous TB were twice as likely to present with drug resistance compared to the subjects without any history of previous TB. Subjects diagnosed outside of Canada were almost three times more likely to have drug resistance compared to the subjects diagnosed in Canada. Reactivated (relapsed) cases are almost three times as likely to present with drug resistance compared to the new cases. Conclusions: In brief, foreign born immigrants, patients with a history of previous TB and reactivated (relapsed) TB cases are at greater risk for TB drug resistance. Efforts should be made to prioritize the development and implementation of effective screening and treatment protocols for these high risk groups in order to avoid the emergence of further resistance tuberculosis. TABLE OF CONTENTS Chapter 1: Background, Rationale and Objectives 1 1.1 Background and Rationale 1 1.2 Objectives 3 Chapter 2: Methodology- design, source of data and its management, variables and analysis 3 2.1 Design 3 2.2 Source of Data 4 2. 3 Databases Management 5 2.4 Variables of interest 10 2.4.1 Outcome (dependent) variable - drug resistance 10 2.4.2 Independent variables 12 2.5 Statistical Analysis 19 Chapter 3: Results 21 3.1 Descriptive Statistics 21 3.2 Assessments of the risk factors in univariate model 28 3.3 Assessments of the risk factors in multiple regression models 30 3.4 Assessments of selected risk factors in polytomous (multi-nominal) logistic regression ". v. 33 3.5 Assessments of selected risk factors in case-control setting 36 3.6 Sensitivity analysis 37 Chapter 4: Discussion ...37 Bibliography 47 Appendix 51 - i i i -LIST OF TABLES Table 1: A n outline of TB data management for drug resistance study .. .8 Table 2: Number of Resistant and Sensitive Cases across the study years in BC 52 Table 3: Number of Resistant Cases with types from 1990-2001 in B C 52 Table 4: Descriptive Statistics of Socio-Demographic variables 53 Table 5: Descriptive Statistics of Medical Variables 54 Table 6: Results of Univariate Analysis for Socio-Demographic variables 55 (Table 7: Results of Univariate analysis for Medical variables 56 Table 8: Multivariate analysis for Socio-Demographic variables (age and gender adjusted) 57 Table 9: Multivariate analysis for Medical variables (age and sex adjusted) 58 Table 10: A comparative analysis of Socio-Demographic variables in Univariate and Multivariate model 59 Table 11: A comparative analysis of Medical variables in Univariate and Multivariate model. ..60 Table 12: Number of cases and controls by strata 61 Table 13: A comparison of socio-demographic variables between first phase (multiple regressions) and second phase (conditional regression) analysis 61 Table 14: A comparison of Medical variables between first phase (multiple regression) and second phase (conditional regression) analysis 63 Table 15: Age, Gender and Ethnicity adjusted analysis of selective variables 63 Table 16: Sensitivity analysis for selective variables (30% or more missing observations) 64 17: Age and gender adjusted effect of selective variables across different categories of drug resistant 65 Table 18: Age, Gender and Birth Country adjusted analysis of selective variables 67 - i v -LIST OF FIGURES Figure 1: Trend of Sensitive and Resistant Cases from 1990-2001 22 Figure 2: Trend of different categories of Resistant Cases from 1990-2001 23 Figure 3: Age distribution of the study population 24 Figure 4: Distribution of the Study Population by Gender 25 Figure 5: Rate of Drug Resistance across age categories 26 -v-LIST OF ABBREVIATIONS AR: active registry. BC: British Columbia. BCCDC: British Columbia Centre for Disease Control. BCG: Bacille Calmette-Guerin. DTB: Division of TB Control of British Columbia. EMB: ethambutol. INH: isoniazid. IPHIS: Integrated Public Health Information System LTBI: latent tuberculosis infection. MOTT: mycobacteria other than M T B . MTB: tubercle bacilli (M. tuberculosis, M. bovis or M. africanum) PZA: pyrazinamide. RM: rifampin. SM: streptomycin. STD: sexually transmitted diseases. TB: tuberculosis; unless specified, this term refers to the disease. TST: tuberculin skin test. -vi-DEDICATION To my father: Late Md. Khalilur Rahman -vii-ACKNOWLEDGEMENTS I offer heartfelt thanks to Dr Mark Fitzgerald, my supervisor who shared so generously of his time, knowledge and general enthusiasm for research. I also wish to thank my co-supervisors, Dr. Michael Sculzer, Dr. Monika Naus and Dr. Kevin Elwood for their invaluable mentorship and support. They consistently provided me with sound advice, great learning opportunities and support in all forms. They really made the completion of this degree a great learning experience. I feel very privileged to have had such an amazing thesis committee and will do my best to follow the example they have set. I would also like to thank the Department of Health Care and Epidemiology for providing a fantastic learning environment filled with opportunities. In particular I would like to thank Laurel Slaney for her friendship, support and never ending help jumping through all the bureaucratic hoops associated with the completion of this degree. I am greatly indebted to Ms. Fay Hutton, data manager at the Division of TB Control in British Columbia, for providing the databases used in this study, as well as for her very efficient and thoughtful help with the corrections and quality control issues that emerged throughout the two years during which the study was carried out. - V l l l -MSc Thesis Chapter 1: Background, Rationale and Objectives 1.1 Background and Rationale According to a 1990 study by the World Health Organization (WHO), approximately 1.7 billion people have been infected with tuberculosis (TB) throughout the world 1. The same report 2 3 showed that about 8 million people develop active disease worldwide every year ' . Among the infectious diseases, TB is considered the second largest contributor to adult mortality and is responsible for approximately two million deaths annually3'4. Another WHO report also found the case fatality rate for TB to be around 23% 4. Among diseases that are caused by a single pathogenic agent, TB ranks the worst in terms of disability and economic cost5'6. Tuberculosis, because of its grave impact on humanity, was the first disease to be declared as a global emergency by the World Health Organization (WHO) in 19937'8. Even before the introduction of chemotherapy in the middle of 20 t h century, the incidence of TB had been steadily declining5'9. Subsequently there has been a further decline, particularly in the developed world 1 0. Despite continued efforts to control the spread of TB, it continues to pose a growing threat throughout the world 1. According to the WHO, the global incidence rate of TB is increasing at an approximate rate of 0.4% per year11. Recent estimates indicate that i f stronger control measures are not implemented, between year 2002 and 2020, approximately 1 billion more people will be infected, over 150 million of them will eventually develop active disease, and 36 million of them will die from T B 1 2 . Two forces have been identified in the published literatures which were strongly associated with this recent increase10. One is a natural phenomenon, the human immune deficiency virus (HIV) which by destroying the T cells-macrophages and CD4 lymphocytes (most important to the containment of tubercle bacilli) rigorously promotes the progression of recent or remotely acquired TB infection to active disease10'13. The second and most important factor is the increase in the incidence of strains that are resistant to standard treatment regimes -so called drug resistance1'5. 1 MSc Thesis Five drugs -streptomycin, isoniazid, rifampin, ethambutol and pyrazinamide have been considered as the first line antibiotics effective against T B 1 0 ' 1 4 ' 1 5 . Mono-resistance is defined as the resistance to only one of the first line antibiotics 1 0 ' 1 4 ' 1 5. Multi-drug resistance (MDR) is defined as resistance to at least isoniaziad and rifampicin- the most powerful anti-tuberculosis agents1 0'1 4'1 5'1 6. Other combinations/patterns of drug resistance are referred to as poly resistance. Although the emergence of drug resistance is a recent development, this form of TB has already been documented in nearly every country surveyed by the World Health Organization (WHO)/International Union Against Tuberculosis and Lung Disease (JJJATLD) Global Drug Resistance Surveillance Project during the period 1994-2000 1 7 ' 1 8 ' 1 9. This project included estimates of the distribution pattern of drug resistant cases by country. The assessment showed that for patients with newly diagnosed tuberculosis, the frequency of resistance to at least one tuberculosis drug ranged from 1.7 percent in Uruguay to 36.9 percent in Estonia (median, 10.7 percent)18. In addition, their estimations showed that about 273,000 (95% CI- 185,000 and 414,000) new cases of multi-drug resistance occurred throughout the world in 2000, which was 3.2% of all new TB cases . The magnitude of the problem becomes more evident from the remarks of Dr. Kenneth G Castro, Director of the TB Elimination Division of the US Centers of Disease Control and Prevention - "We are running out of medications that will work against TB. The scenario threatens to return us to the pre-antibiotic era" Drug resistance is a potential threat to TB control programs throughout the w o r l d 1 5 ' 2 1 ' 2 2 ' 2 3 . Because drug resistant TB cases are less likely to be cured21, and their treatment involves long-term, expensive2 4'2 5 (50-100 times more costly than the cheapest short course regimen used for drug susceptible cases19'25) and toxic drug therapy 2 1 ' 2 5 ' 2 6 ' 2 7 and occasionally, surgery10. Another concern with drug-resistant cases is that they are associated with a higher rate of morbidity and mortality than drug susceptible TB for both HIV-sero-positive and sero-negative subjects13. The cure rate of drug resistant cases varies across countries (from 60% in Hong Kong to as little as 5% in Russia16) depending on treatment strategy and other factors. Although under optimal conditions cure rates of over 60% (in some occasions over 90%) have been obtained in some developed countries, these treatments involve very high cost 2 0 ' 2 S. For example, in London, U K , the cost of managing a case of M D R - T B is in excess o f f 60,000 (about 100,000 U S D ) 2 0 ' 2 9 . 2 MSc Thesis In spite of the rising numbers of resistant TB cases worldwide, the epidemiology of TB drug resistance has not been addressed adequately21'23. For instance, the risk factors for drug resistant TB are not well characterized30, especially at a population level. In spite of evidence that the epidemiology of drug resistant TB is more diverse than was initially thought, the number of studies conducted in this area has been limited. For example, most of the studies are cross-sectional in nature, which provides limited value in terms of understanding causality31. It is well established that retrospective studies have higher credibility than cross sectional design in terms of validity and quality of epidemiological evidence31. So far, very few large population based retrospective studies have been carried out to examine the risk factors of drug resistance in TB. The purpose of this study was to address the need for rigorous research on resistant TB by investigating the epidemiologic risk factors for drug resistance. 1.2 Objectives • To investigate the epidemiological risk factors associated with resistance to anti-tuberculosis drugs in British Columbia (BC). • To determine i f there are differences in risk factor characteristics among different categories of drug resistance (mono-drug, poly and multi-drug resistance) TB cases. \ Chapter 2: Methodology- design, source of data and its management, variables and analysis 2.1 Design A retrospective study design was employed for this research project. The study was carried out in two phases. The first phase of the study was purely retrospective. The data examined in this study were obtained from retrospective review of charts, prescriptions, files of the TB subjects or other clients who attended a provincial TB service - the Division of TB Control. The information 3 MSc Thesis was collected by trained public health nurses and physicians at the time of initial diagnosis. The following selection criteria were used to include the subjects for the study: • Registered with the division of TB Control, British Columbia Center of Diseases Control (BCCDC) • Culture positive for Mycobacterium tuberculosis (Mycobacterium ID) • Conclusive laboratory sensitivity pattern to the first line anti-tuberculosis drugs The second phase involved a matched case control study design (using a sample from the total study population) in order to examine the risk factors for drug resistance as well as to compare the findings with the first phase study. Cases were all resistant TB subjects while controls were selected randomly from the pool of sensitive subjects matched on age group (0-20 years, 21-40 years, 41-60 years, 61-80 years, 81 plus years) and gender. The above selection criteria were applicable for both cases and controls. 2.2 Source of Data The information for this research project was obtained from databases provided by the Division of TB Control (DTB) at the British Columbia Centre for Disease Control (BCCDC). The Division of Tuberculosis Control (TB) serves as a referral centre for the prevention, control and treatment of TB infection and disease occurring in British Columbia (ref). The Division also undertakes programs with a goal of ensuring a progressive reduction in the annual incidence of active TB. In British Columbia (BC), the TB control activities are a centralized activity, therefore, all patients with tuberculosis in British Columbia are assessed and followed by the division of TB control. Although TB is a notifiable disease, the DTB is informed of cases from several sources, including the provincial laboratory located in the B C C D C (British Columbia Center of Diseases Control) and the Pharmacy Division of B C C D C , which is responsible for dispensing all anti-tuberculosis medications throughout British Columbia. These additional safeguards ensure that no case of active or suspect TB can be treated in BC without the DTB being informed. Presently, there are approximately 300 cases of TB reported each year (303 cases in 200232) in British Columbia and the rate for 2002 was 7.3 per 100,00032. A l l the relevant 4 MSc Thesis information of TB cases recorded at the DTB databases from 1990 to 2001 were used for this study. The current research project focuses on three types of information -sensitivity reports, socio-demographic profiles and treatment history. With respect to Drug Sensitivity data, all of those i data was produced by the B C C D C TB Laboratory Services and it is entered into the TB Control Database by the TB Control laboratory clerk who is trained to enter this data. If the client is seen in person in the clinic, then socio-demographic information is usually collected and input by the interview clerk (who functions like an 'intake' clerk). If the client is not seen in person (i.e. TB Control is consulted by private physicians or the patient is in hospital), then usually it is the local Public Health Nurse who collects the socio-demographic information for the patient in their area. A standardized data collection form is used for this purpose. The information is then sent (mailed, faxed) to TB Control and input by Medical Records Staff and Field Operations nursing personnel. In the past, active registry information was completed by a 'Registry Technician' (clerical trained in registry entry). Currently, the data input is completed by a Medical Record staff member with additional training. Any unusual circumstances that present coding difficulties, especially for coding diagnosis, are referred to the TB physician (usually the Director). In addition, since 2001 all active registry information is quality assured by an Epidemiologist who reviews the data input, data quality and produces the TB annual report (epidemiological and statistical report). 