Impact of Statins on Risk of Stroke: A Meta-Analysis
Nickole N Henyan, PharmD; Daniel M Riche, PharmD BCPS; Honey E East, MD; Pamela N Gann, PharmD Ann Pharmacother. 2007;41(12):1937-1945. ?2007 Harvey Whitney Books Company
Abstract and Introduction
Background: Evidence from randomized, controlled trials suggests
that reduction of low-density lipoprotein cholesterol with
hydroxymethylglutaryl coenzyme A reductase inhibitor (statin) therapy in
patients at high risk for cardiovascular disease reduces the incidence of
ischemic stroke; however, data from large epidemiologic observational
studies suggest an inverse relationship between risk of hemorrhagic stroke
and cholesterol levels.
Hydroxymethylglutaryl coenzyme A reductase inhibitors (statins) are well known to block the production of cholesterol and have a profound effect of lowering both total cholesterol and low-density lipoprotein cholesterol (LDL-C). Ample epidemiologic and experimental data have suggested that hypercholesterolemia is a statin-modifiable risk factor for coronary heart disease (CHD). However, the effect of statins on cerebrovascular disease is less clear. Stroke is the leading cause of serious, long-term disability and the third most common cause of death in the US.[3,4] Although several limitations (eg, nonuniform reporting methods and lack of diagnostic differentiation) exist, multiple clinical trials suggest that statins protect against stroke. In fact, several prospective studies have demonstrated that statins reduce overall stroke occurrence by 19-25%.[5-7]
Evidence from randomized controlled trials suggests that cholesterol (particularly LDL-C) reduction by statins in patients at high risk for cardiovascular disease reduces the incidence of ischemic stroke; however, data from large epidemiologic observational studies suggest an inverse relationship between risk of hemorrhagic stroke and serum cholesterol levels, creating a stroke paradox.[8-11] In fact, in vitro analyses suggest that low cholesterol in the absence of modifiers (eg, statins) may impair platelet function or cell membrane integrity.
Epidemiologic data also show that serum cholesterol levels are inversely related to stroke death. The MRFIT (Multiple Risk Factor Intervention Trial) follow-up and the Honolulu Heart Study demonstrated a positive correlation between higher baseline cholesterol levels and nonhemorrhagic stroke; however, there was a negative association between higher baseline cholesterol and intracerebral hemorrhage.[11,13,14] Several Japanese cohorts have reported increases in cerebral hemorrhage in patients with low cholesterol levels.[15-17] Both hemorrhagic stroke and angionecrosis (destruction of blood vessels that may increase hemorrhagic risk) are more prevalent in Japan where residents have lower dietary cholesterol intake versus Japanese residents of Hawaii who have typically higher cholesterol dietary intake, suggesting a J-curve phenomenon with naturally occurring low cholesterol levels.
The mechanism by which statins elicit stroke protection is likely multifactorial, not only due to atheroma stabilization in extracranial arteries and reduction of source of thromboembolism (ie, CHD), but also due to direct neuroprotection (largely independent of cholesterol reduction). Neuroprotection may be linked to reduction of cholesterol-independent (pleiotropic) isoprenoid intermediates in the mevalonate pathway. The role for statins in neuroprotection is still controversial, especially in recovery following acute ischemic stroke.
We conducted a meta-analysis to elucidate the effect of statin therapy on all cerebrovascular events (CVEs), ischemic stroke, and hemorrhagic stroke.
A systematic literature search of MEDLINE, EMBASE, Cumulative Index to Nursing & Allied Health Literature, and Web of Science from June 1975 through September 2006 was performed to identify controlled clinical trials of statin therapy that evaluated stroke as any endpoint (clinical trial acronyms defined in the Appendix). The search strategy is shown in Figure 1. A manual review of abstracts presented at meetings of the American College of Cardiology, the American College of Clinical Pharmacy, and the American Stroke Association from 2001 to 2006 was also conducted. References from review articles were reviewed to identify additional relevant studies.
