key: cord-0714757-lgptgv72
authors: Pépin, Jacques; Labbé, Annie-Claude; Carignan, Alex; Parent, Marie-Elise; Yu, Jennifer; Grenier, Cynthia; Beauchemin, Stéphanie; De Wals, Philippe; Valiquette, Louis; Rousseau, Marie-Claude
title: Does BCG provide long-term protection against SARS-CoV-2 infection? A case–control study in Quebec, Canada
date: 2021-08-11
journal: Vaccine
DOI: 10.1016/j.vaccine.2021.08.019
sha: 5d3cd3048d16d6d16008adc9539b2944888e43e4
doc_id: 714757
cord_uid: lgptgv72

Background Early in the coronavirus disease 2019 (COVID-19) pandemic, before severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) vaccines became available, it was hypothesized that BCG (Bacillus Calmette−Guérin), which stimulates innate immunity, could provide protection against SARS-CoV-2. Numerous ecological studies, plagued by methodological deficiencies, revealed a country-level association between BCG use and lower COVID-19 incidence and mortality. We aimed to determine whether BCG administered in early life decreased the risk of SARS-CoV-2 infection in adulthood and the severity of COVID-19. Methods This case-control study was conducted in Quebec, Canada. Cases were patients with a positive SARS-CoV-2 polymerase chain reaction assay performed at two hospitals between March−October 2020. Controls were identified among patients with non-COVID-19 samples processed by the same microbiology laboratories during the same period. Enrolment was limited to individuals born in Quebec between 1956 and 1976, whose vaccine status was accessible in a computerized registry of 4.2 million BCG vaccinations. Results We recruited 920 cases and 2123 controls. Fifty-four percent of cases (n=424) and 53% of controls (n=1127) had received BCG during childhood (OR: 1.03; 95% CI: 0.89−1.21), while 12% of cases (n=114) and 11% of controls (n=235) had received two or more BCG doses (OR: 1.14; 95% CI: 0.88−1.46). After adjusting for age, sex, material deprivation, recruiting hospital and occupation there was no evidence of protection conferred by BCG against SARS-CoV-2 (AOR: 1.01; 95% CI: 0.84−1.21). Among cases, 77 (8.4%) needed hospitalization and 18 (2.0%) died. The vaccinated were as likely as the unvaccinated to require hospitalization (AOR: 1.01, 95% CI: 0.62−1.67) or to die (AOR: 0.85, 95% CI: 0.32−2.39). Conclusions BCG does not provide long-term protection against symptomatic COVID-19 or severe forms of the disease.

Thus, BCG may theoretically provide some protection against severe acute 24 respiratory syndrome coronavirus 2 (SARS-CoV-2) infection for years after 25 immunisation. Numerous ecological studies published or in preprint, have reported 26 associations between a universal BCG policy and a low incidence of or mortality due to 27 coronavirus disease 2019 at the country level [13] [14] [15] [16] [17] . Ongoing placebo-28 controlled trials of BCG in healthcare workers will address a putative short-term 29 protection [18] . 30 To determine whether BCG administered during infancy/childhood decreases the 41 Cases and controls were identified through the microbiology laboratories of the Hôpital CHUS are 540,000 and 492,000 inhabitants, respectively, and the former covers the 47 eastern part of the island of Montreal while the latter covers adjacent regions (Estrie and 48 parts of Montérégie) east of Montreal. 49 Cases were patients with a positive SARS-CoV-2 nucleic acid amplification test 50 (NAAT) at one of the two participating hospitals between 17 March 2020 and 22 October 51 2020. We initially aimed to select three controls per case, with frequency matching on 52 sex and year of birth, but eventually reduced this to two controls per case for the HMR 53 site, given the high number of cases and a relative paucity of suitable controls. Potential 54 controls were identified through the databases of patients who had a sample other than a 55 SARS-CoV-2 NAAT processed by the same microbiology laboratory during the same 56 period and belonged to the same birth cohorts. Thus, controls were recruited from the 57 same population as cases, as they had access to the same laboratories for an investigation 58 when they got sick. 59 To avoid misclassification of case-control status in the context of a relatively high 60 rate of false negativity of the SARS-CoV-2 NAAT due to pre-analytical issues [21] , we 61 excluded as potential controls patients who had a negative or indeterminate SARS-CoV-2 62 NAAT and those who underwent cultures of blood or respiratory specimens. To obtain 63 controls who were relatively representative of the catchment population at large rather 64 than its sickest fraction, we excluded as controls patients who: had been hospitalized or 4 69 haematology, oncology, radio-oncology, rheumatology, immunology, HIV, renal 70 transplants, or dialysis). Patients whose samples had been sent from a mental health 71 facility or who lived in long-term care settings were excluded as they were deemed 72 unable to give informed consent. To decrease the workload of interviewers, we excluded 73 potential cases and controls whose patronyms indicated that they were very unlikely to 74 have been born in Quebec during 1956−1976, based on the history of immigration into 75 the province [22] . We then randomly selected the controls to obtain the desired numbers 76 in each of the 22 strata based on sex and year of birth (e.g. 1956−57, 1958−59, etc.) .

