Cohort Profile Update: The Québec Birth Cohort on Immunity and Health (CO·MMUNITY)

Key Features.

• The Quebec Birth Cohort on Immunity and Health was originally set up to investigate the non-specific effects of Bacillus Calmette-Guerin (BCG) vaccination on the occurrence of selected autoimmune and inflammatory diseases and included 81 496 persons born in 1974, who were followed until 1994.

• The cohort was expanded to enable studying rarer diseases requiring a larger sample size and follow-up during adulthood. Assembled in 2017, the expanded Quebec Birth Cohort on Immunity and Health (CO·MMUNITY) includes 400 611 individuals born in 1970–74 in the province of Quebec, Canada, and followed until 2014; 2% of the cohort are deceased and 10% were potentially lost to follow-up.

• Administrative data acquired cover perinatal and sociodemographic information, BCG vaccination, medical services, hospitalizations for diabetes, lymphoma, multiple sclerosis, inflammatory bowel disease, asthma and other allergic diseases, prescription drugs and death if applicable.

• A nested case-control study on inflammatory bowel disease was conducted with 2728 participants.

• Although our work is governed by data privacy considerations, we welcome new collaboration opportunities. Please address enquiries to Marie-Claude Rousseau [marie-claude.rousseau@inrs.ca].

The original cohort

The overarching aim of the Quebec Birth Cohort on Immunity and Health, established in 2010, was to assess the occurrence of autoimmune and inflammatory diseases in relation to a non-specific stimulation of the immune function in early age, resulting from Bacillus Calmette-Guerin (BCG) vaccination.¹ It was originally set up to estimate the association of BCG vaccination with childhood diabetes and asthma.

All individuals born in the province of Quebec, Canada, in 1974 after at least 32 weeks of gestation were eligible. Probabilistic linkage allowed the merging of data from the Birth Registry and the Healthcare Registration File, constituting the cohort base.¹ Data were obtained from the Quebec BCG vaccination registry and administrative health databases. In total, 81 496 individuals were successfully linked and included in the cohort (90.5% of those who were eligible). Data were extracted until 1994 and included sociodemographic characteristics, vital status and use of health services for asthma, allergic diseases and diabetes. Validated algorithms were applied to identify disease incidence based on administrative health data. Additional data on potential confounders of the association between BCG vaccination and asthma were collected using a two-stage sampling approach with a balanced design (Survey on Childhood Environment and Allergic Diseases, 2012).² Equal numbers of participants were randomly sampled among four groups defined by BCG vaccination (yes/no) and asthma status (yes/no). Information collected from the 1643 participants included sociodemographic, perinatal, environmental and health- and immunization-related factors unavailable in administrative databases.^1–3

What is the reason for the new focus and new data collection?

The original birth cohort was expanded by adding about 320 000 individuals, and the follow-up lengthened by 20 years, to enable studying other, rarer, inflammatory, autoimmune, metabolic and infectious diseases requiring a larger sample size and follow-up during adulthood.

What will be the new areas of research?

The new focus for the expanded Quebec Birth Cohort on Immunity and Health (CO·MMUNITY) is on studying the associations between BCG vaccination and lymphoma, multiple sclerosis and inflammatory bowel disease, and allowing novel analyses on diabetes and asthma. In addition to providing a larger sample size, the updated cohort enables a new focus on early adulthood (up to 40–44 years of age), specific phenotypes of diabetes and multiple sclerosis based on medication use, potential risk or protective factors for these diseases and patterns of health services use.

We are further broadening the scope of CO·MMUNITY to study the long-term associations between BCG vaccination and mortality, as well as incidence of bacterial and viral infectious diseases, including COVID-19. This will enable us to continue addressing important gaps in knowledge about the non-specific effects of BCG vaccination.

Who is in the cohort?

