Stop playing with data: there is no sound evidence that Bacille Calmette-Guérin may avoid SARS-CoV-2 infection (for now).

Since the beginning of the COVID-19 epidemic, a possible explanation for the high heterogeneity of infection/mortality rates across involved countries was hinted in the prevalence of tuberculosis vaccination with Bacille Calmette-Guérin (BCG). A systematic review was therefore performed on May 2, 2020. A total of 13 articles were ultimately retrieved, 12 of them as preprint papers. All articles were ecological studies of low quality. Most of them did not include main confounding factors (i.e. demographic of the assessed countries, share of peo- ple residing in urban settings, etc.), and simply assessed the differences among incidence/mortality of COVID-19 with vaccination rates or by having vs. having not any vaccination policy for BCG. Even though all studies shared the very same information sources (i.e. international registries for BCG vaccination rates and open source data for COVID-19 epidemics), results were conflicting, with later studies apparently denying any true correlation between COVID-19 occurrence and BCG vaccination rates and/or policies. As a consequence, there is no sound evidence to recommend BCG vaccination for the prevention of COVID-19.

Introduction

The Bacille Calmette-Guérin (BCG) is a live attenuated vaccine against tuberculosis (1,2). Following early reports demonstrating that BCG may reduce infant mortality independent of its effects on tuberculosis (3), some studies have shown that BCG can enhance the reactivity of the innate immune system. Through an increased secretion of pro-inflammatory cytokines (i.e. “trained immunity”), BCG would improve reactivity against tumor cells (e.g. urothelial cancer), but also against some pathogens (e.g. staphylococci, candidiasis, yellow fever) (4), including respiratory viruses such as respiratory syncytial virus and influenza virus (5,6). More precisely, a study by Leentjens et al has shown that sequential BCG – influenza vaccination may elicit a more pronounced antibody response against influenza A(H1N1) (7).

Because of the ongoing unavailability of specific preventive and/or therapeutic measures against SARS-CoV-2 infection, several stakeholders have suggested that the BCG vaccine may be repurposed as a preventive and/or therapeutic option against COVID-19. As some conflicting results have been reported, an updated synthesis of the literature is required in order to better inform health policies and guidelines.

Methods

A systematic review and meta-analysis was undertaken following the “Preferred Reporting Items for Systematic Reviews and Meta-Analysis” (PRISMA) guidelines (8). We searched different scholarly databases (namely, PubMed/MEDLINE and EMBASE) as well as the pre-print server medrxiv.org for relevant studies from inception up to 02/05/2020, without applying any backwards chronological restrictions. The search strategy was a combination of the following keywords (free text and Medical Subject Heading [MeSH] terms, where appropriate): (COVID-19 OR SARS-CoV-2 OR Coronarivus) AND (Bacille Calmette Guérin OR Bacillus Calmette Guérin). Records were handled using a references management software (Mendeley Desktop Version 1.19.5, Mendeley Ltd 2019), and duplicates were removed. Articles eligible for review were original research publications available online or through inter-library loan. A language filter was applied, by retaining articles written in Italian, English, German, French or Spanish, the languages spoken by the investigators. Two independent reviewers reviewed titles, abstracts, and the full-text of articles. Titles were screened for relevance with respect to the subject under study. Any articles reporting original results, which met one or more of the inclusion criteria, were retained for the full-text review. The investigators independently read full-text versions of eligible articles. Disagreements were resolved by consensus between the two reviewers; when it was not possible to reach consensus, input from the main investigator was searched and obtained. Further studies were retrieved from reference lists of relevant articles and consultation with experts in the field.

Results and Discussion

Briefly, a total of 161 entries were ultimately retrieved (more precisely: 4 in PubMed/Medline, 138 in Scopus, 19 in medrxiv.org). After removal of duplicates, and screening of titles and abstracts, a total of 13 full-text articles were retained, 12 of them not peer-reviewed pre-print. As shown in Table 1, all the available studies were ecological ones, with raw data (i.e. COVID-19 epidemics, BCG vaccination coverage) derived from the same institutional sources.

Table 1.

Characteristics of the studies included in the analyses.