2. 3 Databases Management The data collected by the Division of TB Control (DTB) is an ongoing process. Since 1990, the data gathered by the DTB has been arranged in 3 main datasets as follows: a) One for active cases (OAR files). b) One for contacts of TB cases (OCR files). c) The active registry files. 5 MSc Thesis These 3 sets of databases were obtained for each year from 1990 to 2001. The information contained in these databases includes all the TB cases and contacts recorded up to the end of 2001. A brief description of these files and their containing data is discussed below. The OAR files contain the data on all the following patients: • Active cases (all types of TB cases, including extra-pulmonaryTB) • Patients with Mycobacterium other than TB (MOTT) • Patients who received treatment of latent tuberculosis infection (LTBI) The OCR files contain the data pertaining to all the contacts of the pulmonary TB cases diagnosed in the DTB in the corresponding year. The active registry contains information for active cases or other clients on treatment. It includes data on active cases, inactive cases, suspected TB cases, patients who received treatment of LTBI, and Mycobacterium other than tuberculosis (MOTT) during 1990 through 2001. "The active registry files (from 1990 to 2001) were transformed from their original text format into Microsoft Excel, and then to S-Plus and SPSS for data management. Since the data examined in this study were specifically related to TB cases, there is no involvement of contacts, therefore OCR files were not considered for this study. Although each of these databases contain several sub-databases; the files that contain the relevant information on TB drug resistance and its risk factors only, were reviewed for this study. A l l the files were provided by the data manager at the DTB (Ms. Fay Hutton) in the original text format as OCR files, O A R files and an active registry file. These files were given in text format because the TB clinic (where all TB cases and their contacts attended at the DTB) uses the database for clinical and administrative purposes. Hence, the first step was to convert all the files into an Excel format in order to perform the required subsequent data management procedures. In fact, Dr. Moran, a student who recently completed PhD from the Department of Health Care 6 MSc Thesis and Epidemiology, did these initial transformations of the databases. As a result, I received all the databases in excel format, which I later transformed into SPSS and S-Plus format. Several procedures were performed to organize, manage and clean the data in order to prepare them for subsequent analyses. At the first stage, the duplicate information was identified in all relevant databases and several strategies were followed to delete them. In brief, the data cleaning and management procedures were presented in table 1. To maintain confidentiality, the subjects in the databases were represented by two unique identifiers assigned to each individual who attended the Division of TB control. One is a TB number, and the other is a name code. The name codes are generated by computers using 4 letters from the first and last names, 2 digits from the birth date, and 2 digits representing the number of persons in the database which was chosen sequentially by the computer program. On the other hand, the TB number is applied by Medical Records of DTB to registry clients only (those with active TB diagnosis, or treatment for any reason). It is supposed to be unique to the client and a life-long identifier. The TB number was recognized as the most reliable unique ID (revealed after discussion with data analyst Ms. Fay Hutton), therefore, it was used to merge different sub-databases with each other. The file containing sensitivity information was considered to be the most important file. At first, this database was organized and cleaned according to the selection criteria described in the design section. Only the subjects who provided evidence of the strain of Mycobacterium tuberculosis in bacteriological culture were included; atypical mycobacteria and other strains of mycobacteria were not considered for this particular study as they are managed differently. Occasionally drug susceptibility testing is repeated on more than one occasion for a number of individuals depending on their medical requirements and also i f the clinician has concerns regarding the development of drug resistance. But the test result was taken from a single time point in this situation. If the results were consistent over time for the patient, then the report of first test was considered, but i f there were differential lab reports (either sensitive or resistant) across time, the report of resistance was used. Finally, the completeness of the sensitivity report 7 MSc Thesis was checked. Patients whose laboratory information regarding drug susceptibility was inconclusive were excluded from the study. Step 1: Cleaning and Management of Drug Sensitivity Database Major procedures: • Duplicate entries deleted • Atypical mycobacterium excluded • Inconclusive drug susceptibility reports excluded Step 2: Cleaning and Management of Socio-Demographic Database ; Major procedures: • Duplicate entries deleted • Patients on chemoprophylaxis excluded Step 3: Merging of both Drug Sensitivity and Socio-Demographic Database using TB number as the unique identifier Step 4: Cleaning and Management of Active Registry Database Major procedures: • Duplicate entries deleted • Patients on chemoprophylaxis excluded Step 5: Merging of aforementioned combined data files and Active Registry File using TB number as the unique identifier Table 1: An outline of TB data management for drug resistance study In the case of the demographic files, there was information on active TB cases who received drug treatment (chemotherapy) as well as on the subjects who were treated for latent infection (chemoprophylaxis). In this file, there was duplicate information for several TB subjects since the same individual was registered with DTB at different times. As mentioned earlier, the repeated information was identified with first priority and duplicate entries were excluded with caution. After that, this demographic file (OAR01) was merged with the drug sensitivity database (OAR21) using the TB number as a unique identifier. As a result of merging, the new combined file retained the information from both the sensitivity and socio-demographic datasets 8 MSc Thesis on the TB subjects common in both files. Since the drug sensitivity file was used as the primary database, the subjects in the demographic file, who did not have any sensitivity report or who received chemoprophylaxis (treatment of latent TB infection), were excluded in the process of merging. Later, the aforementioned combined file was again merged with the active registry database to capture the treatment history of the TB subjects. As discussed earlier, the active registry database contains every entry for active cases, or other clients on treatment. Once the merging was completed, the new file contained the information regarding sensitivity, demographic and treatment of the TB subjects, common to all three databases currently in use. In this way, the TB subjects primarily enlisted in the sensitivity file were retained in the final database, and the subjects without sensitivity information were subsequently excluded. To assist in compiling national data, the provincial TB service reports all the resistant as well as sensitive cases to Heath Canada. To check whether the new database covered all the resistant cases based on laboratory findings reported to Health Canada over last 12 years, a list containing the resistant cases was obtained from the DTB. There was a discrepancy involving a few cases, which was checked by accessing their personal health files through IPHIS (Integrated Public Health Information System), which contains the current module for case management. Subsequently adjustments or corrections were made accordingly. The sensitivity information was also available for TB subjects or other clients on treatment in the active registry file. The final database was also validated with this file, consistency was checked vigorously and corrections were made accordingly. In the case of inconsistency, the personal health file of TB subjects was considered as the gold standard. For quality control purposes, conservative measures were taken for missing observations. As there was missing information with varying magnitude, emphasis was given to key variables such as age, sex, immigration status, birth country, place of diagnosis and case category to complete the information. The personal health file of TB subjects with missing information was accessed in a random order to verify whether the information was primarily incomplete, or incorrect due to data entry error. From these searches, it was revealed that except for date of 9 MSc Thesis birth, the incompleteness of other variables was mostly primary. There were more than 700 subjects whose date of birth was missing. Therefore, their personal health files were accessed and missing information was filled out accordingly. A similar strategy was applied with the variables containing low missing percentages, subsequently the incomplete information was verified and adjustments were made although the cases of secondary (data entry) error were few. Due to resource constraints, only a randomly selected portion of cases with variables containing many missing values (e.g., history of previous TB infection, weight loss, alcoholism and malnutrition) was examined and corrected. In DTB databases, time data such as date of birth, date of diagnosis, date of immigration was indicated with exact day, month and year. For the sake of simplicity, only the corresponding year of those time variables was extracted. Then, year as a time unit was used in the study to calculate age, duration of immigration since arrival in Canada and other time related variables. Each subject in the database was assigned a cohort year according to the time of year they attended the division of TB control. A number of TB subjects were identified who had been found several times under the same cohort year in the databases. So these cohort years were validated with the year when the TB diagnosis was made in Canada. In a few cases, there was a discrepancy between cohort year and year of diagnosis. Under this condition, the year of diagnosis was considered more reliable and adjustments were made accordingly. In several TB subjects, the year of diagnosis was missing. Under this circumstance, the cohort year was compared with year of laboratory testing for drug resistance and corrections were made. In some cases, their personal health files were accessed, information was verified and adjustments were made. 2.4 Variables of interest 2.4.1 Outcome (dependent) variable - drug resistance In addition to clinical findings, the laboratory tests, in particular, the culture and sensitivity tests were also used to verify a diagnosis of TB. For all culture positive subjects, drug susceptibility testing is routinely performed on the initial isolates of M . tuberculosis10. Based on laboratory 10 MSc Thesis reports, all subjects were primarily categorized into two groups - sensitive or resistant. As the status of resistance pattern was our prime concern, it was considered as the sole outcome variable in this research project. The TB subjects with drug susceptible disease were used as the comparison or reference group. Five drugs- isoniazid (INH), rifampin (RTF), ethambutol (EMB), streptomycin (SM), and pyrazinamide (PZA) are considered as first line antibiotics for the treatment of TB. In Canada, susceptibility testing to rifampin, isoniazid, and ethambutol is performed routinely on all isolates and that to streptomycin and pyrazinamide is done on most isolates10 although there is a discrepancy across the provinces. In British Columbia, routine susceptibility testing is performed for four drugs-rifampin (RIF), isoniazid (INH), ethambutol (EM) and streptomycin (SM). The susceptibility testing for pyrazinamide (PZA) is only done under certain circumstances - i f the subjects are resistant to isoniazid (INH) or rifampin (RM). As a result, the test reports for pyrazinamide (PZA) susceptibility were available for only a few cases and consequently they were not further considered in the current analysis. However, the report on the other four drugs was available in most cases, and was used to determine the status of susceptibility (i.e., whether it should be considered as sensitive or resistant cases). Once the pattern of resistance was determined, three subgroups were created: mono- resistance, multi-drug resistance and poly-resistance. If the individual was found resistant to any of the four first line (although pyrazinamide is one of first line antibiotics, but it was excluded due to unavailability'of test reports) antibiotics mentioned before, he or she was considered as mono-resistant (MR). If the individual was found resistant to more than one drug, the person was considered either poly-resistant or multi-drug resistant (MDR). The distinction between poly resistance and multi resistance was made in the following way. The subjects resistant to at least rifampin and isoniazid are categorized as multi-drug resistant (MDR), while the subjects resistant to all other combinations are called poly-resistant (PR) for example, resistant to INH & S M , INH, E M & S M , R M & SM, etc. There are several methods that are used to determine the susceptibility to anti-tuberculosis drugs. In British Columbia Center of Disease Control (BCCDC) laboratory, the B A C T E C radiometric 11 MSc Thesis system (BACTEC 460, Becton-Dickinson diagnostic instrument systems, Towson, MD) has been used to evaluate the susceptibility to the first line antituberculous drugs using standard techniques. To determine the status of the drug susceptibility test (either sensitive or resistant), the drug concentrations used were as follows: streptomycin (SM)- 2.0 micrograms/micro liter, isoniazid (INH)- 0.1 micrograms/microliter, rifampin (RM)- 2.0 micrograms/microliter and ethambutol (EM)- 4.0 micrograms/microliter. If any of these drugs showed some resistance, the B A C T E C method was repeated for all four drugs. A final report was made conservatively based on primary and confirmatory results. This Bactec method has been used since 1990 and was consistent for the entire study period (1990 to 2001). Two other methods are used in B C C C D C laboratory for susceptibility testing of anti TB drugs-is the enzymatic method and the disk diffusion. The enzymatic method is specifically applied for pyrazinamide (PZA) testing and the disk diffusion is exercised for testing of second line drugs used for TB treatment. Since the current study did not consider pyrazinamide (PZA) and the second line drugs, the test reports obtained by these methods have been automatically excluded from the analysis. The laboratory report of drug susceptibility was originally coded as sensitive, resistant, borderline sensitive and borderline resistant cases. In addition to clearly resistant groups, all borderline categories are also considered as resistant cases. This decision was taken after discussion with Dr. Elwood, the Provincial Director of DTB. 2.4.2 Independent variables Age , Age was calculated for all subjects present in the final database. In the database, the exact age was not available; rather it was measured by subtracting the year of specimen collection from the year of birth. If year of specimen collection was missing, it is replaced by the