To the Editor.
—In their meta-analysis of the published literature on BCG vaccine, Dr Colditz and colleagues1 attempt to summarize BCG's protective effect by building a statistical model with variables for geographic latitude and validity scores. Geographic latitude is a variable without biological cogency (though some stretches have been attempted), yet it is retained in the model because it conveniently explains 41% of the variability. I suspect that its primary function is to attenuate the impact of the largest, most recent, and most damaging prospective study on the thesis that BCG vaccine is effective—the Madras study.2 We should note that the Madras study is larger than the next three studies in size combined1(Table 1) and found no protective effect. The Madras study was felt necessary at the time to finally prove the utility of BCG vaccine, yet its results are ignored without a call for additional study.
Bellin E. Efficacy of BCG Vaccine. JAMA. 1994;272(10):765. doi:10.1001/jama.1994.03520100029014