Study	Study type	Outcome variable (COVID-19)	BCG estimate	Methodology	Control variables	Main findings
Akiyama et al. (9)	Ecological study	Doubling time for mortality	BCG vaccination policies (any vs. none)	Comparison of doubling time for countries with BCG vaccination policies for people aged 0-39 vs. those without	BCG strain	Doubling time of 5.4 days for countries with BCG vaccination policies vs. 4.2 days (p 0.007; Wilcoxon rank-sum test) in countries without BCG vaccination policies
Green C et al. (11)	Ecological study	Mortality	BCG vaccination policies (any vs. none)	Calculation of doubling time for death rate, comparison between countries with BCG vaccination policies vs. those without	TimeAge gratification demographicsReporting timesTimeframe for BCG vaccinations	Doubling time ≥ 5 days for countries with BCG vaccination policies vs. 4 to 1 days in countries without BCG vaccination policies.
Dayal et al. (12)	Ecological study	Case fatality rates	BCG vaccination policies (any vs. none)	Difference between case fatality rates in countries having or not a BCG policy	None	Case fatality rate of 5.2% in countries not having a BCG policy vs. 0.6% in those having a BCG policy
Dolgikh S (10)	Ecological study	Mortality	BCG vaccination policy (4 levels: countries administering BCG; countries never administering BCG; countries that had administer BCG to selected population groups; countries that had administered BCG in the past, but discontinued)	Descriptive analysis	None	All countries with very low levels of mortality have a BCG policy or had it recently
Hegarty et al (4)	Ecological study	IncidenceMortality	BCG policy (any vs. none)	Descriptive analysis	Single day of assessment (March 22, 2020)	Daily incidence of COVID-19 was 0.8/1,000,000 in countries with BCG vaccination vs. 34.8/1,000,000 in countries without such a program.Mortality 0.08/1,000,000 in countries with a BCG vaccination program vs. 34.8/1,000,000 in countries without a program.The crude case fatality rate was 4.1% in countries with BCG vaccination program vs. 5.1% in countries without BCG vaccination program.
Shet et al. (14)	Ecological study	Mortality	BCG policy (any vs. none)	Log-linear regression model using crude COVID-19 mortality as outcome	Gross Domestic ProductDays from 100th COVID-19Percentage of population ≥ 65 years	Mortality of 0.4/1,000,000 inhabitants (IQR 0.06-0.4), 0.65 (IRQ 0.2-2.2), and 5.5 (IQR 1.6-13.9) for low-middle, upper-middle and high income countries.In log-linear regression, COVID-19-attributable mortality was 5.8 times lower in countries with BCG vaccination policies than in countries without.
Sala et al (13)	Ecological study	IncidenceMortalityCase fatality rate	BCG vaccination policy (three levels: countries administering BCG, countries never administering BCG, countries that had administered BCG in the past, but discontinued)	Regression model including	Country life expectancyMean temperature in February/March 2020	Never having had or having discontinued BCG vaccination was effector of incidence (b = 0.6483, p = 0.002, and b = 0.8666, p = 0.0025; R2 = 0.6409) and mortality (b = 0.7262, p = 0.007; b = 1.495, p < 0.0001; R2 = 0.5473). For case fatality rate only never having had a BCG policy was a significant effector (b = 1.1140, p = 0.0267; R2 = 0.1019).
Klinger et al. (15)	Ecological study	IncidenceMortality	BCG policy (having or having had a national policy vs. never)	Correlation between years of BCG policy and incidence/mortality rates.Multivariate model including BCG policy and characteristics of the country in three time lag models (15/20/25 days).	Population >65 yearsEconomic development indicatorsPrevalence of chronic diseasesDeath rate from cardiovascular disorders	Negative correlation between years of BCG vaccination policy and incidence and mortality for COVID-19.In multivariate analysis, beta = -0.67 (p 0.0355) at 15 days; beta = -0.75 (p 0.0123) at 20 days; beta = -1.18 (p = 0.0235) at 25 days.
Berg et al. (16)	Ecological study	IncidenceMortalityGrowth Curves	BCG policy (any vs. none)	Regression analysisMultivariate model including BCG policy and characteristics of the country	Variation of BCG policy during the 20th centuryCultural dimension (individualism vs. collectivism; tightness vs. looseness of social norms)	Growth rate for COVID-19 cases (b = -0.027, p < 0.001) and deaths (b = -0.038, p < 0.001) higher in countries without mandatory BCG vaccination.
Singh S (19)	Ecological study	IncidenceMortalityCase fatality rate	BCG vaccination ratesPrevalence of Latent Tuberculosis infection (LTBI)	Correlation between death rate and LTBI	Healthcare Access and Quality indexIncidence of Neglected tropical diseasesIncidence of tuberculosis	Decline of SARS-CoV-2 infections by increasing LTBI (r2 = 0.6243)Countries with higher LTBI rates show lower COVID-19 case fatality rates, and vice versa, with large heterogeneity and without any correlation
Goswami et al., 2020 (20)	Ecological study	Incidence	BCG vaccination rates	Comparison of COVID-19 data incidence in countries with higher vaccination rates (>95% coverage) vs. countries ≤95% coverage.	Incidence / mortality for Malaria	Higher incidence of COVID-19 (0.002%) for countries with higher vaccination rates compared to lower vaccination sates (0.0003%).In Europe and America, the relationship was inverted (i.e. 0.0002% vs. 0.0029%).
Shivendu et al., 2020 (17)	Ecological study	IncidenceMortality	BCG vaccination rates	Multiple regression models adjusted for economic and demographic factors	Number of COVID-19 tests actually performedPopulation density% population > 65 years of age% population living in urban areasGovernment transparency (Corruption Perception Index)Human development index	No significant differences between countries including BCG in National Immunization Programs (NIP) vs. countries without BCG in NIP
Miller et al. 2020 (18)	Ecological study	IncidenceMortality	BCG vaccination rates	Regression analysis modeled for economic and demographic factors	Dates of initiation of BCG vaccinationGross National Income	No occurrence of COVID-19 deaths in countries of lower income reporting BCG vaccination policies.Mortality for COVID-19 was higher in middle-high income countries that never had BCG vaccination policies compared to lower income countries (i.e. 16.4 ± 7.3/1,000,000 vs. 0.8 ± 4/1,000,000), linearly decreasing with older enforcing of BCG in NIP.

