The association of a casual reading of blood pressure (BP) in 1963 to subsequent 15-year cancer mortality was examined in a cohort of 10,059 middle-aged and elderly men. Systolic BP (SBP), but not diastolic BP, predicted significantly long-term cancer mortality occurring in 369 subjects. The covariate-adjusted relative risk (RR) estimated by Cox's proportional hazard model was 1.10 [95% confidence interval (CI), 1.00-1.21]. In patients aged less than 60 at the beginning of the study, increased cancer mortality was mainly observed in association with SBP of more than 150 mm Hg. In subjects aged 60 or above, the estimated RR was 1.21 (95% CI, 1.03-1.42). Exclusion of 40 men in whom diagnosis made prior to 1963 or death occurred through 1965 did not alter the results. An excess mortality in men who reported pharmacologic treatment for hypertension while under follow-up was fully accountable by their age, BP, and smoking habits. Analysis by site suggested that the association was mainly due to increased mortality from cancer of the digestive and genitourinary organs (estimated RR's, respectively, 1.20 and 1.26; 95% CI's, respectively, 1.03-1.39 and 0.99-1.59). Analysis by histologic subtype suggests an association with adenocarcinoma (RR = 1.19, 95% CI, 1.04-1.37) but not squamous cell or transitional cell carcinomas, myeloma, lymphomas, and leukemias.