corresponding cohort year. At first, age was treated as a continuous variable, and later, it was treated also as a 12 MSc Thesis categorical variable with the following 5 levels: 0-20 years, 21-40 years, 41-60 years, 61-80 years and 81 plus years. For the second phase of the study, these age categories were used to match the cases and the controls. Gender The information about gender was available for almost all cases except two subjects. One was transsexual, and for another, the information on gender was missing. Their personal health file was accessed, and their gender was determined reviewing the name that best describes their gender. The risk for males presenting drug resistance was estimated by comparing them with females (reference category). Immigration status Under this variable, several categories (Canadian born, foreign born Canadian citizen, landed immigrant, visitors, refugee, etc.) were enlisted in the primary database. A l l of these categories were finally collapsed into two groups- Canadian born (Canadian Citizen), foreign-born. The foreign-born group includes landed immigrant, foreign-born Canadian citizen, foreign-born people living on a work permit, minister's permit, or as a refugee, student, visitor, etc. For several subjects, the immigration status was missing or unknown. Attempts were made i f this information could be extracted from other variables. So the subjects with missing or unknown status were reassessed for their immigration information according to their birth country, and adjustments were made accordingly. Furthermore, the foreign-born group was split into two categories -recent immigrants and remote immigrants. Previous report suggests that most cases of TB occur within two years of arriving in Canada. The time point 2 years was used to differentiate the TB patients between recent and remote immigrants. The people who immigrated to Canada within two years were considered recent immigrants and the people who immigrated to Canada more than two years before were considered as remote immigrants. As a result, the immigration variable has three levels- not 13 MSc Thesis immigrants (Canadian born), recent immigrants and remote immigrants. In both the cases, the Canadian born group was considered as reference category. In addition, the remote immigrants were also used as reference to investigate the risk difference between two immigrant groups. Ethnicity In the primary database, several categories were used to determine ethnicity. Among them, there were substantial numbers in certain groups, but few in other groups. In order to keep the number of variables optimum during the analysis, these categories were collapsed into six groups as follows • White- includes Caucasians, used as reference category ( • First Nations (Aboriginal) - includes both registered and non-registered subjects, Eskimo and Metis • Chinese • Southeast Asian- includes East Indian, Punjabi, East Asian, South Asian, Southeast Asian, Cambodian and Vietnamese • Rest of Asia-includes the rest of the ethnicities within Asia • Rest of the World -includes all other ethnicities Birth country As there were exhaustive lists of countries of birth-place for subjects in the current database, it would be difficult to run the analysis for all countries. So, all the listed counties were collapsed into the following six categories-• Canada- used as reference category • Europe and North America- includes all people born in Europe as well as in North America except Canada 14 MSc Thesis • China- including Macau • South East Asia- includes India, Pakistan, Bangladesh, Nepal, Sri Lanka, Thailand, Vietnam, Cambodia, Laos, Burma and Afghanistan • Rest of Asia- includes rest of the countries belonging to Asia • Rest of the world- the remaining countries including African and South American countries ( Last country of residence As with the country of residence, a large list of countries was present in the primary database. In order to keep the number of variables optimum during the analysis, these categories were collapsed into five groups as follows-• Europe and north America- includes all the countries of Europe and North America including Canada (used as reference category) • China- includes mainland China and Taiwan • Southeast Asia- includes India, Pakistan, Bangladesh, Nepal, Sri Lanka, Thailand, Vietnam, Cambodia, Laos, Burma and Afghanistan • Rest Asia- includes rest of the countries belonging to Asia • Rest world- the remaining counties including African and South American countries Place of diagnosis Several countries were mentioned as the place of diagnosis in the primary database. Finally all the countries were collapsed into two groups- Canada and outside the Canada, which includes all other countries except Canada. Canada as a place of diagnosis was used as the reference category. 15 MSc Thesis Education level The levels of education contained in the primary questionnaire were as follows "Completed Grade 10", "Elementary", "No Education", and "Post-Secondary". At first, the categories no education and elementary education were combined into one group and the remaining two were considered as separate groups. Finally, elementary, no education and grade 10 were collapsed into one group and named as below secondary. There was one more group in the database "not asked' which later considered as missing. In both stages, analysis was carried out using post secondary as the reference group. Marital status There were several categories enlisted in the primary database. A l l the categories were finally collapsed into two following groups-• Single-includes separated, divorced, single and widowed • Family -includes common law groups and married groups Like other variables, 'not asked' was considered a missing observation. The category "family" was used as reference group during the analysis. Previous TB This variable was recorded as yes or no. "Yes" means history of previous TB, while "No" means absence of previous TB. The category "no" was used as the reference category. There were two more answer options in the original questionnaire - not asked and uncertain. These two answers were treated as missing observations as they did not convey any clear message favoring yes or no. 16 MSc Thesis Diabetes It is a dichotomous variable, and was recorded as a yes or no. "Yes" means history of Diabetes while "No" means the absence of such a diagnosis. The category "No" was used as reference category. In addition, there were two more categories (uncertain and not asked) which latter were treated as missing observations. Alcoholism This variable is recorded as yes or no. "Yes" means a history or current intake of alcohol to excessive amount (subjective judgment), while "No" means the absence of such an intake of alcohol. The category "no" was used as a reference category. There were two more options in the primary questionnaire - not asked and uncertain. These two answers were treated as missing as they did not convey any clear message favoring yes or no. Malnutrition This variable was recorded as yes or no. "Yes" means presence of malnutrition, while "No" means the absence of such a symptom. The category "no" was used as a reference category. There were two more answering options in the original questionnaire - not asked and uncertain. These two answers were treated as missing observations because they did not provide any evidence in favor of yes or no. Weight loss The categories listed in the primary database were yes, no, not asked and uncertain. "Yes" means presence of weight loss, while "No" means absence of such symptom. The category "no" is used as a reference category. As usual, the categories "not asked" and "uncertain" were treated as missing observations due to lack of evidence in favoring yes or no. 17 MSc Thesis Case type It has been recommended by Health Canada that all the subjects who are diagnosed with TB in Canada need to be recorded under the following two headings according to their criteria as follows-33 • New case: no documented evidence or history of previously active TB • Reactivated (relapsed) case: documented evidence or history of previously active tuberculosis which became inactive In fact, several additional categories were listed in the primary database which were eventually collapsed into two major groups to be consistent with Health Canada requirements-• New- includes new active/notified ex province and "other X-ray abnormality" in addition to new category (used as reference category) • Reactivated- includes reactivated/previous ex-province, diagnosed outside Canada, and presumed TB inactive in addition to reactivated category Site of involvement Lists of several sites or organs were shown up in the primary database under the site of involvement. This information was presented in the database into four separate columns- one for pulmonary, the second and third for non-pulmonary and the fourth for other pulmonary such as pleurisy, tracheal and bronchial TB. A l l this information was grouped into two broad categories as follows: • Pulmonary- includes "other pulmonary" in addition to pulmonary (used as reference category) • Non-pulmonary includes genitourinary, abdominal, meningeal, bones/joints, etc 18 MSc Thesis Symptom Summary This information was presented in the database as a dichotomous variable- yes and no. Yes means presence of symptom summary and no means absence of symptom summary. Yes was used as the reference category. As usual, the unknown information was considered missing. H I V status The Division of STD/AIDS Control of B C at the B C C D C records information regarding HIV status (confirmed by the provincial laboratory) of all the subjects tested in the province, as well as the date of the first positive HIV test from 1988 through 2002. However, the information contained in the current database contains reliable personal identification only since 1995. Since the information of the databases ranges from 1990-2001, this variable is not considered in the final analysis. 2.5 Statistical Analysis An initial descriptive analysis of all the variables of interest including frequencies of missing observations was performed. Two different types of regression modeling were carried out to assess the risk of developing drug resistance to TB over a period of 12 years. Since the outcome variable is dichotomous, logistic regression models were employed in initial stage. A univariate analysis was carried out in order to identify the potential risk factors accounting for drug resistance. This analysis was performed for each individual socio-demographic and medical risk factor. The strength and magnitude of the association was estimated for each variable from the corresponding univariate model and was expressed in terms of an odds ratio. Subsequently, multiple regression analyses were carried out with the same set of variables that were used in the univariate models. Maximum efforts were made to include the potential risk factors for drug resistance in the model with a multivariate setting. No multiple regressions (multivariate model) 19 MSc Thesis were performed including all variables of interest because when all the variables were included simultaneously in one model, an unexpected change of parameters of the potential variables was observed. The reasons, anticipated, for this dramatic change might include-• Too many variables • The frequency of missing observations associated with each variable • The issue of collinearity Consequently the multivariate analysis was performed with a meaningful combination of potential variables. It is believed age and sex are potentially universal confounders. Moreover, the information about age and gender were complete, so these two variables were included in all multivariate models. The conventional significance value of p (less than 0.05) was used to assess individual variables. Furthermore, the 95 percent confidence interval was calculated for the odds ratios to assess the significance of the effect. Initially there were two levels of the resistance status. Later on, this variable was regrouped into four levels in order to examine the difference across subgroups with drug resistance. When outcome variables have more than two categories, it is not appropriate to use ordinary logistic regression. Therefore, polytomous (multi-nominal) logistic regression analysis was carried out. The odds ratio was estimated to show the effect of independent variables on the outcome of drug resistance status. Similar procedures were used as discussed earlier for model building and the selection of variables in the multi-nominal regression. In the second phase of the study, a matched case control design was employed. In a matched case control study, the traditional logistic regression did not apply due to violation of independence condition. So, conditional logistic regression was performed. Odds ratios with corresponding confidence intervals were calculated to evaluate the significance of independent variables. Both univariate and multivariate (multiple regression) analyses for the models that included all the variables of interest was performed with SPSS and S-Plus statistical software. The SPSS directly produces the odds ratio for each coefficient with 95 percent confidence interval, but S-20 MSc Thesis Plus has limitations in this regard. So the odds ratio was calculated for each coefficient separately in S-plus, and later, this output was compared with SPSS output. There were occasionally some differences in the second or third decimals in the coefficients, but the results were practically the same for the odds ratios. The tables presented with the results of the logistic model were primarily from SPSS, with the exception of the descriptive statistics of several variables, which were derived using S-Plus 6.1. For polytomous (multi-nominal) logistic regression, the SPPS 11.0 software was used to perform the desired analysis. In the case of conditional logistic regression, the^software SAS was used for convenience in comparison to SPSS and S-plus. Chapter 3: Results The results are presented in 6 sections: • Descriptive statistics • Assessment of the risk factors in the univariate model • Assessment of risk factors in the multiple regression model • Assessment of selected risk factors in polytomous (multi-nominal) logistic model • Assessment of selected risk factors in the case-control setting • Sensitivity analysis In fact, the results were presented in 17 tables which were shown up in the appendix. 3.1 Descriptive Statistics The outcome variable was presentation of TB subjects with drug resistance. According to the selection criteria set for this study, a total of 3041 subjects were found eligible for the twelve years study period (1990-2001). Among those, there were 295 (10%) resistant cases and 2746 (90%>) sensitive cases. Table-2 describes the distribution of resistant and sensitive cases across 21 MSc Thesis the study years. The highest number of culture positive TB cases (324) was observed in 1997, while the maximum number of sensitive (290) and resistant cases (36) was found in 1997 and 2001 respectively. 1 0 0 ~\ P e r c e n t g a e of S e n s i t i v e C a s e s 0 i 1990 1992 1994 1996 1998 2000 2002 Y e a r Figure 1: Trend of Sensitive and Resistant Cases from 1990-2001 In contrast, the lowest number of subjects (195) was observed in 1990, while the lowest number of sensitive (176) and resistant (13) cases was found in 1990 and 1994 respectively. A graphical presentation (Figure-1) shows the trends of sensitive and resistant cases over the 12 years. Although there were variations in the percentages of sensitive and resistant cases across the study years, no significant pattern was observed at a l l However, the percentage of resistant cases was within 9%-l 1% for most of the years except for the year 1994, when only 5% of the cases were found to be resistant and the remaining (95%) cases were sensitive. 