Comparison of baseline characteristics and mortality experience of participants and nonparticipants in a randomized clinical trial: the Physicians' Health Study

Although subjects randomized into clinical trials tend to have different mortality experiences from those not randomized, few studies have examined how baseline characteristics may explain these differences. We used the recruitment experience of the Physicians' Health Study (PHS) to describe and compare subgroups of subjects ultimately randomized-or not-into the PHS. A total of 112,528 male physicians ages 40-84 years responded to letters of invitation and baseline questionnaires. sent to 261,248 subjects. Baseline information was collected on eligibility criteria, plus lifestyle and clinical risk factors. Total, cardiovascular, cancer, and other mortality were determined through the National Death Index after a mean follow-up of 5.39 years. Respondents had 19 and 19% lower total and cardiovascular mortality rates than nonrespondents. Similarly, willing respondents were 19 and 16% less likely to die than unwilling respondents. However, much of this difference in mortality was explained by disease and lifestyle factors. Respondents who were eligible for the PHS had significantly lower age-adjusted relative risks (RRs) that were attenuated but remained substantially below 1 upon multivariate adjustment in models for total (RRs from 0.48 to 0.79), cardiovascular (from 0.40 to 0.85), and cancer mortality (from 0.55 to 0.87). Finally, a nearly halving in the age-adjusted risk of total and cause-specific mortality among men completing a run-in and randomized into PHS compared with those not randomized was nominally altered upon adding all covariates into multivariate models. In conclusion, a difference in mortality rates according to willingness to participate in a trial was explained by disease and lifestyle factors. In contrast, diseases and risk factors explain some, but not all, of the lower mortality rates of physicians based on eligibility status and their ability to complete a run-in phase. (C) 2002 Elsevier Science Inc. All rights reserved.