We compared observed and predicted changes in serum cholesterol in women with mammographic dysplasia who participated for 12 mo in a randomized, controlled trial of a low-fat, high-carbohydrate diet, in which total fat intake was reduced from an average of 37% of calories to 21% and carbohydrate intake increased from 44% to 52% of calories. Changes observed in serum cholesterol were greater than those predicted (by the formulas of Hegsted and Keys) for subjects with initial serum cholesterol values in the upper tertile of the population, were not significantly different from those predicted for subjects with baseline values in the middle tertile, and were significantly less than those predicted for subjects with initial values in the lower tertile. These results show that the usefulness of serum cholesterol as a marker of change in dietary fat intake in women depends on the distribution of serum cholesterol values in the population studied.