22 MSc Thesis 2D H 1 S 1 0 5 P e r c e n t a g e M o n o R e s i s t a n c e P o l y R e s i s t a n c e N / _ SNNV^s,;> ^ ^ / M u l t i - d r V g _ 1 9 9 0 1 9 9 2 1 9 9 4 1 9 9 6 1 9 9 8 2 0 00 2 0 02 Figure 2: Trend of different categories of Resistant Cases from 1990-2001 As was mentioned earlier the resistant cases were further reclassified into three categories: mono-resistance, poly-resistance and multi-drug resistance cases. The distribution of these resistant types is presented in Table 2. The highest incidence of mono resistance (total-23) and multi-drug resistance (total-7) was in 2001, while the highest incidence of poly-resistance (total-11) was in 1998. On the other hand, the lowest incidence (total-7) of mono-resistance was in 1996, while the lowest incidence of poly-resistance (total-1) was in 1994. For multi-drug resistance cases, there was no case in 1991 and 1999. When expressed as a rate, the overall proportion of mono-resistance was 6.7%; the proportion of poly resistance was 2%, while the proportion of multi-drug resistance was less than 1 percent. The highest percent of mono-resistance (8.54%) and poly-resistance (4.33%) cases were found in 2000 and 1998 respectively. The lowest percent of mono-resistance occurred (2.95%) in 1996, while the lowest percent of poly-resistance (0.38%) was in 1994. In terms of multi-drug resistance, the highest percent (2.2%) was observed in 2001; while zero (0) percent was found in 1991 and 1998. A graphical presentation (Figure-2) is made to show the trend of drug resistant case types across the study y ears and no significant pattern was found. There were variations both ups and down, but it did not show any particular trend based upon the type of drug resistance. 23 MSc Thesis When the descriptive analysis was carried out for each individual variable, it was found that the youngest TB subject was a one-year old, while the oldest TB subject was 102 years old. The mean age of the study population was 48 years with a standard deviation of 21 years, while the median age was 51 years. The distribution of age is shown in Figure 3 which was slightly bimodal. In addition, age was grouped into five categories. The highest proportion (32%) of TB cases was found in age group 21-40 years, while the lowest proportion (9%) was observed in older age group (81 years plus). In terms of gender, there were more males than females. The ratio of males and females was 57: 43 which is shown in a pie chart (Figure 4). The information about age and gender was complete for all subjects. Table-4 describes the frequency of all variables of interest with corresponding missing values. 0.000 0.015 H 0.012 H 0.009 H 0.006 H 0.003 H 0 20 40 60 80 100 120 A g e Figure 3: Age distribution of the study population 24 MSc Thesis F E M A L E M A L E Figure 4: Distribution of the Study Population by Gender The rate of TB drug resistance across the age groups is shown in Figure 5. The rate was low in two extreme age groups (0-20 and 81 plus) than the overall resistance rate (10%) although it was lowest in the senior age group. The age group 21-40 had the highest incidence of drug resistance and then it started to fall in the next age groups and this continued until the oldest age group. The rate of drug resistance was very similar for both males and females which were consistent with the overall TB resistance rate (10%). The incidence of TB resistance was 9 (per 100) in the case of males and 10 (per 100) in the case of females. 25 MSc Thesis 4 H AGE 0-20 AGE 21-40 AGE 41-60 AGE 61-80 AGE 81 PLUS AGE Figure 5: Rate of Drug Resistance across age categories As mentioned earlier, the total subjects were classified into -Canadian-born (32%) and foreign-born (68%>) based on immigration status. It indicated that most of the TB subjects were immigrants. When foreign-born people were categorized into recent immigrants (< 2 years) and remote immigrants, most of the TB subjects were remote immigrants (48%) who had been in Canada for more than two years. The proportion of recent immigrants was also 20 percent and the remaining 32 percent were Canadian-born (not immigrants). About 6 percent of all subjects failed to provide any information regarding their immigration status. With respect to ethnicity, the majority of TB subjects were of Chinese origin (26%) followed by Caucasians (White), and the Southeast Asian, both with a percentage of about 20%. The total number o f First Nations (Aboriginals) subjects was 428 (14% o f total study population). The number of people in the remaining two groups was 310 (Rest Asian-10%) and 306 (Rest of the world-10%). Only 25 (1%) people did not report their ethnic status. In terms of country of birth, more than 99 percent of the subjects provided information regarding country of origin. About thirty percent of the subjects repeated Canada as their birth country. 26 MSc Thesis Nevertheless, this proportion was the highest compared to other groups. The next biggest group (20%) of people was from the region of Southeast Asia followed by the rest part of Asia (19%) except China. The lowest number of people (6%>) was from Europe and North America. When asked about the country of last residence, 44%> people did not disclose this information. Thirty eight percent of the rest of the people, who disclosed information on the country of residence, were from the region of Rest of Asia. The percentages of other groups were China (with Taiwan) 15%, South East Asia 25%, Rest Asia 4%, etc. In the case of education status, the percentage of missing observations was high (63%) among all socio-demographic data. Among the rest, 35 percent reported that they had post secondary education, 35 percent had grade-10 education, and the remaining 30 percent had only elementary education or no education at all. When the last two groups were merged into one group, it was observed that about 65 % people had an education level below secondary education. In respect to marital status, the overall response rate was 67%>. Among the responders, 46 percent were in the single category, while the remaining 54 percent were in the family category. In the case of place of diagnosis, information was available for most of the subjects (98%). Among the respondents, 94%> of the TB diagnosis occurred in Canada, while in the remaining 6% the diagnoses were made outside Canada. In the case of medical variables (table 5), there was a high proportion of missing information (about 50 percent or more). However, several variables such as case type, symptom summary, site of involvement, etc were associated with few missing observations. In fact, the physician or other trained individuals completed the information about case type, site of involvement (pulmonary/non-pulmonary involvement) and symptom summary. As a result, the percentage of missing observation for case type, pulmonary involvement and symptom summary was 1%, 1% and 5% respectively. In the case of other medical variables, the highest percentage was associated with malnutrition (65%) followed by weight loss. In the case of diabetes and alcoholism, 60% observations were missing. The missing observations were 47% for previous TB status, and 50%> for smoking status. Among the available information, 18 percent answered 27 MSc Thesis that they had history of previous TB. In case of diabetes, 18% had a history of diabetes. Among the respondents to the question regarding alcohol intake, 17%> had a history of alcoholism. The proportions of "yes" were same as alcoholism for weight loss question. In the case of smoking status, 23%) had a history of smoking. 3.2 Assessments of the risk factors in univariate model The results from the univariate model have been presented in two tables- one for socio-demographic variables (Table-6) and another for medical variables (Table-7). The effect size for each individual variable was measured in terms of odds ratio (exponential of coefficient), which is presented in the table with 95% confidence intervals, and the correspondingp value. The age variable was analyzed initially as a continuous variable. A significant protective effect was found for age. The odds ratio of 0.98 implies a 2% reduction in the risk of drug resistance for each year of increase in age. Consequently, age was analyzed as a categorical variable using youngest age (0-20) as a reference, but it did not show a similar effect. Only the oldest age category (81 plus) was statistically significant, while other groups had non-significant p values. Although being male had a higher chance (1.08) of presenting with drug resistance, but the effect was not statistically significant. Subjects born outside Canada (foreign-born) had a 3 times higher chance of presenting with drug resistance than Canadian-born TB subjects. The 95%> confidence interval (2.26, 4.49) was found to be statistically significant. The same variable was analyzed using three levels. The subjects who immigrated recently (within 2 years) and who immigrated more than 2 years before had a significantly higher chance for drug resistance than the Canadian-born subjects (4 times and 3 times respectively). In addition, recent immigrants had also a greater chance of drug resistance than remote immigrants, but the odds ratio was not statistically significant. The variable ethnicity was analyzed using Caucasians as a reference category. Except for Aboriginals (First Nations), all the other categories were found to be significant for the risk of 28 MSc Thesis TB drug resistance. Chinese had twice as high, Southeast Asian had three times, and the Rest of Asian had a four-fold higher chance of presenting with drug resistance than Caucasians. Although the First Nations had a lower risk for drug resistance, the odds ratio was not significant. The variable country of birth was analyzed using Canada as a reference category. The people who were born in China (3 times), Southeast Asia (3 times) and the Rest of Asia (4 times) had an elevated risk of presenting with drug resistance compared to people born in Canada. These results were found to be statistically significant. People born in North America had no statistically significant increased risk of resistance. The heterogeneous group Rest of the world also had twice the risk of presenting with drug resistance compared to TB subjects born in Canada. In spite of the high proportion of missing observations (44%), the variable country of residence was analyzed using people resided in North America (including Canada) and Europe as the reference category. A l l the categories had an elevated risk of presenting with drug resistance compared to the reference group. The people who resided in China and Taiwan, in Southeast Asia, in Rest of Asia and in Rest of the world- were twice as likely to present drug resistance compared to the reference subjects. Education level was analyzed using "post secondary" as the reference group. The TB subject with no education or elementary education had a statistically significant protective effect (odds ratio-0.58, 95% CI-0.36, 0.94) on the risk of drug resistance, while grade-10 education did not show such significant effect. When the no-education and grade-10 education were collapsed together and analyzed in the same way, the odds ratio (odds ratio-0.70, 95%> CI-0.48, 1.02) was not found to be significant. Marital status was analyzed using family group as a reference. Although the single group (odds ratio- 0.75) had a reduced risk for drug resistance, the result was not statistically significant. In the case of medical variables, factors that showed a significant impact on drug resistance were alcoholism, weight loss and case type. The subjects who were alcoholics were less likely (odds 29 MSc Thesis ratio-0.34) to present with drug resistance compared to patients without such a history. However, the reactivated (relapsed) cases had an elevated risk (odds ratio-1.86) for drug resistance compared to the subjects diagnosed as new cases. In addition, the subjects who had history of weight loss also had a reduced risk (odds ratio-0.36) for drug resistance than those who had no history of weight loss. The other medical variables - history of previous TB, diabetes, malnutrition, smoking, pulmonary vs. non-pulmonary involvement, and symptom summary were not significant in the univariate model. Subjects with a history of previous TB had a higher risk compared to the patients without any history of TB; subjects who did not present with symptoms had a higher chance than those who presented with symptoms; subjects with a history of diabetes are less likely to present with drug resistance compared to the subjects without diabetes; subjects who had malnutrition had a lower risk of drug resistance than those who did not have. However, these results were not statistically significant. 3.3 Assessments of the risk factors in multiple regression models As with the univariate analysis, the multiple regression analysis (multivariate model) using the same set of variables and also similar reference categories was carried out in two steps. Initially, age and gender were used in addition to each individual variable in the model to adjust for their effect. Table 8 and 9 shows the results of the multiple regression models for step one. Table 8 describes the results for socio-demographic variables and Table 9 describes the effect size for medical variables after adjustment for age and gender. In the second step, ethnicity was used as an additional adjustment factor with age and gender. These multiple regression models were run only for specific variables, in particular, the significant variables in either univariate or multivariate analysis. However, a few non-significant variables (e.g. history of previous TB) were also considered as their effect was found significant in previous studies. The results of the effect size for the variables after adjustment of age, sex and ethnicity are presented in Table 13. Like the uni-variate analysis, the magnitude of the association (effect size) for each individual 30 MSc Thesis variable was measured in terms of the odds ratio (exponential of coefficient), which is presented in the table with 95% confidence intervals and a corresponding p value. The proportion of missing observations was similar as in the univariate model because the adjusted variables- age and gender had complete information. The addition of ethnicity in step 2 did not raise the number of missing values significantly as only 1% (25) information was missing about ethnicity in the final database. The inclusion of ethnicity as another adjusting factor in this multiple regression added up to at most 25 extra values to their corresponding missing percentage. The variable immigration status was significant in the multivariate model. There was a marginal increase in the odds ratio for foreign-born subjects compared to the univariate model (from 3.18 to 3.46). In the case of recent immigrants, the odds ratio was similar to the univariate model; but in t he c ase o f r emote i mmigrants ( who w ere C anada i n m ore t han t wo years), t he o dds r atio increased from 3.06 to 3.48 with a wider confidence interval. Although both recent and remote immigrants were more likely to present with drug resistance compared to subjects born in Canada, there was no significant difference in risk between these two groups (recent immigrants and remote immigrants). After adjustment by age and gender, all categories of ethnicity except First Nations (Aboriginals) were found to be significant. People of Chinese ethnicity were twice, Southeast Asians were three times, Rest of the Asian were four times and the Rest of the World were twice as likely to present with drug resistance compared to Caucasians after adjusting for the effect of age and gender. These results were consistent with the univariate analysis, in spite of a discrepancy in value of odds ratio and confidence interval. In the multiple regression models, all the categories