Given the absence of a unique identification number, a probabilistic record linkage using five nominal identifiers (surname and given name, date of birth, sex, father’s given name) was realized between the Birth Registry and the Healthcare Registration File. This process estimates an overall matching weight for each pair of records, based on agreement and data frequency.⁴ Then, pairs with weights below a lower threshold are rejected, those with weights above an upper threshold are linked, and pairs with in between matching weights are evaluated manually. The linkage was conducted with the Generalized Record Linkage System and the G-Link v2.4 software, both developed by Statistics Canada.⁵

CO·MMUNITY includes persons born in the province of Quebec, Canada, from 1 January 1970, to 31 December 1974, after at least 32 weeks of gestation. Of the 443 045 individuals meeting these inclusion criteria, the probabilistic record linkage was successful for 400 611 persons, representing 90.4% of those who were initially eligible in the provincial population, fostering high representativeness (Figure 1). There were no substantial differences between the cohort and the provincial population in terms of sociodemographic or perinatal characteristics. The proportion of deceased individuals was 2% in the cohort, compared with 3% in the provincial population, indicating that some deceased individuals may have been missed. This is likely due to the archiving of their data during an update of the health care data management system in 1996–97.

Figure 1.

The expanded Quebec Birth Cohort on Immunity and Health (CO·MMUNITY): overview of record linkage and data sources. Adapted from Rousseau et al.¹ Compilation based on data from the ©Government of Quebec, Statistics Quebec, 2017. Statistics Quebec is not responsible for compilations or interpretation of results. BCG, Bacillus Calmette-Guérin; CO·MMUNITY, expanded Quebec Birth Cohort on Immunity and Health; INRS, Institut National de la Recherche Scientifique; ISQ, Institut de la Statistique du Québec (Statistics Quebec); MED-ECHO, Maintenance et Exploitation des Données pour l’Étude de la Clientèle Hospitalière (Meta-data on hospitalizations from the public health care provider); MSSS, Ministère de la Santé et des Services Sociaux (Ministry of Health and Social Services); RAMQ, Régie de l’Assurance Maladie du Québec (public health care provider). Administrative databases are indicated by a white background and datasets based on self-reported information are depicted with grey shading

In 2021, we also conducted a case-control study, nested within CO·MMUNITY: the Life History Intestinal Health Study. Persons identified with Crohn’s disease or ulcerative colitis using validated algorithms based on administrative health data,^6,7 and a random sample of potential controls, were invited to participate. Response rates were 47% (n = 946), 52% (n = 1212) and 55% (n = 570), respectively, among controls, cases of Crohn’s disease and cases of ulcerative colitis.

What has been measured?

There were two sources of data for subjects in CO·MMUNITY: (i) follow-up through data extraction from administrative sociodemographic and health databases; and (ii) self-reported information among subgroups, including the Life History Intestinal Health Study (M.-C.R. and P.J., manuscript in preparation). The cohort establishment and subsequent analyses were approved by the Commission of Access to Information (Commission d’Accès à l’Information du Québec), the governmental body overseeing access to and protection of personal information, and the research ethics committees of Institut National de la Recherche Scientifique (INRS) and Statistics Quebec.

As previously described,¹ perinatal and sociodemographic information was extracted from the Birth Registry. Data on BCG vaccination were retrieved from the provincial registry. CO·MMUNITY will allow addressing hypotheses on the long-term non-specific effects of BCG vaccination, for which some features of the Quebec BCG vaccination programme provide important advantages. The cost of the vaccine was paid by the public health system, but vaccination was not compulsory. As a result, a large enough proportion of the population remained unvaccinated, allowing sufficiently large numbers of persons in the unexposed group. Regional differences in BCG vaccination rates seemed to depend on organizational factors rather than on individual preferences for or against BCG vaccination.⁸ Self-selection for BCG vaccination is thus unlikely to bias the associations of interest.

Other data acquired from administrative databases included physician and prescription drug claims, hospitalizations and date and cause of death, if applicable. In CO·MMUNITY, the administrative health data were expanded by 20 years (until 31 December 2014) and included information on previous and new diseases of interest.

