Physical Activity and Risk of Stroke in Women

Context Persuasive evidence has demonstrated that increased physical activity is associated with substantial reduction in risk of coronary heart disease. However, the role of physical activity in the prevention of stroke is less well established.

Objective To examine the association between physical activity and risk of total stroke and stroke subtypes in women.

Design and Setting The Nurses' Health Study, a prospective cohort study of subjects residing in 11 US states.

Subjects A total of 72,488 female nurses aged 40 to 65 years who did not have diagnosed cardiovascular disease or cancer at baseline in 1986 and who completed detailed physical activity questionnaires in 1986, 1988, and 1992.

Main Outcome Measure Incident stroke occurring between baseline and June 1, 1994, compared among quintiles of physical activity level as measured by metabolic equivalent tasks (METs) in hours per week.

Results During 8 years (560,087 person-years) of follow-up, we documented 407 incident cases of stroke (258 ischemic strokes, 67 subarachnoid hemorrhages, 42 intracerebral hemorrhages, and 40 strokes of unknown type). In multivariate analyses controlling for age, body mass index, history of hypertension, and other covariates, increasing physical activity was strongly inversely associated with risk of total stroke. Relative risks (RRs) in the lowest to highest MET quintiles were 1.00, 0.98, 0.82, 0.74, and 0.66 (P for trend=.005). The inverse gradient was seen primarily for ischemic stroke (RRs across increasing MET quintiles, 1.00, 0.87, 0.83, 0.76, and 0.52; P for trend=.003). Physical activity was not significantly associated with subarachnoid hemorrhage or intracerebral hemorrhage. After multivariate adjustment, walking was associated with reduced risk of total stroke (RRs across increasing walking MET quintiles, 1.00, 0.76, 0.78, 0.70, and 0.66; P for trend=.01) and ischemic stroke (RRs across increasing walking MET quintiles, 1.00, 0.77, 0.75, 0.69, and 0.60; P for trend=.02). Brisk or striding walking pace was associated with lower risk of total and ischemic stroke compared with average or casual pace.

Conclusion These data indicate that physical activity, including moderate-intensity exercise such as walking, is associated with substantial reduction in risk of total and ischemic stroke in a dose-response manner.

Persuasive evidence has demonstrated that increased physical activity is associated with substantial reduction in risk of coronary heart disease.¹ However, the role of physical activity in the prevention of stroke is less well studied, and results from epidemiological studies have been inconsistent. A significant inverse association between increasing physical activity and risk of stroke has been observed in some studies^1-6 but not in others.^7-10 Also, the dose-response relationship between physical activity and stroke has not been well characterized. Some studies have demonstrated a monotonic decreasing risk with increasing physical activity,^2-4 while others have indicated a U-shaped relationship.^6,11 In addition, few studies have examined the effects of physical activity on subtypes of stroke.^1,2,6 Furthermore, most previous studies have focused on men^2,3,5-8,11; data on women are sparse.^1,12 In the Framingham Heart Study,¹² high levels of physical activity were protective against total stroke risk in men but not in women.

Current guidelines from the Centers for Disease Control and Prevention¹³ and the National Institutes of Health¹⁴ recommend that Americans should accumulate at least 30 minutes of moderate-intensity physical activity on most, preferably all, days of the week. However, the role of low- and moderate-intensity activities such as walking, compared with vigorous exercise, in the prevention of cardiovascular disease remains controversial. If walking is confirmed to provide the same benefits as more vigorous forms of physical activity, it will have important public health implications because walking is the most popular form of physical activity, especially among middle-aged and older women.¹⁵

In this study, we examined in detail the relationship between physical activity and risk of stroke in a large prospective cohort of women. We specifically assessed the role of walking compared with vigorous activities in the prevention of stroke. We also examined the influence of change in activity level on subsequent risk of stroke.

The Nurses' Health Study cohort was established in 1976, when 121,700 female registered nurses aged 30 to 55 years and residing in 1 of 11 US states responded to mailed questionnaires regarding their medical history and health practices.¹⁶ In 1986, 82,409 women responded to a physical activity questionnaire. Of these women, 90% repeated the same physical activity questionnaire in 1988, and 89% in 1992. After exclusion of women who were diagnosed as having myocardial infarction, stroke, angina, or other cardiovascular disease or coronary bypass surgery (n=4470) or cancer (n=5451) through 1986, the analysis included 72,488 women aged 40 to 65 years.

