Current issues of ACP Journal Club are published in Annals of Internal Medicine


Serum cholesterol level and mortality in U.S. men

ACP J Club. 1993 Jan-Feb;118:22. doi:10.7326/ACPJC-1993-118-1-022

Related Content in the Archives
Correction: Serum cholesterol level and mortality in U.S. men

Source Citation

Neaton JD, Blackburn H, Jacobs D,et al. Serum cholesterol level and mortality findings for men screened in the Multiple Risk Factor Intervention Trial. Arch Intern Med. 1992 Jul;152:1490-500.



To determine whether total serum cholesterol levels are associated with causes of death.


Cohort of men screened for the Multiple Risk Factor Intervention Trial (MRFIT) and followed for a mean of 12 years.


22 clinical centers in 18 cities in the United States.


350 977 men aged 35 to 57 years with no previous hospitalizations for myocardial infarction and not taking medication for diabetes. Recruitment was from 1973 to 1975.

Assessment of risk factors

Blood samples taken from nonfasting participants assessed total serum cholesterol levels at study start. Cholesterol levels were stratified into 4 levels (< 4.14, 4.14 to 5.15, 5.17 to 6.18, and ≥ 6.21 mmol/L [< 160, 160 to 199, 200 to 239, and ≥ 240 mg/dL]).

Main outcome measures

All deaths listed in the National Death Index (1979 to 1986) and Social Security Administration (1973 to 1986). Death certificates were coded using the International Classification of Diseases, Ninth Revision.

Main results

Mean level of serum cholesterol was 5.53 mmol/L. 6%, 31%, 39%, and 24% of the study population were in the 4 serum cholesterol level groups as listed above. 21 499 deaths occurred. A serum cholesterol level of 3.15 mmol/L was associated with a minimum risk for death, with mortality increasing for lower and higher levels. Relative risk (RR) for death from cardiovascular disease for those in the highest level compared to the lowest was 2.26 (95% CI 2.00 to 2.56). For death from coronary heart disease, the association was stronger (RR 3.03, CI 2.58 to 3.55). The risk for death from nonhemorrhagic stroke was greatest among those in the highest cholesterol level, whereas men in the lowest cholesterol group were at greatest risk for intracranial hemorrhage. The lowest level was also associated with a higher risk for death from cancer of the liver, pancreas, lung, and lymphatic and hematopoietic systems. Overall, the inverse association between serum cholesterol levels and most cancers weakened with increasing follow-up but did not disappear. Low cholesterol level was not associated with death from accidents or homicide, but the risks were increased for deaths from suicide and the alcohol-dependence syndrome.


The association of total serum cholesterol levels with causes of mortality in men varied in direction and strength.

Source of funding: National Heart, Lung, and Blood Institute.

For article reprint: Dr. J.D. Neaton, Division of Biostatistics, School of Public Health, University of Minnesota, 2221 University Avenue, S.E., Suite 200, Minneapolis, MN 55414, USA. FAX 612-626-0660.


If the 1980s brought a consensus that lowering serum cholesterol reduces one's risk for coronary heart disease, the 1990s have heralded a re-appraisal of the presumed safety of cholesterol-lowering treatments. The possibility that cholesterol reduction may increase mortality from causes other than heart disease began to receive serious attention in 1990 when meta-analyses of cholesterol-lowering trials showed that coronary heart disease benefits may be offset by increased rates of noncoronary heart disease deaths (1, 2). These studies found that cholesterol-lowering treatments were associated, albeit inconsistently, with increased cancer mortality and death from suicides and violence. The absence of preformed hypotheses and biological mechanisms raises suspicion that such findings may be spurious, but the strong design of the clinical trials encourages attribution of increased noncoronary heart disease mortality to the treatment prescribed. The 3 studies abstracted here pertain to these controversial issues.

The report of the Coronary Primary Prevention Trial by the Lipid Research Clinics Investigators provides an additional 6 years of follow-up past the planned 7.4 year duration of the trial. Because the randomized treatments (cholestyramine and placebo) were no longer being provided, the extended follow-up can only help to determine whether beneficial or adverse effects increase or diminish over time. Although the in-trial reduction in coronary heart disease disappeared with follow-up, the absence of any increase in cancer mortality or death from suicide or trauma is reassuring. Nevertheless, the men who were treated experienced small but nonsignificant increased rates for certain tumors.

The other 2 papers use the alternative strategy of examining epidemiologic associations between serum cholesterol concentration and mortality from various causes. Neaton and colleagues' study of screenees for the Multiple Risk Factor Intervention Trial is the largest of its kind, having several hundred thousand participants and over 20 000 deaths to study. Lindberg and colleagues provide the most thorough examination to date of the relation between serum cholesterol and non-illness-related causes of death (suicides, accidents, and violence). Nevertheless, because cholesterol levels were not manipulated, these epidemiologic studies cannot establish whether any observed associations between cholesterol level and mortality are causally linked.

The 2 papers by Neaton and Lindberg and their colleagues, as well as a recent meta-analysis of 18 other epidemiologic studies (3), focus on the relation between low serum cholesterol and increased mortality from causes other than coronary heart disease. They permit refinement of our emerging understanding of this unanticipated phenomenon. Specifically, the increased risk for cancer death in men with low cholesterol is reproducible and appears to be specific for lung, pancreas, and liver cancer; leukemia; and lymphoma (and not colon or breast cancer). Low serum cholesterol is associated with increased risk for death from hemorrhagic stroke, chronic obstructive lung disease, and cirrhosis. Finally, men with low cholesterol die at increased rates from other causes, principally suicide. The importance of many of these adverse associations is limited, however, by recognition that they appear to exist only when the cholesterol level is below 4.14 mmol/L (160 mg/dL).

Jacobs (3) and Lindberg and their colleagues provide information on similar relations in women. Notably, high serum cholesterol in women is a less potent predictor of cardiovascular disease, yet low cholesterol, as in men, is associated with an approximately 40% increase in death from assorted noncoronary causes.

Although it is widely agreed that high serum cholesterol promotes atherosclerosis and heart disease, Neaton, Jacobs (3), and their colleagues feel that low cholesterol is not causally related to deaths from various noncoronary causes. Instead, as described at length, the investigators postulate that various confounding factors explain these adverse associations. Among these confounding factors, however, preclinical cancers, alcohol consumption, and smoking are shown not to account for low cholesterol. Data, including mortality associations, are not available to support other postulated mediating factors (e.g., nutritional status, dietary β-carotene, and socioeconomic factors).

Together, the findings of the epidemiologic studies indicate that associations between low cholesterol and increased noncoronary heart disease mortality are robust. Although low cholesterol may be only indirectly related to cancer, liver and chronic lung disease, and trauma- or suicide-related mortality, a causal role cannot be ruled out at present. The findings of clinical trials, despite some reassuring data from the new Coronary Primary Prevention Trial, also indicate that cholesterol reduction may increase noncoronary heart disease mortality. It is becoming increasingly difficult, therefore, to assert that serum cholesterol reduction is always safe, and, indeed, Hulley and colleagues (4) have called for a suspension of screening and treatment for primary heart disease prevention (i.e., they suggest that cholesterol-lowering treatment should be reserved for very-high-risk individuals). Nonetheless, a new consensus regarding the wisdom of widespread cholesterol reduction has not been reached. For the time being, a reasonable approach might involve continued general adherence to present treatment recommendations but with added hesitancy to treat isolated hypercholesterolemia.

Matthew F. Muldoon, MD
University of Pittsburgh Pittsburgh, Pennsylvania, USA