Covid-19 vs BCG: Statistical Significance Analysis

Abstract

We present an updated time-adjusted dataset and conclusions at Covid19 Time Zero + 5 month (04.06.2020). The conclusions of the original analysis reviewed and mostly maintained at this time point. With the data accumulated to date a statistical significance of the BCG immunization correlation hypothesis is evaluated with the conclusion that it has achieved the level of confidence. Several specific cases are discussed with respect to the induced immunity hypothesis.

Covid-19 vs BCG Universal Immunization: Statistical Significance at Six Months of Exposure

“…In addition to convincing, in our view, qualitative arguments in support of the correlation hypothesis, a statistical analysis of the correlation between UBIP record and current epidemiological impact in Section 5 confirms statistical significance of the correlation hypothesis with a confidence of at least 0.0001. The result is consistent with the analysis of statistical significance at the time of exposure of 3 months [6] and increases the confidence in the overall conclusion because under the null hypothesis samples taken at different time intervals should be considered as independent as well.…”

“…A similar pattern can be seen in the cases of Portugal (Group A) and Spain (Group B2), with the relative impact, at the time, of 0.064 and 0.22, respectively. These cases provide anecdotal but consistent over the period of four months [2,6] and clear support for the correlation hypothesis.…”

“…The null hypothesis in this case would dictate that immunization should carry no statistical significance for the epidemics impact, and therefore distributions in all of BCG group sample points (A, B, C) as defined above were described by a single distribution with, possibly, time-dependent parameters μ(t), σ(t) that can be estimated from the overall dataset. The basis for the analysis that follows is the observation of a strong disparity between the sample means in groups A and C as first reported in [1,2,6]. Under the null hypothesis, these cases should be treated as the difference between the means of randomly drawn samples of a given size, for which distribution parameters can be estimated with the sample-mean law.…”

“…To satisfy the null hypothesis, the means of samples A -C would need to independently, satisfy the normal distribution laws with the same mean μS that can be assumed to be equal to the overall dataset mean and σG defined by (1). Then following the arguments outlined in [6] one obtains:…”

“…In this research we compile and provide an analysis of a dataset of cases with different time of local arrival of the epidemics as defined in [2], adjusted and aligned at the time point of approximately six months after the first local exposure. The intent of this work was to analyze publicly available Covid-19 epidemiological data by reporting national and subnational jurisdictions with respect to the hypothesized induced immunity population-scale protection resulting from a universal BCG vaccination policy (UBIP), current or previous, and attempt both qualitative and quantitative analysis of the hypothesis of correlation between a current or previous UBIP in the jurisdiction and a milder scenario of Covid-19 epidemics; to verify the assumptions, results and conclusions of the earlier studies [1,2,[4][5][6] with a specific objective to determine, in a quantitative analysis, the constraints and confidence of the correlation and null hypotheses.…”