Quantitative methods in the tuberculosis epidemiology and in the evaluation of BCG vaccination programs

Abstract

Controversies concerning the protective efficacy of the BCG vaccination result mostly from the fact that quantitative methods have not been used in the evaluation of the BCG programs. Therefore, to eliminate the current controversy an unconditional requirement is to apply valid biostatistical models to analyse the results of the BCG programs. In order to achieve objective statistical inferences and epidemiological interpretations the following conditions should be fulfilled: data for evaluation have to be taken from epidemiological trials exempt from sampling error, since the morbidity rates are not normally distributed an appropriate normalizing transformation is needed for point and confidence interval estimations, only unbiased point estimates (dependent variables) could be used in valid models for hypothesis tests, in cases of rejected null hypothesis the ranked estimates of the compared groups must be evaluated in a multiple comparison model in order to diminish the Type I error in the decision. The following quantitative methods are presented to evaluate the effectiveness of BCG vaccination in Hungary: linear regression analysis, stepwise regression analysis and log-linear analysis.