In summary, three different blueprints were identified: a) studies comparing COVID-19 rates (i.e. incidence, mortality and case fatality rates, in absolute terms or focusing on their doubling time) in countries having a BCG vaccination policy vs. countries having not and/or having discontinued (No. = 5) (4,9–12); b) studies assessing through regression models how a BCG vaccination policy may have affected the ongoing epidemic in terms of incidence and/or mortality (No. = 4) (13–16); c) assessing how BCG vaccination rates may have affected incidence and mortality for COVID-19 (No. = 4) (17–20). All studies were based on the same institutional databases, and differed on the modelling of the statistical analyses. Even among studies reporting an analogous study design, models were quite heterogeneous, particularly when focusing on control variables. For instance, only 4 studies (11,14,15,17) included in their modeling the demographic of the study population, with a further study that rather controlled the analyses for life expectancy (13). Moreover, only 6 studies included in the statistical models socio-economic factors, either as a raw data (e.g. Gross Domestic Income) or a synthetic index (e.g. Human development index; Healthcare Access and Quality Index, etc.) (12,14–17,19). Eventually, only two studies controlled the analyses for a critical factor such as the risk of social interaction, either through a proxy (i.e. percentage of total population residing in urban areas) (17), or by means of synthetic indices (16).

Not coincidentally, the results are conflicting. Despite the substantial sharing of the information sources, while certain studies hinted towards higher incidence and/or mortality rates in countries having and/or having implemented BCG vaccination policies, studies performed with different design and modeling substantially denied such a correlation (17, 20). As the latter studies were performed more recently, it is reasonable that the timeframe and the dynamics of the COVID-19 epidemic may a common distorting bias from the original reports. However, available estimates should be cautiously retained, as all studies were affected by similar shortcomings.

Firstly, most of reports assessed the effector variable BCG vaccination as the presence/absence of BCG policies. Even though some countries are deprived of national policies for BCG vaccine, certain population groups such as healthcare workers (HCWs) may be selectively vaccinated following specific recommendation (21). Despite nearly all Italian HCWs who began their formal education before 2000 were actively vaccinated against BCG, Italian National Institute for Occupational Injuries (INAIL) recently reported that around 10% of all COVID-19 positive cases had occurred among HCWs, with 150 deaths among medical professionals, most of them aged 65 years or more (22-24).

Second, it should be stressed that all estimates depended on the quality of raw data on COVID-19 (22). On the one hand, some countries may have unreliably reported the number of cases for a variety of reasons, including the availability and the reliability of diagnostic tests, and the locally implemented policies (25-27). On the other hand, the incidence rate is rather a notification rate, whose significance is better understood when compared with the total number of processed samples – a figure that only few international authorities properly record (22, 25-28).

Eventually, the COVID-19 pandemic had a distinctive progression, with an initial outbreak in mainland China. Even though China had a universal BCG policy since 1950s (29), extensive BCG vaccination did not impaired Wuhan to becoming the initial epicenter of COVID-19 pandemic. Moreover, after a significant but limited involvement of nearby countries such as South Korea and Vietnam, COVID-19 pandemic had a sudden spread to Western Europe (22, 28). As Western Europe has progressively become a low-notification area for tuberculosis, also national vaccination policies have progressively restrained the official recommendations for BCG (2). In other words, as the global diffusion of SARS-CoV-2 has clearly followed economic highways connecting mainland China with high-developed areas in Europe and North America, we are dealing with higher occurrence of COVID-19 in highly-developed, highly interconnected areas rather than in countries deprived of vaccination policies for BCG. Similarly, as estimates from countries implementing a national policy for BCG vaccine (e.g. Russian Federation, African countries) have only recently showed the exponential increase that has previously affected Europe and North America, available figures are presumptively underestimating the actual morbidity of COVID-19 in countries with a national BCG policy (30).

In conclusion, there is no evidence that the BCG may protect people against infection with SARS-CoV-2. As a consequence, there is no evidence to recommend BCG vaccination for the prevention of COVID-19.