In 1986, 1988, and 1992, participants were asked the average amount of time they spent per week doing each of the following physical activities: walking, jogging, running, bicycling, calisthenics/aerobics/aerobic dance/rowing machine, lap swimming, squash/racquetball, and tennis. They were also asked about their usual walking pace, specified as easy (<2.0 mph), moderate (2.0-2.9 mph), brisk (3.0-3.9 mph), or very brisk (≥4 mph). From this information, energy expenditure in metabolic equivalent tasks (METs), measured in hours per week, was calculated.¹⁷ Because only 2% of the subjects reported very brisk pace, we combined the brisk and very brisk categories in the analyses of walking pace and stroke risk. We defined any physical activity requiring 6 METs or more (≥6-fold increase from resting metabolic rate) as vigorous. These activities included jogging, running, bicycling, calisthenics/aerobic/aerobic dance/rowing machine, lap swimming, squash/racquetball, and tennis. In contrast, walking requires an energy expenditure of only 2.0 to 4.5 METs, depending on pace; we therefore considered it to be a moderate-intensity activity.

On the 1980 questionnaire, women were asked to report the average number of hours they spent each week during the past year on moderate and vigorous recreational activities, such as heavy gardening, vigorous sports, jogging, brisk/very brisk walking, bicycling, and heavy housework. On the 1982 questionnaire, women were asked a slightly different question: "For how many hours per week, on average, do you engage in activity strenuous enough to build up a sweat?" To use this information, we created a variable representing average hours per week spent doing moderate or vigorous recreational activities (all activities described heretofore except for hours spent walking at an easy or normal pace) across 1980, 1982, 1986, 1988, and 1992. To examine the effects of change in physical activity levels on subsequent risk of stroke, we calculated the differences in average hours per week of moderate or vigorous activities between 1980 and 1986 for women who reported on these activities in both periods (n=62,983).

The detailed physical activity questionnaire that was used in this study has been validated in a number of settings. In a representative sample (n=147) of participants in the Nurses' Health Study II cohort, this questionnaire was completed on 2 occasions 2 years apart, in conjunction with past-week activity recall and 7-day activity diaries completed 4 times during a 1-year period.¹⁸ The 2-year test-retest correlation for activity was 0.59. The correlation between physical activity reported on 1-week recalls and that reported on the questionnaire was 0.79. The correlation between activity reported in diaries and that reported on the questionnaire was 0.62. In a separate study on a population aged 20 to 59 years recruited from a university community (n=103), the correlation between physical activity score on a very similar questionnaire and maximum oxygen consumption was 0.54.¹⁹ These data indicate relatively good validity and reproducibility for the questionnaire.

The end point was incident stroke occurring between return of the baseline questionnaires in 1986 and June 1, 1994. Women who reported stroke on follow-up questionnaires were asked for permission to review medical records; these were reviewed by a physician without knowledge of the participant's exposure status. Cerebrovascular pathology due to infection, trauma, or malignancy was excluded. Incident strokes were confirmed by medical record review according to National Survey of Stroke criteria,²⁰ requiring a constellation of neurologic deficits, sudden or rapid in onset and lasting at least 24 hours or until death.

Strokes were classified as subarachnoid hemorrhage, intracerebral hemorrhage, ischemic stroke (embolic infarction or thrombotic infarction), or stroke of undetermined type, according to Perth Community Stroke Study criteria and based on computed tomography (CT), magnetic resonance imaging (MRI), or autopsy findings.²¹ Subarachnoid hemorrhage was defined as hemorrhage in the subarachnoid space, usually caused by rupture of a saccular aneurysm of the cerebral arteries or, less commonly, by arteriovenous malformations or other causes. Intracerebral hemorrhage was defined as hemorrhage in intracerebral regions of the brain not due to an aneurysm or arteriovenous malformation.

Hospital records were available for approximately 88% of nonfatal cases; among these subjects, the percentage of cases with CT or MRI documentation increased over time, reaching 82% for CT, 28% for MRI, and 93% for CT and/or MRI findings in 1988-1994. Overall, CT/MRI findings were present for 88% of women for whom hospital records were obtained. If no records could be obtained, strokes were considered probable if they were corroborated by additional information provided by letter or interview and the subject required hospitalization. Analyses that excluded probable cases yielded similar results.

Deaths were reported by next of kin, coworkers, postal authorities, or the National Death Index. Using all sources combined, we estimated that follow-up for deaths was more than 98% complete.²² Fatal stroke was confirmed using medical records or autopsy reports (74%) or considered probable if these records were not obtainable but stroke was listed as the underlying cause on the death certificate.

