Cholesterol-lowering drugs reduce the incidence of nonfatal and fatal myocardial infarction in primary and secondary prevention [6, 7]. Whether cholesterol-lowering interventions affect the incidence of stroke remains uncertain. A meta-analysis of randomized trials found no reduction in stroke-related morbidity or mortality rates, but this review did not include more recent trials that used HMGcoA reductase inhibitors (3-hydroxy-3-methylglutaryl coenzyme A) [8].

We conducted a systematic review of randomized, controlled trials to assess the efficacy of HMGcoA reductase inhibitors compared with the efficacy of other cholesterol-lowering interventions to reduce the incidence of stroke.

Methods

Top Methods Discussion Author & Article Info References

We did a systematic search of the MEDLINE and EMBASE databases and previously published meta-analyses [9-14]. We searched the reference lists of identified papers for citations to additional relevant articles, and we identified all randomized, controlled trials published through October 1996 that compared dietary or pharmaceutical cholesterol-lowering interventions with placebo or usual diet.

We considered studies to be eligible for our meta-analysis if they randomly assigned participants to active treatment or placebo and reported the occurrence of nonfatal and fatal strokes. We included trials regardless of whether they focused on primary or secondary prevention of coronary heart disease and regardless of whether they used unifactorial or multifactorial interventions. We excluded trials that were restricted to patients who had previously had stroke because these trials provided insufficient data on the underlying pathophysiologic cause of stroke and therefore had study samples that were too heterogenous. We also excluded trials of cholesterol-lowering interventions in heart transplant recipients [15]. We did not consider transischemic attacks as study end points because they were of limited importance and because consistent criteria for ascertainment across trials were lacking.

We assessed the methodologic quality of included trials with respect to the following variables: 1) proportion of participants with complete followup, 2) concealment of randomization, 3) blinding of patients and caregivers, and 4) blinding for outcome assessment. We created a dichotomous variable for each of the four quality variables and added scores to yield a summary scale ranging from 0 to 4, with 0 as the lowest and 4 as the highest quality score. Because too few studies reported data on death from stroke, we combined fatal and nonfatal strokes into a single category.

To pool treatment effects across studies, we calculated a weighted average risk ratio of all outcomes in the treatment and control groups by using a random-effects model [16]. We tested for heterogeneity with the Breslow-Day test [17]. Because a previous meta-analysis (Bucher and colleagues. In preparation) suggested that HMGcoA reductase inhibitors had a greater beneficial effect than other antilipidemic interventions on major morbidity and mortality, we examined the effects of major drug classes separately. However, for completeness of data presentation, we report overall estimates and results of overall tests of heterogeneity. We tested the difference in combined estimates of intervention type and used the z-score for each type of intervention by dividing the difference in the subgroup summary log relative risk by the SE of the difference. We grouped trials into four categories: trials of HMGcoA reductase inhibitors (n = 8 [6, 7, 18-23]), trials of fibrates (n = 5 [24-28]), trials of resins (n = 3 [29-31]), and trials of dietary interventions (n = 10 [31-40]). Five trials studied niacin [25], the combination of niacin and clofibrate [41], hormones [42-44], fish oil [45], and ileal bypass surgery [46].

Data Synthesis

We identified 64 randomized, controlled trials that reported data on mortality. Of these 64 trials, 30 provided data on the incidence of nonfatal and fatal stroke. Two trials were excluded because they studied patients who had previously had stroke [47-49]; this left 28 trials for analysis. Two of these 28 trials had multiple treatment arms [25, 26, 42-44]. In total, the intervention group consisted of 49 477 patients and the control group consisted of 56 636 patients. Table 1 and (Table 3) describes the 28 included trials.

