Physical Activity and Stroke Risk A Meta-Analysis Chong Do Lee, EdD Aaron R. Folsom, MD Steven N. Blair, PED Background and Purpose���Whether physical activity reduces stroke risk remains controversial. We used a meta-analysis to examine the overall association between physical activity or cardiorespiratory fitness and stroke incidence or mortality. Methods���We searched MEDLINE from 1966 to 2002 and identified 23 studies (18 cohort and 5 case-control) that met inclusion criteria. We estimated the overall relative risk (RR) of stroke incidence or mortality for highly and moderately active individuals versus individuals with low levels of activity using the general variance��� based method. Results���The meta-analysis documented that there was a reduction in stroke risk for active or fit individuals compared with inactive or unfit persons in cohort, case-control, and both study types combined. For cohort studies, highly active individuals had a 25% lower risk of stroke incidence or mortality (RR 0.75 95% CI, 0.69 to 0.82) compared with low-active individuals. For case-control studies, highly active individuals had a 64% lower risk of stroke incidence (RR 0.36 95% CI, 0.25 to 0.52) than their low-active counterparts. When we combined both the cohort and case-control studies, highly active individuals had a 27% lower risk of stroke incidence or mortality (RR 0.73 95% CI, 0.67 to 0.79) than did low-active individuals. We observed similar results in moderately active individuals compared with inactive persons (RRs were 0.83 for cohort, 0.52 for case-control, and 0.80 for both combined). Furthermore, moderately and highly active individuals had lower risk of both ischemic and hemorrhagic strokes than low-active individuals. Conclusions���We conclude that moderate and high levels of physical activity are associated with reduced risk of total, ischemic, and hemorrhagic strokes. (Stroke. 2003 34:2475-2482.) Key Words: incidence meta-analysis mortality physical activity stroke Sin troke is a leading cause of death and long-term disability the United States. Approximately one half million US adults suffer from stroke (first attack) each year, with 100 000 recurrent attacks.1 Of these, 160 000 are fatal.2 Since effec- tive treatments for stroke are limited and many stroke survivors require lengthy rehabilitation and chronic care, primary prevention of stroke is imperative to avoid the burden of this disease. Hypertension and cardiac disease are the primary risk factors for stroke.3,4 Physical activity may modify these risk factors for stroke and may have more direct effects to lower stroke risk as well. A systematic review reported that physical activity is inversely associated with incidence of hypertension and coronary heart disease.5 However, whether physical activity alters stroke risk was less clear, partly because there were fewer stroke events in some studies. Some investigators found an inverse association between physical activity and stroke risk.6 ���21 Other studies have shown a U-shaped associ- ation, no association, or a positive association between physical activity and stroke incidence or mortality.22���27 To See Editorial Comment, page 2481 address stroke prevention strategies, it is important to inves- tigate whether physical activity reduces stroke incidence or mortality. We therefore investigated the overall association between physical activity and stroke incidence or mortality from published studies between 1966 and 2002. Materials and Methods Study Selection We searched MEDLINE from January 1966 through July 2002, using the medical subject headings physical activity, exercise, leisure-time activity, stroke, and cardiovascular disease. We also searched the Surgeon General���s report on physical activity and health.28 We reviewed all relevant articles and identified 31 pub- lished epidemiological studies of physical activity and stroke, 23 of which (18 cohort and 5 case-control) met our inclusion criteria. Inclusion criteria were English language reports of any cohort or case-control study in which physical activity (leisure-time activi- ty)6 ���27 or cardiorespiratory fitness29 was classified as low, moderate, or high6,8 ���10,12���21,23���27 or was classified as the least to the highest intensity7,11,22,26 of physical activity. We excluded 8 studies from the Received March 31, 2003 final revision received June 3, 2003 accepted June 24, 2003. From the Department of Sports and Exercise Sciences, West Texas A&M University, Canyon (C.D.L.) Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis (A.R.F.) and Division of Epidemiology and Clinical Applications, The Cooper Institute, Dallas, Tex (S.N.B.). Reprint requests to Dr Chong Do Lee, Department of Sports and Exercise Sciences, West Texas A&M University, Canyon, TX 79016. E-mail CLee@wtamu.edu �� 2003 American Heart Association, Inc. Stroke is available at http://www.strokeaha.org DOI: 10.1161/01.STR.0000091843.02517.9D 2475

