Hawaiian men of Japanese ancestry followed up for 18 or more years after a baseline examination showed a quadratic distribution of death rates at different levels of serum cholesterol. Mortality from cancer progressively decreased and mortality from coronary heart disease progressively increased with rising levels of serum cholesterol. There was a positive association between baseline serum cholesterol levels and deaths from coronary heart disease at 0 to 6 years, 7 to 12 years, and 13 years and longer after examination. The inverse relationship between cancer and serum cholesterol levels was stronger in the first 6 years than in the next 6 years and, although still inverse, lost statistical significance after 13 years. Cancers of the colon and lung showed the strongest association with low baseline serum cholesterol levels, while gastric or rectal cancer failed to show this association. Organ specificity and persistence of the inverse association beyond 6 years suggest that the nutritional demands of cancers may not entirely explain the inverse association with some cancers. The quadratic distribution of deaths in this cohort remained after coronary heart disease, stroke, and cancer were removed from the analysis. For the entire period of observation, the lowest mortalities were found in men with serum cholesterol levels between 4.65 and 6.18 mmol/L (between 180 and 239 mg/dL). Manipulation of serum cholesterol levels below this level would not be desirable if this were to result in increased risk of death from cancer or other disease. This study does not rule out this possibility.