This study evaluated the association between changes in serum levels of high-density lipoprotein (HDL) cholesterol that occur under bezafibrate therapy and cardiac mortality in patients with coronary heart disease (CHD) who were enrolled in the Bezafibrate Infarction Prevention trial. We compared serum levels of HDL cholesterol in 1,509 patients in tertiles of on-treatment increments with those of 1,517 patients in the placebo group. Long-term follow-up showed that cardiac mortality decreased significantly with increasing tertiles of on-treatment change in HDL cholesterol (9.5%, 6.6%, and 6.3% of patients tertiles 1, 2, and 3, respectively, died of cardiac causes, p for trend = 0.02). In multivariate analysis, the magnitude of on-treatment increment of HDL cholesterol was independently associated with a decreased risk of cardiac death (hazard ratio 1.05, 95% confidence interval 0.74 to 1.47, for tertile 1; hazard ratio 0.73, 95% confidence interval 0.50 to 1.07, for tertile 2; hazard ratio 0.65, 95% confidence interval 0.43 to 0.97, for tertile 3, compared with placebo-allocated patients, p for trend = 0.02). Analyzing the association with change in HDL cholesterol as a continuous variable showed that the risk of cardiac mortality was decreased by 27% for every 5-mg/dl increase in on-treatment change in HDL cholesterol (p <0.001). In conclusion, although a definite secondary prevention effect of bezafibrate could not be found when examining the intervention group as a whole, our findings are consistent with a possible independent association between an increase in HDL cholesterol with bezafibrate therapy and a decrease in cardiac mortality. In appropriately selected patients, monitoring short-term response to bezafibrate therapy through change in HDL cholesterol may indicate the potential long-term benefit of the drug.