Does Latent Tuberculosis Lead to a Spurious Correlation between BCG and COVID-19 Mortality?


 Background: Many factors have been suggested to confound coronavirus disease 2019 (COVID-19) studies, and BCG studies have been criticized for not adjusting for many confounders. We conducted this study to analyze the presumed effectiveness of the Bacillus Calmette–Guérin (BCG) vaccine in decreasing the COVID-19 mortality rate, and to answer the question of whether this is confounded by latent tuberculosis (LTB) prevalence.Materials and methods: We chose sixty-nine malaria-free countries with different BCG vaccination policies. TB prevalence was considered as a proxy for LTB. The BCG, TB prevalence, and COVID-19 mortality data are publically available. Contingency coefficients (C.C.) and a ROC analysis were used to assess the relationship between TB prevalence and BCG status, and identify cutoff points in each BCG group category. A stem–leaf plot was also used to explore the data’s apparent behavior concerning COVID-19 in relation to the BCG groups.Results: TB prevalence was significantly associated with BCG status. The BCG vaccination status apparently had a relationship with BCG status.Conclusions: TB is suggested to have a confounding effect on BCG results, leading to a spurious correlation between BCG and COVID-19 mortality.


Introduction
The WHO currently recommends that, in countries with a high tuberculosis (TB) burden, a single dose of the Bacillus Calmette-Guérin (BCG) vaccine should be provided to all infants as soon as possible after birth as part of childhood immunization programs. In countries with low TB incidence rates, the provision of the BCG vaccine may be limited to neonates and infants in recognized high-risk groups, or older children who are negative for TB infection according to the tuberculin skin test (TST). Despite clear evidence and the WHO's recommendations, however, global BCG administration practices appear to be arbitrary. , Among 180 countries, 154 reported universal BCG vaccination, 20 reported having had a national BCG policy for everyone in the past, and the remaining six reported selective vaccination for persons in high-risk groups. BCG coverage ranged from 53-99%; the coverage was <80% in six high-incidence countries.
Studies on the BCG vaccine's e cacy against TB have been confounded by the cross-reactivity of antigens and absence of measures for excluding latent infection. Early in this pandemic, it was also suggested by a few studies that latent tuberculosis (LTB) could mitigate COVID-19 morbidity and mortality. 11, 16, 17, 18, 19, 20, 21, Some tested TB prevalence 16,20,22 as a proxy for LTB, while others considered LTB estimates that might cross-react with BCG and other mycobacteria. Other studies considered positivity in the TST/interferon-gamma release assay (IGRA) as a measure of the potential protective effect of the resident populations' exposure to Mycobacterium spp., whether from BCG vaccination or as a result of exposure to environmental mycobacteria. 17 BCG studies are likely to be confounded by LTB, since low-TB countries do not implement vaccination, and countries with high TB prevalence implement BCG programs. About one-quarter of the world's population has LTB, 24,25 making the latter an important factor to adjust for.
Furthermore, malaria's confounding effect was not adjusted for in most of these studies; this is of special importance because malaria is highly endemic in most high-TB countries.
The study of risk and confounding factors is very important since previous studies have shown hugely disparate case-fatality rates, in spite of some reporting signi cant ndings. For example, disparities exist among both countries with low LTB incidence or BCG coverage and those with high LTB incidence or BCG coverage. 23 Regarding malaria's in uence on COVID-19 mortality, it was suggested by a few preliminary studies to reduce it. 22,26 In the context of the con icting evidence regarding BCG's signi cant correlations, the growing evidence of a role of LTB and malaria, and a lack of literature de ning LTB's confounding effect in malaria-free countries, we suggest that a confounding effect exists, as a background hypothesis for this study. This study is important at this time because several clinical trials are underway to evaluate the e cacy of BCG vaccination for COVID-19.

