Does Prior Use of Aspirin Affect Outcome inIschemic Stroke?

Lalit Kalra, MD, PhD, Inigo Perez, MD, David G. Smithard, MD, David Sulch, MD

BACKGROUND: Large intervention studies suggest that aspi-rin may reduce mortality when given to patients who presentwith strokes or transient ischemic attacks. We sought to deter-mine whether patients who were already using aspirin at thetime of an ischemic stroke had a lower mortality than those whowere not.METHODS: A prospective cohort study was undertaken in pa-tients (mean age 76 6 15 years) with acute ischemic stroke.Detailed information on demography, stroke characteristics,and aspirin use prior to the stroke was collected from patients,medical records, and other sources. Patients were classified bycause and subtype of stroke using standard criteria. Mortalitywas measured 4 weeks after the initial episode.RESULTS: Of the 1,457 patients, 650 (45%) were using aspirin(median dose 75 mg; range 75 to 300 mg) prior to the stroke.

Prior use of aspirin was associated with lower 4-week mortality(14% versus 20%, P ,0.01). Beneficial effects of prior aspirinuse on mortality were seen in patients with atheroscleroticstrokes (15% versus 21%, P ,0.05) and with cardioembolicstrokes (21% versus 34%, P ,0.05), but not among patientswith strokes due to small vessel occlusion (10% versus 11%, P 50.8). Prior aspirin use was also associated with lower mortalityin patients in whom the cause of ischemic stroke could not bedetermined (15% versus 22%, P ,0.01). The effect of prioraspirin use on mortality was independent of age, gender, otherrisk factors, and use of other medication.CONCLUSIONS: Prior use of low-dose aspirin may be associ-ated with a small but significant reduction in stroke mortality.Am J Med. 2000;108:205–209. q2000 by Excerpta Medica, Inc.

Several studies have established the role of aspirin inthe prevention of stroke in patients with cardiovas-cular disease (1). Recently, two large clinical studies

have shown that early use of low-dose aspirin signifi-cantly reduces mortality in the first few weeks after stroke(2,3), which has led to speculation that prior use of aspi-rin may potentially reduce stroke severity and improveoutcome. There is some support for this hypothesis fromhistorical reports to suggest that prior use of aspirin isassociated with less severe stroke (4,5). However, morerecent studies failed to show significant differences instroke severity or mortality between patients on aspirinand those not on any antiplatelet medication at the timeof stroke (6 – 8). It is possible that the type of stroke maybe important. The objective of this study was to comparethe type, clinical extent, and outcome of ischemic strokein patients using aspirin or other antiplatelet agents at thetime of a stroke with those who were not using aspirin.

METHODS

Patients and SettingThe study included all patients with a diagnosis of acutehemispheric stroke in a white, largely middle class, sub-

urban population of 530,000 in South East England be-tween 1995 and 1997. Patients were identified using aprospective community-based stroke registry and in-cluded in the study whether they were managed in thehospital or at home. Physicians, nurses, and therapistswere asked to report all stroke patients to a central office,which triggered review by the stroke team and initiationof specialist multidisciplinary stroke services. Case ascer-tainment was undertaken by regular cross checks againsthospital admission databases in local and neighboringhospitals, general practice records, and community nurs-ing or therapy referrals.

Assessments and Data CollectionThe World Health Organization (WHO) definition ofstroke was used, and initial diagnosis was based on his-tory and clinical examination (9). The diagnosis and pa-thology of stroke was confirmed by computed tomo-graphic (CT) scanning undertaken within 48 hours ofadmission in hospitalized patients. Further CT scans ormagnetic resonance (MR) images were obtained within 4to 7 days of the presenting episode in patients in whominitial scan findings were unhelpful or who had not yethad them performed.

Data were collected prospectively on patient andstroke characteristics, risk factors, and outcome. Hospitaland primary care treatment records prior to the strokewere reviewed to collect information on risk factors andtheir management. Medications at the time of assessmentwere recorded with particular emphasis on prior treat-ment with aspirin and the dose of aspirin used. All pa-tients were directly questioned about the use of nonpre-scription medication including aspirin. In addition, pa-

From the Guy’s, King’s, and St Thomas’s School of Medicine (LK, IP,DS), King’s College, London; and the Stroke Unit (DGS), William Har-vey Hospital, Ashford, Kent, United Kingdom.