Potential participants were contacted by phone. After explaining the study's goals 78 and procedures, verbal consent was sought to administer the questionnaire and link up the 79 person's data with the BCG registry at INRS after verifying whether they were indeed 80 born in Quebec. Demographic information was collected, including the six-digit postal 81 code which was used to obtain a census-based material deprivation index [23] . We asked 82 questions about occupation (healthcare or other frontline workers with exposure to the 83 public during the lockdown), and whether participants remembered having received the 84 BCG vaccine or having a vaccine scar. For the controls, we asked two additional 85 questions to assess whether they might in fact have been an undiagnosed case: whether 86 they had close contact with a COVID-19 case and/or experienced a recent episode of 87 anosmia or dysgeusia [24] . For cases, we determined whether they had required 88 hospitalization for COVID-19 or had died, as per hospital records. For the deceased 89 participants, the institutional review boards waived the requirement of informed consent 90 from the next of kin, and we collected data from hospital records. given name, we proceeded to identify whether each participant had received BCG and the 104 age at vaccination. The registry was designed to store information on vaccinees and has 105 been verified as highly complete and accurate; individuals not found in the registry were 106 considered unvaccinated [20] . Probabilistic data linkage was performed with the fastLink 107 package in R (R Foundation, Vienna, Austria) [26] . Manual verification of matches 108 below a predefined threshold was done to look for spelling and other data-entry errors.

Ninety-five percent of linkages were qualified as definite; the remainder was considered 110 as probable.

At the design stage, we aimed to recruit 900 cases and 2700 controls. This 112 corresponded to 80% power (with alpha error = 0.05) to detect a vaccine effectiveness of 113 20%. When this study was initiated in the spring of 2020, no SARS-CoV-2 vaccine was on the horizon, and we believed that a vaccine effectiveness of at least 20% would be a 115 useful contribution. age at first vaccination. Analyses were performed for all subjects, and then stratified into 121 four age categories to look for effect modification and potential waning of immunity.

Analyses were carried out for both sexes combined, then for males and females 123 separately. In multivariate models, adjustments were made for age, sex, hospital, 124 occupation (healthcare setting, essential worker or contact with public, and all others), 125 urban vs. rural residence, and material deprivation (in quintiles) as potential confounders. to a more stringent pre-selection based on patronyms amongst the former. Participation 137 rates were high, 95% and 88% among eligible cases and controls at CHUS, and 86% and 138 78% at HMR, respectively. 139 We recruited 920 cases and 2123 controls. Out of 3043 participants, 1545 (51%) 140 were considered to have been definitely vaccinated, 78 (3%) were probably vaccinated, 141 while 1420 (47%) were unvaccinated. Given the small number of participants with a 142 probable rather than definite match with the BCG registry, analyses did not differ 143 whether the former were excluded or not, and we will present results based on definite 144 and probable matches. Compared with the registry data, self-reported BCG vaccination 145 was considered unreliable as was self-report of a vaccine scar (data not shown), many of 146 which were presumably provoked by smallpox vaccine rather than BCG, and these were 147 not analysed further. Table 3 displays the frequency of hospital admission, and Table 4 shows the case-171 fatality ratio according to BCG status. For the latter, the small number of deaths 172 precluded full adjustment for confounders. As COVID-19 mortality is strongly related to 173 age, and the vaccinated were older than the unvaccinated (mean 55.7 vs. 52.9 years), we 174 present age-and sex-adjusted odds ratios. There was no evidence that BCG conferred any 175 protection against more severe forms of COVID-19. 

This large case-control study showed that BCG vaccination in infancy or childhood does 179 not provide long-term protection against COVID-19 or lessen illness severity. When it 180 was designed in May 2020, experts predicted that developing a specific vaccine would 181 require at least 12−18 months. The availability less than a year later of several marketed 9 182 vaccines with efficacy ranging between 70% and 95% [28] [29] [30] was beyond the most 183 optimistic expectations of a North American expert panel [31] . Access to COVID-19 184 vaccines remains uneven across and within countries and identifying all potential 185 prevention tools and measures remains valuable.

In that context, there has been much interest in the hypothesis that BCG might 187 confer some protection against COVID-19, due to its non-specific effect on innate 188 immunity. More than twenty ecological studies were deposited as preprints on MedRxiv 189 or published after peer review with most claiming that countries using BCG in infancy or 190 childhood experienced a lower incidence of COVID-19 or a lower mortality [13] [14] [15] [16] [17] .

Ecological studies are useful for the preliminary testing of hypotheses. With the power of 192 the internet and publicly available data, they can now be carried out in a matter of days.

