To calculate the estimated relative and absolute risks for cardiovascular diseases according to alcohol consumption increase, we performed a cross-sectional study of men aged 40-79 years using the data from the National Survey of Circulatory Disorders 1990 in Japan. First, we investigated the association between alcohol intake and two types of biochemical variables: serum lipids and blood pressure, by using analysis of variance. Then, concerning the subjects aged 40-59 years, linear regression analyses between alcohol intake and these biochemical markers were done to calculate regression coefficients adjusted for confounding factors. Finally, we substituted these regression coefficients for previously reported regression coefficients in Cox proportional hazard models to calculate relative and absolute risks for coronary heart disease and stroke. Alcohol intake positively correlated to systolic and diastolic blood pressure (SBP and DBP) and high-density lipoprotein cholesterol (HDLC). However, there was no relation between alcohol intake and serum total cholesterol. Linear regression coefficients of SBP, DBP and HDLC were 2.25, 1.43, 2.70, respectively. We used two proportional hazard regression formulas of Japanese population: one for coronary disease (Circulation 89, 2533-39) and the other for stroke (NIPPON DATA80, a 14-year cohort study of randomly selected Japanese). The estimated relative risks of coronary heart disease and stroke were 0.89 and 1.06 for 1 "go" (23 g of ethanol) increment of alcohol intake. The results indicate that the mortality rates shift from 0.44 to 0.39 per 1,000 person-years for coronary heart disease and from 0.76 to 0.81 per 1,000 person-years for stroke in the representative Japanese population aged 40-59 years. The absolute risk of cardiovascular disease may be unchanged if we assume 23 g increment of ethanol intake in Japan because the risk reduction in coronary heart disease was nearly equal to the risk increment in stroke.