OBJECTIVE--To examine the relationship between plasma cholesterol concentration and mortality from major causes of death.
Design: --Cohort study. SETTING--Civil service offices in London, England. PARTICIPANTS--There were 17,718 male civil servants aged 40 through 64 years at the time of study entry between 1967 and 1969. MAIN OUTCOME MEASURE--Mortality from major cause groups. RESULTS--There were 4022 deaths in the cohort over the 18 years of follow-up. Total mortality increased with cholesterol level, although mortality in the small group with very low cholesterol levels (5% of study population) was nonsignificantly higher (P greater than .5) than that of the remainder of the lowest quintile cholesterol group. Coronary heart disease mortality increased with increasing cholesterol concentration from the lowest levels (P less than .001 for trend). The cancer mortality rate in the group below the fifth centile of the cholesterol distribution was higher than in the remainder of the cohort for lung (P less than .001), pancreas (P = .05), liver (P = .09), and all smoking-related cancers (P = .02). Only for lung cancer was there a consistent inverse trend with cholesterol level (P less than .01). Rates of mortality due to non-neoplastic respiratory disease were inversely related to cholesterol level (P less than .001). Health state at the time of examination and socioeconomic position were related to cholesterol concentration--subjects in lower employment grades, with disease at baseline, with a history of recent unexplained weight loss, or who had been widowed had lower initial cholesterol levels. These associations largely accounted for the relationships between cholesterol level and noncardiovascular mortality. CONCLUSIONS--The inverse associations between plasma cholesterol concentration and mortality from certain causes of death seen in cohort studies could be because the participants with low cholesterol levels possess other characteristics that place them at an elevated risk of death.