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


Cholesterol lowering and mortality: a meta-analysis

ACP J Club. 1993 Nov-Dec;119:65. doi:10.7326/ACPJC-1993-119-3-065

Source Citation

Davey Smith G, Song F, Sheldon TA. Cholesterol lowering and mortality: the importance of considering initial level of risk. BMJ. 1993 May 22,306:1367-73.



To evaluate the effectiveness of cholesterol-lowering treatment in reducing mortality, as a function of risk for coronary heart disease (CHD) mortality, by using meta-analysis.

Data sources

Relevant citations were identified by searching MEDLINE and Bath Information Data Services, scanning bibliographies of previous reviews, and consulting experts.

Study selection

Randomized, controlled, single-factor trials of cholesterol-lowering treatment with ≥ 6 months follow-up in which ≥1 death occurred were selected. 35 trials met the selection criteria.

Data extraction

Data on cholesterol-lowering treatment used, patient baseline characteristics including serum cholesterol level, and death from all causes and CHD were extracted. Pooled treatment effects were estimated by calculating a weighted average of odds ratios (ORs) for deaths in the treatment and control groups using the random effects method. Death rates in the control groups provided an index of the baseline risk for death from CHD.

Main results

A reduction in total mortality from cholesterol lowering was only seen in studies that included patients at high risk for CHD mortality. The ORs for total death in the high-, medium-, and low-risk groups (> 50, 10 to 50, and < 10 CHD deaths per 1000 person-y, respectively) were 0.74 (95% CI 0.60 to 0.92), 0.96 (CI 0.84 to 1.09), and 1.22 (CI 1.06 to 1.42), respectively. Weighted linear regression analysis showed that total mortality benefit increased with increasing baseline CHD mortality risk. At > 3% CHD mortality annually, the net benefit was positive. The increase in non-CHD mortality was confined to drug intervention trials (drug trials OR 1.21, CI 1.05 to 1.39; nondrug trials OR 1.02, CI 0.88 to 1.19). A chronologic, cumulative meta-analysis for total mortality by strata of risk for CHD death showed stable results over time.


Cholesterol-lowering treatments were effective in patients with a high initial risk for death from coronary heart disease and not in patients with medium or low initial risks. Total mortality was increased in trials involving low-risk patients. The increased mortality from causes other than coronary heart disease was evident only in drug intervention trials.

Source of funding: Department of Health, United Kingdom.

For article reprint: Dr. G. Davey Smith, Department of Public Health, University of Glasgow, Glasgow, United Kingdom, G128RZ. FAX 44-41-330-5018.


The crucial question for any therapy is "when does the benefit outweigh the risk?" This meta-analysis by Davey Smith and colleagues suggests net benefit for groups with a CHD mortality greater than 3% but a net harm less than this rate. However, the 95% confidence interval for this threshold is from {2% to 4% per year}. (Information supplied by author.)

To make the cost and inconvenience of therapy worthwhile, the risk for CHD mortality must be somewhat greater than this threshold. Davey Smith and colleagues show that a high-risk group with at least 5 deaths per 100 per year had a significant 26% reduction in all-cause mortality. For this group the number needed to treat is about 1 death averted per 16 people treated for 5 years; this effort seems worthwhile for both patient and clinician. Groups attempting to formulate practice guidelines should now focus on deciding who should be included in this high-risk group. The absolute level of CHD risk must be calculated, taking into account not just the cholesterol level but also risk factors such as history of CHD, age, gender, smoking, hypertension, and diabetes.

Two further points emerge from this meta-analysis. First, as was shown recently by Holme (1), greater reductions in cholesterol give greater benefit, hence once the decision is made to treat, it should be done well. Second, the apparent increase in non-CHD mortality appears only to occur for drug therapy at low levels of risk.

The cholesterol debate is not over, but this meta-analysis suggests that those at high risk benefit from treatment, those at low risk may be harmed by treatment, and those at moderate risk might best be enrolled in 1 of the large, randomized trials that are underway or planned.

Paul Glaziou, MBBS, PhD
University of Queensland Brisbane, Australia