Recommendations for identifying persons at high risk for coronary heart disease are based primarily on levels of total and low-density lipoprotein cholesterol. We examined whether, given knowledge of these levels, information on the high-density lipoprotein cholesterol level would improve the prediction of arteriographically documented coronary artery disease among 591 men. We found that even at levels of total and low-density lipoprotein cholesterol considered desirable, high-density lipoprotein cholesterol was inversely related to disease severity. For example, among the 112 men with a total cholesterol level <180 mg per dl, the mean occlusion score (representing the overall severity of disease) was 107 among men with a high-density lipoprotein cholesterol level ≤30 mg per dl vs a mean score of 52 among men with levels ≥45 mg per dl. Furthermore, men with low levels of both low-density lipoprotein cholesterol (<110 mg per dl) and high-density lipoprotein cholesterol (≤30 mg per dl) had as much occlusive disease as did men with high levels of both lipoprotein fractions. Given information on the ratio of high-density lipoprotein cholesterol to total cholesterol, the actual levels of the lipoprotein fractions did not improve disease prediction. Our results emphasize the importance of considering high-density lipoprotein cholesterol when assessing coronary heart disease risk.