Two cohorts of men aged 40-59 in geographically defined areas of Croatia, former Yugoslavia, were enrolled in 1958 and cardiovascular risk factors were measured. The cohort of Dalmatia, on the Adriatic coast, was made up of 671 men (participation rate 98%), that of Slavonia, in the inland plain, of 696 men (participation rate 91%). Similar examinations were repeated 5 and 10 years after the entry one. A complete follow-up for vital status and causes of death was run for 25 years. Death rates in 25 years from coronary heart disease (CHD) were 90 per 1000 in Dalmatia and 148 per 1000 in Slavonia where also all other major causes of death and all-cause mortality rates were higher (642 vs 465 per 1000 in 25 years). Univariate and multivariate analysis relating ten risk factors to CHD mortality, showed that age and systolic blood pressure were significant predictors in Dalmatia; age, subscapular skinfold and body mass index (inverse) in Slavonia; and age, systolic blood pressure and subscapular skinfold in a multivariate model with lumped cohorts, where dummy variables identifying cohorts indicated a large unexplained extra-risk in Slavonia. In the pooled analysis cigarette smoking and serum cholesterol carried non significant coefficients. Blood pressure change in the first 10 years of follow-up added predictivity to a model exploring the delayed CHD morality occurring between year 10 and 25 of follow-up. Among the base-line measurements only serum cholesterol was significantly different between the two groups (188 mg/dl in Dalmatia and 199 mg/dl in Slavonia).(ABSTRACT TRUNCATED AT 250 WORDS)