Aspirin for the Primary Prevention of Stroke and Other Major Vascular Events: Title and subTitle BreakMeta-analysis and Hypotheses FREE

ONE HUNDRED years after its introduction, aspirin is the world's most commonly used drug. About 75,000 pounds of aspirin (about 100 million tablets) are consumed each day in the United States alone.¹ Among Americans older than 65 years, about 20% use aspirin regularly, mostly for the relief of pain.^{2 - 7}

The regular use of aspirin reduces the risks of myocardial infarction (MI), stroke, and vascular death for patients with clinically manifest vascular disease (ie, those at high risk). An influential meta-analysis of over 100 randomized clinical trials involving more than 60,000 participants showed that aspirin therapy reduces stroke by about 25% in most high-risk patients.⁸ Paradoxically, aspirin use was associated with small increases in the incidences of stroke in 2 large clinical trials of primary prevention included in this meta-analysis although these results were not statistically significant.^{8 - 10}

An estimated 8 million elderly Americans without apparent vascular disease consume aspirin regularly,¹¹ and its effect on stroke in such persons is important to characterize. With the recent publication of 4 studies addressing the effect of aspirin use on stroke in low-risk persons,^{11 - 14} we undertook a critical review of this issue. Our objectives were to determine the effect of the regular use of aspirin on stroke for persons without clinically manifest atherosclerotic vascular disease (ie, primary prevention) as well as its effect on other major vascular events in this population.

We sought to identify all published randomized clinical trials assessing the effect of aspirin in any dose vs placebo for primary prevention of stroke and other major vascular events in persons without clinically manifest vascular disease. Vascular diseases included MI; stroke; transient ischemic attack; angina pectoris; intermittent claudication; atrial fibrillation; or revascularization of carotid, coronary, or peripheral arteries for atherosclerosis. Clinical trials in which all participants had major atherosclerotic risk factors (ie, advanced age, hypertension, or diabetes mellitus) were included, but considered separately in secondary analyses. Clinical trials in which more than 20% of the participants had clinically diagnosed vascular disease, those who did not report stroke outcomes, and those without observed strokes were not analyzed. Whenever possible from published results, subgroups of participants in primary prevention trials who had vascular disease were removed for secondary analyses.¹⁵ Clinical trials in which aspirin was combined with another antiplatelet agent were excluded.

Computerized medical databases were searched for the years 1980 through 1998 using the following key words: "clinical trial," "aspirin," and "stroke." In addition, the Cochrane Collaboration Registry and published lists of the Antiplatelet Trialists' Collaboration were reviewed seeking randomized clinical trials meeting these specifications. Of 6 clinical trials identified, 5 were included in the main analyses,^{9 - 10 ,13 - 14 ,16} as the remaining small pilot trial did not report strokes according to treatment assignment.¹⁷ The large, rigorously conducted Antiplatelet Trialists' Collaboration meta-analysis was used to assess the effect of aspirin therapy in patients with established vascular disease.⁸ In addition to randomized clinical trials, large, prospective observational studies were analyzed separately to assess the effect of the regular use of aspirin on stroke in a broader spectrum of persons taking widely different aspirin doses (although distinguishing association from cause-effect is impossible with confidence in these nonrandomized studies). A computerized literature search using the following key words: "aspirin," "stroke," "population-based," and "epidemiological" was done, the reference lists of recent review articles were also screened, and 4 studies were identified.^{5 ,11 - 12 ,18}

Results of the randomized clinical trials were combined using a stratified Mantel-Haenszel estimate of the odds ratio and 95% confidence intervals (CIs).¹⁹ The effect of aspirin therapy vs placebo was compared using a stratified Mantel-Haenszel-Cochran statistic. Homogeneity of the combined results was evaluated by the Breslow-Day test,²⁰ and confirmed using a random effects model for both rate differences and odds ratio since event rates were low. Computations were performed using SAS (SAS Institute, Cary, NC).

