Statistical design of the Women's Health Trial

The National Cancer Institute has initiated a randomized trial to determine whether a low fat diet can reduce the incidence of breast cancer among women at increased risk for this disease. A feasibility trial involving 303 women has been conducted to examine recruitment strategies, study short-term compliance and, more generally, develop and refine trial procedures. The feasibility trial group also developed a detailed full-scale trial design plan, and randomization of participants to such a trial is currently underway. The purpose of this report is to describe the major design features of this Women's Health Trial, with particular emphasis on the statistical aspects of the design. The trial is planned to last 10 years and to include 32,000 participants. Of these 32,000 women, 12,800 will be assigned to a low fat diet intervention, and the other 19,200 will constitute a control group. The sample size of 32,000 arises from a range of estimates and assumptions pertaining to (a) the incidence of breast cancer at enrollment corresponding to selected eligibility criteria, (b) the relative risk of breast cancer as a function of a woman's history of dietary fat intake, (c) compliance assumptions in terms of average percent fat in the intervention and control groups as a function of time from randomization, and (d) rates of competing causes of death. These estimates and assumptions will be discussed, as will the robustness of the intended sample sizes to departures from such design assumptions.