Table of Contents

Most risk factors for non-hemorrhagic stroke are common knowledge for internists, although we may not know their relative importance. Independent risk factors for brain infarction found in the Framingham Heart Study are age, hypertension (systolic and diastolic), diabetes, cardiovascular disease, smoking, atrial fibrillation and left ventricular hypertrophy (LVH) by ECG. Utilizing Framingham data, a point system has been developed which allows us to calculate the 10-year stroke risk for an individual patient with these risk factors (Appendix 1).[ii] Of these risk factors, only hypertension and cigarette smoking are directly modifiable; however, since hypertension is far and away the most important risk factor for stroke apart from age, such a calculation can add relevance to anti-hypertensive treatment. The authors given the example of a 70 year-old man who is diabetic, smokes and has a systolic blood pressure (SBP) of 180 mmHg despite being on treatment, giving him a score of 20 points: his 10-year risk for stroke is about 37 percent. A 70-year old diabetic who does not smoke and has a SBP of 120 has a score of 11 points and a 10-year risk for stroke of 10 percent. If our patient stops smoking and his SBP is better controlled, his risk will trend downward towards this lower number. 

Hypertension is the most important determinant of stroke risk apart from age, and treatment has been shown to modify this risk. The relationship between hypertension and stroke grows stronger with age relative to most other risk factors, even as the incidence of hypertension markedly increases.[iii] Meta-analysis of observational studies comprising over 400,000 subjects demonstrates a continuous decrease in risk of stroke directly related to diastolic blood pressure (DBP), with no evidence of increased risk of DBP down to 69 mmHg (i.e., no evidence of a “J-shaped” curve).[iv] Thus, risk is seen to decrease even among DBPs within the normotensive range. Notably, the association of DBP and stroke is stronger and more responsive than that of DBP and coronary heart disease. This association was found to be similar in women in the studies that enrolled female subjects. Furthermore, meta-analysis of randomized trials of drug treatment of hypertension comprising over 37,000 subjects finds that stroke risk (and vascular mortality) is reduced within two to three years to levels commensurate with DBPs found in the meta-analysis of observational studies.[v] This potential for risk reduction remains for older persons with isolated systolic hypertension, as shown in the SHEP study (Systolic Hypertension in the Elderly Program), where treatment with chlorthalidone +/- atenolol reduced risk of stroke and of cardiovascular death.[vi] SHEP data suggest that 30 to 40 elderly people would have to be treated for five years to prevent one stroke; most would agree that the high prevalence of this condition makes this a clinically relevant effect.

Cigarette smoking is another modifiable risk factor that has been studied. Both Framingham and the Nurses Health Study find a dose-dependent relationship of cigarette smoking to stroke risk.[vii]^,[viii]Roughly, smokers with a one to two pack per day habit have a risk of stroke twice that of smokers who smoke a half-pack or less per day. Framingham data allows calculation of stroke risk in people who stop smoking, and find that stroke risk decreases substantially within two years of cessation, and returns within five years to that of people who never smoked.

Other potentially modifiable risk factors include hypercholesterolemia, diabetes, and alcohol. Stroke risk attributable to alcohol use has not been quantified, presumably due to confounding variables. Modification of stroke risk by treatment of diabetes mellitus has not been studied. Hypercholesterolemia is an independent risk factor for stroke in people under 55 years of age, but the overall risk of stroke in this group is quite low. The Multiple Risk Factor Intervention Trial, enrolling over 350,000 men from 35 to 57 years of age, found an increased risk of death due to non-hemorrhagic stroke varying directly with total serum cholesterol levels above 200 mg/dl; however it also found an increased risk in hemorrhagic stroke with cholesterol levels less than 160 mg/dl.[ix] The relevance of this result is tempered by the fact that even for hypertensive men with total cholesterol over 280 mg/dl, the risk of stroke death is less than 0.2 percent; non-fatal stroke data was not collected. Framingham data suggest that neither total cholesterol nor lipoprotein fractions remain independent risk factors for stroke beyond age 55, although total cholesterol does predict development of carotid stenosis.^3,[x]

