Among 1,433 men of Japanese ancestry living in Hawaii with blood pressure measured at four different physical examinations over a 10-year period, 110 events of definite coronary heart disease (CHD) occurred during 11.6 years of subsequent follow-up. Each subject's mean blood pressure, the slope of the regression of his blood pressure on age, and the variance of blood pressure about this regression line were tested for association with subsequent incident definite CHD. Adjusted for mean systolic blood pressure (SBP), the variance of SBP was significantly associated with CHD (p < 0.001); however, the slope was not significantly associated with CHD. Variation in body weight was an independent risk factor for CHD. The effect of variation in SBP was significantly higher among men not taking antihypertensive medication; among men taking antihypertension medication, the standardized relative risk was 1.00. Comparing men in the highest quintile of SBP variation with those in the lowest quintile, the relative risk of CHD was 2.0 among all subjects and 5.3 among the 1,007 men not taking antihypertensive medication (95% confidence interval 1.8-15.4). Some of the beneficial effect of taking antihypertensive medication may have been due to reducing the effect of SBP variance rather than simply lowering the average SBP.