Nonfatal and Fatal Stroke

The risk ratio for nonfatal and fatal stroke for all 28 trials was 0.95 (95% CI, 0.86 to 1.05; test of heterogeneity, P > 0.2) (Table 2). For HMGcoA reductase inhibitors, the risk ratio for nonfatal and fatal stroke was 0.76 (CI, 0.62 to 0.92; test of heterogeneity, P > 0.2) (Figure 1). This estimate was statistically significantly different (P = 0.01) from the summary estimate for all other cholesterol-lowering interventions (risk ratio, 1.02 [CI, 0.91 to 1.15]). Pooled estimates for fibrates (risk ratio, 1.12 [CI, 0.84 to 1.48]; test of heterogeneity, P > 0.2), resins (risk ratio, 1.07 [CI, 0.57 to 2.00]; test of heterogeneity, P > 0.2), and dietary interventions (risk ratio, 0.98 [CI, 0.82 to 1.18]; test of heterogeneity, P > 0.2) suggested that these interventions had no effect on the incidence of nonfatal and fatal stroke.

View larger version (14K): [in this window] [in a new window]   Figure 1. Risk ratio and summary estimates with 95% CIs for nonfatal and fatal stroke in randomized, controlled trials comparing HMGcoA (3-hydroxy-3-methylglutaryl coenzyme A) reductase inhibitors with placebo.P > 0.2 (Breslow-Day test for heterogeneity). 4S = Scandinavian Simvastatin Survival Study; CARE = Cholesterol and Recurrent Events Trial; KAPS = Kuopio Atherosclerosis Prevention Study; PMG = Pravastatin Multinational Study Group for Cardiac Risk Patients; REGRESS = Regression Growth Evaluation Statin Study; WOSCOPS = West of Scotland Coronary Prevention Study Group.

Death from Coronary Heart Disease and Overall Mortality

For death from coronary heart disease; the risk ratio for all trials was 0.87 (CI, 0.79 to 0.94; test of heterogeneity, P = 0.003). With HMGcoA reductase inhibitors, the risk ratio was 0.69 (CI, 0.59 to 0.80; test of heterogeneity, P > 0.2); this was significantly different from the risk ratio for all other trials of cholesterol-lowering interventions (risk ratio, 0.90 [CI, 0.83 to 0.98]) (P = 0.003 for the difference). For groupings of fibrates, resins, and dietary interventions, risk ratios for death from coronary heart disease showed no statistically significant reduction of coronary heart disease mortality.

The risk ratio for overall mortality was 0.94 (CI, 0.84 to 1.06; test of heterogeneity, P < 0.001). With HMGcoA reductase inhibitors, the risk ratio for death from all causes was 0.80 (CI, 0.71 to 0.90; test of heterogeneity, P > 0.2); this was statistically significantly different from the risk ratio for death from all causes with all other cholesterol-lowering interventions (risk ratio, 0.97 [CI, 0.86 to 1.10]) (P = 0.03 for the difference).

Discussion

Top Methods Discussion Author & Article Info References

This systematic review shows that decreasing cholesterol levels with HMGcoA reductase inhibitors reduces the risk for nonfatal and fatal stroke. Our inference that this class of drugs results in larger reductions in the incidence of nonfatal and fatal strokes compared with other antilipidemic interventions is strong [50]. The reasons for this include the fact that our hypothesis about differential drug class effects preceded our analysis and was the only subgroup analysis tested. The difference between the summary estimates for HMGcoA reductase inhibitors and the summary estimates for other cholesterol-lowering interventions is moderate, clinically important, and statistically significant. The treatment effect of HMGcoA reductase inhibitors was consistent across studies. Furthermore, HMGcoA reductase inhibitors are more potent in their cholesterol-lowering effect than are older drugs. This provides a biological rationale for their greater effect with respect to risk for stroke.

Our findings should nevertheless be regarded with some caution. Our evaluation of HMGcoA reductase inhibitors relative to other cholesterol-lowering interventions is based on between-study rather than within-study treatment comparisons. Differences in efficacy that are apparently due to treatment, when inferred from between-study comparisons, may actually be due to other factors, including differences in study design or populations. For example, on average, study participants in trials of HMGcoA reductase inhibitors were older and had higher event rates for nonfatal and fatal strokes compared with participants in trials of other interventions. This may explain, at least in part, apparent differences between HMGcoA reductase inhibitors and other interventions.