Table 1 provides an overview of the type of information extracted from the birth and death registries, health care registration file, prescription drug claims, physician claims, hospitalization data, Quebec Cancer Registry, Quebec BCG vaccination registry and the Canadian Census.

Table 1.

Data acquired from administrative sociodemographics, health databases and the Canadian Census available in CO·MMUNITY, 1970–2014

Data source
Database (years covered)	Domain	Variables
MSSS/Statistics Quebec
Birth Registry (1970–1974)	Sociodemographic	Sex, date of birth, place of residence (regional district), place of birth (municipality of hospital)
	Birth	Birthweight, gestational age, birth type (singleton, multiple)
	Family and parents	Number of live births/stillbirths, total number of deliveries, parental age at child’s birth, parent’s province/country of birth
Death Registry (1970–2014)	Death	Age of deceased person, date and cause of death
RAMQ
Healthcare Registration File (1983–2014)	Sociodemographic	Birth date, sex, three-digit postal codes (1987–2014), indices of material and social deprivation based on postal code (1987–2014), year of death (if applicable)
Prescription Drugs Claims (1983–2014)	Medication	Drug insurance coverage (start/end dates), date of service, AHFS class, code for generic name, form code, dosage code, prescription type, substitution code, duration of treatment, quantity of medication, specialty of prescriber, specialty of referring professional
Physician Claims (1983–2014)	Medical services use	Category and specialty of health professional, medical act code, role of health professional in the intervention, date of service, number of services, category and specialty of referring health professional, selected ICD-9 and ICD-10 diagnostic codes (see Supplementary Table S1)
MSSS
MED-ECHO Hospitalization data (1987–2014)	Hospitalization	Sequential number of hospitalization(s), admission and discharge dates, type of admission, date of registration at the emergency department, total number of days in hospital, type of diagnosis, selected ICD-9 and ICD-10 diagnostic codes (see Supplementary Table S1), characteristics of diagnosis, consultation domain, total number of consultations, hospital service codes, number of days within the service, category and specialty of health professional (service, intervention), date/place/type of intervention, total number of interventions, type of intensive care unit, number of days in intensive care
Quebec Cancer Registry (1984–2010)	Cancer diagnosis	Dates of diagnosis, histological confirmation, tumour declaration, laterality, mode of diagnosis, morphology, tumour status, topography, date of death (if applicable), place of death
INRS
Quebec BCG Vaccination Registry (1970–92)	BCG vaccination	BCG vaccination status, year/type/reaction (mm) of pre-vaccination test, year/type/institution of BCG vaccination
Statistics Canada
Canadian census (1991, 1996, 2001, 2006)	Sociodemographic	Median income—all census families 1991, median income—all census families 1996, median income—all census families 2001, median income—all census families 2006
National Household Survey (2011)	Sociodemographic	Median income—all census families 2011

Adapted from Rousseau et al.¹ AHFS, American Hospital Formulary Service; BCG, Bacillus Calmette-Guerin; CO·MMUNITY, expanded Quebec Birth Cohort on Immunity and Health; ICD-9, International Classification of Diseases—9th Revision; ICD-10, International Classification of Diseases—10th Revision; INRS, Institut National de la Recherche Scientifique; MED-ECHO, Maintenance et Exploitation des Données pour l’Etude de la Clientèle Hospitalière (Meta-data on hospitalizations from the public health care provider); MSSS, Ministère de la Santé et des Services Sociaux (Ministry of Health and Social Services); RAMQ, Régie de l’Assurance Maladie du Québec (Québec Public Health Insurance). 

Table 2 presents a summary of the data collected among 2728 participants in the Life History Intestinal Health Study, 2021. The information included sociodemographic characteristics, diet (including alcohol, tea and coffee consumption), physical activity, and environmental, health-related and psychosocial factors over the life course.

Table 2.