Our primary analyses used 1986 as baseline. Person-time for each participant was calculated from the date of return of the 1986 questionnaire to the date of confirmed stroke, death from any cause, or June 1, 1994, whichever came first. Relative risks (RRs) were computed as the incidence rate in a specific MET quintile divided by that in the lowest quintile, with adjustment for 5-year age categories. Tests of linear trend across increasing MET quintiles were conducted by treating the quintiles as a continuous variable and assigning the median score for each quintile as its value. To best represent long-term physical activity levels of individual subjects and reduce measurement error, we created measures of cumulative average METs from all available questionnaires up to the start of each 2-year follow-up interval. A similar method for analyzing repeated dietary measurements has been described in detail elsewhere.²³ We also used restricted cubic spline transformations with 4 knots to flexibly model the relation between physical activity level and stroke risk, avoiding the need for prior specification of the risk function or the location of a threshold exposure value.²⁴

In a secondary analysis, we used 1980 as baseline. We used the continuous values of hours per week to compute cumulative averages of physical activity at each period and categorized the hours per week into 5 levels (<1.0, 1.0-1.9, 2.0-3.9, 4.0-6.9, and ≥7.0 h/wk) after averaging. To examine the effects of change in physical activity on risk of stroke, we related the difference in hours spent on moderate and/or vigorous activities between 1980 and 1986 to incident cases of stroke occurring between 1986 and 1994.

We used pooled logistic regression across the five 2-year intervals,²⁵ which is asymptotically equivalent to Cox regression, to adjust simultaneously for potential confounding variables such as age (5-year categories); smoking status (never, past, or current smoking of 1-14, 15-24, or ≥25 cigarettes/d); alcohol consumption (0, 1-4, 5-14, or ≥15 g/d); body mass index (quintiles); menopausal status (premenopausal, postmenopausal without hormone replacement therapy, postmenopausal with past hormone replacement therapy, or postmenopausal with current hormone replacement therapy); aspirin use (<1 time/wk, 1-6 times/wk, or ≥7 times/wk); parental history of myocardial infarction before age 60 years; and history of diabetes, hypercholesterolemia, or hypertension at baseline. Because fruit and vegetable intake was inversely associated with risk of stroke in our cohort,²⁶ we adjusted for fruit and vegetable intake (both in quintiles) in a secondary analysis.

During 8 years (560,087 person-years) of follow-up, we documented 407 incident cases of stroke (258 ischemic strokes, 67 subarachnoid hemorrhages, 42 intracerebral hemorrhages, and 40 strokes of unknown type). As described elsewhere,²⁷ women who were more physically active tended to be leaner and were less likely to be current smokers. Increasing total physical activity level was strongly associated with progressively lower risk of total stroke (Table 1). Age-adjusted RRs of total stroke across increasing MET quintiles for total physical activity were 1.00, 0.87, 0.68, 0.57, and 0.49 (P for trend <.001). Further adjustment for smoking, body mass index, and other covariates only somewhat attenuated the association for total stroke (RRs across increasing MET quintiles, 1.0, 0.98, 0.82, 0.74, and 0.66; P for trend=.005). Additional adjustment for intake of fruits and vegetables did not materially alter the results (corresponding RRs, 1.0, 0.97, 0.80, 0.71, and 0.63; P for trend=.003). Further adjustment for antihypertensive, cholesterol-lowering, or hypoglycemic medications did not change the results. The inverse association was primarily observed for ischemic stroke (multivariate RRs across increasing MET quintiles, 1.0, 0.87, 0.83, 0.76, and 0.52; P for trend=.003). Significant trends indicate an overall linear relationship between physical activity level and risk of total and ischemic stroke. Spline regression analysis demonstrated a dose-response relationship between physical activity level and incidence of ischemic stroke (Figure 1). Physical activity level had no significant relationship with either subarachnoid hemorrhage (RR for lowest vs highest MET quintile, 0.77; 95% confidence interval [CI], 0.36-1.66; P for trend=.64) or intracerebral hemorrhage (RR for lowest vs highest MET quintile, 1.20; 95% CI, 0.45-3.19; P for trend=.34). The wide CIs of these estimates are in part due to the small number of cases. Thus, we combined intracerebral hemorrhage and subarachnoid hemorrhage in subsequent analyses (Table 1).