The results of the overall test for heterogeneity were not statistically significant, raising questions about the appropriateness of a subgroup analysis. Because of limitations in the power of overall tests of heterogeneity, however, important differences between subgroups may exist and yet not be evident [51]. The statistically significant difference between the effect of HMGcoA reductase inhibitors and the effect of other agents emphasizes the limitations of the overall test of heterogeneity. In addition, we made our decision to examine subgroups before we examined the data.

A further limitation of our analysis is that most of the trials of antilipidemic agents did not provide data about the incidence of fatal and nonfatal stroke. We identified five trials with HMGcoA reductase inhibitors that did not explicitly report data on the incidence of stroke. With one exception, these trials included relatively few participants. All studies had short follow-up periods and reported very low event rates for death from coronary disease and overall mortality. The remaining trials, which did not report data on stroke, were primarily small studies that achieved only moderate reductions in cholesterol levels and had shorter follow-up periods and older publication dates (Bucher and colleagues. In preparation). Thus, underreporting of stroke events in these studies is unlikely to have biased our estimate of HMGcoA reductase inhibitor trials. If this bias is present, it is probably conservative.

Previous studies have left some doubt about whether elevated cholesterol levels are a risk factor for stroke. The finding that HMGcoA reductase inhibitors reduce the incidence of stroke suggests that elevated cholesterol levels not only are a risk factor for stroke but may play a causal role in the pathogenesis of stroke. It is possible, however, that lipid-lowering agents decrease the incidence of stroke through some mechanism that is independent of their lipid-lowering effect. At the same time, strong biological and epidemiologic links exist between cholesterol reduction and a decreased incidence of clinical manifestations of atherosclerosis other than stroke, and a biological link exists between atherosclerosis and stroke. This suggests that HMGcoA inhibitors do reduce the incidence of stroke by reducing cholesterol levels.

Could cholesterol reduction with HMGcoA reductase inhibitors increase risk for hemorrhagic stroke? Unfortunately, we do not have the data necessary to assess this possibility. However, indirect evidence from the included trials of HMGcoA reductase inhibitors suggests that in populations with a baseline risk for cardiovascular disease of 1% to 3%, any possible risk is outweighed by the beneficial effects reflected in the statistically significant and clinically important reduction of overall mortality. Further trials are needed to discover whether persons with intermediate elevations in cholesterol level have an excess risk for hemorrhagic stroke.

How do HMGcoA reductase inhibitors compare with antiplatelet agents for the prevention of stroke? A systematic review that included various populations at high risk for stroke, such as patients with previous myocardial infarction, previous stroke, or transient ischemic attacks, found that pooled odds reductions for nonfatal strokes ranged between 12% and 23% [52]. Thus, in populations with manifest systemic atherosclerosis, HMGcoA reductase inhibitors may achieve a risk reduction similar to that produced by antiplatelet agents.

The favorable effect of HMGcoA reductase inhibitors on death from coronary heart disease and overall mortality was statistically significantly different from the effect of all other cholesterol-lowering interventions. In an additional analysis that included data on mortality from 64 randomized, controlled trials, we confirmed this finding and showed that heterogeneity for both death from coronary heart disease and overall mortality was primarily explained by type and extent of cholesterol-lowering interventions (Bucher and colleagues. In preparation).

In conclusion, the results of this systematic review suggest that HMGcoA reductase inhibitors used for primary prevention of stroke reduce the incidence of nonfatal and fatal stroke. Other, less potent antilipidemic drugs and dietary interventions are not efficacious for stroke prevention. These findings provide further evidence that clinicians who treat hyperlipidemic patients at moderate to high risk for important atherosclerotic events should offer aggressive lipid-lowering treatment, including HMGcoA reductase inhibitors.

Detailed information on the data from all included trials can be obtained from the authors.

Presented in part at the 20th Meeting of the Society of General Internal Medicine, Washington, D.C., 1-3 May 1997.

Ms. Griffith and Dr. Guyatt: Department for Clinical Epidemiology and Biostatistics, McMaster University, 1200 Main Street West, Hamilton, Ontario L8N 3Z5, Canada.