Data collected in the nested case-control study—Life History Intestinal Health Study, 2021

Domain	Variables
Personal sociodemographic characteristics	Education level attained, longest full-time occupation
Family sociodemographic characteristics	Parent’s place of residence at participant’s birth, parental age at participant’s birth, parent’s ethnicity, parent’s occupation at participant’s birth, parent’s longest occupation during participant’s childhood, parent’s longest occupation during participant’s adolescence, parent’s education level, number of siblings
Early life	Type of delivery (natural, caesarean), breastfeeding, introduction of solid foods, daycare attendance
Anthropometry	Silhouettes at age 3, 10, 15, 25 and 35a
Environment	Domestic pet ownership from birth to 2014
Smoking	Personal history of cigarette smoking until 2014 (duration, quantity, frequency), parental history of cigarette smoking during pregnancy, history of smoking in the participant’s household (from birth to age 18 and in adulthood), history of smoking in the workplace in adulthood
Diet	Dietary habits at age 10 and 20, alcohol, tea and coffee drinking in adulthoodb
Physical activity	Leisure physical activity in childhood, adolescence, and adulthood,c occupational physical activity in adulthoodd
Intestinal health	Personal and family history of inflammatory bowel disease, history of other intestinal diseases
Reproductive history and medication	Age at puberty, history of pregnancies and childbirth, history of hormonal contraceptive use,e history of hormonal treatments, history of antibiotic treatments
Psychosocial factors	Occurrence, age at occurrence and perceived effect (on a scale of most negative to most positive) of a list of psychosocial circumstances including physical or sexual abuse, difficulties in school, parental neglect, divorce or separation (parental/self), death of a loved one, financial difficulties and depressionf

Adapted from Rousseau et al.¹

^aSilhouettes for children²⁷ and adults²⁸ used with the authors’ authorizations and validated for use in epidemiological studies.^29,30

^bInspired by Hosking et al.’s Lifetime Diet Questionnaire.³¹

^cInspired by the Godin-Shepard Leisure-Time Exercise Questionnaire.^32,33

^dAdapted from the Canadian Community Health Survey.³⁴

^eAdapted from the Women’s Interview Study of Health Questionnaire.^35,36

^fInspired by the Social Readjustment Rating Scale.³⁷

Characteristics of the CO·MMUNITY study population are shown in Table 3. The CO·MMUNITY subjects were distributed equally in each birth year (1970–74), and half were men (51%). Most subjects had Quebec-born parents (88% of the subjects’ mothers and 87% of their fathers); 46% of the cohort subjects were vaccinated with BCG, 88% (156 513/178 352) in the first year of life. After the exclusion of persons who died over the follow-up (n = 7894) and those who may have been lost to follow-up (n = 38 814), defined as such due to missing yearly postal codes for ≥5 years likely indicating temporary or permanent emigration, the characteristics of the remaining subjects are similar to those of the baseline study population. The only minor differences are 2% increases in the proportions of subjects with parents born in Quebec and of those who received the BCG vaccine. At the end of follow-up in 2014, subjects were aged 40–44 years.

Table 3.

Characteristics of the CO·MMUNITY study population (1970–2014) at baseline and end of follow-up in 2014