of birth country variable, except the category Europe and North America (excluding Canada) were found to be significant, as in the univariate model, although there were discrepancies in the odds ratios and confidence intervals. People born in North America (except Canada) and Europe were equally likely to develop drug resistance compared to people born in Canada. But people born in China, people born in 31 MSc Thesis Southeast Asia and people born in Rest of Asia were four times more likely to present TB drug resistance while people born in the Rest of the World had twice the chance of drug resistance compared to people born in Canada, controlling for age and gender. In multivariate models, there was a discrepancy in the significance of the categories of the variable "country of residence" from the univariate model. The categories- Europe & North America, China with Taiwan, Southeast Asia and Rest of Asia were found to be significant in both univariate and multiple regression models. But the category Rest of the World was found to be significant in the univariate model and non-significant in the multivariate model. However, people in theses significant categories were twice as likely to present with drug resistance as the reference group. When the variable education level (with three categories) was analyzed in the multivariate model, none of the two categories was found to be significant, although the education level "no or e lementary e ducation" w as found s ignificant i n t he m ultiple r egression analysis. W hen t he same variable was analyzed using two levels, it remained non-significant as it was in the univariate model. Although the variable marital status was found non-significant in the univariate model, it was significant in the multivariate model. Single people were less likely to present with drug resistance compared to those who lived with a family adjusting for the effect of age and gender. The variable "place of diagnosis" was found significant in both univariate and multivariate models. There was an increase in odds ratio value in the multivariate model (from 2.86 to 3.48). The subjects diagnosed outside of Canada were almost four times more likely to present drug resistance compared to the subjects diagnosed in Canada. In the case of medical factors, the variables that were significant in the univariate analysis were also found to be significant in the multiple regression models. These variables were alcoholism, weight loss and case type (new cases vs. reactivated cases). The other variables - previous TB, diabetes, smoking, malnutrition, pulmonary TB, and symptom summary were not significant. 32 MSc Thesis These findings were consistent with the univariate analysis. In the case of alcoholism and weight loss, the risk of drug resistance was reduced (70% reduction with alcoholism and 60% reduction with weight loss) compared to the reference group. But in the case of case type; the risk was elevated-1 he reactivated (relapsed) c ases w ere t wice a s 1 ikely top resent w ith d rug r esistance compared to the new cases after adjusting for the effects of age and gender. In the second step of the multiple regression (age, gender and ethnicity adjusted), the following variables were analyzed- education level, marital status, place of diagnosis, previous TB, alcoholism, weight loss, smoking, case type and pulmonary involvement. The risk factors- case type (new vs. reactivated) and place of diagnosis (Canada vs. outside of Canada) were also significant in this multivariate analysis. Although alcoholism and weight loss were significant in both univariate and multivariate model (age and sex adjusted), after adjustment for ethnicity, they were no longer found to be statistically significant. Marital status as a risk factor was not significant in this ethnicity adjusted model in spite of its significance in previous multiple regression. Education level, smoking and pulmonary vs. non-pulmonary- all these variables were found non-significant as they were in previous multivariate model. H owever, t he risk factor, h istory o f p revious T B w as found i nsignificant i n p revious univariate and multivariate model (age and gender adjusted), but it became, significant after adjusting for the effect of age, gender and ethnicity. The subjects with a history of previous TB were twice as likely to present with drug resistance compared to subjects without a history of previous TB. 3.4 Assessments of selected risk factors in polytomous (multi-nominal) logistic regression A multi-nominal logistic regression analysis was performed for selected variables (significant in both univariate and multivariate model, and with few numbers of missing observations) to see i f there were differences in risk factor characteristics between mono-drug, poly and multi-drug resistance. The variables analyzed under this regression using the same reference category were-age, gender, immigration status, recent immigration, ethnicity, place of diagnosis and birth 33 MSc Thesis country. Age (as a continuous variable) and gender were used as the adjustment factors for all these variables. Therefore, the effect size for age and gender was not reported although it was consistent with the results of univariate and multivariate models. The results obtained for each individual r isk f actor from t he c orresponding m odel a re p resented i n T able-17. Like p revious analyses, the magnitude of the association (effect size) for each individual variable was calculated in terms of the odds ratio (exponential of coefficient) and was presented in the table with 95% confidence intervals and corresponding p values across different categories of TB drug resistance. The risk factor, immigration status was found to be significant in all three categories of drug resistance. The risk for presenting with mono-drug resistance was three-fold higher for foreign-born TB subjects compared to the subjects born in Canada. Although the risk was very high for the subjects with poly resistance (17 times), it had a very wide confidence interval (4.13, 69.83). In the case of TB subjects with multi-drug resistance (MDR), the risk was also high (5 times) with wider confidence interval (1.48, 16.62). In a sample of 64 poly-resistant cases and 27 multi-drug resistant cases (out of 3041 total TB subjects),, this type of wide confidence interval is to be expected. The variable recent immigration was found to be significant in all three categories of drug resistance. The risk of presenting with drug resistance for recently immigrated people (within 2 years) was twice as high in mono-resistant subjects, 16 times higher in poly-resistant subjects and 7 times higher in multi-drug resistant subjects compared to TB subjects born in Canada (no immigration). The remote immigrants (immigrated before two years) were two times and 17 times more likely to present with drug resistance compared to subjects born in Canada in mono-resistant and poly-resistant category respectively. The confidence intervals were relatively small for mono-resistance, wider for multi-drug resistance, and were very wide in poly-resistance cases. When the multinomial regression analysis was performed for ethnicity after adjusting for the effect of age and gender, several categories of ethnicity were significant while others were not found to be significant. Subjects belonging to the ethnic groups: Aboriginals (First Nations) and 34 MSc Thesis Rest of the World were found insignificant across all categories. The ethnic groups- Chinese, Southeast Asian and Rest of the Asian were found significant in both mono-resistant and poly-resistant cases. None of the ethnic categories were statistically significant in multi-drug resistant cases. In the case of significant ethnic groups, the risk was higher (two to three times) with tighter confidence intervals in mono-resistant subjects and very high (six to sixteen times) with much wider confidence intervals in poly-resistant subjects compared to Caucasians. In the case of country of birth, the results of polytomous regression analysis were heterogeneous in terms of significance for individual category. Categories belonging to subjects born in China and the subjects born in the Rest part of Asia were significant, while the category belonging to subjects born in North America (except Canada) & Europe was insignificant across all categories of drug resistance. Subjects born in Southeast Asia was significant in mono-resistant and poly-resistant subjects for presenting with drug resistance, but was non-significant in multi-drug resistant cases. In the case of all significant categories, the risk was high in the mono-resistant group, was moderately high in multi-drug resistant, and very high in poly-resistant group compared to reference group (subjects born in Canada). The confidence intervals for each individual category were relatively tighter in mono group, wider in multi-drug group, and much wider in poly group. The risk factors- "place of diagnosis" and "case type" were found significant across all the categories of drug resistance when polytomous logistic regression was performed for these variables. The subjects, who were diagnosed outside Canada, were more likely to be drug resistant than those who were diagnosed in Canada. The risk was twice as high of mono-resistant cases, three times higher in poly-resistant cases, and 20 times higher in multi-drug resistant cases. For case types, the reactivated (relapsed) cases were more likely to have drug resistance than the new cases. The risk was twice as high as in both mono and poly resistant, and 8 times as in multi-drug resistant group. In the case of these two risk factors, the confidence interval was relatively tighter in both mono and poly r esistant group, but was much wider in multi-resistant group, which was not consistent with the findings of previous factors. Nevertheless, the confidence intervals were found to be much wider in the poly-resistance group in all the variables except these two factors (place of diagnosis, case type). 35 MSc Thesis 3.5 Assessments of selected risk factors in case-control setting As mentioned before, a restrictive analysis was performed using a sample of the total study population in a case-control setting. The number of cases included all the resistant cases (295 resistant cases). Controls were chosen randomly from the sensitive group. Since three controls were chosen per case, the total number of controls was 885. The new sample size was 1180 out of 3041. The controls were matched by age group and sex. There were five age categories (0-20, 21-40, 41-60, 61-80, 81 plus) for age, and two categories for gender, which ultimately led to a total of ten strata. In fact, controls were matched according to these ten strata. The distribution of cases and controls are shown up in the table 12. The majority number of subjects (71 cases and 213 controls-total 284) was in strata three (age 21-40, female), and the least number of subjects (2 cases and 6 controls-total 8) was in strata nine (age 81 plus, female). Since it was a matched case control study, conventional (unconditional) logistic regression could not be employed in this analysis. Instead, a conditional logistic regression was performed for all potential individual variables under this circumstance. It was a type of univariate analysis; no additional adjusting variables were used since it already accounted for the effect of stratification variable -age and sex. The analysis was carried at first for socio-demographic variables, then for the medical variables. Two tables were constructed to show the effect of all potential variables-Table-13 for socio-demographic factors and Table-14 for medical factors. In fact, the effect size for each of the variable is shown in the Tables in terms of odds ratio with corresponding confidence intervals and p values. In brief, the results of this special regression analysis were very much consistent with the age and gender adjusted multiple regression analysis. The variables that were found to be significant in the multiple regression models were also found to be significant in this conditional regression. Although there were discrepancies in the odds ratio, the p values and confidence intervals closely matched the findings of multivariate (multiple) regression analysis. The findings of both regression models (multiple logistic regression and conditional logistic regression) are presented side by side in Table-13 and 14 to compare the results. The consistency of the results was expected as in both instances we were dealing with a large sample size. / 36 MSc Thesis 3.6 Sensitivity analysis A sensitivity analysis was carried for dichotomous variables, which had at least 30 percent missing observations. The factors analyzed were- education level, marital status, history of previous TB, diabetes, alcoholism, malnutrition, w eight loss and smoking. Age and sex were used as adjustment factors for each individual variable. Analysis was performed in two stages. At first, all missing data was considered as reference groups, and then age and sex adjusted logistic regression was performed. In the second stage, all missing observations were considered as outcomes of interest, and similar analysis was carried out accordingly. The results of the sensitivity analysis are shown up in Table-16. The effect size for each variable was measured in terms of odds ratio with confidence intervals and p values, and are presented in the table. From the results, only three factors were found to be significant in one direction in the sensitivity analysis. These were marital status, alcoholism and weight loss, where the risks were reduced compared to the reference group. This implies that missing observations have minimum influence in the significance and direction of odds ratio of these variables. When these three variables are analyzed in age, sex and ethnicity adjusted models, they were not found to be significant to exert an influence in the development of drug resistance. Chapter 4: Discussion This study identified several independent risk factors for drug resistance among TB patients in British Columbia. Variables that were identified as significant risk factors in the multiple regression analysis were immigration status, ethnicity, birth country, country of residence, a history of previous TB, place of diagnosis and case category. Age and gender were used as adjusting variables for all these risk factors. With respect to immigration, foreign born people are four times more likely to present with drug resistance than Canadian born subjects. Previous reports suggest that immigrants developing TB after moving to Canada are more likely to harbor drug resistant baci l l i 1 3 ' 3 4 ' 3 5 . Therefore, the 37 MSc Thesis foreign born people were categorized into recent immigrants (short-term residents: within 2 years) and remote immigrants (long-term residents of more than two years) to examine the difference between these two groups. Although the short-term residents had a greater chance for drug resistance than long-term residents when analyzed in uni-variate and multivariate models; the results were not statistically significant. However, both levels were significant when compared with a reference group of non-immigrants (i.e., local residents). One factor affecting the significance of this variable may be the cutoff time (2 years) used to differentiate between recent and remote immigrants. This cut point may be too low to separate the higher risk for recent immigrants compared to remote immigrants. A similar analysis was repeated using a 10 year cut point. Recent immigrates were found to have significantly higher risk in the univariate model, however, the effect was not significant in the multivariate model (multiple regression). The study conducted by Long et. a l . 