OBJECTIVE--To study the association between number of cups of coffee consumed per day and coronary death when taking other major coronary risk factors into account. DESIGN--Men and women attending screening and followed up for a mean of 6.4 years. SETTING--Cardiovascular survey performed by ambulatory teams from the National Health Screening Service in Norway. PARTICIPANTS--All middle aged people in three counties: 19,398 men and 19,166 women aged 35-54 years who reported neither cardiovascular disease or diabetes nor symptoms of angina pectoris or intermittent claudication. MAIN OUTCOME MEASURE--Predictive value of number of cups of coffee consumed per day. RESULTS--At initial screening total serum cholesterol concentration, high density lipoprotein cholesterol concentration, blood pressure, height, and weight were measured and self reported information about smoking history, physical activity, and coffee drinking habits was recorded. Altogether 168 men and 16 women died of coronary heart disease during follow up. Mean cholesterol concentrations for men and women were almost identical and increased from the lowest to highest coffee consumption group (13.1% and 10.9% respectively). With the proportional hazards model and adjustment for age, total serum and high density lipoprotein cholesterol concentrations, systolic blood pressure, and number of cigarettes per day the coefficient for coffee corresponded to a relative risk between nine or more cups of coffee and less than one cup of 2.2 (95% confidence interval 1.1 to 4.5) for men and 5.1 (0.4 to 60.3) for women. For men the relative risk varied among the three counties. CONCLUSIONS--Coffee may affect mortality from coronary heart disease over and above its effect in raising cholesterol concentrations.