	Baseline (n=400 611)		End of follow-up a (n=353 903)
Characteristic	n	%	n	%
Birth year
1970	83 197	20.8	72 961	20.6
1971	80 847	20.2	71 247	20.1
1972	76 730	19.2	67 681	19.1
1973	77 209	19.3	68 723	19.4
1974	82 628	20.6	73 292	20.7
Sex
Men	205 276	51.2	179 341	50.7
Women	195 335	48.8	174 563	49.3
Birthweightb
Extremely low, very low, and low	28 598	7.1	25 468	7.2
Normal	360 320	90.0	318 252	89.9
High	11 660	2.9	10 160	2.9
Missing	33		24
Gestational agec
Premature	61 460	15.3	54 413	15.4
At term	329 701	82.3	291 289	82.3
Post-term	9 443	2.4	8 195	2.3
Missing	7		7
Number of older siblings
0	167 664	43.1	147 629	42.9
1	118 937	30.6	104 762	30.5
>1	102 306	26.3	91 503	26.6
Missing	11 704		10 010
Mother’s age at childbirth (years)
<35	357 561	91.4	315 828	91.4
≥35	33 630	8.6	29 894	8.6
Missing	9 420		8 182
Father’s age at childbirth (years)
<40	355 654	93.0	314 792	93.1
≥40	26 611	7.0	23 264	6.9
Missing	18 346		15 848
Mother’s birthplace
In Québec	341 471	87.6	309 899	90.0
Outside Québec	48 117	12.4	34 583	10.0
Missing	11 023		9 422
Father’s birthplace
In Québec	329 409	86.6	299 108	88.8
Outside Québec	51 065	13.4	37 545	11.2
Missing	20 137		17 251
Area of residenced
Rural	126 667	31.9	115 997	32.8
Urban	270 886	68.1	237 294	67.2
Missing	3 058		613
Family incomee (CAN$/year)
<35 333	95 467	24.4	86 915	24.8
35 333–40 666	99 951	25.6	91 332	26.0
40 667–47 058	97 568	25.0	88 261	25.2
≥47 059	98 010	25.1	84 039	24.0
Missing	9 615		3 357
BCG vaccination status
Not vaccinated	209 736	54.0	178 460	52.1
Vaccinated	178 352	46.0	164 044	47.9
Uncertain	12 523		11 400
Age at BCG vaccination (years)
Not vaccinated	209 736	54.0	178 460	52.1
<1	156 513	40.3	144 140	42.1
≥1	21 839	5.6	19 904	5.8
Uncertain	12 523		11 400

Compilation based on data from the ©Government of Quebec, Statistics Quebec, 2017. Statistics Quebec is not responsible for compilations or interpretation of results.

BCG, Bacillus Calmette-Guerin; CO·MMUNITY, expanded Quebec Birth Cohort on Immunity and Health.

^aAll cohort members, excluding those who were deceased over the follow-up and those who had a missing postal code for a minimum of 5 years at the end of follow-up (considered as possibly lost due to emigration).

^bBirthweight: extremely low (<1000 g), very low (1000 to <1500 g), low (1500 to <2500 g), normal (2500 to ≤4200 g), and high (>4200 g).

^cGestational age: premature (<37 weeks), at term (37–41 weeks) and post-term (>41 weeks).

^dDetermined using the second character of the subjects’ postal code in 1987 (0: rural, ≠0: urban).

^eEstimated by ‘median household income’ from the 1991 Canadian census using the first three characters of subjects’ postal code.

The main outcomes under study were diabetes, lymphoma, multiple sclerosis, inflammatory bowel disease (Crohn’s disease and ulcerative colitis) and asthma. Health services for the main and other secondary outcomes were retrieved from medical service claims, prescription drugs and hospitalization data. The extracted International Classification of Diseases (ICD) codes are listed in Supplementary Table S1. Table 4 shows the algorithms applied to identify the main outcomes, and their period prevalence in CO·MMUNITY.

Table 4.

Period prevalence of the diseases of interest based on validated algorithms applied to health services data (n = 400 563), CO·MMUNITY 1983–2014^a