Risk of stroke attributable to carotid artery stenosis can be modified by surgery in some cases. Population-based studies and medical treatment arms of studies of asymptomatic carotid artery stenosis indicate that overall risk of stroke is somewhere around two percent per year for people with stenoses greater than 70 percent.[xi]^,[xii]^,[xiii] If a person has a stroke or TIA, this risk increases to around 13 percent per year. Three large studies have confirmed both a stroke and a mortality benefit of carotid endarterectomy over aspirin in asymptomatic patients with carotid artery stenosis > 70 percent, with absolute risk reductions of around 17 percent over two years, implying that five patients would need to have surgery to prevent stroke during this time.[xiv]^,[xv]^,[xvi] No studies have employed warfarin in the medical treatment arm.

Most vexing is the question of asymptomatic carotid artery stenosis: first, because it must be identified and second, because the benefits to be gained from this risky procedure are much smaller. Typically, we think of cervical bruits as being predictive of carotid artery stenosis; however a review of the literature published as part of “The Rational Clinical Examination” series finds that, although there is good inter-rater reliability as to the presence or absence of bruits, in people with a history of TIAs, both the sensitivity and the specificity of this finding are quite poor, around 60 to 70 percent, when compared to a gold standard of angiography.[xvii] In the Asymptomatic Carotid Atherosclerosis Study (ACAS), about 75 percent of enrolled subjects had ipsilateral cervical bruits, raising the question of how the rest were identified and whether or not the presence of a cervical bruit predicts a better or worse outcome.[xviii] In that study and the other major recent study of endarterectomy in “asymptomatic” patients, the carotid stenoses of almost one-third of enrolled subjects were identified during evaluation for TIA or stroke affecting the contralateral circulation; therefore the definition of “asymptomatic” must be regarded carefully when applying the findings of these studies.[xix]

Both of these studies found a benefit to carotid endarterectomy of asymptomatic stenoses, but due to the relatively low baseline risk, small methodological concerns may greatly affect our ability to apply the results. The smaller VA study could only demonstrate a benefit when all neurological events, including TIAs, were taken into account, and a mortality benefit could not be demonstrated over a mean of four years of follow-up. ACAS was able to demonstrate a six percent absolute risk reduction for ipsilateral stroke or death over a mean of 2.7 years (number needed to treat = 17), and it was projected that this would be clinically significant over a five year span. They did not find a significant risk reduction in women. ACAS also points out that angiography is not an entirely benign procedure: subjects were randomized before angiography, so that only those going to surgery would bear the increased risk, and there was a 1.2 percent rate of minor stroke due to angiography. While ACAS appears to have been well-designed and implemented, and its results are supported by the VA study, it remains a difficult task to weigh risks and benefits of surgery for an asymptomatic patient.

Stroke or TIA due to any cause increases the risk of subsequent stroke and there is some evidence for the effectiveness of secondary prophylaxis. Annual risk of stroke after an initial cerebrovascular event is 4.5 to 6 percent, although this is as high as 10 percent the first year, and is increased by the presence of other risk factors, especially hypertension.[xx]^,[xxi] In the Northern Manhattan Stroke Study about 30 percent of people with stroke had prior stroke or TIA.¹Anticoagulation has not been shown to decrease the risk of stroke or death in this scenario. A review of the literature and pooled analysis of the data found an increase in stroke and death with secondary anticoagulation, although data from studies done after 1974 tends towards an insignificant benefit.[xxii] We must wait for the results of the ongoing Warfarin vs. Aspirin in Recurrent Stroke Study (WARSS) for more substantial data.. Aggregated results of trials using aspirin alone as secondary prophylaxis after TIA finds a relative risk reduction of about 15 percent, but the rate of recurrent stroke in these trials tends to be less than that seen in observational studies, so that it appears that around 20 to 30 people need to be treated for five years to prevent one stroke.^19,[xxiii] The dose of aspirin has not been shown to be important.¹⁹ Ticlopidine has been touted as being more effective than aspirin, but the large trial that compared the two found only a 12 percent relative risk reduction of ticlopidine as compared with aspirin (i.e. 12 percent of a 15 percent risk reduction), and the 95 percent confidence interval ranged from -2 to 26 percent.[xxiv] The authors of this study have since tried to enhance its validity by multiple subgroup analyses.[xxv] Beware.

