A. Original Cohort Methodology.
Given the significant public health burden of autoimmune or inflammatory diseases, both causative and preventive agents of those conditions have been paid a significant attention. The modulation of the immune maturation and function, which is derived from the Bacillus Calmette-Guérin (BCG) vaccination, and its role in the occurrence of autoimmune or inflammatory disease is one subjects among others1. Notably, BCG vaccine is characterized by the effect on long-lasting Th1 Th17 lymphocytes and cell-mediated immunity unlike Th2 lymphocytes and antibody production, which is induced by majority of vaccines currently in use2,3. A predominant Th1 and/or Th17 immunity is considered to the characteristic condition of several autoimmune disease such as multiple sclerosis and type 1 diabetes1. Nonetheless, there were inconclusive results between BCG vaccination and autoimmune and inflammatory diseases including asthma and diabetes1,4. An overarching purpose of the Québec Birth Cohort on Immunity and Health, herein referred to as (QBCIH), therefore is to assess the occurrence of autoimmune and inflammatory diseases in relation to a non-specific stimulation of the immune function in early age, as resulting from BCG vaccination5. In particular, the QBCIH was initially set up in 2010 aimed to address three objectives: (i) to estimate the association between BCG vaccination and childhood diabetes; (ii) to assess the relation-ship between BCG vaccination and childhood asthma, adjusting for individual and area-level potential confounders obtained from provincial administrative demographic and health databases; and (iii) given the multifactorial etiology of asthma and hence a greater likelihood of confounding, to evaluate the impact of further adjusting the measure of association between BCG vaccination and childhood asthma by collecting data on additional potential confounders unavailable in administrative databases5. All persons born in 1974 in the province of Québec, Canada, after at least 32 weeks of gestation, were eligible when the birth cohort was established. Due to the absence of a unique identification number, probabilistic record linkage was carried using five basic nominal identifiers (surname and given name, date of birth, sex, father’s given name). Linkage between the Québec BCG Vaccination Registry and the administrative databases has been validated to be highly feasible and successful6. As a result, the QBCIH included 81, 496 subjects, representing 90.5% of those who were initially eligible after linkage with the Québec BCG Vaccination Registry. Major contribution from the QBCIH includes the study of determinants of BCG vaccination7, BCG vaccination and pediatric diabetes8, BCG vaccination and asthma9, and validity of the two-stage sampling method10.
B. Adding a New Module: Research Design and Methods
Multiple sclerosis (MS) is the most common chronic inflammatory and degenerative disease of the central nervous system11. It affects more than 2 million people worldwide with geographic variation11. The prevalence of MS in Canada and the northern United States are among the highest worldwide11,12. There is currently no treatment that fully prevents or reverses the progressive neurologic deterioration, including impaired ambulation, loss of bladder control, and slowed cognitive processing11. The typical onset of MS is between the age of 20 and 40 years11,13, which results in the significant human and societal cost, and both direct and indirect costs associated with MS are expected to increase over the next two decades in Canada13. Although it is unknown whether multiple sclerosis is attributable to a single or multiple causes, various genetic and environmental risk factors have been identified11. Known risk factors include Epstein-Barr Virus and mononucleosis, some other virus infection, risk genes, temperate latitude, fibrinogen, toxins, trauma, low vitamin D, smoking, obesity, early adulthood, and female sex11. As of 2017, a systematic review found only seven case-control studies, in which the number of cases varied from 36 to 140, and one cross-over trial, which examined the effect of BCG vaccination on the course of MS patients14. Although all seven studies failed to find the association between the BCG vaccine and the MS onset, all studies were a small study and largely depending on the elf-reported exposure assessment and/or outcome14. Of note, the trial suggested a decreased relapse risk following the BCG vaccination15. Given the public health burden, which MS puts on society, and the potential of BCG vaccination to prevent the or modulate the risk of MS, a rigorous examination on the association between BCG vaccination and MS is warranted. Thus, the proposed module aims to explore the association between BCG vaccination and risk of MS within the QBCIH.
Study Population and Study Design
This module will be incorporated within QBCIH. Information on potential confounders of the association between BCG vaccination and MS will be obtained from the Birth Registry, as well as the Healthcare Registration File (universal public health system), and the the Québec BCG Vaccination Registry5.
Independent Variable: BCG vaccination
The exposure measure is defined as having been vaccinated with BCG or not in the Québec BCG Vaccination Registry. Before the province-wide BCG vaccination program ended at the end of 1974 in Québec, the centralized registry of vaccine record was integral part of the provincial government funded BCG vaccination program7,16.The Québec BCG Vaccination Registry has been validated to be highly complete, accurate and linkage with administrative data was reasonably feasible6. Thus, cohort members who were not documented in the registry as having been vaccinated will be assigned to the unvaccinated according to the previous work of QBCIH5. As the province-wide BCG vaccination program ended in 1974, subjects who vaccinated in 1975 or after will be excluded from the analytic sample (n = 2890, 3.5 of the QBCIH cohort)8.
