We conducted a prospective investigation of the effect of coffee consumption on coronary heart disease in 1130 male medical students who were followed for 19 to 35 years. Changes in coffee consumption and cigarette smoking during follow-up were examined in relation to the incidence of clinically evident coronary disease in comparisons of three measures of coffee intake--base-line intake, average intake, and most recent intake reported before the manifestation of coronary disease. Clinical evidence of coronary disease included myocardial infarction, angina, and sudden cardiac death. In separate analyses for each measure of coffee intake, the relative risks for men drinking five or more cups of coffee per day, as compared with nondrinkers, were approximately 2.80 for all three measures in the univariate analyses (maximum width of 95 percent confidence intervals, 1.27 to 6.51). After adjustment for age, current smoking, hypertension status, and base-line level of serum cholesterol, the estimated relative risk for men drinking five or more cups of coffee per day (using the most recent coffee intake measure), as compared with those drinking none, was 2.49 (maximum width of 95 percent confidence interval, 1.08 to 5.77). The association between coffee and coronary disease was strongest when the time between the reports of coffee intake and the coronary event was shortest. These findings support an independent, dose-responsive association of coffee consumption with clinically evident coronary heart disease, which is consistent with a twofold to threefold elevation in risk among heavy coffee drinkers.