Myocardial infarction (MI) is a risk factor for stroke, especially within the first month after the event, when overall rates are about 3 percent.¹⁹ For MI overall,

two randomized double-blinded placebo-controlled studies with a total of over 2000 subjects demonstrated a significant mortality benefit attributable to the use of anticoagulation (number needed to treat = 20 to 25) and, taken together, also demonstrated a significant decrease in stroke of around 40 percent over two to three years (number needed to treat = 25-35).[xxvi]^,[xxvii]^,[xxviii] A third study of 3400 subjects, excluding people with other cardiac risk factors for stroke (thrombus, aneurysm, atrial fibrillation), did not look for a mortality benefit, but found a similar risk reduction over three years despite the lower event rate (number needed to treat = 50).[xxix] Presumably, these findings were not widely applied due to an increase in complications in the treatment groups and extra costs of monitoring for anticoagulation, but the mortality benefit is clear. Notably, they all aimed for an INR of 2.5 to 4.5, reportedly recommended for prevention of arterial thromboses. As these studies were done before the post-MI-aspirin era, it would be interesting to know how aspirin would modify this risk reduction. INRs over 3.5 likely carry unacceptable risk of hemorrhage.

Various subgroups have been proposed to account for the higher risk of stroke following MI, such as those with mural thrombus, atrial appendage thrombus, ventricular aneurysm and congestive heart failure. For NYHA Class II to III congestive heart failure, retrospective analysis of V-HEFT data demonstrates only a moderately increased risk for any thromboembolism: about 2.5 events per 100 patient-years; similarly a rate of 1.7 strokes per 100 patient-years was found in 264 outpatients with CHF.[xxx]^,[xxxi] Although many patients in either series were taking aspirin or warfarin, there was no evidence for a treatment effect of either agent. The literature of risk with left-sided thrombus is quite disorganized, with many small series; in studies of effectiveness of anticoagulation, extensive use of surrogate endpoints further complicates the search for a reliable morbidity/mortality benefit. Nevertheless, it seems that a cogent argument could be made for anticoagulation after any MI on the basis of the aforementioned studies. 

Atrial fibrillation is a well-known risk factor for stroke. This issue has been discussed in Chapter 11.

Patients at high risk for stroke can be reliably identified. Lowering blood pressure and smoking cessation can provide the largest reductions in risk without adding risks of their own. If there is a history of stroke or TIA on the same side as a high grade carotid stenosis, carotid endarterectomy provides a clear benefit to patients at low surgical risk; however, for patients with a high grade stenosis but no ipsilateral ischemia/infarction, the benefit is much smaller, subject to controversy and, thus far, has not been demonstrated in women. Aspirin appears to reduce the risk of stroke somewhat in people who have already had TIAs or stroke. Warfarin reduces stroke and death in patients who have had a myocardial infarction, but the effect is small and must be weighed against the risk of major hemorrhage and adherence/quality of life issues with regard to monitoring prothrombin time. There is no clear evidence that warfarin or aspirin reduces the risk of stroke attributable to congestive heart failure, but this risk may not be as great as previously believed. Warfarin is clearly superior to aspirin in reducing the risk of stroke due to atrial fibrillation.

Numbers needed to treat in some cases are rough estimates based on the available data, but I felt that some imprecision was allowable in order to increase the clinical relevance and applicability of these data.

Table of Contents