To evaluate prospectively the effect on mortality of wine drinking in Eastern France, we conducted an analysis on 34,014 consecutive middle-aged men coming for a comprehensive health appraisal between 1978 and 1983. We evaluated education, physical activity, smoking, and drinking habits by a questionnaire. Electrocardiogram, blood pressure, serum total cholesterol, and gamma-glutamyltransferase level were routinely measured. Seventy-seven per cent of the subjects drank wine; there was little difference between social classes in this proportion. We evaluated mortality over 10-15 years of follow-up. We estimated the relative risk (RR) of death by Cox proportional hazard models using nondrinkers as the reference and adjusting for six covariables. For an intake of 22-32 and 33-54 gm of alcohol per day, the RR of all-cause death was 0.70 [95% confidence interval (CI) = 0.59-0.82] and 0.76 (95% CI = 0.66-0.87), respectively. The lower mortality resulted from fewer deaths from cardiovascular disease and cancer. Above 128 gm per day of alcohol consumption, the RR was 1.37 (95% CI = 1.16-1.61). A moderate intake of wine (2-5 glasses per day) was associated with a 24-31% reduction in all-cause mortality, a proportion that was similar for smokers, ex-smokers, and nonsmokers.