Materials And Methods
It was not appropriate or possible to involve patients or the public in this work; we used data summated at the general practice level and related, publically published morbidity and mortality statistics.
The main objectives of this study were based on the hypothesis that the decreased mortality rate in BCG studies is related to the in uence of LTB rather than the BCG effect, and that BCG studies have been confounded by LTB prevalence. We designed this study to look for an association between TB prevalence (re ected as LTB prevalence) and BCG policy status in the absence of a confounding effect from malaria, through restricting the sample to malaria-free countries. Furthermore, the stem-leaf graphical plot method was proposed to illustrate (apparent) BCG group behavior regarding COVID-19 mortality.
We selected countries that have achieved at least three consecutive years of zero indigenous cases of malaria.
The total number of countries was 69, as shown in Appendix A. Countries eligible to apply for WHO certi cations of malaria-free statuses were included. 27 Countries and territories with populations of less than 1 million were excluded and are listed in Appendix B.
The data for the TB prevalence, BCG, malaria, and COVID-19 mortality are publicly available (references are listed BCG with a previous booster (JSC2-BCG), multiple current BCGs (MC-BCG), one previous BCG (JP-BCG), and no previous or current BCG (NP/C-BCG or BCG: 0 or the BCG control group) (Appendix A).
Contingency coe cients (C.C.) and ROC analysis were used to test the relationship between TB prevalence and BCG status, and to identify the cutoff points in each BCG group category. A stem-leaf plot was also used to explore the data's apparent behavior concerning COVID-19 mortality rate for different BCG groups. All the statistical operations were performed using the statistical package SPSS, ver. 22.
Results And Findings Table 1 shows the distribution of TB prevalence during 2011-2018 in three categorized BCG groups, as well as a comparison that is signi cant in terms of the contingency coe cient of the re ected relationship between the preceding factors; they had either a random or constrained distribution. The results show a highly signi cant relationship at P<0.01, indicating that a meaningful constrained distribution is accounted for regarding the studied factors.
Regarding the low-TB-prevalence category the distribution of the NP/C-BCG group was 83.30% (highest), followed by 81.0% of the JP-BCG group, and the lowest was 6.3% of the JSC2-BCG group.
Regarding the moderate-TB-prevalence category, the distributions were 75.0% of the JSC2-BCG group (highest), and both JP-BCG and MC-BCG were 14.3% (lowest).
Regarding the high-TB-prevalence group, the distributions were 71.4% for MC-BCG and 0.00% for NP/C-BCG (BCG: 0, or the BCG control group).   Figure 1 show estimates of the area of the trade-off between the sensitivity and speci city; sensitivity is plotted against a complementing speci city outcome to examine the trade-off, which is called the ROC curve.
The signi cance level for the testing area was under 50% guideline, with a 95% con dence interval of all the probable combination pairs for the four statuses of the BCG groups under the proposed guideline group (the control) due to NP/C-BCG group status.  Figure 1).
Sig. (*) Asymp. 95% C.I. The stem-leaf graphical plots clearly illustrate the apparent behavior of COVID-19 mortality within the BCG groups ( Figure 2); they show that the ranking of mortality rates by BCG group status gives the impression of high mortality within the no-BCG-vaccination groups, and low mortality in countries with BCG vaccination.

Discussion
Many BCG studies showing a signi cant relationship between BCG and a reduction in COVID-19 mortality and/or morbidity have been criticized for not considering confounding factors, and simply assessing the differences in the incidence/mortality of COVID-19 based on having or not having BCG vaccination policy, as well as sharing the same sources of information with questionable data accuracy. 25,28,29,30 These studies were, therefore, A confounding factor may mask an actual association or falsely demonstrate an apparent association between a study's variables where no real association between them exists. 36 This confounder may lead to the overestimation of the true association between an exposure and outcome. 37 One of limitations of this study is that it did not address whether there is an overestimation of the BCG effect, since it focused on the relationship between TB prevalence and BCG status Furthermore, it is limited by not measuring BCG vaccination coverage rate, stage of epidemic, socio-economic differences, and differences in practicing of preventive measures to contain the disease, etc.
The e cacy and effectiveness of BCG vaccination against TB have been found to differ considerably between studies and populations. 38 The BCG vaccine has a documented protective effect against TB meningitis and disseminated TB in children, but it prevents neither primary infection nor, more importantly, the reactivation of a latent pulmonary infection, which is the principal source of bacillary spread in the community. The impact of BCG vaccination on the transmission of Mycobacterium TB is, therefore, limited. 39 Despite BCG being effective in 50% of the target population, much controversy surrounds its effect on mild forms of infection, as well as the duration of its effect. 25 According to the WHO's recommendations, countries with low TB burdens may limit BCG vaccination to infants in high-risk groups (or TST-negative older children) and adults at high risk for occupational TB exposure and who are TST negative. 38 Table 1 shows a highly signi cant association between TB prevalence and certain BCG groups: countries not implementing BCG vaccinations had low TB prevalence, and vice versa (p value = 0.000). Table 2 and Figure 1 show the ROC analyses indicating that BCG group is signi cantly associated with corresponding TB prevalence.
These results con rm an association between TB prevalence and BCG status.
The nding of this study that BCG status is highly associated with TB prevalence leads us to conclude that BCG studies can be easily confounded by LTB. The ranking of mortality rates within BCG group statuses shown in a stem-leaf plot ( Figure 2) follows the rank of association between TB prevalence and BCG status. This gives the impression of high mortality within groups with no BCG vaccination and low mortalities within countries with BCG vaccination.
This could apply to all BCG studies not adjusting for LTB.
In countries that do not undertake vaccination, confounding occurs simply because of a possible low TB prevalence, giving a false impression that not administering BCG is the cause of high mortality. Another possible confounder is previous TB prevalence, since the immunity generated by TB lasts for a certain period of time. We tried to control for this by considering the highest available TB prevalence during 2011-2018. Availability of data and materials: We used publically available data. Patients were not involved.
All data generated or analysed during this study are included in this published article [and its supplementary information les].
• Competing interests: There is no con ict of interest to be declared.
• Funding: No funding source to be declared.
• Author contributions : The author wrote the manuscript and provided data, conducted the statistical analyses and reviewed the nal manuscript.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. 021221appeneces.docx