To be included in this meta-analysis, studies had to meet the following criteria: (1) controlled clinical trials versus placebo, (2) well-described protocol, and (3) data reported on incidence of all CVEs, ischemic stroke, or hemorrhagic stroke. Since major studies that included cerivastatin were terminated prior to completion, studies were excluded if cerivastatin was the active treatment (n = 2). Studies were also excluded from each analysis if there were no events in either group (n = 7) or if the control group included an active therapy or standard of care (n = 16). Abstracts not reporting stroke (n = 2) were also excluded. In cases in which there was more than one published report on the same population or group of patients (n = 4), the original article was selected for analysis. Two of the included trials reported transient ischemic attacks (TIAs) in combination with all CVEs. Considering the inadequacy of TIAs as a solid cerebrovascular endpoint, data on TIAs were not included in the analysis of all CVEs when reported separately.
Assessment of Literature Quality
The following methodological features most relevant to the control of bias were assessed: randomization, concealment, masking of treatment allocation, and withdrawals. All trials were independently reviewed by 3 investigators (NNH, DMR, PNG), with disagreement resolved by consensus.
The following information was collected from each article: study design, sample size, follow-up duration, statin and dose evaluated, number and type of CVE, comorbid disease states, age, sex, race, concomitant medications, cholesterol levels, and smoking history.
This meta-analysis was completed through the use of StatsDirect statistical software, version 2.4.5 (www.statsdirect.com). Summary statistics were combined and weighted averages were calculated using a random effects (Der-Simonian and Laird methodology) model. Statistical heterogeneity was evaluated via the Q statistic (p < 0.1 was considered representative of significant statistical heterogeneity). Publication bias was assessed through visual inspection of funnel plots and the Egger weighted regression method, with a p value less than 0.05 considered representative of statistically significant publication bias. Data are reported as relative risks with 95% confidence intervals.
A total of 56 full-text reports of trials and 3 abstracts were identified and retrieved. Upon review and evaluation, 27 trials (and 1 contributory abstract) met all of the inclusion criteria ( Table 1 ).[5-7,22-45] The majority of patients were white males with a history of hyperlipidemia. Most studies were a mixture of primary and secondary cardiovascular disease prevention trials with a placebo run-in period. Approximately 21% of the overall population were current smokers.
All Cerebrovascular Events
A total of 26 trials (N = 100,560) reported data on all CVEs.[5-7,22-44] Upon meta-analysis, statin therapy significantly reduced the risk of all CVEs (RR 0.83; 95% CI 0.76 to 0.91; Figure 2). Statistical heterogeneity was not observed among the trials (Q statistic; p = 0.19). Publication bias was not evident via Egger weighted regression (p = 0.35).
Six trials (n = 37,292) were included in the analysis of the effect of statin therapy on ischemic stroke.[6,23-25,44,45] Upon meta-analysis, statin therapy was shown to significantly reduce the risk of ischemic stroke (RR 0.79; 95% CI 0.63 to 0.99; Figure 3). Significant statistical heterogeneity was observed among trials (Q statistic; p = 0.0229). Publication bias was not evident via Egger weighted regression (p = 0.9848).
Nine trials (n = 57,895) reported data on hemorrhagic stroke.[5-7,23-27,44] Upon meta-analysis, statin therapy was shown to nonsignificantly increase the risk of hemorrhagic stroke (RR 1.11; 95% CI 0.77 to 1.60; Figure 4). Statistical heterogeneity was not observed among the trials (Q statistic; p = 0.15). Publication bias was not evident via Egger weighted regression (p = 0.24).
The results of our meta-analysis were not altered via methodological changes. Use of a fixed-effects model (Mantel-Haenszel, Rothman-Boice) did not alter the statistical significance of reduced risk for all CVEs (RR 0.82; 95% CI 0.77 to 0.88) or ischemic stroke (RR 0.76; 95% CI 0.69 to 0.84) or the nonsignificance of increased risk of hemorrhagic stroke (RR 1.15; 95% CI 0.90 to 1.45).