Requests for reprints should be addressed to Lalit Kalra, MD, Depart-ment of Medicine, GKT School of Medicine, Denmark Hill Campus,Bessemer Road, London SE5 8PJ, United Kingdom.

Manuscript submitted April 22, 1999, and accepted in revised formSeptember 27, 1999.

q2000 by Excerpta Medica, Inc. 0002-9343/00/$–see front matter 205All rights reserved. PII S0002-9343(99)00431-3

tients’ spouses, family, and caregivers were questionedabout aspirin use and asked to list all tablets or medica-tion used by the patient. This information was comparedwith hospital records and prescriptions issued by thefamily practitioner during the previous 6 months to re-duce the possibility of differential recall bias.

Ischemic strokes were classified as those due to largeartery atherosclerosis, small artery occlusion, or cardio-embolic disease based on the TOAST (Trial of Org 10172in Acute Stroke Treatment) subtype classification system(10). Categorization as large vessel atherosclerosis wasbased on clinical findings of cerebral cortical impairmentwith infarcts greater than 1.5 cm in diameter, clinical his-tory of hypertension, diabetes, hyperlipidemia, or smok-ing, and evidence of atherosclerotic internal carotid ar-tery disease (stenosis or heterogeneous plaque) on ca-rotid duplex examination. The cardioembolic categoryincluded patients with known cardiac disease, evidence ofprevious strokes or transient ischemic attacks in differentvascular territories, valvular abnormalities, left atrial en-largement, or dyskinetic myocardial segments who wereat medium to high risk and had no other cause of strokeon investigations, including carotid duplex studies (10).All of these patients underwent transthoracic echocardi-ography, with transesophageal studies used as needed.Small artery occlusion was considered to be present inpatients with lacunar syndromes who had no corticalsigns, lesions ,1.5 cm on CT scans or MR images, ab-sence of carotid disease on carotid duplex studies, andnormal cardiac and electrocardiographic findings. Theincident strokes due to large vessel atherosclerosis andcardiac emboli were divided into total (cortical and sub-cortical involvement or large subcortical infarct) or par-tial (mainly cortical involvement) lesions (11). Classifica-tion into different causal and clinical subtypes was under-taken by a stroke specialist who was unaware of the use ofaspirin. The accuracy of classification was verified by anindependent review by another specialist who was un-aware of the stroke subtype allocated to the patient. Thetwo physicians independently agreed on the final subtypeof stroke in 1,608 (93%) of the 1,736 patients assessed(kappa 5 0.89). Consensus on subtype was reached in afurther 67 patients after discussions. Patients in whomconsensus could not be reached were assigned as “unde-termined cause.”

All patients with stroke due to large vessel atheroscle-rosis or small vessel disease (lacunar infarcts) receivedaspirin for secondary prophylaxis within the first 48hours of the acute episode regardless of prior use of aspi-rin. Patients with cardioembolic strokes were started onanticoagulation or aspirin as appropriate 2 weeks afterthe acute episode. Thrombolytic or neuroprotective in-terventions were not used in any of the patients. All pa-tients were managed using similar acute care protocolsfor prevention of stroke-related complications, treat-

ment of other risk factors, management of comorbidity,and early rehabilitation therapy, as appropriate.

The primary outcome was mortality due to any causeat 4 weeks after the initial episode.

Data AnalysisAnalyses of the effect of aspirin on patients with differentsubtypes of ischemic stroke excluded patients with hem-orrhagic strokes (but not hemorrhagic transformation),unusual causes of stroke (eg, vasculitis, dissection,thrombotic disorders), stroke of undetermined cause,and those on antithrombotic measures other than aspirin(eg, warfarin, dipyridamole). Patients under 60 years ofage with no vascular risk factors or with atypical neuro-logical features were also excluded from this analysis. Inaddition, the effects of prior aspirin use were examined inthe entire cohort.