Dependent Variable: Risk of developing Multiple Sclerosis
MS cases will be identified according to the validated definition used in Canada. Participants will be considered as havening MS if they have had one hospitalization or five physician billings coded for MS, within a 2-year period17. The validity of identifying MS cases based on Canadian administrative health data has previously been shown to be highly sensitive and specific (a sensitivity of 84.2% (95% CI 79.7 – 88.8%), specificity of 100% (95% CI 99.9 – 100%), in the original study17) , and used in the government official statistics17,18. Among MS cases, age at first occurrence of MS will be defined as the earliest of either the age at first MS-related hospitalization or medical visit. The time at risk was considered to be from 1 January 1974 until occurrence of MS, death, or 31 December 2017, whichever came first.
Information on potential confounders of the association between BCG vaccination and MS occurrence for all cohort members were extracted from the provincial birth registry, as well as the utilization of medical resources databases. From the birth registry sex, birthweight for gestational age in line with a Canadian sex-specific reference, the number of older siblings (used as a surrogate measure of infections during childhood, considered to be a risk factor of MS), parental age at childbirth, and parental place of birth. Residential postal code in 1991 will be used to determine (i) area of residence using Canada Post’s definition for mail delivery and (ii) family income estimated by median household income in the neighborhood from the 1991 Canadian census5.
The prevalence of MS according to the definition based on health administrative data will be calculated. We will utilize Cox proportional hazards regression analyses to estimate hazard ratios (HR) and 95% confidence intervals (CI) for the effect of BCG vaccination on MS occurrence. Regression models will be adjusted for the following a priori defined covariates: sex, birth weight for gestational age, number of older siblings, maternal age at childbirth, family income, area of residence, and parental place of birth5.
C. Power Calculations
Calculate the anticipated power (or required sample size) for your new study using a publicly-available software package. Please list the assumptions that went into these calculations and remember to cite the software package that you used. Please interpret your power calculations in words. (1 table or figure, with 1 paragraph description anticipated power – 5 points)
D. Strengths and Limitations
Strengths of this proposed module include its large sample size with few loss-to-follow-up, reduced misclassification of both exposure and outcome, and readily available information including provincial health registries, which would reduce resources and time required to collect data. As a result, this module has a unique potential to contribute to expand the knowledge of MS etiology. As mentioned before there has been only case-control studies with relatively a relatively small study population and a cross-over trial14. Examining the effect of BCG vaccination during a relevant for the immune maturation period on MS in a large cohort may lay the foundation for future research in the field of MS.
Among several potential limitations of the proposed module, confounding is a noteworthy threat. In QICBH, BCG vaccination was offered with no charge and not mandatory5, and previous research has shown that vaccinated people were different than their non-vaccinated counterparts with respect to gestational age, birthweight, parental age, parental birthplace, and income level in the QICBH7. Thus, in this proposed module, information on these factors will be collected as potential confounders and taken account for in the analyses. As with all observational study, however, we are not able to rule out bias due to unmeasured confounders. Although the rarity of MS in population would put a challenge regarding the sample size and power, conducting a two-stage sampling could be an efficient alternative way to collect information on important but lacking in the entire cohort and evaluate the potential magnitude and direction of unmeasured confounding19. When to conduct a two-stage sampling, covariates collected could include family history of MS and autoimmune diseases such as thyroid disease, type 1 diabetes or inflammatory bowel disease, history of infectious mononucleosis, exposure history to toxic agents, history of trauma, smoking history, the timing of puberty, breastfeeding, daycare attendance, parental education, and occupation5,11.
Selection bias would be another significant threat to internal validity due to the retrospective design of the module. In the retrospective cohort study, differential participation jointly affected by combination of exposure status and outcome could be more likely to occur than prospective cohort design. Nonetheless, the selection bias would be minimized in the module since both exposure and outcome are defined according to the the registry data, which has been highly validated and complete. Further, temporary interruptions in health coverage or lost to follow-up happened among only 4% of the entire cohort participants, thanks to the high coverage by the public health care program in Canada5.
Misclassification of exposure and outcome, and confounders is another limitation of the study. Misclassification in the exposure and outcome of interest would be minimized since because these variables were retrieved from administrative and medical registries, in which data are routinely and prospectively recorded even though the original cohort study as well as proposed module are retrospective cohort design. As mentioned above, the Québec BCG Vaccination Registry has been shown to be highly accurate and complete6. Nonetheless, misclassification of exposure status will be likely to occur and be nondifferential resulting in bias towards the null since exposure status is dichotomous. Since the definition of MS has been validated in Canada and showed high sensitivity and specificity (84.2% and 100%, respectively)17 and high specificity of outcome could help to obtain an unbiased estimates of relative risk assuming nondifferential misclassification20, the potential misclassification of outcome would not be a huge concern in the proposed module. Similarly, dependent misclassification is less likely to occur since both exposure and outcome were obtained from administrative registries. Misclassification of confounders would likely to occur as in all observational studies, and unmeasured confounders would make the situation complicated21.
Final consideration with the proposed module is the generalizability of the results. When we believe the study results are internally valid, we can consider the extent to which they are generalizable, or in other words externally valid. Given the high variability of MS incidence across the regions, ethnicity or country of origin could be an effect modifier of the association between BCG vaccination and MS11. In fact, a different prevalence of MS has been reported between two generation of Manitoba, Canada, which suggests that both genetic predisposing and other environmental factors might play a role in disease occurrence of MS22. Nonetheless, the proposed module could add a significant knowledge and open the window of potential research in the field regardless of whether this proposed module will find the null association or not.