Cause-specific mortality-rates were calculated in 17 718 men aged 40-64 years who participated in the Whitehall Study. Over a 7 1/2 year follow-up, total mortality showed a J-shaped relation to the plasma cholesterol concentration measured at entry to study. This shape resulted from a strong positive relation of plasma cholesterol with deaths from coronary heart-disease (CHD) combined with an opposite (inverse) relation between plasma cholesterol and deaths from other causes. Cancer mortality was 66% higher in the group with the lowest plasma cholesterol than in the group with the highest plasma cholesterol. Further analysis showed that this inverse association between plasma cholesterol and non-CHD deaths was confined to the first 2 years of follow-up; beyond this time total mortality and cholesterol level were evenly and positively correlated. Analysis of data from the Framingham study revealed the same phenomenon, which is presumed to result from the metabolic consequences of cancer which was present but unsuspected at the time of examination.