3 4 in Alberta suggested that immigrants presenting with TB within 5 years of immigration should automatically receive an initial four drug regimen, as their risk of having resistant diseases is five-fold higher than the patients who migrated more than 5 years before diagnosis34. In light of the findings of Long et al, the analysis was re-run again using a 5 year cut point, however the results did not change from the original 2-year cut point analysis in this study. Another study in Alberta suggested that the increased risk for drug resistance was apparent in those who have lived in Canada for up to 15 years13. Given the belief that TB occurring in immigrants to Canada represents reactivation of disease acquired in their country of birth, one would expect that resistance rate for TB subjects years after immigration should reflect the 1 ^  situation seen in their birth country prior to their immigration to Canada . In the case of ethnicity, people belonging to several ethnic groups had an increased risk for drug resistance when compared to Caucasians. Although it was presumed that the Aboriginals (First Nations) might have a higher risk for drug resistance (the rate of tuberculosis among aboriginals is also much higher than the provincial rate36), their risk was found to be similar to non Aboriginal C anadians. T his m ight b e d ue t o t he f act t hat T B c ontrol p rograms a nd t reatment regimes have improved over time in this population (the aboriginals either take anti-tuberculosis 38 MSc Thesis medication appropriately or not at all ). However, among other ethnic groups, the Chinese, Southeast Asian and the Rest of the Asian were two to three times more likely to present drug resistance than Caucasians. Furthermore, the most heterogeneous group, the Rest of the World including South Americans and Africans also had an elevated risk compared to the white population. Country of birth had a significant impact on the risk of drug resistance. It was found that people who were born in North America and Europe did not pose any elevated risk than Canadian-born people. But the people, who were born in other geographical areas, specifically, in China, Southeast Asia, and other parts of Asia, were four times more likely to present with drug resistance than Canadian-born people. This finding is to be expected as most of the Asian countries have a high incidence of TB and also a high incidence of drug resistance. When data related to previous country of residence was analyzed, it was found that people, who resided in China (including Taiwan), Southeast Asia, and other parts of Asia, were twice as likely to present with drug resistance as the people who resided in Europe and North America, after controlling for the effect of age and gender. Although country of residence was significant in the analysis, this information was based on data with many missing values. Among the demographic variables, country of residence was associated with a high percentage (about 44%) of missing observations. Further questions concerning the validity of this finding also arose when the results for this variable were compared with other variables that reflected similar information (e.g., birth country). For example, 913 patients who reported that they were born in Canada, only 145 people reported Canada as their country of residence. Similar inconsistencies were found for other categories of the country of residence variable. Additionally, country of residence does not incorporate the duration of residence in that particular country which might influence the effect. The results for country of residence should therefore be interpreted with caution. The results of this study suggest that after controlling for the effects of age, gender and ethnicity, people who were diagnosed with TB outside Canada were three times more likely to present with drug resistance compared to patients who were diagnosed inside Canada. The same level of risk was also observed for reactivated (relapsed) TB cases compared to the new TB cases. The risk of 39 MSc Thesis drug resistance in patients with a previous history of TB was twice as high as in those known to have no history of previous TB, after adjusting for the effect of age, sex and ethnicity. The management of these patients (previous TB, reactivated cases) may be more difficult (i.e., they have a greater chance of developing drug resistance) and should concentrate on the use of rapid diagnostic tests for the identification of the mycobacterium species as well as sensitivity testing for first line anti-tuberculosis drugs. The administration o f appropriate and strictly supervised treatment is most likely to succeed in this situation when treatment protocols are tailored to individual needs. The findings on risk factors for resistance are helpful and have direct implications for TB control programs. The findings should help clinicians as well as policy makers to identify high risk groups and prioritize the development of more effective strategies for targeting these groups. The study further highlights the emphasis that should be given to the monitoring of foreign-born (except Europe and North America) people who bear a substantial health burden because of drug resistance. Another important finding of this study was the significance of several factors that have not been found to be significant in previous studies. Marital status, alcoholism and weight loss were all found to exert a significant contribution to the development of drug resistance in uni-variate model and multivariate model (age and sex adjusted). The effects of alcoholism, weight loss and marital status (being single) were protective. These results appear counterintuitive, and may be ethnicity driven. If people belonging to a Caucasian category have higher levels of alcoholism and are more likely to be single, then these risk factors may be identified as being protective against drug-resistance. Support for this proposition was found in that when ethnicity was used as an additional adjusting factor, these variables were no longer found significant in the model. Differences in the social desirability bias may also have affected the results. Asian respondents may be less willing than Caucasian respondents to report alcoholism. Further research on these particular findings is definitely required before examining potential implications of the results. Throughout the study, odds ratios were used as the measure of association. Since the study design was retrospective in nature, one could not calculate directly risk ratio from this type of 40 MSc Thesis study. However, one can obtain a very good estimate of the relative risk from a retrospective study using odds ratio under certain circumstances37. It is common practice to approximate the relative risk by odds ratio when the risk of outcome or prevalence of disease is low 3 7 . In this study project, the overall prevalence of drug resistance was less than 10 percent. Therefore, it seems reasonable to use the risk ratio as an alternative of odds ratio in this study. Previous studies suggest that males generally have a higher level of drug resistance than females5'14. This relationship was not found to be significant in any of the analyses in the current study. The relationship .between age and risk of drug resistance is not as clear-cut13. Data from some studies have suggested no direct correlation between age and drug resistance38'39. However, other studies have demonstrated a significantly higher level of resistance in younger patients14' 1 3 ' 4 0 . In this study, a protective effect was observed when age was analyzed as a continuous variable. Consequently, when age was analyzed as a categorical variable (using 0 to 20 years as a reference group), the risk was higher in 21 to 40 years age group, and lower in other age groups, although the effects were not significant. Only being in the oldest age group was significantly associated with reduced risk when compared to being in the youngest age group. This was a population-based study that captured all the eligible TB patients for the last 12 years (from 1990 to 2001) who were registered with the Division of TB control (DTB) of British Columbia. It is unlikely that DTB missed any potential subjects who developed TB while living in British Colombia because TB control activities in this province are a centralized activity. Therefore, all patients with TB in British Columbia are assessed and followed by the division of TB control. In addition, TB is a notifiable disease; the DTB is informed of cases from several sources, including the provincial laboratory located in the B C C D C (British Columbia Center of Diseases Control) and the Pharmacy Division of B C C D C , which is responsible for dispensing all anti-tuberculosis medications throughout British Columbia. These additional safeguards ensure that no case of active or suspect TB can be treated in B C without the D TB being informed. Additional support for the robustness of the findings can be found in that the results were derived from a sample size of over three thousand (3041) patients. As a result of the large sample size, the confidence intervals for most findings were relatively small. 41 J MSc Thesis Previous studies have shown that a history o f previous anti-tuberculosis treatment indicates a higher risk for the development of drug resistance5'1 0'1 3'4 0'4 1. However, this information was not directly available in the current B C database. There were several factors that reflect this information in an indirect way. For example, a history of previous TB, case type and place of diagnosis are three examples that could be used as markers of previous TB treatment. It can be presumed that subjects who have a history of previous TB are more likely to receive the anti-tuberculosis treatment in a previous setting. Under the category of Case type, all the TB subjects in DTB were recoded into two broad categories according to criteria set by Health Canada- New case: no documented e vidence or history of previously active tuberculosis , and Reactivated (relapsed) case: documented evidence or history of previously active TB which became inactive3 3. According to the definition, it can be presumed that new cases did not have a history of. any previous anti-tuberculosis treatment while reactivated cases should have a history of previous anti-tuberculosis treatment. I In addition, place of diagnosis was categorized under two headings- inside Canada and outside Canada. The patients who were diagnosed outside Canada were more likely to receive any anti-tuberculosis drugs once a diagnosis was made while living in that country. These three factors were found to be significant in the multivariate setting after adjusting for age, sex and ethnicity. For the history of previous TB, the risk is twofold higher while the risk is three-fold higher than the reference group for case type and place of diagnosis. There was a close relationship between previous TB and Case type. However, previous TB is self reported but the evidence for case type is based on documents. Since the case type more closely matches the history of previous anti-tuberculosis treatment and might be considered as an alternative to capture that particular factor. Nevertheless, information regarding completeness of previous treatment, in particular treatment received outside Canada was not available and this important factor might influence the effect of these risk factors. The information regarding completeness of previous treatment would also be useful to differentiate primary and acquired (secondary) resistance that was not possible to be accounted in this study. The TB control program should consider collecting this information for future research on risk factors of TB drug resistance. 42 MSc Thesis The findings of the population study (first phase study) were then compared with the results of a study utilizing a different type of analysis using a sample (1180 subjects out of 3041) of total TB population in a case control setting (three controls per case) matched by age and gender. The results of this conditional logistic regression were very consistent with the age and gender adjusted multivariate regressions conducted in the initial retrospective population study. The same variables were found to be significant in both sets of analyses. There were minor discrepancies in the specific values of odds ratios; however, the trends of p values and confidence intervals were similar. These results add to the robustness of the findings. The consistency of results was expected as we were dealing with a large sample in both instances. However, the findings depending on the total TB population should be considered more reliable and decisions are probably best based on those particular findings. This research project contained data over a study period of 12 years. One might express concern about technological changes (e.g. laboratory methodology) that might occur during this period. A l l the laboratory tests for sensitivity were carried out in the B C C D C laboratory, and the B A C T E C radiometric system has been used to evaluate the susceptibility to the first line anti-tubercular drugs using standard techniques. The cutoff values of drug concentration used to determine resistant status in the B C C D C laboratory were consistent over the study period. If any of these drugs showed some resistance, the B A C T E C method was repeated to confirm the resistance. Another important feature of this research project is that all the potential factors were analyzed using b oth u hi-variate a nd m ulti-variate m odels. T he findings w ere c onsistent i n b oth m odels although there were discrepancies in the case of a few variables. The most significant example is the history of previous TB. This variable was not significant in both the uni-variate and multi-variate (age and sex adjusted) analyses, but the relationship became significant when ethnicity was used as an additional adjusting factor in the model. One reason for this change in significance may be related to the fact that information was obtained through self-reports. No 43 MSc Thesis supporting document was required and different groups may have felt less or more comfortable disclosing previous TB infection. For example, people who belong to Asian ethnic groups may have a higher risk for resistance development however; they also may have withheld information on their previous TB status13. Such a scenario might explain why after adjusting for ethnicity the history of previous TB became significant. In addition, polytomous regression (age and gender adjusted) was performed in this study for selected variables to see i f there were differences in risk factor characteristics between mono-drug, poly and multi-drug resistance. The factors immigration status, place of diagnosis and case category were found to be significant risk factors across all the categories of resistance. In all these cases, the risk was elevated compared to reference group. Overall, the confidence intervals for each individual category were relatively smaller in the mono group, wider in multi-drug group, and widest in the poly group. However, the number of poly resistant and multi-drug resistant cases were 64 and 27 respectively (in a population of 3041 TB subjects). With such a small sub-group, a wide variation in confidence intervals is expected; however, one should be careful to interpret the findings observed in this particular type of subgroup analysis. One of the limitations of this research project was the percentage of missing observations. The percentage o f i ncomplete i nformation w as 1 ow (01 o 6 %) f or t he m ost s ignificant r isk f actors which is expected given a TB population of over 3000 patients. Maximum precautions were taken to minimize the frequency of missing values for the variables with low missing percentage. For example, the personal health file of TB subjects with missing information was accessed to verify