In secondary analyses, the inverse association between physical activity and risk of total stroke persisted in subgroup analyses according to body mass index, current smoking status, and parental history of myocardial infarction. Multivariate RRs of total stroke comparing the extreme MET quintiles were 0.61 for current smokers and 0.68 for nonsmokers; 0.64 for women with a body mass index of 29 kg/m² or less and 0.61 for women with a body mass index of more than 29 kg/m²; and 0.64 for women without parental history of myocardial infarction and 0.57 for women with parental history of myocardial infarction. To avoid potential bias due to preclinical conditions, we eliminated stroke cases that occurred in the first 2 years of follow-up, and the results did not appreciably change (390 incident cases were included in the analyses; multivariate RRs across MET quintiles, 1.0, 0.92, 0.84, 0.76, and 0.59; P for trend=.004).

To evaluate the long-term effects of physical activity, we examined the cumulative averages of physical activity level from 1980, 1982, 1986, 1988, and 1992 in relation to incident stroke from 1980 to 1994 (695 stroke cases with 1,168,015 person-years of follow-up). Multivariate RRs across categories of average hours per week spent on moderate/vigorous physical activity (<1 hour, 1-1.9 hours, 2-3.9 hours, 4-6.9 hours, and ≥7 hours) were 1.0, 0.83, 0.90, 0.79, and 0.60, respectively (P for trend=.01). This inverse association was primarily observed for ischemic stroke.

After adjusting for age, walking was associated with a graded reduction in risk of total stroke (RRs across increasing MET quintiles, 1.0, 0.71, 0.66, 0.54, and 0.49; P for trend <.001) (Table 2). This association was somewhat attenuated after adjustment for other risk factors and vigorous physical activity (multivariate RRs across increasing MET quintiles, 1.0, 0.76, 0.78, 0.70, and 0.66; P for trend=.01). To minimize residual confounding by vigorous physical activity, we conducted an additional analysis excluding women who performed vigorous physical activity and obtained similar results (RRs across increasing MET quintiles, 1.0, 0.71, 0.78, 0.74, and 0.64; P for trend=.10). As with total physical activity level, the inverse association for walking was primarily observed for ischemic stroke (Table 2).

Independent of the number of hours spent walking, walking pace was strongly associated with risk of stroke. Compared with women whose usual walking pace was easy, multivariate RRs of total stroke were 0.81 (95% CI, 0.63-1.03) for women with moderate pace and 0.49 (95% CI, 0.36-0.68) for women with brisk/very brisk pace (Table 3). This inverse association appeared to be particularly strong for ischemic stroke. The reduction in risk of hemorrhagic stroke associated with brisk/very brisk walking pace was not statistically significant.

To assess whether more vigorous activity had an increased benefit beyond that of walking, we examined risk of stroke according to the joint distribution of METs from walking and nonwalking vigorous physical activity. Equivalent energy expenditure from walking and vigorous physical activity resulted in comparable risk reductions. Women in the highest categories (≥7 METs) of both vigorous physical activity and walking had an RR of 0.30 (95% CI, 0.16-0.57) for ischemic stroke compared with the most sedentary women (0 METs from vigorous physical activity and <0.7 METs from walking). When METs for both walking and vigorous activity were entered into the model as continuous variables simultaneously, RRs of ischemic stroke associated with 10-MET increases in energy expenditure were 0.82 (95% CI, 0.69-0.97) for vigorous activity and 0.83 (95% CI, 0.69-0.99) for walking.

We examined changes in physical activity between 1980 and 1986 in relation to incidence of stroke from 1986 to 1994 (Table 4). After adjustment for baseline physical activity level and other covariates, each 3.5-h/wk increase in moderate/vigorous physical activity from baseline was associated with a 19% reduction in total stroke and a 29% reduction in ischemic stroke. Increasing physical activity levels were not significantly associated with risk of hemorrhagic stroke. Compared with women who were consistently sedentary (<1 h/wk of moderate/vigorous physical activity in both 1980 and 1986), those who were consistently active (≥4 h/wk of physical activity in both 1980 and 1986) had the lowest risk of ischemic stroke (RR, 0.46; 95% CI, 0.22-0.96).

In this large prospective cohort study of women, greater leisure-time physical activity was associated with reduced risk of stroke in a dose-response manner. Independent of vigorous physical activity, walking was associated with a substantial reduction in stroke risk, and brisk/very brisk usual walking pace was independently associated with decreased risk compared with normal or easy pace. We observed comparable magnitudes of risk reduction with equivalent energy expenditures from walking and vigorous physical activity.

