The UGDP Study

To the Editor.— It would appear from the Cornfield-Schor exchange (217:1671-1687, 1971) that the major statistical controversy over the UGDP study is whether there is enough evidence of an imbalance between the placebo and tolbutamide patient groups at baseline to account for the imbalance in death patterns that appeared after five years. Schor has examined the published data and believes that the baseline characteristics were sufficiently different. Cornfield has examined both the published data and additional material in the study files and says they were not.I would like to point out that Cornfield has displayed highly significant evidence in favor of Schor's hypothesis.It is fairly easy to pursue imbalances in patient populations that have been randomly assigned to different treatments. One can always find some peculiar ancillary statistic (like the ages at death of the patients' grandmothers) for which the study was not balanced. However, there are