whether the information was incomplete primarily, or incorrect due to data entry error. From these searches, it was revealed that except for birth date, the incompleteness of other variables was mostly primary (i.e., the original provincial databases were missing information). Personal health files were accessed and attempts were made to use these files to complete missing information in the TB database. As mentioned earlier, last country of residence (44%) and history of previous TB (47%>) were associated with high percentage of missing values. Due to resource constraints, only a randomly selected portion of cases for these two variables was examined and corrected. However, there 44 MSc Thesis was only a few data entry errors. Failure to check all the incomplete or incorrect information is unlikely to have much influence on the overall findings. In addition, sensitivity analyses were carried out in order to evaluate the impact of missing observations for the dichotomous factor with high percentage of incomplete information. Since country of residence has more than two levels, no sensitivity analysis was done for this particular factor. However, sensitivity analysis was done for previous TB status. The findings showed that incomplete information might have be an impact on the effect size for this particular factor. Care should be taken to interpret the findings of this risk factor. Previous reports suggest that patients with HIV infection are at increased risk for the development of drug resistance 1 ' 1 0 , 2 1 ' 4 1 ' 4 2. However, this factor could not be examined in the current study due to the absence of HlV-related information. The Division of STD/AIDS Control of BC located at the B C C D C records information regarding HIV status (confirmed by the provincial laboratory) of all the subjects tested in the province, as well as the date of the first positive HIV test from 1988 and onward. Nevertheless, the information contained in the current database only allows reliable personal identification from 1995 onwards. Since the time period of this study ranged from 1990-2001, this variable was not considered in the final analysis. As mentioned earlier, reporting bias may have occurred in this study as information on several factors was obtained through self-reports. There was incomplete information regarding immigration, birth country, country of residence, history of previous TB, etc. The foreign born person is more likely to withhold these sorts of information and missing.information might be more associated with that particular group. It might weaken the effect size between resistance and these factors (i.e., it would most likely bias the results toward the null hypothesis). Moreover, bias in abstracting record and bias in interviewing might be associated with this study as data on several demographic and medical records were also collected through these processes. However, given the strength of association and adequate number of subjects, we do not consider it likely that they are simply due to these potential biases. In conclusion, the major findings of this study are that foreign born immigrants, patients with the history of previous TB and reactivated (relapsed) TB cases are at greater risk for TB drug 45 MSc Thesis resistance. Because resources and infrastructure for TB-control programs are severely limited in many foreign countries, persons treated for TB in these countries may receive inadequate or incomplete treatment. This puts foreign-born persons at greater risk for disease recurrence with drug-resistant strains, which complicates and lengthens the course of treatment. Some persons with multiple resistant strains are chronically i l l and persistently infectious. Treating these patients can severely strain local health department resources, especially because foreign-born populations are disproportionately underinsured or uninsured. Therefore, efforts should be made to prioritize the development and implementation of effective screening and treatment protocols for these high risk groups in order to avoid the emergence of further resistance tuberculosis. 46 MSc Thesis Bibliography 1. T. C. Y . Tsao , W. Chiou, H . Lin, T. Wu, M . Lin, P. Yang, Y . Tsai. Change in Demographic Picture and Increase of Drug Resistance in Pulmonary Tuberculosis in a 10-Year Interval in Taiwan. Infection 2002; 30:75-80 2. Snider DE and Roper WL. The new tuberculosis. N Eng J Med 1992; 326: 703-705 3. Primary Multidrug-Resistant Tuberculosis — Ivanovo Oblast, Russia, 1999. M M W R -Morbidity & Mortality Weekly Report 1999; 48(30):661-4 4. Dye C, Scheele S, Dolin P, Pathania V , Raviglione M C . Consensus statement. Global burden of tuberculosis: estimated incidence, prevalence, and mortality by country. \ WHO Global Surveillance and Monitoring Project. JAMA 1999; 282(7):677-686. 5. VP Keane, N de Klerk, T Krieng, G Hammond, and A W Musk. Risk factors for the development of non-response to first-line treatment for tuberculosis in southern Vietnam. Int. J. Epidemiol. 1997 26: 1115-1120 6. Porter J D H , McAdam K P W J. The re-emergence of tuberculosis. Ann Rev Public Health 1994; 15: 303-23 7. Mather D, Raviglione M C . The global epidemic of tuberculosis: A World Health Organization perspective. In: Friedman L N , ed. Tuberculosis. Current Concepts and Treatment. 2nd ed. Boca Raton: CRC Press L L C ; 2000:104-115 8. Canadian Tuberculosis Standards 5th Edition, 2000. Chapter I-A: The epidemiology of Tuberculosis in Canada. http://www.lung.ca/tb/TBStandards Eng.pdf. Accessed on January, 2004 9. LancstarH O. Tuberculosis. In: expectations of life. New York: Springer-Verlag, 1990. pp.81-10. Canadian Tuberculosis Standards 5th Edition, 2000. Chapter-II-F: Drug-Resistant Tuberculosis. http://'www.lung.ca/'tb/TBStandards Eng.pdf. Accessed on January, 2004 11. WHO. Global Tuberculosis Control: Surveillance, Planning, Financing. WHO Report 2003. Geneva 2003. WHO/CDS/TB/2003.316. 12. WHO, Tuberculosis, Fact sheet No 104. Revised August 2002. http://www.who.int/mediacentre/factsheets/vvhol04/en/. Accessed October, 2003 13. Manns B J, Fanning E A , Cowie R L. Antituberculosis drug resistance in immigrants to Alberta, Canada,, with tuberculosis, 1982-1994. International Journal of Tuberculosis and Lung Disease 1997, 1(3): 225-230. 47 MSc Thesis 14. Djuretic T, Herbert J, Drobniewski F, Yates M , Smith E G, Magee J G, Williams R, Flanagan P, Watt B , Rayner A , Crowe M , Chadwick M V , Middleton A M , and Watson J M . Antibiotic resistant tuberculosis in the United Kingdom: 1993-1999. Thorax 2002; 57: 477-482 15. Farzad E, Holton D, Long R, FitzGerald M , et al. Drug resistance study of Mycobacterium tuberculosis in Canada, February 1, 1993 to January 31, 1994. Canadian Journal of Public Health 2000;91:366-370 16. Peter D.O. Davies. Drug-Resistant Tuberculosis: Concluding Remarks. Ann N Y Acad Sci 2001;953:241-246 17. The Stop TB Working Group on DOTS-Plus for MDR-TB. A prioritised research agenda for DOTS-Plus for multidrug-resistant tuberculosis (MDR-TB). International Journal of Tuberculosis and Lung Disease 2003, 7(5): 410-414. 18. Espinal M A , Laszlo A , et al. World Health Organization-International Union against Tuberculosis and Lung Disease Working Group on Anti-Tuberculosis Drug Resistance Surveillance. Global Trends in Resistance to Antituberculosis Drugs. N Engl J Med. 2001; 344 (17): 1294-303 19. Dye C, Espinal M A , et al. Worldwide incidence of Multidrug-Resistant Tuberculosis. The Journal of Infectious Diseases 2002; 185:1197-202 20. Alimuddin Zumla, John M Grange. Multidrug-resistant tuberculosis—can the tide be turned? The Lancet Infectious Diseases 2001; 1: 199-202 21. Pablos-Mendez A , Raviglione M C , Laszlo A, Binkin N , Rieder H L , Bustreo F, Cohn DL, Lambregts-van Weezenbeek CS, K im SJ, Chaulet P, Nunn P. Global surveillance for antituberculosis-drug resistance, 1994-1997. World Health Organization-International Union against Tuberculosis and Lung Disease Working Group on Anti-Tuberculosis Drug Resistance Surveillance. N Engl J Med. 1998; 338:1641-9. 22. Kochi A , Vareldzis B, Styblo K . Multidrug-resistant tuberculosis and its control. Res Microbiol 1993 ; 144: 104-10 23. Espinal M A . The Global Situation of MDR-TB. Tuberculosis 20037 83:44-51 24. Pritchard A J , Hayward A C , Monk PN, Neal K R . Risk factors for drug resistant tuberculosis in Leicestershire - poor adherence to treatment remain an important cause of resistance. Epidemiology and Infection 2003; 130: 481 -483 25. World Heath Organization. WHO > WHO sites > Media centre > Fact sheets. Tuberculosis, Drug-resistant TB. http://www.who.int/mediacentre/factsheets/fs 104/en/. Accessed June 2004. 48 MSc Thesis 26. Viskum K , Kok-Jensen A. Multidrug-resistant tuberculosis in Denmark. International Journal of Tuberculosis and Lung Disease 1997, 1(4): 299-301. 27. Mahmoudi A , Isemen M D . Pitfalls in the care of patients with tuberculosis: common errors and their association with the acquisition of drug resistance. J A M A 1993; 270: 65-8 28. Bastian I, Colebunders R. Treatment and prevention of multi-drug resistant tuberculosis. Drugs 1999; 58:633-61 29. White V L, Moore-Gillon J. Resource implications of patents with multidrug resistant tuberculosis. Thorax 2000; 55: 962-63 30. World Heath Organization. WHO > WHO sites > Database on global MDR-TB research activities > MDR-TB Research Project Detail. http://www^who.int/tb_search/detail/en/index.html?proiect id=13. Accessed October 2004. 31. Gordis L. Epidemiology, 2 n d edition: W.B Saunders Company. Chapter 9. Case-Control and Cross-Sectional Studies. Page 155 32. B C Center for Disease Control. Tuberculosis Control, 2002 TB Annual Report. Accessed June 2004. 33. Tuberculosis in Canada -1997. Health Canada. Definition of terms. Page 57 http://www.hc-sc.gc.ca/pphb-dgspsp/publicat/tbcan97/pdf/tbcan97e.pdf. Accessed January 2003 34. Long R, Manfreda J, Mendella L , Wolfe J, Parker S, Hershfield E. Antituberculous drug resistance in Manitoba from 1980 to 1989. Can Med Assoc J. 1993; 148:1489-1495 35. Eidus L, Jessamine A G, Hershfield E S, Helbecque D M . A national study to determine the prevalence of drug resistance in newly discovered previously untreated tuberculosis patients as well as in retreatment cases. Can J Public Health 1978; 69: 146-153 36. Elwood K R. "TB in BC-Who gets it?". BC Medical Journal 2002, 44:493 37. Gordis L. Epidemiology, 2 n d Edition: W.B Saunders Company. Chapter 10. Estimating risk: is there an association? Page 165 38. Wang J S, Allen E A , Chao C W et al. Tuberculosis in British Columbia among immigrants in five Asian countries, 1982-1985. Tubercle 1989; 70:179-186 39. Centers for Disease Control. Primary resistance to antituberculous drugs- united states. M M W R 1983; 32:521-523 40. Alrajhi A A , AbdulwahabS, Almodover E, Al-Abdely H M . . Risk factors for drug-resistant , Mycobacterium tuberculosis in Saudi Arabia. Saudi Medical Journal. 2002; 23(3): 305-10 49 MSc Thesis 41. Hayward A C , Bennett DE, Herbert J, Watson JM. Risk factors for drug resistance in patients with tuberculosis in England and Wales 1993-1994. Thorax 1996; 51 (Suppl):s31 42. Moore M , Onorato I M , McCray E. et al. Trends in drug-resistant tuberculosis in the United States, 1993-96. J A M A 1997; 278: 833-7 \ 50 MSc Thesis APPENDIX 51 MSc Thesis Table 2: Number of Resistant and Sensitive Cases across the study years in BC Year Culture positive TB cases by year Sensitive cases (percentage) Resistant cases (percentage) 1990 195 ' 176 (90%) 19 (10%) 1991 214 197 (92%) 17 (8%) 1992 232 207 (89%) 25 (11%) 1993 254 225 (89%) 29 (12%) 1994 265 252 (95%) 13 (5%) 1995 239 212 (89%) 27(11%) 1996 237 222 (94%) 15 (6%) 1997 324 290 (90%) 34(11%) 1998 254 222 (87%) 32 (13%) 1999 260 239 (92%) 21 (8%) 2000 246 219 (89%) 27(11%) 2001 321 285 (89%) 36(11%) Grand total 3041 2746 (90%) 295 (10%) Table 3: Number of Resistant Cases with types from 1990-2001 in BC Year Mono resistant (percentage) Poly Resistant (percentage) Multi Drug Resistant (percentage) Any Resistant cases (percentage) TB cases 1990 13 (7%) 5 (3%) 1 (0.5%) 19(10%) 195 1991 13 (6%) 4 (2%) 0 (0%) 17 (8%) 214 1992 16(7%) 8 (3%) 1(0.5%) 25(11%) 232 1993 19(8%) 8 (3%) 2 (1%) 29 (12%) 254 1994 10(4%) 1 (.5%) 2(1%) 13 (5%) 265 1995 20 (8%) 4 (2%) 3(1%) 27(11%) 239 1996 7 (3%) 6 (3%) 2(1%) 15 (.6%) 237 1997 27 (8%) 4 (1%) 3(1%) 34(11%) 324 1998 18(7%) 11 (4%) 3(1%) 32 (13%) 254 1999 17 (7%) 4 (2%) 0 (0%) 21 (8%) 260 2000 21 (9%) 3 (1%) 3(1%) 27(11%) 246 2001 23 (7%) 6 (2%) 7 (2%) 36(11%) 321 Grand total 204 (7%) 64 (2%) 27 (1%) 295 (10%) 3041 52 MSc Thesis Table 4: Descriptive Statistics of Socio-Demographic variables Name of variables Levels of categorical variables Frequency with percentage Missing observations with percentage Age as continuous Mean age- 48 years Median age -51 years M i n age-1 year Maximum age-102 years SD-21 years 3041 (100%) None Age with categories Age 1 (0-20 years) 160 (5%) None Age 2 (21-40 years) 980 (32%) Age 3 (41-60 years) 787 (26%) Age 4 (61-80 years) 843 (28%) Age 5 ( 81 plus) 271 (9%) Gender Female 1321 (43%) None Male 1720 (57%) Immigration status Canadian born 913 (32%) 171 (6%) Foreign born 1957 (68%) Recent immigration (<2) No immigration 913 (32%) 234 (8%) Recent immigrants 542 (20%) Remote immigrants 1352 (48%) Recent immigration (<10) No immigration 913 (32%) 234 (8%) • Recent immigrants 1206 (43%) Remote immigrants 688 (25%) Ethnicity Caucasian 604 (20%) 25 (1%) Aboriginals 428 (14%) Chinese 778 (26%) Southeast Asian 590 (20%) Rest Asian 310(10%) Miscellaneous (rest world) 306(10%) -Place of diagnosis Canada 2821 (94%) 51 (2%) Outside of Canada 169 (6%) Birth country Canada 908 (30%) 15 (.5%) Europe and north America 196 (6%) China 460 (15%) South east Asia 603 (20%) Asia rest 572 (19%) Rest world 287 (10%) Last country of residence Europe and North America 310(18%) 1335 (44%) 53 MSc Thesis China with Taiwan 256(15%) South east Asia 423 (25%) Rest Asia 646 (38%) Rest world 71 (4%) Education level No or below secondary 327 (30%) 1928 (63%) Grade 10 398 (36%) Post secondary 388 (35%) Education level No or below secondary 725 (65%) 1928 (63%) Post secondary 388 (35%) Marital status Single 936 (46%) 997 (33%) Family 1108 (54%) Table 5: Descriptive Statistics of Medical Variables Name of variables Levels of categorical variables Frequency with percentage Missing observations with percentage Previous TB Yes 295 (18%) 1431 (47%) No 1315(82%) Diabetes Yes 213(18%) 1834 (60%) No 994 (82%) Alcoholism Yes 208 (17%) 1824 (60%) No 1009 (83%) Malnutrition Yes 42 (4%) 1962 (65%) No 1037(96%) Weight loss Yes 179(17%) 1953 (64%) No 909 (83%) Smoking Yes 345 (23%) 1514(50%) No 1182 (77%) Case type , New 2513 (84%) 42 (1%) Reactivated 486(16%) v Diagnosis type Pulmonary 2217(74%) 42 (1%> Non-pulmonary 782 (26%) Symptom summary Yes 2739 (94%) 140 (5%) No 162 (6%) 54 MSc Thesis Table 6: Results of Univariate Analysis for Socio-Demographic variables Name of variables Levels of categorical variables Odds ratio 95% confidence interval (upper, lower) p value Age as continuous .98 0.98,0.99 <.000 Age with categories Age 1 (0-20 years) Reference Age 2 (21-40 years) 1.27 0.75,2.14 .368 Age 3 (41-60 years) .78 0.45, 1.35 .380 Age 4 (61-80 years) .64 0.37, 1.11 .111 Age 5 (81 plus) .21 .08, 0.51 .001 Gender Female Reference Male 1.08 0.85, 1.37 .549 Immigration status Canadian born Reference Foreign born 3.18 2.26, 4.49 <.000 Recent immigration (<2) No immigration Reference Recent immigrants 3.78 2.54, 5.62 .