The chi-square test was used to compare categoricaldata. Continuous variables were compared using the t testfor unpaired data. Relative risks (RR) with 95% confi-dence intervals (CI) were calculated for mortality in var-ious stroke subtypes for comparisons between patientswho received aspirin and those who did not receive aspi-rin prior to stroke.

The effects of potential confounders were adjusted us-ing logistic regression models, with the backward elimi-nation technique. Potential confounders included age,gender, systolic and diastolic blood pressure, plasma glu-cose levels at admission, and histories of prior hyperten-sion, diabetes, smoking, hypercholesterolemia, previousstrokes, atrial fibrillation, and ischemic heart disease.Treatment variables included use of antihypertensivemedication, hypoglycemic agents, and cholesterol-lower-ing agents. Odds ratios (OR) with 95% confidence inter-vals were used to compare patients who received aspirinwith those who did not in these adjusted analyses.

RESULTS

Of the 2,431 patients who suffered a stroke during thestudy, 2,236 (92%) were admitted to the hospital. Initialanalysis excluded 731 (30%) of the 2,431 patients becauseof unclear evidence about stroke subtype or other factors,including hemorrhagic stroke on CT scanning (n 5 353[14%], 17 on aspirin), inadequate investigations (n 5 267[11%]), prior treatment with warfarin (n 5 66 [3%]),and stroke of other determined cause (n 5 9 [vasculitis 4,carotid dissection 3, thrombotic disorders 2]). A further243 (10%) patients were excluded because a definitecause for the infarct could not be established, there weretwo or more equally likely causes of stroke, or there wasdisagreement in classification between the two adjudica-tors. In addition, 36 patients on dipyridamole (n 5 32)alone or in combination with aspirin (n 5 4) were ex-cluded from the analyses of stroke of determined cause.

Does Prior Aspirin Use Affect Stroke Outcome/Kalra et al

206 February 15, 2000 THE AMERICAN JOURNAL OF MEDICINEt Volume 108

Of the 1,457 patients with ischemic stroke fulfilling thecriteria for inclusion (1,426 treated in the hospital, 51treated at home), 1,206 (82%) had one or more risk fac-tors for stroke (Table 1). Aspirin was used regularly be-fore the stroke by 650 (45%) patients. The median dailydose of aspirin was 75 mg, with a range from 75 to 300mg. None of the patients was taking more than 300 mg ofaspirin per day.

The age and sex distribution of the aspirin-treatedpatients and the patients not using aspirin were similar(Table 1). There were significantly more patients withhypertension in the nonaspirin group and significantlymore patients with a history of previous cerebrovascularor ischemic heart disease in the aspirin group (Table 1).Hormone replacement therapy was being used by 44(13%) of the 342 women in the aspirin-treated group and66 (14%) of the 452 women in the nonaspirin-treatedgroup. There were no significant differences in the pro-portions of patients being treated for vascular risk factorsbetween the two groups (Table 2).

Type and Size of InfarctsThere were no significant differences in the proportionsof patients with strokes due to large vessel atherosclerosis(57% in the aspirin-treated versus 59% in the nonaspirin-treated patients), small vessel occlusion (29% versus26%), or cardioembolic disease (14% versus 15%) be-tween the aspirin and nonaspirin group. The proportionof patients with large infarcts was lower in those who usedaspirin prior to the stroke, especially among patients withatherosclerotic strokes (34% versus 52%, P ,0.001).

Effects of Aspirin Use on 4-Week MortalityPatients with large vessel atherosclerotic or cardioem-bolic strokes who were using aspirin prior to stroke hadsignificantly lower mortality compared with those not us-ing aspirin (Table 3), especially in those with total in-farcts. Causes of mortality included stroke progression(n 5 158), recurrent stroke (n 5 25), acute myocardialinfarction (n 5 13), pulmonary embolism (n 5 14), as-piration pneumonia (n 5 18), and other unrelated causes

Table 1. Demographic and Clinical Characteristics of Patients with Ischemic Stroke ofDetermined Cause (n 5 1457) by Prior Use of Aspirin