Disease (ICD-9/ICD-10/ICD-O-3) b	Definition	n	Period prevalence,/1000 (95% CI)
Diabetes (250/E10-E14)	≥2 diabetes-related physician claims within 2 years or ≥1 diabetes-related hospitalization38 c	12 975	32.4 (31.8–32.9)
Type 1 diabetes in childhood (250)	≥4 diabetes-related physician claims within 2 years, age <19 years10	807	2.0 (1.9-2.2)
Lymphoma (200, 201, 202/C81-C86, C88/959, 965-972)	≥2 physician claims or hospitalizations within 2 months or listed in the Quebec Cancer Registry with a diagnosis of Hodgkin’s or non-Hodgkin’s lymphoma18	1 526	3.8 (3.6–4.0)
Hodgkin’s (201/C81/965-966)	Same as above with ≥80% of health services related to Hodgkin’s lymphoma18	433	1.1 (1.0-1.2)
Non-Hodgkin’s (200, 202/C82-C86, C88/959, 967-972)	Same as above with ≥80% of health services related to non-Hodgkin’s lymphoma18	951	2.4 (2.2-2.5)
Multiple sclerosis (340/G35)	≥3 multiple sclerosis-related hospital or physician claims over the follow-up39–41	1 659	4.1 (3.9–4.3)
Inflammatory bowel disease (555, 556/K50, K51)	Paediatric: ≥1 sigmoidoscopy/colonoscopy and (≥4 IBD-related physician claims or 2 hospitalizations within 3 years) or 0 sigmoidoscopy/colonoscopy and (≥7 IBD-related physician claims or 3 hospitalizations within 3 years)6	405	1.0 (0.9–1.1)
	Adult: ≥5 IBD-related physician claims or hospitalizations within 4 years7	3 317	8.3 (8.0–8.6)
Crohn’s disease (555/K50)	Same as above (paediatric and adult), with Crohn’s disease identified based on the 5 most common diagnoses among the 9 latest6,7	2 546	6.4 (6.1–6.6)
Ulcerative colitis (556/K51)	Same as above (paediatric and adult), with ulcerative colitis identified based on the 5 most common diagnoses among the 9 latest6,7	1 134	2.8 (2.7–3.0)
Asthma (493/J45)	≥2 asthma-related physician claims within 2 years or ≥1 asthma-related hospitalization42,43	48 671	121.5 (120.5–122.5)

Adapted from Rousseau et al.¹ Compilation based on data from the ©Government of Quebec, Statistics Quebec, 2017. Statistics Quebec is not responsible for compilations or interpretation of results.

IBD, inflammatory bowel disease; CI, confidence interval; CO·MMUNITY, expanded Quebec Birth Cohort on Immunity and Health.

^aGiven that health services were available only from 1983, 48 participants who were deceased before 1983 were excluded from these analyses.

^bInternational Classification of Diseases Ninth/10th Revision/Oncology Third Revision.

^cTo exclude gestational diabetes, evidence of diabetes in women aged 10–54 years was excluded if it occurred 120 days before or 180 days after hospital records containing any pregnancy-related or obstetric code (ICD-9: 641–676, V27; ICD-10: O1, O21-95, O98, O99, Z37).

What has it found? Key findings and publications

BCG vaccination and incidence of type 1 diabetes in adolescence

Following our previous research on BCG vaccination and type 1 diabetes in childhood,⁹ we assessed whether BCG vaccination was associated with the incidence of type 1 diabetes during the adolescence period, applying disease identification algorithms validated in the Quebec population.¹⁰ The risk of type 1 diabetes was similar in vaccinated compared with unvaccinated individuals (adjusted hazard ratio, HR_adj = 1.06, 95% confidence interval (CI): 0.88–1.29). There was no association with age at vaccination, and results did not differ by sex.¹¹

BCG vaccination and incidence of diabetes in adulthood

Despite evidence that BCG vaccination can have a long-term impact on sugar metabolism and immune mechanisms involved in autoimmunity,^12–15 no previous study had addressed the association between early life BCG vaccination and type 1, type 2 and latent autoimmune diabetes in early adulthood. We found that BCG vaccination was not associated with type 1 diabetes up to 30 years of age but was related to a decreased incidence at 30–44 years of age (HR_adj = 0.65, 95% CI: 0.44–0.95). BCG vaccination was associated with a decreased risk of type 2 diabetes (HR_adj = 0.85, 95% CI: 0.79–0.92) but not with latent autoimmune diabetes in early adulthood (HR_adj = 1.30, 95% CI: 0.71–2.38).¹⁶ The novel use of drug prescriptions to define the three main clinical phenotypes in early adulthood allowed, for the first time, to document inverse associations between BCG vaccination and adult onset of types 1 and 2 diabetes. Further studies should be performed with follow-up extending later in adulthood, particularly for type 2 diabetes and latent autoimmune diabetes in adults.