analysis. Four studies either classified or analyzed physical activity only as low versus other30 ���33 2 studies analyzed physical activity on a continuous scale34,35 and 2 studies reported death rates or relative risks (RRs) without CIs, and therefore the variances of the RRs were inestimable.36,37 Data Extraction All the data were independently abstracted by 1 investigator (C.D.L.). Measures of association reported within a single study separately for different ethnic groups, sexes, age groups, or outcome measures were analyzed as separate units. For instance, in the National Health and Nutrition Examination Survey (NHANES) I epidemiological follow-up study, we included 5 different data units: white women aged 45 to 64 years, white women aged 65 to 74 years, white men aged 45 to 74 years, white men aged 65 to 74 years, and blacks aged 45 to 74 years.13 For Honolulu Heart Study men, we used 2 data units with different age groups (aged 45 to 54 and 55 to 68 years).6 For the Framingham Study, we separated 2 data units (men and women).12 In the Established Populations for Epidemio- logic Studies of the US Elderly, we included 3 data units: Boston (Mass) elderly, New Haven (Conn) elderly, and Iowa elderly.27 We obtained 23 studies that met inclusion criteria, yielding a total of 31 data units. The studies��� characteristics were recorded as follows: author���s name, publication year, study population (sample size, age, sex, and ethnicity), physical activity classification (low, moderate, high activity intensities), activity type (leisure-time only), follow-up years (cohort studies), outcome measure (stroke incidence or mor- tality), RR (or odds ratio) and CI, and covariates. Statistical Analysis The RR or odds ratio was used to estimate the risk ratio of stroke incidence or mortality for moderately or highly active individuals TABLE 1. Characteristics of 18 Cohort Studies of Physical Activity and the Risk of Stroke Incidence or Mortality Study (Reference) Year Study Population Exposure Follow-Up (Y) Outcome (No. of Events) Covariates Lee and Blair29 2002 Aerobics Center Longitudinal Study (16 878 US men age, 40���87 y) Low fitness 10 Stroke deaths (n 32) Age, examination year, smoking, alcohol intake, BMI, hypertension, diabetes, and parental history of coronary heart disease Ellekjaer et al9 2000 Nord-Trondelag Survey (14 101 Norwegian women age 50 y) Low activity 10 Stroke deaths (n 457) Age, smoking, diabetes, BMI, antihypertensive medication, systolic blood pressure, angina pectoris, MI, illness, and education Hu et al11 2000 Nurses��� Health Study (72 488 US women age, 40���65 y) Low activity 8 Stroke incidence (n 407) Age, time, cigarette smoking, BMI, menopausal status, parental history of MI before age 60 years, alcohol intake, aspirin use, history of hypertension, diabetes, and hypercholesterolemia Lee et al26 1999 Physicians��� Health Study (21 823 US men age, 40���84 y) Low intensity 11.1 Stroke deaths (n 533) Age, treatment assignment, cigarette smoking, alcohol intake, history of angina, parental history of MI at 60 years, BMI, history of hypertension, high cholesterol, and diabetes mellitus Agnarsson et al7 1999 Reykjavik Study (4484 Icelandic men age, 45���80 y) Low intensity 10.6 Stroke incidence (n 249) Age, blood glucose, smoking, hypertension, and ventilatory function Evenson et al10 1999 ARIC Study (6279 US men and 8296 US women age, 45���64 y) Low activity 7.2 Stroke incidence (n 189) Age, sex, race-center, education, and smoking Bijnen et al8 1998 Zutphen Elderly Study (802 Dutch men age, 64���84 y) Low activity 10 Stroke deaths (n 47) Age, baseline stroke, smoking, and alcohol consumption Lee and Paffenbarger22 1998 Harvard Alumni Study (11 130 men age, 43���88 y) Low intensity Stroke deaths (n 378) Age, smoking, alcohol intake, and early parental death Nakayama et al25 1997 Shibata Study (961 Japanese men and 1341 women age 40 y) Light activity 15.5 Stroke incidence (n 141) Age Gillum et al13 1996 NHANES I Follow-up Study (7895 white and black men and women age, 45���74 y) Low activity 11.6 Stroke incidence (n 623) Age, smoking, history of diabetes, history of heart disease, education, systolic blood pressure, total cholesterol, BMI, and hemoglobin Abbott et al6 1994 Honolulu Heart Program (7530 men, Japanese ancestry age, 45���68 y) Inactivity 22 Stroke incidence (n 537) Age Kiely et al12 1994 Framingham Study (1897 US men and 2299 US women age, 28���62 y) Low activity 32 Stroke incidence (n 427) Age, systolic blood pressure, cholesterol, smoking, vital capacity, BMI, left ventricular hypertrophy, atrial fibrillation, valvular disease, history of congestive heart failure, history of ischemic heart disease, and occupation Simonsick et al27 1993 Epidemiologic Studies of the Elderly (4840 US men and women age 65 y) Inactivity 6 Stroke incidence (n ?) Age, sex, education, work status, smoking, respiratory symptoms, MI, diabetes, angina, self-rated health, and modified depression score Haheim et al14 1993 Oslo Study (14 403 Norwegian men age, 40���49 y) Sedentary 12 Stroke incidence (n 81) None Wannamethee and Shaper15 1992 British Regional Heart Study (7735 British men age, 40���59) Inactivity 9.5 Stroke incidence (n 128) Age, social class, smoking, heavy drinking, and BMI Lindsted et al23 1991 Seventh-Day Adventist (9484 men age 30 y) Low activity 26 Stroke deaths (n 410) Race, smoking, education, medical illness, BMI, marital status, and dietary pattern Folsom et al16 1990 Iowa Women���s Health Study (41,837 Iowa women age, 55���69 y) Low activity 2 Stroke incidence (n 218) Age Menotti and Seccareccia24 1985 Italian Railroad Worker (99,029 men age, 40���59) Sedentary 5 Stroke deaths (n 187) Age BMI indicates body mass index MI, myocardial infarction. 2476 Stroke October 2003