CharacteristicAspirin

(n 5 650)No Aspirin(n 5 807) P Value

Number (Percent), orMean 6 SD

Age (years) 76 6 12 77 6 13 0.7Female sex 342 (53) 452 (56) 0.6Systolic blood pressure (mm Hg)* 165 6 18 166 6 21 0.3Diastolic blood pressure (mm Hg)* 96 6 11 94 6 13 0.3Plasma glucose (mmol/L)* 6.5 6 2.7 5.9 6 3.1 0.1Current smoker 114 (18) 169 (21) 0.4History of

Hypertension 171 (26) 364 (45) ,0.01Diabetes mellitus 94 (14) 145 (18) 0.3Hypercholesterolemia 96 (15) 89 (11) 0.7Previous strokes or TIA 323 (50) 186 (23) ,0.01Atrial fibrillation 146 (22) 161 (19) 0.2Ischemic heart disease† 316 (49) 149 (14) ,0.001Peripheral vascular disease 86 (13) 97 (12) 0.8

* At the time of initial presentation with stroke. To convert glucose levels to mg/dL, multiply by 17.9.† Includes patients with a history of angina, myocardial infarction, coronary revascularization, or chronicnonvalvular heart failure.TIA 5 transient ischemic attack.

Table 2. Treatment with Other Medications in Patients withIschemic Stroke of Determined Cause (n 5 1457), by Use ofAspirin

Type of Medication Aspirin No Aspirin P Value

Number Treated/Number with Condition

(Percent)

Antihypertensivemedication

101/171 (59) 207/364 (57) 0.3

Diuretics (72) (77) 0.2Beta blockers (18) (21) 0.3Calcium channel

antagonists(31) (37) 0.2

ACE inhibitors (47) (49) 0.5Others (16) (14) 0.6

Hypoglycemic agents 86/94 (92) 136/145 (94) 0.8Oral agents (79) (77) 0.7Insulin (21) (23) 0.6

Statin 52/96 (54) 42/89 (47) 0.1

ACE 5 angiotensin converting enzyme.

Does Prior Aspirin Use Affect Stroke Outcome/Kalra et al

February 15, 2000 THE AMERICAN JOURNAL OF MEDICINEt Volume 108 207

(n 5 26). There were no significant differences in theproportions of deaths caused by these conditions be-tween aspirin-treated and untreated patients.

There was a similar reduction in mortality in patientswith ischemic stroke of undetermined cause who usedantiplatelet treatment (Table 4). The difference in mor-tality between the aspirin and the nonaspirin group con-tinued to be significant, even when patients with hemor-rhagic and other strokes were included (Table 4).

Gender, systolic and diastolic blood pressure on ad-mission, history of hypertension, diabetes mellitus or hy-percholesterolemia, and previous treatment with antihy-pertensive, hypoglycemic, or cholesterol-lowering agentswere not associated with mortality. Age above 75 years(OR 5 1.6; 95% CI: 1.1 to 2.3; P ,0.01), plasma glucoselevel .8 mmol/L [.140 mg/dL] on admission (OR 51.1, 95% CI 1.01 to 1.2; P 5 0.02), history of ischemicheart disease (OR 5 1.2, 95% CI: 1.01 to 1.4; P 5 0.04),atrial fibrillation (OR 5 1.3, 95% CI: 1.1 to 1.6; P 50.01), and prior use of aspirin (OR 5 0.7, 95% CI: 0.6 to0.9; P 5 0.01) were associated with mortality.

DISCUSSION

We found that prior treatment with low-dose aspirin wasassociated with a small but significant reduction in mor-tality in stroke patients, particularly in those with largevessel atherosclerosis or cardioembolic disease. This ef-fect of aspirin was not seen in patients with ischemic in-farcts due to small vessel occlusion.