Five randomized clinical trials included 52,251 participants with a mean age of 57 years who were followed up for about 240,000 patient-years observation (mean follow-up = 4.6 years per patient, Table 1). Three clinical trials excluded women^{9 - 10 ,14} ; women constituted 20% of the participants, contributing 17% of total exposure. Recruits varied widely between clinical trials, from healthy male physicians to patients with diabetes mellitus or hypertension or men with coronary risk factors (Table 1). Mortality rates ranged from 0.4% to 3.0% per year, averaging 1% per year. In comparison, US mortality rates for men and women aged 55 to 59 years, average 0.7% and 0.4% per year, respectively. Although considered primary prevention trials, about 5% of the participants (range, 1%-20%) had clinically manifest vascular disease. Aspirin doses were 75 mg/d (21,330 participants), 325 mg every other day (22,071 participants), and 500 to 650 mg/d (8850 participants; Table 1). The stroke rate in pooled analysis of the 5 clinical trials averaged 0.3% per year (range, 0.2%-0.8% per year, Table 2).

Aspirin therapy was associated with modest increases in the rate of stroke in 3 clinical trials^{9 - 10 ,16} a decrease in 1 small study,¹⁴ no appreciable effect in the large remaining clinical trial,¹³ but was not statistically significant in any individual clinical trial or their pooled results (relative risk [RR] = 1.08; 95% CI, 0.95-1.24, Figure 1). This contrasted with a highly significant reduction in MI in these clinical trials, with an overall RR reduction in MI of 26% (RR, 0.74; 95% CI, 0.68-0.82; P<.001; Table 2). Secondary analyses of the Physicians' Health Study restricted to highly adherent participants widened the contrasting effects of aspirin therapy on stroke (a 29% increase P = .14) vs MI (a 51% decrease, P<.001).²¹ Pooled analysis of all 5 clinical trials revealed a RR of 0.93% (95% CI for RR, 0.83-1.03) for deaths categorized as vascular and of 0.94 (95% CI, 0.87-1.01) for all-cause mortality, with both trends fairly consistent across the clinical trials although not statistically significant in any individual clinical trial or when pooled.

Three^{13 - 14 ,16} of these 5 clinical trials recruited only participants with vascular risk factors. Participants in these 3 clinical trials had a mean age of 59 years, 42% were women, and rates of death and stroke averaged 1.5% per year and 0.5% per year, respectively. In contrast, the 2 remaining clinical trials^{9 - 10} recruited healthy male physicians of whom only a minority harbored vascular risk factors. In these 2 clinical trials, the mean age averaged 55 years with about half the rates of death and stroke (0.6% per year and 0.2% per year, respectively) compared with clinical trials that included participants with risk factors. No net effect of aspirin therapy on stroke was noted in the clinical trials of participants with vascular risk factors (RR = 1.02; 95% CI, 0.86-1.21) compared with a possible small increase (RR = 1.20; 95% CI, 0.96-1.49) in 2 clinical trials of men without risk factors although of borderline statistical significance and not statistically significant in either clinical trial individually (Figure 2).

Pooled experience in participants with manifest vascular disease (ie, high risk) by the Antiplatelet Trialists' Collaboration yielded a 27% RR reduction (RR, 0.73; 95% CI, 0.67-0.79) in stroke by antiplatelet therapy (Figure 2).^{8 ,22} This effect was incompatible with the effect of aspirin therapy in the 5 clinical trials of primary prevention (P = .001). In contrast, the effect of aspirin therapy for prevention of MI is similar for those with vs without vascular disease (about 25%) and for participants with vs without vascular risk factors in the 5 primary prevention clinical trials (23% vs 26% reductions, respectively).