However, nothing comes easily in science, and ecological studies are plagued with 194 multiple deficiencies. In this case, a major flaw is that access to a diagnosis of laboratory- [3]. We could not address this question in the current study, as BCG was infrequently 204 used in Quebec after 1976. Ongoing randomized trials, mostly in healthcare workers and the elderly, will evaluate putative short-term protection. Second, this trained immunity 206 should persist for a very long period. For how many years does the BCG-induced trained 207 immunity last, providing a protection against respiratory pathogens? A case in point is its 208 effect against the pathogen for which BCG was originally developed, Mycobacterium 209 tuberculosis. Long-assumed to be reflective of cell-mediated adaptive immunity, recent 210 work suggests that some of the BCG-induced protection is derived from innate trained Our study, the first that was specifically designed to address this issue with any effect of the vaccine will be missed. Conversely, the classic case-control design using 261 diseases targeted by the vaccine as cases and healthy controls is appropriate for testing 262 the hypothesis of a protection generated by unspecific innate mechanisms.

Recruitment of a suitable control group requires great care, and this requirement 264 was complexified by the pandemic context. Since cases were identified through hospital 265 microbiology laboratories, controls were selected from the same source. Given the 266 lockdown, it was necessary to ensure that selected controls were not much sicker than the 267 general population, which led us to make some exclusions (persons hospitalized, with an 268 emergency room visit, with infections highly related to hospitalizations, or who were 269 likely immunosuppressed). The decision to exclude potential controls who had a negative 270 or indeterminate SARS-CoV-2 NAAT was due to a concern about false-negative results. with a negative or indeterminate NAAT result were excluded from potential controls to 282 prevent misclassification of disease status. There is a legitimate concern that this may 283 have introduced selection bias. 284 We attempted to assess the extent of selection bias that could have resulted from (Table S4 ). This is congruent with the relative 288 risk estimated in the Quebec population during the first COVID-19 wave (RR=9) in the 289 province of Quebec [52], during which the vast majority of cases and controls were 290 recruited into our study. Although it is not possible to entirely rule out selection bias, this 291 argues toward the lack of a sizeable bias due to the selection of the control group.

A study limitation is that, given time and budgetary constraints, we elected not to 293 collect data on chronic co-morbidities. We believed that chronic diseases were very 294 unlikely to be confounders, given that this would require them to be associated with 295 exposure to BCG several decades earlier. Halfway into the study, we had to reduce the association between BCG exposure and SARS-CoV-2 did not differ by recruitment site.

The higher refusal rate at HMR compared to CHUS might reflect a large-city effect 299 where people are more suspicious of phone calls from unknown persons.

Another limitation of the study is that there was substantial evolution of BCG 301 strains, propagated in culture media in several laboratories between 1921−1961, when 302 laboratories started using -80 o C freezers to store seed lots and standardize their products. ⌧The authors declare that they have no known competing financial interests or personal 607 relationships that could have appeared to influence the work reported in this paper.

Vaccinations against smallpox and tuberculosis are associated with better long-413 term survival: a Danish case-cohort study 1971-2010

BCG) vaccination predicts flattened curves for the spread of 417

Further evidence of a possible correlation between the severity of 419 COVID-19 and BCG immunization

COVID-19 across 173 Countries: An Ecological Study

BCG vaccine protection from 425 severe coronavirus disease 2019 (COVID-19)

Is BCG 428 associated with reduced incidence of COVID-19? A meta-regression of global 429 data from 160 countries

Controls were identified among patients with non-631

COVID-19 samples processed by the same microbiology laboratories during the same 632 period. Enrolment was limited to individuals born in Quebec between 1956 and 1976, 633 whose vaccine status was accessible in a computerized registry of 4.2 million BCG 634 vaccinations

Results: We recruited 920 cases and 2123 controls. Fifty-four percent of cases

After adjusting for age, sex, material 639 deprivation, recruiting hospital and occupation there was no evidence of protection 640 conferred by BCG against SARS-CoV-2 (AOR: 1.01; 95% CI: 0.84−1.21). Among cases

Conclusions: BCG does not provide long-term protection against symptomatic COVID-645

Keywords: BCG, SARS-CoV-2, COVID-19, vaccine effectiveness

Data sharing statement: The anonymized datasets used and analysed during the current 656 study are available from the corresponding author on reasonable request

Does BCG provide long-term protection against SARS-CoV-2 infection? 660 A case-control study in

673 d Centre Armand-Frappier Santé Biotechnologie

Centre Armand-Frappier Santé Biotechnologie 681 Institut national de la recherche scientifique (INRS) 682 531, boul. des Prairies

Email: marie-claude.rousseau@inrs.ca

SARS-CoV-2 infection; Controls: non-SARS-CoV-2 samples in same

We found no evidence of BCG-induced long-term protection against SARS-CoV

 BCG does not confer long-term protection against severe forms of COVID-19

AC 702 designed the study, analysed and interpreted the data. JP, ACL, CG and SB supervised 703 data acquisition. MCR, MEP and JY carried out the data analysis. AC, LV, ACL 704 contributed to data interpretation. JP and MCR wrote the first draft. All authors 705 contributed to subsequent revisions of the manuscript