Pike M C, Bernstein L
Cancer. 1985 Apr 15;55(8):1855-8. doi: 10.1002/1097-0142(19850415)55:8<1855::aid-cncr2820550835>3.0.co;2-9.
The conclusion drawn in a recent paper by Minton and associates (Cancer 1983; 51:1249-1253), that caffeine and an unsaturated fat diet significantly promoted dimethylbenz(a)anthracene (DMBA)-induced breast cancer in rats, is based on fallacious statistical reasoning. Minton and associates based their conclusion on the mean latency to first tumor appearance for rats diagnosed with tumors. However, evaluation of percentages of tumor-bearing rats yields contrary results. Using data from a totally negative hypothetical two-group experiment, we demonstrate how such "conflicting" results could arise from differential intercurrent mortality in the two groups. The correct statistical analysis of this hypothetical experiment allows for differential intercurrent mortality and no conflict arises; the data of Minton and associates need to be analyzed by these methods.
明顿及其同事在最近一篇论文（《癌症》，1983年；51卷：1249 - 1253页）中得出结论，即咖啡因和不饱和脂肪饮食显著促进了二甲基苯并(a)蒽（DMBA）诱发的大鼠乳腺癌，该结论基于错误的统计推理。明顿及其同事的结论是基于被诊断出患有肿瘤的大鼠首次出现肿瘤的平均潜伏期。然而，对患肿瘤大鼠百分比的评估却得出了相反的结果。利用一个完全阴性的假设性两组实验的数据，我们展示了两组间不同的并发死亡率是如何导致这种“相互矛盾”的结果的。对这个假设性实验进行正确的统计分析时考虑到了不同的并发死亡率，就不会出现矛盾；明顿及其同事的数据需要用这些方法进行分析。
