Non-high-density lipoprotein (HDL) cholesterol (total cholesterol [TC] minus HDL cholesterol) has been suggested as the preferred lipid fraction to predict cardiovascular disease. We compared the ability of lipids, lipoproteins, the ratio of total to HDL cholesterol (TC/HDL), and non-HDL cholesterol to predict fatal coronary heart disease (CHD) and cardiovascular disease in 1,386 women and 1,094 men (mean age 69 years). After 10 years, there were more deaths in men (n = 310) than women (n = 268), but the proportions of deaths attributed to CHD (23% and 25%, respectively) and cardiovascular disease (48% and 47%) were similar. In men, age-adjusted values for non-HDL cholesterol, TC/HDL ratio, and triglycerides each predicted a significantly increased risk of CHD and cardiovascular disease; none of these associations was independent of pack-years of smoking, systolic blood pressure, fasting plasma glucose, body mass index, and physical activity. In women, age-adjusted non-HDL cholesterol levels did not predict CHD or cardiovascular disease events before or after adjusting for these covariates and for estrogen replacement therapy. In women, only the ratio of TC to HDL cholesterol predicted CHD and cardiovascular disease deaths independent of estrogen use and other risk factors. Observed associations were sensitive to time, being evident in women at 3 and 5 years, and lost thereafter, but not apparent before 10 years in men. Thus, non-HDL cholesterol is not superior to individual lipids, lipoproteins, or their ratios in the prediction of cardiovascular death in older adults.