Estimating the effect of the run‐in on the power of the physicians' health study

Abstract

In the Physicians' Health Study, a randomized, placebo-controlled, double-blind trial of aspirin in the reduction of cardiovascular mortality and beta-carotene in decreasing cancer incidence, 33,223 subjects were eligible and willing to enter the trial. Instead of randomizing this group immediately, all participants received identical calendar packs that contained active aspirin and beta-carotene placebo. Following an 18-week run-in, only 22,071 subjects who remained eligible and willing and had taken at least 2/3 of their pills were randomized. We estimated the effect of the run-in as follows: pill taking compliance increased 20-41 per cent; sample size decreased 34 per cent; duration of follow-up decreased 7 per cent which resulted in a 7 per cent decrease in the expected event rate for the placebo group. To estimate these changes, we made assumptions about compliance and outcome risk for those excluded by the run-in. Our conclusion, however, about the net effect of the run-in on the power of the study remains constant across variations in a number of those assumptions. The power with the run-in, with 22,071 good compliers was typically higher, and never more than negligibly lower, than the power without the run-in, with 33,223 good and poor compliers. In addition, savings from enrolling 11,152 fewer subjects in the trial resulted from the use of the run-in.