Background: A generally held belief is that cholesterol concentrations should be kept low to lessen the risk of cardiovascular disease. However, studies of the relation between serum cholesterol and all-cause mortality in elderly people have shown contrasting results. To investigate these discrepancies, we did a longitudinal assessment of changes in both lipid and serum cholesterol concentrations over 20 years, and compared them with mortality.
Methods: Lipid and serum cholesterol concentrations were measured in 3572 Japanese/American men (aged 71-93 years) as part of the Honolulu Heart Program. We compared changes in these concentrations over 20 years with all-cause mortality using three different Cox proportional hazards models.
Findings: Mean cholesterol fell significantly with increasing age. Age-adjusted mortality rates were 68.3, 48.9, 41.1, and 43.3 for the first to fourth quartiles of cholesterol concentrations, respectively. Relative risks for mortality were 0.72 (95% CI 0.60-0.87), 0.60 (0.49-0.74), and 0.65 (0.53-0.80), in the second, third, and fourth quartiles, respectively, with quartile 1 as reference. A Cox proportional hazard model assessed changes in cholesterol concentrations between examinations three and four. Only the group with low cholesterol concentration at both examinations had a significant association with mortality (risk ratio 1.64, 95% CI 1.13-2.36).
Interpretation: We have been unable to explain our results. These data cast doubt on the scientific justification for lowering cholesterol to very low concentrations (<4.65 mmol/L) in elderly people.