SUMMARY The associations observed in the Coronary Drug Project between two baseline factors serum cholesterol level and number of cigarettes smoked per day and 5-year mortality in men who had a history of myocardial infarction are used to derive required sample sizes for future risk factor intervention trials in the secondary prevention of coronary heart disease. Consider a trial in which it is anticipated that the baseline cholesterol level will be 250 mg/dl in the control group and that a 20% reduction in this level to 200 mg/dl will be experienced by the treated group. Let it also be assumed that this reduction in cholesterol level will have the effect of immediately reducing the mortality risk to that corresponding to a patient with a baseline level of 200 mg/dl. Given a type I error rate, a, of 0.05 for a two-sided test, a type II error rate, fi, of 0.10, and a follow-up period of 5 years, the required size of the trial would be over 10,000 patients. A similar sample size would be required for a single-factor trial focusing on cigarette smoking cessation. The sample size might be reduced to 6000 for a two-factor trial with intervention on cholesterol and cigarette smoking simultaneously. These numbers increase two-to threefold when dropout patients and "dropin" patients (i.e., those in the control group who decide on their own to begin the intervention therapy) are considered and when other assumptions are made concerning the anticipated time required for the treatment to achieve maximum benefit with respect to the mortality or morbidity end point.MANY FACTORS -demographic, historical, clinical, biochemical, electrocardiographic and hygienic -are associated with the development of first and recurrent events of coronary heart disease (CHD). Factors that can be modified by diet, drugs or other prophylactic or therapeutic measures have been the foci during the past 2 decades for clinical trials in the primary and secondary prevention of CHD. Although the rationale for such trials lies largely in the observed association between a particular risk factor and manifestation of the disease, rather infrequently has information on the strength of such an association been used in deriving the required sample size for the trial. Rather, it has been customary to assume that the intervention will produce a 25%, 35% or even 50% reduction in the rate of CHD events compared with the control group, and then to derive the sample size required to detect such a reduction.Data from the Framingham study on the relationship of serum cholesterol level to first events of CHD have been used to determine how large the sample sizes would have to be for trials of dietary intervention in the primary prevention of CHD