BCG vaccination and risk of lymphoma

We conducted the largest study to date addressing the association between BCG vaccination and incidence of lymphoma.¹⁷ We identified cases of non-Hodgkin’s and Hodgkin’s lymphoma with a validated algorithm¹⁸ complemented by tumour registry data. No association was observed with non-Hodgkin’s lymphoma (HR_adj = 0.99, 95% CI: 0.86–1.13). For Hodgkin’s lymphoma, hazards were not proportional over time, thus we stratified the analysis according to follow-up. Risk of lymphoma before 18 years of age was increased among vaccinated compared to unvaccinated individuals (HR_adj = 2.26, 95% CI: 1.39–3.69). After 18 years of age, no association was observed (HR_adj = 0.93, 95% CI: 0.75–1.15).¹⁹ These results need confirmation in other populations.

BCG vaccination and risk of multiple sclerosis

Previous studies on BCG vaccination and multiple sclerosis did not consider clinical phenotypes. Since aetiology could potentially differ by phenotype, we used prescription of immunomodulatory drugs to identify individuals with relapsing-remitting multiple sclerosis. No association was found between BCG vaccination and this clinical phenotype (HR_adj = 1.01, 95% CI: 0.85–1.20), whereas multiple sclerosis of unknown phenotype was positively associated with BCG vaccination (HR_adj = 1.25, 95% CI: 1.10–1.41). These original findings highlight the need to conduct further research based on data sources allowing the identification of phenotype, such as clinical cohorts or disease registries.²⁰

Influence of age at diagnosis on the use of health services for multiple sclerosis

We studied the use of health services among persons living with multiple sclerosis in a real-world setting. We aimed at determining whether health services utilization differed if persons were diagnosed before or after 29 years of age, which was until recently the average age at diagnosis.²¹ In the year following diagnosis, persons who were diagnosed younger had a higher rate of visits to a neurologist, a lower rate of visits to a general practitioner and a higher rate of hospitalization compared with those who were diagnosed at or after 29 years of age.²² The year after diagnosis was the period during which health services use was the highest, and no striking differences by age at diagnosis were observed in other time periods.²²

Appendectomy and risk of inflammatory bowel disease

Appendectomy induces gut microbiome changes that may affect the subsequent risk of inflammatory bowel disease, namely Crohn’s disease and ulcerative colitis. In this cohort, Crohn’s disease risk was increased among those who had an appendectomy (HR_adj = 2.02, 95% CI: 1.66–2.44), particularly when the procedure was performed during young adulthood (18–29 years vs 8–17 years: HR_adj = 2.81, 95% CI: 1.65–4.78). This excess risk was observed in the 2 years after appendectomy but not thereafter, suggesting a possible detection bias. In contrast, an inverse association was observed with ulcerative colitis (HR_adj = 0.39, 95% CI: 0.22–0.71), which was strongest ≥5 years compared with 0–4 years after the appendectomy (HR_adj = 0.21, 95% CI: 0.06–0.72).²³

What are the main strengths and weaknesses?

There are inherent limitations to using administrative databases for epidemiological research, the main one being the lack of some information on potential confounders and of clinical data. We have addressed the former issue in sub-studies on asthma² and on inflammatory bowel disease (M.-C.R. and P.J., manuscript in preparation) by enriching the administrative databases with self-reported data collected directly from subgroups of the study population.

A second weakness relates to the unavailability of prescription drug claims data for a portion of the study population until 1997. Indeed, from 1983 to 1996, public medication insurance was only available to persons on social assistance, their dependents and persons aged ≥65 years; 5% of CO·MMUNITY was covered in this period. In January 1997, public medication coverage became mandatory for anyone who did not have private medication insurance through employment, and 70% of the cohort subjects had at least one period of public medication insurance coverage over the follow-up.