These findings contrast with those of two recent re-ports (7,8). These earlier studies had small sample sizesand may not have been able to detect relatively small dif-ferences in stroke severity or outcome. A nonsignificantreduction in mortality of about 4% was seen in one of thestudies, which included 125 patients on aspirin (7). Inaddition, previous studies did not differentiate betweenstrokes due to different causes. This may be an importantconsideration because of the different pathogenic mech-anisms involved. Some of these mechanisms (eg, thosedependent upon platelet aggregation and activation) maybe modified by aspirin use, whereas others (eg, lipohyali-nosis) may be relatively immune to its effect (12). Differ-

Table 3. Comparison of Mortality between Aspirin-Treated and Aspirin-Untreated PatientsAccording to Stroke Subtype among Patients with Causally Defined Stroke (n 5 1457)

Stroke Subtypes Aspirin No AspirinRelative Risk

(95% Confidence Interval)*

Number Who Died/Numberwith Stroke Subtype (Percent)

Large vessel atherosclerosis 55/372 (15) 98/474 (21) 0.7 (0.5–0.9)Total infarct 41/246 (17) 74/228 (30) 0.6 (0.5–0.8)Partial infarct 14/126 (11) 24/246 (10) 1.1 (0.6–2.3)

Small vessel occlusion 18/187 (10) 23/214 (11) 0.9 (0.5–3.0)Cardioembolic 19/91 (21) 41/119 (34) 0.6 (0.4–0.9)

Total infarct 12/49 (29) 29/57 (46) 0.6 (0.4–0.8)Partial infarct 7/42 (14) 12/62 (21) 0.6 (0.03–4.2)

Total 92/650 (14) 162/807 (20) 0.7 (0.6–0.9)

* Relative risks compare mortality in patients using aspirin at the time of the stroke with those not using aspirin.

Table 4. Comparison of Mortality between Aspirin-Treated and Aspirin-Untreated Patients in AllPatients with Ischemic Stroke, Regardless of Cause (n 5 2431)

Group (Number) Aspirin No AspirinRelative Risk

(95% Confidence Interval)*

Number Who Died/Numberwith Stroke Subtype (Percent)

Ischemic strokeDetermined cause (1,457) 92/650 (14) 162/807 (20) 0.7 (0.6–0.9)Undetermined cause (555) 44/303 (15) 55/252 (22) 0.7 (0.5–0.9)Anticoagulant therapy (66) — 7/66 (11) —

Hemorrhagic stroke (353) 4/17 (24) 76/336 (23) 1.0 (0.3–3.3)All stroke patients (2,431)† 140/970 (14) 300/1461 (21) 0.7 (0.5–0.8)

* Relative risks compare mortality in patients using aspirin at the time of the stroke with those not using aspirin.† Includes 36 patients treated with dipyridamole.

Does Prior Aspirin Use Affect Stroke Outcome/Kalra et al

208 February 15, 2000 THE AMERICAN JOURNAL OF MEDICINEt Volume 108

ences in mortality associated with the prior use of aspirinin one type of stroke may not be detected because of thediluting effect of a lack of change in other types. We in-cluded a large sample and categorized stroke patients intocausal subtypes using a validated tool.

Studies such as this one, however, are susceptible tobias because of inconsistencies in subtype classification,inadvertent errors in data collection, and the possibleconfounding effect of early treatment with low-dose as-pirin. Bias in the classification of stroke subtype and ex-tent was reduced in this study by excluding patients withincomplete investigations or in whom there was disagree-ment about causal subtype between observers. Othermeasures to reduce bias included blinding the observerswho categorized the types of stroke to aspirin use, andusing mortality due to any cause as the outcome measure.Unlike neurological or functional scores, mortality is notsusceptible to patient or observer bias.

There were significantly more hypertensive patients inthe group not using aspirin and significantly more pa-tients with ischemic heart disease in the aspirin group,which probably reflects the use of aspirin in clinical prac-tice. These differences had the potential to confound themortality results (13,14). However, aspirin continued tohave an independent effect on mortality even after adjust-ing for these differences. It is also possible that aspirin usewithin the first 48 hours after a stroke may have dilutedthe effect of prior use of aspirin, but this was unavoidablegiven the strong evidence to support the use of aspirin inpatients after a stroke (2,3).