This meta-analysis illustrates that statins significantly reduce the risk of all CVEs (17%) and ischemic stroke (21%), although there was a nonsignificant increased risk of hemorrhagic stroke (11%). Although the increased risk of hemorrhagic stroke seen in this meta-analysis was driven predominately by results of the SPARCL trial, the LIPID, CARDS, and MEGA trials each contributed a nonsignificant increase in incidence of hemorrhagic stroke to the totality of data. Whether increased risk of hemorrhagic stroke is a direct class effect of statins or is dependent on the degree of cholesterol reduction remains unknown.
A meta-analysis performed by Amarenco et al. concluded that statins may reduce the incidence of all strokes without any increase in hemorrhagic strokes. In that metaanalysis, not all trials were placebo-controlled (eg, ACAPS, ALLHAT-LLT, KLIS). To minimize the potential for bias introduced with active control groups, these trials were excluded in our meta-analysis. Amarenco et al. used a fixed-effects model versus a random-effects model. Since the choice between fixed and random effects remains debatable, we reported our data using both models. Additionally, several trials (eg, SPARCL, CARDS, MEGA) were not published at the time of the previous meta-analysis.
Consistent with previously published observations, risk of all CVEs and ischemic stroke is significantly reduced with statin use. Both the 4D and ALERT trials evaluated patients with severe renal complications and reported an increase in the incidence of all CVEs in patients receiving statin therapy.[24,34] The 4D study further defined specific types of stroke, indicating an increase in ischemic stroke with statin therapy, which contributed to the observation of statistical heterogeneity in the ischemic analysis. The 4D investigators did not have a basis of explanation for this finding; however, routine use of statins in patients receiving hemodialysis is not currently recommended.
Among trials exhibiting a nonsignificant increase in risk of hemorrhagic stroke, LIPID and MEGA evaluated pravastatin, while SPARCL and CARDS evaluated atorvastatin.[7,23,25,44] Among trials with a nonsignificant reduction in risk of hemorrhagic stroke, MIRACL and 4D evaluated atorvastatin, and CARE evaluated pravastatin.[24,26,27] Both trials evaluating simvastatin (4S and HPS) demonstrated a nonsignificant reduction in risk of hemorrhagic stroke.[5,6] The SPARCL and LIPID trials included patients with a previous history of CVE.[7,23] Since SPARCL evaluated statins as secondary prevention of stroke, the majority of patients were receiving antiplatelet therapy (87%). Although LIPID included only 4% of patients with a previous CVE, the greater part of this population (83%) was also receiving antiplatelet treatment for secondary cardiovascular disease prevention. Interestingly, only SPARCL and LIPID demonstrated an increase of 40% or more in hemorrhagic stroke compared with the increase shown with placebo.[7,23] On the other hand, CARE and MIRACL (evaluating the same statins at equivalent doses) also included a large percentage of patients (> 80%) on antiplatelet therapy and demonstrated a nonsignificant reduction in hemorrhagic stroke.[26,27] Therefore, an increased risk of hemorrhagic stroke with concurrent antiplatelet and statin therapy cannot be ruled out.
A recent study by Bang et al. evaluated the effect of statin use and cholesterol level on the risk of symptomatic hemorrhagic transformation (sHT) in patients post acute ischemic stroke. Multivariable analysis did not find an association between statin use and increased likelihood of sHT (p = 0.566) but did demonstrate an association between lower LDL-C levels (mean ? SD 77.9 ? 40.5 mg/dL) and sHT (p = 0.02). The SPARCL and CARDS trials reached similar LDL-C levels and demonstrated an increased risk of hemorrhagic stroke.[23,25] Of note, study patients in the lower third percentile of LDL-C values were more likely to be on statins and antiplatelet/anticoagulant therapy, possibly increasing sHT risk. Although this study was retrospective and had a relatively small sample size, it does provide some interesting insights into a possible association of hemorrhagic stroke and lower cholesterol levels in patients immediately following an ischemic stroke.