A third weakness pertains to the slightly lower mortality rate in the cohort compared with the provincial population. We suspect that some individuals who died in their youth may have been missed due to the archiving of their data during an update of the health care data management system in 1996–97. However, given that CO·MMUNITY includes 90.4% of the source population, it is highly representative of the Quebec population.

Despite these limitations, our cohort has several strengths. These include a large sample size, population coverage and representativeness. Additionally, it benefits from a long follow-up, and we have defined a diverse range of health outcomes using validated algorithms. The addition of two instances of data collection, to document factors of primary interest and potential confounders in sub-samples of the initial study population, distinguishes this cohort from others based solely on either administrative or self-reported data. Furthermore, an important strength of this cohort is that BCG vaccination status is determined from reliable registry records. This is crucial, since self-report of BCG vaccination has low validity^24,25 and may result in substantial misclassification.²⁶

Can I get hold of the data? Where can I find out more?

Although our work is governed by data privacy considerations, we support expanding the capacity and diversity of our research team and encourage collaborative research. Interested researchers are invited to contact the corresponding author and principal investigator of the study [marie-claude.rousseau@inrs.ca]. News and results from the Epidemiology and Biostatistics Unit at INRS—Centre Armand-Frappier Santé Biotechnologie can be followed at [www.epi.inrs.ca].

Ethics approval

The Commission of Access to Information (Commission d’Accès à l’Information du Québec), the governmental body overseeing access to and protection of personal information, authorized the cohort establishment and subsequent analyses (#110267-S, #1023063-S, #1021593-S). The research ethics committees of Statistics Quebec (Comité d’Ethique: #19–07) and of INRS (Comité d’Ethique en Recherche avec des Etres Humains: CER-09–196, CER-09–203, CER-09–204, CER-15–375, CER-15–377, CER-15–379, CER-18–489) approved the projects.

Data availability

See ‘Can I get hold of the data?’ above.

Supplementary data

Supplementary data are available at IJE online.

Author contributions

M.-C.R. designed the study with collaboration from M.-E.P., P.J. and A.B. M.-C.R. directed the study’s implementation. All authors contributed to designing the analytical strategy and helped to interpret the findings. F.C. and H.R. cleaned, recoded and managed the data. M.-C.R., P.C., C.S., M.M., C.F., F.C., H.R. conducted the analyses. P.C.’s contribution was part of his PhD project at Institut National de la Recherche Scientifique (INRS). C.S. and C.F.’s contributions were part of their respective Master’s projects conducted at INRS. M.M.’s contribution was part of her PhD project at Université de Montréal. M.-C.R. drafted all sections of the manuscript. All authors critically revised the text, read and approved the final version.

Funding

This work was supported by an infrastructure grant from the Canada Foundation for Innovation & the Quebec Ministry of Education, Leisure and Sports [grant number 12532] and research grants from the Canadian Institutes of Health Research [grant numbers MOP-97777, MCH-97593, MOP-142705, PJT-159791], Fonds de recherche du Québec-Santé (FRQS) [grant number 16227], the Multiple Sclerosis Society of Canada [grant number 2435], the Canadian Cancer Society [grant number 703309] and through a partnership with Statistics Quebec. M.-E.P. [award number 15868], P.J. [award number 283723] and A.B. [award number 254051] were supported by career awards from the FRQS. P.C. was supported by a FRQS fellowship [fellowship number 272756]. C.S. [scholarship number 44302] and C.F. [scholarship number 55612] received scholarships from Aquimob. M.M. was supported by scholarships from FRQS [scholarship number 217312], Fondation Armand-Frappier (Beaulieu-Saucier scholarship), and Centre de Recherche du Centre Hospitalier de l'Université de Montréal.

Conflict of interest

None declared.