Patients with incomplete investigations or more thanone cause of stroke (stroke of undetermined cause) ac-counted for about one quarter of the patients with isch-emic strokes, similar to previous results (15). However,the proportion of patients with ischemic stroke who hadlarge vessel atherosclerosis (40%) appears to be greaterthan that recorded in other stroke registries (15). This isprobably because the exclusion of patients with no vascu-lar risk factors, equivocal features of stroke, or prior treat-ment with warfarin may have reduced the proportion ofpatients with lacunar and cardioembolic strokes. Al-though differences in categorization may bias the relativecontribution of aspirin in reducing mortality in differentstroke subtypes, this does not affect the overall findingthat prior use of aspirin was associated with a small butsignificant reduction in mortality when all subgroupswere combined (Table 3). In addition, selection bias dueto exclusion from the initial analysis seems unlikely, asthe apparent benefits of aspirin were seen in analyses ofthe whole cohort, regardless of the type or cause of thestroke (Table 4).

That prior use of aspirin may influence outcome inacute ischemic stroke has important implications. Manyacute intervention studies have not adjusted for prior useof aspirin. This interpretation of the results of the studiesmay be affected by aspirin imbalance between placeboand active groups. This imbalance was present in the Na-tional Institute of Neurological Diseases trial on throm-bolysis in acute stroke, where more patients in the activetreatment group were on aspirin compared with the pla-cebo group (16). Further research is needed to define theclinical effectiveness of aspirin in the prevention andtreatment of different types of stroke.

REFERENCES1. Antiplatelet Trialists Collaboration Group. Collaborative overview

of randomized trials of antiplatelet therapy I. Prevention of death,myocardial infarction and stroke by prolonged antiplatelet therapyin various categories of patients. BMJ. 1994;308:81–106.

2. International Stroke Trial Collaborative Group. The InternationalStroke Trial (IST). Lancet. 1997;349:1569 –1581.

3. CAST (Chinese Acute Stroke Trial) Collaborative Group. CAST:randomized placebo-controlled trial of early aspirin use in 20,000patients with acute ischemic stroke. Lancet. 1997;349:1641–1649.

4. Grotta JC, Lemak NA, Gary H, et al. Does platelet antiaggreganttherapy lessen the severity of stroke? Neurology. 1985;35:632– 636.

5. Joseph R, Han E, Grunfeld S, Welch KMA. Prior use of aspirin mayreduce stroke severity. Ann Neurol. 1989;26:131.

6. De Keyser J, Herroelen L, De Kippel N. Early outcome in acuteischemic stroke is not influenced by the use of low dose aspirin.J Neurolog Sci. 1997;145:93–96.

7. Karepov V, Bornstein NM, Hass Y, Korczyn AD. Does daily aspirindiminish severity of first-ever stroke? Arch Neurol. 1997;54:1369 –1371.

8. Sivenius J, Cunha L, Diener H-C, et al. Antiplatelet therapy does notaffect subsequent stroke severity. J Neurolog Sci. 1997;150:S84 –S85.

9. WHO Special Report. Stroke—1989: recommendations on strokeprevention, diagnosis and therapy. Stroke. 1989;20:1407–1431.

10. Adams HP, Bendixen HB, Kappelle JL, et al. Classification of sub-type of acute ischemic stroke: definitions for use in a multicentreclinical trial. Stroke. 1993;24:35– 41.

11. Bamford J, Sandercock P, Dennis M, et al. Classification and natu-ral history of clinically identifiable subtypes of cerebral infarction.Lancet. 1991;337:1521–1526.

12. Bamford JM, Warlow CP. Evolution and test of the lacunar hypoth-esis. Stroke. 1988;19:1074 –1082.

13. Bornstein NH, Karepov VG, Aronovich BD, et al. Failure of aspirintreatment after stroke. Stroke. 1994;25:275–277.

14. Puranen J, Laasko M, Riekkinen P, Sivenius SJ. Risk factors andantiplatelet therapy in TIA and stroke patients. J Neurolog Sci. 1998;154:200 –204.

15. Rothrock JF, Lyden PD, Brody ML, et al. An analysis of ischemicstroke in an urban southern California population. The Universityof California, San Diego, Stroke Data Bank. Arch Intern Med. 1993;153:619 – 624.

16. The National Institute of Neurological Disorders and Stroke rtPAStroke Study Group. Tissue plasminogen activator for acute isch-emic stroke. NEJM. 1995;333:1581–1587.

Does Prior Aspirin Use Affect Stroke Outcome/Kalra et al

February 15, 2000 THE AMERICAN JOURNAL OF MEDICINEt Volume 108 209