The most recent guidelines from the National Cholesterol Education Panel state that patients with coronary heart disease (CHD) or a CHD risk equivalent should achieve a goal LDL-C less than 100 mg/dL, with a therapeutic option of less than 70 mg/dL in patients with significant CHD or CHD risk. Although symptomatic carotid artery disease is a CHD risk equivalent, patients with a history of cerebrovascular disease (regardless of type) are not specifically defined as a high-risk group, and LDL-C goals for these patients can be based on Framingham risk calculation. The results of SPARCL and meta-analyses comparing high- and low-dose statins are likely to change LDL-C goals and increase the use of statins in patients with a history of ischemic stroke.[50-52] To achieve LDL-C levels lower than those currently recommended, aggressive statin therapy may be necessary. Considering epidemiologic evidence of an association between the increased risk of intracerebral hemorrhage and naturally occurring low cholesterol levels, the use of high-dose statins with concomitant antiplatelet therapy and risk of hemorrhagic stroke (especially in patients with lower baseline total cholesterol levels) should be assessed on an individual basis. Gradual reduction in LDL-C shortly after stroke may provide typical long-term benefit against recurrent ischemic stroke, while minimizing risk of intracranial hemorrhage.
In addition to lowering cholesterol levels, statins have cholesterol-independent effects known as pleiotropism.[20,53,54] Pleiotropic effects can be grouped into 4 categories: (1) effects explained by the influence of statins on nitric oxide bioavailability, (2) antiinflammatory effects, (3) antithrombotic effects, and (4) antioxidant effects. The critical pleiotropic effect of statins relating to hemorrhagic complications is their influence on both platelet activity and fibrinolysis. Statins elicit an increase in tissue-type plasminogen activator levels promoting fibrinolysis, while platelet activator inhibitor-1 levels fall, possibly destabilizing the extracellular matrix.[20,21] Unfortunately, limited evidence exists to suggest the extent of pleiotropic effect among statins.
Despite the large size of the meta-analysis (n = 100,660), we recognize several limitations. First, several studies reported data on all CVEs, but fewer than half reported the incidence of hemorrhagic or ischemic stroke. Data for hemorrhagic and ischemic stroke from all included studies would have provided additional power and potentially added to a more conclusive interpretation of the results. However, there were more than 37,000 patients in the ischemic analysis and more than 57,000 in the hemorrhagic analysis, suggesting adequate power. Secondly, the definition of stroke, fatal stroke, and CVE was not uniform across all studies, potentially introducing investigator bias, especially in trials with an open-label design. Finally, publication bias is a concern with any meta-analysis. Publication bias results from the greater ease of finding studies that yielded significant or positive results, potentially leading to overrepresentation of benefit in systematic reviews. In light of the Egger weighted regression p values, the potential for publication bias appears minimal.
Lipophilic statins cross the blood-brain barrier, while hydrophilic statins may not.[56,57] Whether degree of penetration of the blood-brain barrier alters the neuroprotective effects of statins remains unclear. Future trials (eg, FASTER and J-STARS; clincaltrials.gov identifiers NCT00109382 and NCT00221104, respectively) should help further define the role of statins in overall stroke prevention.
The results of this meta-analysis illustrate the benefit of statins in prevention of all CVEs or ischemic stroke but a potential for increased risk of hemorrhagic stroke.
Table 1. Characteristics of Selected Studies
Appendix: Acronyms of Clinical Trials
Dr. Riche, The University of Mississippi Medical Center, 2500 N. State St., Jackson, MS 39216, fax 601/984-2751. E-mail: email@example.com .
Nickole N Henyan, PharmD,1 Daniel M Riche, PharmD BCPS,2 Honey E East, MD,3 Pamela N Gann, PharmD4
1School of Pharmacy, University of Mississippi, Jackson, MS
2Schools of Pharmacy and Medicine, University of Mississippi, Jackson, MS
3School of Medicine, University of Mississippi, Jackson, MS
4University of Mississippi Medical Center, Jackson, MS
Disclosure: Dr. East is a member of the Speaker's Bureau for Pfizer and Sankyo-Daiichi and has received research funding from Abbott and Novo Nordisk.