The influence of aspirin therapy on stroke in randomized clinical trials of participants who had low rates of stroke (≤1% per year) despite manifest vascular disease is difficult to estimate. These clinical trials usually involved young patients with coronary syndromes during an era when stroke was often undetected or underreported (based on high ratios of fatal to nonfatal strokes),^{23 - 25} with notable exception.²⁶ Pooled analysis of 4 such clinical trials show stroke rates averaging 0.7% per year, and aspirin treatment associated with a 35% reduction in the incidence of stroke (95% CI for RR, 0.49-0.88).^{23 - 26}

Four large, prospective, observational studies assessed the relation between self-selected regular use of aspirin and stroke in low-risk participants.^{5 ,11 - 12 ,18} In each, aspirin use was associated with increases in the incidence of stroke, statistically significant in 2 studies after adjustment for other stroke risk factors.^{11 - 12} In 1 study, the association of the regular use of aspirin with increased incidence of ischemic stroke was observed in women but not men.¹¹

Long-term use of aspirin therapy increased the incidence of hemorrhagic stroke (RR = 1.35, P = .03 from pooled analysis of all randomized clinical trials) both for those with and without manifest vascular disease (Table 3). No apparent relation was noted between aspirin dose and the incidence of hemorrhagic stroke based on indirect comparisons from randomized clinical trials (Table 3), and incidence based on direct comparisons are limited by the lack of hemorrhagic events.^{29 - 30} In 1 case-control study, aspirin in doses of 1225 mg/wk or more was associated with a 3-fold increased rate of intracerebral hemorrhage (P = .05), while no increase was observed with lower doses.³¹ In clinical studies of occidental cohorts, 10% to 15% of all strokes are hemorrhagic. Hence, a 35% increase in the rate of hemorrhagic stroke by aspirin therapy (Table 3) would raise the overall stroke rate of aspirin users by about 3% to 5%—accounting theoretically for the small trends in excess stroke observed among those assigned aspirin in the pooled analysis of primary prevention trials (Table 2), assuming no reduction in the incidence of ischemic strokes occurred in those given aspirin therapy.

It is unclear whether aspirin therapy can paradoxically increase the risk of ischemic stroke in certain individuals at low intrinsic risk. This was initially suggested by the findings of 2 large primary prevention trials involving male physicians (Table 4).^{9 - 10} In the US Physicians' Health Study,³² most strokes were categorized as ischemic or hemorrhagic based on neuroimaging or autopsy. There was a small excess of ischemic strokes in this randomized clinical trial (91 strokes among those assigned aspirin vs 82 among those assigned placebo). In the British Doctors Study,⁹ lack of neuroimaging makes data regarding ischemic stroke weaker. Together, a small increase in the incidence of ischemic stroke associated with aspirin therapy was seen, but the overall 14% increase does not reach statistical significance (Table 4). In contrast, 2 clinical trials of patients with vascular risk factors taking an aspirin dose of 75 mg/d showed no increase in the incidence of ischemic stroke although the frequency of neuroimaging was not reported (Table 4).^{13 - 14} Considering pooled data from all 4 randomized primary prevention trials that separated ischemic and hemorrhagic stroke, there was no effect on the incidence of ischemic stroke by aspirin therapy (Table 4). This finding contrasts significantly with the effect of aspirin therapy to reduce the incidence of ischemic stroke for those with vascular disease (P = .001).

Pooled analysis of 5 randomized clinical trials involving over 50,000 participants testing aspirin therapy for primary prevention shows no overall effect on stroke, a result incompatible (P = .001) with the unequivocal effect of aspirin therapy to reduce stroke by about 25% for those with vascular disease. If the effect of aspirin therapy on stroke differs based on the presence of clinically recognized vascular disease, then there could logically exist an intermediate group of patients with vascular risk factors in whom disease is not manifest. While available data are potentially confounded by differences in aspirin doses and sex distribution, they support the hypothesis that in those persons with major vascular risk factors, the effect of aspirin therapy on stroke may be intermediate between those with recognized vascular disease and those with neither manifestations of nor major risk factors for vascular disease (Figure 2).

