Over twenty years ago, we evaluated diet, serum cholesterol, and other variables in 1900 middle-aged men and repeated the evaluation one year later. No therapeutic suggestions were made. Vital status was determined at the 20th anniversary of the initial examination. Scores summarizing each participant's dietary intake of cholesterol, saturated fatty acids, and polyunsaturated fatty acids were calculated according to the formulas of Keys and Hegsted and their coworkers. The two scores were highly correlated, and results were similar for both: there was a positive association between diet score and serum cholesterol concentration at the initial examination, a positive association between change in diet score and change in serum cholesterol concentration from the initial to the second examination, and a positive association prospectively between mean base-line diet score and the 19-year risk of death from coronary heart disease. These associations persisted after adjustment for potentially confounding factors. The results support the conclusion that lipid composition of the diet affects serum cholesterol concentration and risk of coronary death in middle-aged American men. (N Engl J Med. 1981; 304:65–70.)