Decisive factors in the tolbutamide controversy

J. B. O'Sullivan and R. B. D'Agostino

No increased mortality trend attributable to tolbutamide is shown by an analysis of variance on logit-transformed data from the University Group Diabetes Program (UGDP) study. The UGDP's controversial finding of an increased rate with mortality subgrouped by "cardiovascular" causes is confirmed by the Biometric Committee's report, with reservations that failed to include overriding decisive factors. The basic problem is that inspected data set up the hypothesis (the increased cardiovascular mortality), and that the same data were used to test the hypothesis, so that resulting probability values no longer have the usual meaning. The problem was compounded by multiple testing of the data without adjusting the probability levels. When cardiovascular deaths were redefined as myocardial infarcts and sudden deaths, in an attempt to test a proposed etiologic inotropic hypothesis, no significant increase in cardiovascular mortality was found.