In the 4 observational studies, the consistent association between the regular use of aspirin and increases in stroke risk in both women and men is notable. Aspirin use in these studies was largely self-selected and confounding by uneven distribution of other stroke precipitants cannot be discounted, even after statistical adjustment for recognized differences.³³ Considered in the context of the randomized comparisons, however, these observational studies raise hypothetical concerns about higher doses of aspirin used for analgesia and accentuation of stroke in elderly persons and in women without manifest vascular disease, underrepresented in randomized clinical trials to date.³⁴

It is remarkable that the expected decrease in the incidence of ischemic stroke was not observed in randomized clinical trials of primary prevention. Aside from inhibition of platelet thromboxane synthesis underlying aspirin's antithrombotic effect, other competing actions of aspirin could be relevant. Aspirin therapy has been reported to increase systemic blood pressure (especially if measured while the patient is supine) and to antagonize the effect of some antihypertensive drugs.³⁵ Hypertension is a strong, prevalent risk factor for both hemorrhagic and ischemic stroke. Thrombogenic effects of aspirin possibly related to inhibition of endothelial prostacyclin synthesis have been demonstrated experimentally, particularly at high doses.^{36 - 38} In patients without symptomatic atherosclerotic vascular disease, inhibition of prostacyclin synthesis may predispose to thrombosis (the "aspirin dilemma"),^{39 - 40} but the clinical relevance of this effect of aspirin therapy on prostacyclin is unknown.³⁴

Given the consistent effect of aspirin therapy to reduce MI by about 25% and the strong trend for reduction of all-cause mortality, is the regular use of aspirin beneficial for healthy elderly persons? Among healthy elderly persons (without recognized vascular disease), the incidence of stroke equals or exceeds that of MI,^{5 ,11} in contrast with younger cohorts in whom the reverse holds.^{9 - 10 ,14 ,16} It is clear that aspirin therapy increases major extracranial bleeding, with an average RR of 1.5 in clinical trials testing doses between 75 mg/d and 325 mg every other day.^{10 ,13 - 14} Rates of major extracranial hemorrhage in placebo-assigned patients vary between primary prevention clinical trials involving middle-aged persons (0.15%-0.5% per year), in part due to differing criteria. The risk of major extracranial hemorrhage during aspirin use is likely age related.⁴¹

For those without manifest vascular disease, the risk-benefit analysis of aspirin therapy is complex and must balance a reduction in MI and possibly death against small increases in major extracranial and intracranial bleeding (Table 5). In our view, the benefits of the regular use of aspirin for primary prevention of vascular events in healthy elderly persons have not been sufficiently established to warrant a general recommendation for widespread use. Available data most clearly favor aspirin use for middle-aged persons at special risk for MI. Individual values and preferences may bear importantly on the risk-benefit equation.

Several limitations of these analyses warrant comment. Some fraction of participants in the 5 key primary prevention trials at the core of these analyses had recognized vascular disease (Table 1), and it is unclear how they influenced our estimates of the effect of aspirin therapy in those without vascular disease. Evaluation of the importance of vascular risk factors was confounded by segregation on the basis of aspirin dose and sex; this may be clarified when the results of the Women's Health Study⁴² become available. Other potential limitations are those common to meta-analysis: Pooling results of several studies to yield a statistically significant result when no individual clinical trial showed statistical significance is problematic; such a result should be considered hypothesis-generating and not hard evidence. We attempted to identify all relevant clinical trials but, although it seems unlikely that large clinical trials were overlooked, omission of studies could introduce bias.

The effect of aspirin therapy on stroke may differ for individuals based on the presence or absence of atherosclerotic vascular disease. The overall effect of the regular use of aspirin for primary prevention of major vascular events in elderly persons is unknown since randomized clinical trials to date have included mostly middle-aged persons who are more likely to suffer MI than stroke. Long-term aspirin use increases the incidence of hemorrhagic stroke, but the magnitude of increase is small (about 1/2000 aspirin users per year for elderly persons). Because of an unequivocal effect to reduce MI in a wide range of doses, aspirin use for primary prevention would benefit most those at special risk for MI (such as patients with diabetes mellitus).¹³ Current evidence favors low-dose aspirin (75-81 mg/d) when given for primary prevention. The influence of aspirin therapy on stroke is complex, differing in different patient populations, and further understanding of the mechanisms underlying those differences would allow refinement of therapy with this most widely-used medication.

Accepted for publication June 11, 1999.

Corresponding author: Robert G. Hart, MD, Department of Medicine (Neurology), University of Texas Health Science Center, 7703 Floyd Curl Dr, San Antonio, TX 78284 (e-mail: hartr@uthscsa.edu).