The BCG controversy. A methodological and statistical reappraisal.

Because of the conflicting results of eight major controlled trials, BCG vaccination against tuberculosis remains controversial despite more than 50 years of use. Suspecting a methodological source for the controversy, we reviewed the scientific and statistical quality of each trial. The analysis showed that (1) although biased allocation of the vaccine appeared an unlikely explanation for the disparate results, adequate demonstration of unbiased detection of tuberculosis was available only for the three trials reporting 75% or greater protective efficacy; and (2) in most trials reporting low efficacy, the results had wide confidence intervals that could not exclude high efficacy, but the trials reporting high efficacy all had narrow confidence intervals that excluded low efficacy. Because the trials with the best methodological quality and greatest statistical precision reported high efficacy, the evidence suggests that BCG can confer a high degree of protection against tuberculosis and that bias or inadequate statistical power may have contributed to the conflicting data.