Findings from previous prospective cohort studies on regular physical activity and risk of stroke have been inconsistent. A significant inverse association between increasing physical activity and stroke has been observed in some studies^1-6 but not in others.^7-10 Also, the dose-response relationship between physical activity and stroke has not been well characterized. Some studies have demonstrated a monotonic decreasing risk with increasing physical activity,^2-4 while others have indicated a U-shaped relationship.^6,11 Small sample sizes and inadequate physical activity assessment may partially account for these discrepancies. The NIH Consensus Development Panel on Physical Activity and Cardiovascular Health¹⁴ concluded that "data are inadequate to determine whether stroke incidence is affected by physical activity or exercise training." Similarly, the surgeon general's report on physical activity and health²⁸ concluded that "it is unclear whether physical activity plays a protective role against stroke." With large sample size and detailed and repeated measures of physical activity, our study provides strong evidence for a graded inverse relationship between physical activity levels and risk of stroke.

Few previous studies have examined the effects of physical activity on ischemic and hemorrhagic stroke separately. Sacco and colleagues⁴ observed a strong dose-response relationship between leisure-time physical activity and risk of ischemic stroke, but the study did not examine the association with hemorrhagic stroke. Gillum and colleagues¹ found a significant positive association between lower levels of nonrecreational activity and increased risk of total and nonhemorrhagic stroke. In contrast, 2 other studies^5,12 suggest a stronger inverse association with hemorrhagic stroke than ischemic stroke. However, neither study obtained detailed or repeated measurements of physical activity. Our data support the role of physical activity, including walking, in the prevention of ischemic stroke rather than hemorrhagic stroke. The risk reduction in ischemic stroke observed in our study was similar to that for coronary heart disease, reflecting shared risk factors and atherothrombotic origin of these 2 diseases. Although hypertension is also a risk factor for both intracerebral and subarachnoid hemorrhage, the pathophysiology of these events is not predominantly atherogenic in nature. However, our power for detecting an association with hemorrhagic stroke is limited due to a smaller number of cases.

Our data suggest that similar energy expenditure from walking and vigorous physical activity confer similar reductions in stroke risk and that substantial reduction in stroke risk appears achievable through a moderate amount of walking. This finding is reassuring, since walking is a physical activity that is highly accessible, readily adopted, inexpensive, and rarely associated with exercise-related injury. In our previous studies, METs for walking and brisk walking pace were independently associated with lower risk of type 2 diabetes²⁹ and coronary heart disease.²⁷ In the Honolulu Heart Study, walking distance was inversely associated with risk of coronary heart disease³⁰ and total mortality.³¹ These findings have important public health implications because walking is the most popular form of physical activity, especially among middle-aged and older women.¹⁵

Another important finding of our study is that sedentary women who became active in middle and later adulthood had lower stroke risk than their counterparts who remained sedentary. This implies a relatively prompt effect of physical activity—older adults can enjoy the benefit of exercise even if they were sedentary for a long time.

The protective effect of physical activity may be partly mediated through its effects on various risk factors for stroke.³² Physical activity lowers blood pressure and increases high-density lipoprotein cholesterol concentration. It has been associated with reductions in plasma fibrinogen level and platelet aggregation and elevations in plasma tissue plasminogen activator activity.³² Physical activity facilitates weight loss and weight maintenance.³³ It can increase insulin sensitivity because of increased number and activity of glucose transporters, both in muscle and adipose tissue.^34,35 Convincing epidemiological data demonstrate the beneficial effects of physical activity on risk of type 2 diabetes,³⁶ an important risk factor for stroke.

Our physical activity questionnaire has been validated against a physical activity diary,¹⁸ and similar questionnaires have correlated well with measured oxygen consumption.³⁷ Although some error in self-report is inevitable, because of the prospective design of this study, misclassification would be nondifferential with respect to the outcome and would bias the results toward the null. The present study is the largest cohort for whom data on physical activity and health outcomes were collected prospectively, and it is the only large study examining the association between physical activity and stroke in a female population. The present study is also the only one in which physical activity exposures were updated after the initial assessment, with detailed examination of moderate vs vigorous activity.

In conclusion, increasing physical activity levels are associated with substantial reductions in risk of total and ischemic stroke in women. We observed comparable magnitudes of risk reduction with similar energy expenditure from walking and vigorous physical activity. Our findings lend further support to current guidelines from the Centers for Disease Control and Prevention¹³ and the National Institutes of Health¹⁴ that promote regular moderate-intensity physical activity for prevention of chronic diseases.