<000 Remote immigrants 3.06 2.14, 4.36 <.000 Recent immigration (<10) No immigration Reference Recent immigrants 3.83 2.69, 5.46 <.000 Remote immigrants 2.32 1.54, 3.48 <.000 Recent immigration (<2) Remote immigrants Reference Recent immigrants 1.24 0.93,1.65 .147 No immigration 0.33 0.23, 0.47 <.000 Recent immigration (£10) Remote immigrants Reference Recent immigrants 1.65 1.23, 2.23 .001 No immigration 0.43 0.29,0.65 <.000 Ethnicity Caucasian Reference Aboriginals (First nations) 0.65 0.34, 1.24 .187 Chinese 2.32 1.51,3.57 <.000 South east Asian 2.92 1.88,4.52 <.000 Rest Asian/west Asian 4.40 2.77, 7.01 .016 Miscellaneous (rest world) 1.93 1.13,3.29 <.000 Place of diagnosis Canada Reference Outside of Canada 2.86 1.94,4.21 <.000 Birth country Canada Reference Europe and north America 1.42 0.73,2.75 .306 China 2.89 1.89,4.41 <.000 55 MSc Thesis South east Asia 3.46 2.34, 5.13 <.000 Asia rest 3.63 2.45, 5.38 .019 Rest world 1.89 1.11,3.21 <.000 Last country of residence Canada with Europe and north America Reference China with Taiwan 2.30 1.29, 4.09 .005 South east Asia 2.40 1.41, 4.07 .001 Rest Asia 2.17 1.30,3.60 .003 Rest world 2.38 1.06,5.33 .036 Education level Post secondary Reference Grade 10 0.81 0.53, 1.23 .322 No or elementary 0.58 0.36,0.94 .027 Education level Post secondary Reference No or below secondary 0.70 0.48, 1.02 .065 Marital status Family Reference Single 0.75 0.56, 1.02 .067 Table 7: Results of Univariate analysis for Medical variables Name of variables Levels of categorical variables Odds ratio 95% confidence interval (upper, lower) p value Previous TB No v . Reference Yes 1.11 0.74, 1.67 .620 Diabetes No Reference Yes 0.94 0.57, 1.53 .793 Alcoholism No Reference Yes 0.34 0.17, 0.68 .002 Malnutrition No Reference Yes 0.41 0.10, 1.70 .217 Weight loss No Reference Yes 0.36 0.17, 0.74 .006 Smoking No Reference Yes 0.73 0.48, 1.10 .129 Case type New Reference Reactivated 1.86 1.40, 2.47 <.000 56 MSc Thesis Diagnosis type Pulmonary Reference Non-pulmonary 0.96 0.73, 1.27 .792 Symptom summary Yes Reference No 1.25 0.76, 2.04 .39 Table 8: Multivariate analysis for Socio-Demographic variables (age and gender adjusted) Name of variables Levels of categorical variables Odds ratio 95% confidence interval (upper, lower) p value Immigration status Canadian born Reference Foreign born 3.46 2.44,4.90 <.000 Recent immigration (<2) No immigration Reference Recent immigrants 3.59 2.41, 5.35 <.000 Remote immigrants 3.48 2.42, 5.00 <.000 Recent immigration (<10) No immigration Reference Recent immigrants 3.80 2.66,5.44 <.000 Remote immigrants 2.90 1.90,4.41 <.000 Recent immigration(<2) Remote immigrants Reference Recent immigrants 1.03 0.77, 1.39 .827 No immigration 0.29 0.20, 0.41 <.000 Recent immigration (<10) Remote immigrants Reference Recent immigrants 1.31 0.95, 1.80 .096 No immigration 0.35 0.23, 0.53 <.000 Ethnicity Caucasian Reference Aboriginals (First Nations) 0.54 0.28, 1.04 .067 Chinese 2.39 1.55,3.70 <.000 Southeast Asian 2.66 1.71,4.16 <.000 Rest Asian/west Asian 3.72 2.31,5.98 <.000 Miscellaneous (rest world) 1.72 1.01,2.96 .048 Place of diagnosis Canada Reference Outside of Canada 3.48 2.33,5.17 <.000 Birth country Canada Reference Europe and North America 1.75 0.90, 3.43 .101 China 3.75 2.42,5.81 <.000 Southeast Asia 3.71 2.49,5.51 <.000 Asia rest 3.63 2.44,5.40 <.000 Rest world 2.10 1.23,3.59 .006 Last country of residence Europe and North Reference 57 MSc Thesis America China with Taiwan 2.45 1.37,4.38 .002 Southeast Asia 2.38 1.40,4.06 .001 Rest Asia 2.23 1.34,3.72 .002 Rest world 1.93 0.86,4.37 .113 Education level Post secondary Reference Grade 10 0.80 0.52, 1.22 :297 No or elementary 0.69 0.42, 1.13 .140 Education level Post secondary Reference No or below secondary 0.76 0.52, 1.10 .144 Marital status Family Reference Single 0.64 0.47, 0.89 .007 Table 9: Multivariate analysis for Medical variables (age and sex adjusted) Name of variables Levels of categorical variables Odds ratio 95% confidence interval (upper, lower) p value Previous TB No Reference Yes 1.24 0.82, 1.88 .315 Diabetes No Reference Yes 1.35 0.79, 2.30 .276 Alcoholism No Reference Yes 0.32 0.16, 0.66 .002 Malnutrition No Reference Yes 0.43 0.10, 1.82 .253 Weight loss No Reference Yes 0.38 0.18,0.79 .009 Smoking No Reference Yes 0.66 0.43,1.01 .053 Case type New Reference Reactivated 2.44 1.81,3.30 <.000 Diagnosis type by involvement Pulmonary Reference Non-pulmonary 0.89 0.67, 1.18 .434 Symptom summary Yes Reference No 1.29 0.78,2.12 .318 58 MSc Thesis Table 10: A comparative analysis of Socio-Demographic variables in Univariate and Multivariate model Name of Levels of Odds ratio 95% CI p value Odds ratio 95% CI p value variables categorical in (upper, in in (upper, for variables univariate lower) univariate Multi- lower) multi-model model variate variate model model Immigration Canadian born Reference Reference status Foreign born 3.18 2.26, 4.49 <.000 3.46 2.44, 4.90 <.000 Recent No immigration Reference Reference immigration (<2) Recent 3.78 2.54,5.62 <.000 3.59 2.41, 5.35 <.000 immigrants Remote 3.06 2.14, 4.36 <.000 3.48 2.42, 5.00 <.000 immigrants Recent Remote Reference immigration immigrants (<2) Recent 1.24 0.93, 1.65 .147 1.03 0.77, 1.39 .827 immigrants No immigration 0.33 0.23, 0.47 <.000 0.29 0.20,0.41 <.000 Ethnicity Caucasian Reference Reference Aboriginals 0.65 0.34, 1.24 .187 0.54 0.28, 1.04 .067 Chinese 2.32 1.51,3.57 <000 2.39 1.55, 3.70 <.000 South east 2.92 1.88,4.52 <.000 2.66 1.71,4.16^ <.000 Asian Rest Asian/west 4.40 2.77, 7.01 .016 3.72 2.31, 5.98 <.000 Asian Miscellaneous 1.93 1.13,3.29 <.000 1.72 1.01, 2.96 .048 (rest world) Place of Canada Reference Reference diagnosis Outside of 2.86 1.94,4.21 <.000 3.48 2.33, 5.17 <.000 Canada Birth country Canada Reference Reference Europe and north 1.42 0.73, 2.75 .306 1.75 0.90, 3.43 .101 America China 2.89 1.89,4.41 <.000 3.75 2.42, 5.81 <.000 South east Asia 3.46 2.34, 5.13 <.000 3.71 2.49, 5.51 <.000 Asia rest 3.63 2.45, 5.38 .019 3.63 2.44, 5.40 <.000 Rest world 1.89 1.11,3.21 <.000 2.10 1.23,3.59 .006 Last country Europe and Reference Reference of residence North America China with 2.30 1.29,4.09 .005 2.45 1.37,4.38 .002 Taiwan 59 MSc Thesis South east Asia 2.40 1.41,4.07 .001 2.38 1.40, 4.06 .001 Rest Asia 2.17 1.30,3.60 .003 2.23 1.34,3.72 .002 Rest world 2.38 1.06, 5.33 .036 1.93 0.86, 4.37 .113 Education level Post secondary Reference Reference Grade 10 0.81 0.53, 1.23 .322 0.80 0.52, 1.22 .297 - No or elementary 0.58 0.36,0.94 .027 0.69 0.42, 1.13 .140 Education level. Post secondary Reference Reference J No or below secondary 0.70 0.48, 1.02 .065 0.76 0.52, 1.10 .144 Marital status Family Reference Reference Single 0.75 0.56, 1.02 .067 0.64 0.47, 0.89 .007 Table 11: A comparative Analysis of Medical variables in Univariate and Multivariate model Name of variables Levels of categorical variables Odds ratio in univariate model 95% CI (upper, lower) p value in univariate model Odds ratio in Multi-variate model 95% CI (upper, lower) P value for multi-variate model Previous T B No Reference Reference Yes 1.11 0.74, 1.67 .620 1.24 0.82, 1.88 .315 Diabetes No Reference Reference Yes 0.94 0.57, 1.53 .793 1.35 0.79, 2.30 .276 Alcoholism No Reference Reference Yes 0.34 0.17,0.68 .002 0.32 0.16, 0.66 .002 Malnutrition No Reference Reference Yes 0.41 0.10, 1.70 .217 0.43 0.10, 1.82 .253 Weight loss No Reference Reference Yes 0.36 0.17,0.74 .006 0.38 0.18, 0.79 .009 Smoking No Reference Reference Yes 0.73 0.48, 1.10 .129 0.66 0.43, 1.01 .053 Case type New Reference Reference Reactivated 1.86 1.40, 2,47 <.000 2.44 1.81,3.30 <.000 Diagnosis type Pulmonary Reference Reference Non-pulmonary 0.96 0.73, 1.27 .792 0.89 0.67, 1.18 .434 Symptom summary Yes Reference Reference No 1.25 0.76, 2.04 .390 / 1.29 0.78,2.12 .318 60 MSc Thesis Table 12: Number of cases and controls by strata Strata Age and sex group Number of Original Number of Total numbers in cases number of controls each strata (col 3 controls chosen plus col 5) 1 Age 0-20, female 6 70 18 24 2 Age 0-20, male 12 72 36 48 3 Age 21-40, female 71 387 213 284 4 Age 21-40, male 65 457 195 260 5 Age 41-60, female 27 289 81 108 6 Age 41-60, male 44 427 132 176 7 Age 61-80, female 27 331 81 108 8 Age 61-80, male 36 449 108 144 9 Age 80 plus, female 2 111 6 8 10 Age 80 plus, male 5 153 15 20 Grand total 295 2746 885 1180 Table 13: A comparison of socio-demographic variables between first phase (multiple regressions) and second phase (conditional regression) analysis Name of variables Levels of categorical variables Odds ratio in Multiple regression model 95% CI (upper, lower) p value for multi-variate model Odds ratio in conditional model 95% CI * (upper, lower) p value for conditional model Immigration status Canadian born Reference Foreign born 3.46 2.44,4.90 <.000 3.45 2.39,4.98 <.000 Recent immigration (<10) No immigration Reference Recent immigrants 3.80 2.66,5.44 <.000 3.71 2.53, 5.42 <.000 Remote immigrants 2.90 1.90,4.41 <.000 3.03 1.94,4.75 <.000 Ethnicity Caucasian Reference Aboriginals 0.54 0.28, 1.04 .067 0.56 0.29, 1.10 .092 Chinese 2.39 1.55,3.70 <.000 2.49 1.56,3.96 <.000 Southeast Asian 2.66 1.71,4.16 <.000 2.68 1.66,4.32 <.000 Rest Asian 3.72 2.31, 5.98 <.000 4.19 2.49, 7.06 <.000 Miscellaneous (rest world) 1.72 1.01, 2.96 .048 2.12 1.19,3.80 .011 61 "1 MSc Thesis Place of diagnosis Canada Reference Outside of Canada 3.48 2.33, 5.17 <.000 3.35 2.07, 5.41 <.000 Birth country Canada Reference Europe and North America 1.75 0.90, 3.43 .101 1.70 0.83,3.47 0.149 China 3.75 2.42, 5.81 <.000 3.98 2.48,6.40 <.000 Southeast Asia 3.71 2.49, 5.51 <.000 3.46 2.26, 5.30 <.000 Asia rest 3.63 2.44, 5.40 <.000 3.50 2.28, 5.35 <.000 Rest world 2.10 1.23,3.59 .006 2.22 1.25,3.94 .007 Last country of residence Europe and North America Reference China with Taiwan 2.45 1.37,4.38 .002 2.17 1.16,4.08 .015 Southeast Asia 2.38 1.40,4.06 .001 2.28 1.28,4.07 .005 Rest Asia 2.23 1.34,3.72 .002 2.18 1.26,3.77 .006 Rest world 1.93 0.86, 4.37 .113 2.26 0.91, 5.61 .078 Education level Post secondary Reference Grade 10 0.80 0.52, 1.22 .297 0.86 0.53, 1.37 .516 No or elementary 0.69 0.42, 1.13 .140 0.69 0.33, 1.10 .098 Education level Post secondary Reference No or below secondary 0.76 0.52, 1.10 .144 0.76 0.49, 1.18 .224 Marital status Family Reference Single 0.64 0.47, 0.89 .007 0.66 0.46, 0.95 .025 62 MSc Thesis Table 14: A comparison of Medical variables between first phase (multiple regression) and second phase (conditional regression) analysis Name of variables Levels of categorical variables Odds ratio in Multivariate model 95% CI (upper, lower) p value for multi-variate model Odds ratio in conditional model 95% CI (upper, lower) p value for conditional model Previous TB No Reference Yes 1.24 0.82, 1.88 .315 1.28 0.81,2.03 .296 Diabetes No Reference Yes 1.35 0.79, 2.30 .276 1.60 0.87, 2.95 .132 Alcoholism No Reference Yes 0.32 0.16,0.66 .002 0.31 0.15,0.64 .001 Smoking No Reference Yes 0.66 0.43, 1.01 .053 0.66 0.42,1.06 .083 Case type New Reference Reactivated 2.44 1.81,3.30 <.000 2.76 1.94, 3.92 <.000 Diagnosis type Pulmonary Reference Non-pulmonary 0.89 0.67, 1.18 .434 0.92 0.68, 1.26 .615 Table 15: Age, Gender and Ethnicity adjusted analysis of selective variables Name of variables Levels of categorical variables Odds ratio in age and sex adjusted model 95% CI (upper, lower) p value Odds ratio in age, sex and ethnicity adjusted model 95% CI (upper, lower) p value Education level Post secondary Reference No or below secondary 0.76 0.52, 1.10 .144 0.85 0.57, 1.26 .413 Marital status Family Reference Single 0.64 0.47,0.89 .007 0.83 0.59, 1.16 .276 Place of diagnosis Canada Reference 63 MSc Thesis Outside of Canada 3.48 2.33, 5.17 <.000 2.77 ' 1.84,4.17 <.000 Previous TB No Reference Yes 1.24 0.82, 1.88 .315 1.70 1.09,2.64 .020 Alcoholism No Reference Yes 0.32 0.16,0.66 .002 0.67 0.29, 1.54 .341 Weight loss No Reference Yes 0.38 0.18,0.79 .009 0.53 0.25, 1.12 .097 Smoking No Reference Yes 0.66 0.43, 1.01 .053 0.95 0.60, 1.50 .827 Case type New Reference Reactivated 2.44 1.81,3.30 <.000 2.51 1.83, 2.44 <.000 Diagnosis type Pulmonary Reference Non-pulmonary 0.89 0.67, 1.18 .434 0.78 0.58, 1.04 .086 Table 16: Sensitivity analysis for selective variables (30% or more missing observations) Name of variables Levels of categorical variables Missing observ. with percent Odds Ratio in multi-variate model (age and sex adjusted) 95% CI for odds ratio A l l missing as ref-Odds ratio A l l missing as ref-95% CI A l l missing as outcome-Odds ratio A l l missing as outcome-95% CI Education level Post secondary 1928 (63%) Ref Below secondary 0.76 0.52, 1.10 1.01 0.77, 1.34 0.70 0.51,0.97 Marital status Family 997 (33%) Ref Single 0.64 0.47, 0.89 0.63 0.47, 0.84 0.78 0.60,1.00 Previous TB No 1431 (47%) Ref Yes 1.24 0.82, 1.88 1.24 0.84, 1.83 1.05 0.82, 1.34 Diabetes No 1834 (60%) Ref Yes 1.35 0.79,2.30 1.35 0.83, 2.18 0.95 0.73, 1.22 64 MSc Thesis Alcoholism No 1824 (60%) Ref Yes 0.32 0.16,0.66 0.39 0.20, 0.77 0.74 0.58, 0.95 Mal-nutrition No 1962 (65%) Ref Yes 0.43 0.10, 1.82 0.48 0.12, 2.00 0.84 0.66, 1.08 Weight loss No 1953 (64%) Ref Yes 0.38 0.18, 0.79 0.43 0.21, 0.89 0.78 0.60,1.00 Smoking No 1514 (50%) Ref Yes 0.66 0.43, 1.01 0.85 0.57, 1.26 0.67 0.52, 0.85 17: Age and gender adjusted effect of selective variables across different categories of drug resistant Name of variables Level Mono resistant Poly resistant Multi-drug resistant OR (95% CI) P value OR (95% CI) p value OR (95% CI) p value Immigration status Canadian born Reference Foreign born 2.56 (1.75, 3.74) <.000 16.99 (4.13, 69.83) <.000 4.96 (1.48, 16.62) .010 Recent immigration (<2) No immigration Reference Recent immigrants 2.61 (1.67, 4.09) <.000 15.54 (3.57, 67.68) <.000 7.40 (2.07, 26.50) <.000 Remote immigrants 2.61 (1.75, 3.88) <.000 ) 18.21 (4.38, 75.76) <.000 3.82 (1.06, 13.77) .040 Recent immigration (<10) No immigration Reference Recent immigrants 2.82 (1.91, 4.17) <.000 17.40 (4.19, 72.34) <.000 6.32 (1.87, 21.39) .003 Remote immigrants 2.13 (1.32, 3.43) .002 17.12 (3.93, 74.66) <.000 2.07 (0.40, 10.60) .385 Ethnicity Caucasian Reference Aboriginals (First Nations) 0.60 (0.30, 1.19) .141 0.63 .707 undetermined 65 MSc Thesis Chinese 2.01 (1.24, 3.27) .005 6.02 (1.36, 26.67) .018 3.23 (0.88, 11.92) .078 Southeast Asian 1.75 (1.05, 2.94) .033 15.81 (3.72, 67.14) <.000 1.75 (0.41, 7.53) .453 Rest Asian 2.85 (1.67, 4.88) <.000 15.74 (3.53, 70.15) <.000 3.18 (0.74, 13.76) .122 Miscellaneous (rest world) 1.61 (0.89, 2.91) .119 2.93 (0.49, 17.69) .242 1.73 (0.34, 8.74) .507 Place of diagnosis Canada Reference Outside of Canada 2.48 (1.48, 4.14) .001 2.96 (1.30, 6.73) .009 19.63 (8.52, 42.26) <.000 Birth country Canada Reference Europe and North America 1.82 (0.90, 3.68) .094 2.76 (0.25, 30.71) .410 undetermined China 2.83 (1.72, 4.63) <.000 11.72 (2.49, 55.14) .002 10.91 (2.84, 41.84) <.000 Southeast Asia 2.35 (1.49, 3.71) <.000 29.65 (7.06, 124.60) <.000 1.98 (0.40, 9.93) .406 Asia rest 2.79 (1.79, 4.33) <.000 16.27 (3.75, 70.66) <.000 5.14 (1.34, 19.67) .017 Rest world 1.87 (1.04, 3.37) .037 1.79 (0.16, 19.83) .636 \ 5.25 (1.16, 23.74) .031 Case category New Reference Reactivated 2.12 (1.47, 3.05) <.000 2.26 (1.22, 4.20) .010 7.52 (3.29, 17.17) <.000 66 MSc Thesis Table 18: Age, Gender and Birth Country adjusted analysis of selective variables Name of variables . Levels of categorical variables OR in age and sex adjusted model 95% CI (upper, lower) OR in age, sex and ethnicity adjusted model 95% CI (upper, lower) OR in age, sex and birth country adjusted model 95% CI (upper, lower) Education level Post secondary Reference No or below secondary 0.76 0.52, 1.10 0.85 0.57, 1.26 0.75 0.50, 1.12 Marital status Family Reference Single 0.64 0.47, 0.89 0.83 0.59, 1.16 0.86 0.62, 1.61 Place of diagnosis Canada Reference Outside of Canada 3.48 2.33, 5.17 2.77 1.84,4.17 2.65 1.76,4.00 Previous TB No Reference Yes. 1.24 0.82, 1.88 1.70 1.09, 2.64 1.57 1.01, 2.42 Alcoholism No Reference Yes 0.32 0.16, 0.66 0.67 0.29, 1.54 1.49 0.67, 3.35 Weight loss No Reference Yes 0.38 0.18, 0.79 0.53 0.25, 1.12 0.51 0.24, 1.10 Smoking No Reference Yes 0.66 0.43, 1.01 0.95 0.60, 1.50 1.02 0.65, 1.61 Case type New Reference Reactivated 2.44 1.81,3.30 2.51 1.83, 2.44 2.35 1.72,3.21 Diagnosis type Pulmonary Reference Non-pulmonary 0.89 0.67, 1.18 0.78 0.58, 1.04 0.78 0.59, 1.05 67 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0091623/manifest

Comment

Related Items