The association between serum uric acid level and risk of coronary heart disease (CHD) over 21 years was investigated among 6411 middle-aged Japanese-American men who were participants in the Honolulu Heart Program. In an age-stratified Cox regression model, high serum uric acid (quartile 4 [>6.7 mg/dl], relative to quartile 1 [<5.0 mg/dl]) was a significant predictor of definite CHD (RR = 1.33; 95% confidence interval = 1.08-1.63; p = 0.006). However, when adjustment for confounders (body mass index, heavy alcohol consumption, triglycerides, diastolic blood pressure, blood glucose, and the ratio of animal to vegetable protein) was made, the association of high uric acid with coronary events was substantially reduced and became nonsignificant (RR = 1.14; 95% confidence interval = 0.92-1.42; p = 0.21). There was a significant interaction between serum uric acid and drinking status (P = 0.03). Thus, the risk of definite CHD associated with high urate levels (quartile 4), relative to low levels (quartile 1), was elevated in the abstainers (RR = 1.40; 95% confidence interval = 1.01-1.93; p = 0.02), but not in light and moderate drinkers (RR = 1.1 1; 95% confidence interval = 0.79-1.55; p = 0.58) or among the heavy drinkers (>40 ml of ethanol/day; RR = 0.57; 95% confidence interval = 0.27-1.21; p = 0.08). It is concluded that elevated uric acid may be associated with higher CHD among alcohol abstainers. Whether raised urate is an etiological factor for CHD or a manifestation of existing arterial disease in nondrinkers deserves further investigation.