Physical Activity and Stroke Incidence: The Harvard Alumni Health Study

I-Min Lee and Ralph S. Paffenbarger, Jr

Stroke. 1998;29:2049-2054
doi: 10.1161/01.STR.29.10.2049
Stroke is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231
Copyright 1998 American Heart Association, Inc. All rights reserved.
Print ISSN: 0039-2499. Online ISSN: 1524-4628

The online version of this article, along with updated information and services, is located on the
World Wide Web at:
http://stroke.ahajournals.org/content/29/10/2049

Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally published
in Stroke can be obtained via RightsLink, a service of the Copyright Clearance Center, not the Editorial Office.
Once the online version of the published article for which permission is being requested is located, click
Request Permissions in the middle column of the Web page under Services. Further information about this
process is available in the Permissions and Rights Question and Answer document.
Reprints: Information about reprints can be found online at:
http://www.lww.com/reprints
Subscriptions: Information about subscribing to Stroke is online at:
http://stroke.ahajournals.org//subscriptions/

Downloaded from http://stroke.ahajournals.org/ by guest on November 4, 2014

Physical Activity and Stroke Incidence

The Harvard Alumni Health Study
I-Min Lee, MBBS, ScD; Ralph S. Paffenbarger, Jr, MD, DrPH
Background and PurposePhysiologically, it appears plausible for physical activity to decrease stroke risk; however,
epidemiological studies have produced mixed findings. Furthermore, few studies have examined specific kinds and
intensities of activities. The purpose of this study was to examine the association between physical activity, including
its various components (walking, climbing stairs, participation in sports and recreational activities), and stroke risk.
MethodsThis was a prospective cohort study of 11 130 Harvard University alumni (mean age, 58 years) without
cardiovascular disease and cancer at baseline. Men reported their walking, stair climbing, and participation in sports or
recreation on baseline questionnaires in 1977. Stroke occurrence was assessed with another questionnaire in 1988. Death
certificates were obtained for decedents through 1990 to determine strokes not previously reported (total strokes5378).
We used Cox proportional hazards regression to estimate the relative risks and 95% CIs for stroke occurrence associated
with physical activity.
ResultsAfter adjustment for age, smoking, alcohol intake, and early parental death, the relative risks of stroke associated
with ,1000, 1000 to 1999, 2000 to 2999, 3000 to 3999, and $4000 kcal/wk of energy expenditure at baseline were 1.00
(referent), 0.76 (95% CI, 0.59 to 0.98), 0.54 (0.38 to 0.76), 0.78 (0.53 to 1.15), and 0.82 (0.58 to 1.14), respectively;
P50.05 for linear trend. Walking $20 km/wk was associated with significantly lower risk, independent of other
physical activity components. Climbing stairs and activities of at least moderate intensity ($4.5 METs, or multiples of
resting metabolic rate) each showed U-shaped relations to stroke risk, with the risk being significantly lower at the nadir
of the curve. Light intensity activities (,4.5 METs), however, were unrelated to stroke risk.
ConclusionsPhysical activity is associated with decreased stroke risk in men. A decreased risk was observed at energy
expenditures of 1000 to 1999 kcal/wk, with further risk decrement seen at 2000 to 2999 kcal/wk but not beyond.
Confirmation of the U-shaped relation observed in these data requires similar observations in other populations. (Stroke.
1998;29:2049-2054.)
Key Words: epidemiology n exercise n risk factors n stroke prevention

troke ranks as the third leading cause of death in the

United States, following coronary heart disease and
cancer.1 Every year, .150 000 persons die from this disease,
while many more are disabled.1 It primarily afflicts older
individuals: .130 000 of the 150 000 stroke deaths each year
occur among those aged $65 years.1 Currently, there is
limited treatment for most types of stroke. Thus, its prevention is of public health importance. Two major causes of
stroke are atherosclerosis of intracranial or extracranial vessels and high blood pressure.2 Therefore, in searching for
prevention strategies, physical activity is promising because it
has beneficial effects on the atherosclerotic process and
reduces blood pressure.3,4 Several studies indeed show that
physical activity is associated with lower stroke risk517;
however, other studies do not support this hypothesis.18 24 In
fact, the Surgeon Generals report on physical activity and

health concluded that it is unclear whether physical activity

plays a protective role against stroke.25
To provide further information, we examined data from an
ongoing study of college alumni. Previous investigation
revealed that alumni who had been varsity athletes experienced less than half the risk of subsequent stroke death.5 In
this study we extend our observations regarding physical
activity and stroke incidence among these alumni later in life.

Subjects and Methods

Participants
The Harvard Alumni Health Study is an ongoing study of the
predictors of chronic diseases among men matriculating as undergraduates at Harvard University between 1916 and 1950. We first
mailed a health questionnaire to surviving alumni in either 1962 or
1966. After this initial survey, we periodically resurveyed all

Received April 7, 1998; final revision received July 13, 1998; accepted July 13, 1998.
From the Department of Epidemiology, Harvard School of Public Health, Boston, Mass (I-M.L., R.S.P); Division of Preventive Medicine, Department
of Medicine, Brigham and Womens Hospital and Harvard Medical School, Boston, Mass (I-M.L.); and Division of Epidemiology, Stanford University
School of Medicine, Stanford, Calif (R.S.P.).
This is report No. LXI in a series on chronic disease in former college students.
Correspondence to I-Min Lee, MBBS, ScD, Brigham and Womens Hospital, 900 Commonwealth Ave E, Boston, MA 02215. E-mail
i-min.lee@channing.harvard.edu
Reprint requests to I-Min Lee, MBBS, ScD, Department of Epidemiology, Harvard School of Public Health, 677 Huntington Ave, Boston, MA 02115.
1998 American Heart Association, Inc.

2049
Downloaded from http://stroke.ahajournals.org/
by guest on November 4, 2014

2050

Physical Activity and Stroke Incidence

surviving alumni in the relevant classes to update information on

health habits and medical history.
For this study we were interested in information from a mail
survey in 1977. Potentially eligible subjects were 17 835 alumni who
returned a questionnaire then. We excluded men reporting physiciandiagnosed cardiovascular disease or cancer (n52863) to avoid
potential bias from men who recently may have changed their
activity level because of these diseases. We further excluded men
with missing data on the variables of interest: physical activity,
cigarette smoking, alcohol consumption, and ages of parental death
(n5673). Of the remaining 14 299 men, we successfully followed
11 130 (ie, they returned another questionnaire in 1988 or were
known to have died by 1990; details below), or 78%. These 11 130
men formed the study group.

Assessment of Physical Activity and Other

Predictors of Stroke
We asked alumni on the 1977 questionnaire (baseline) to estimate the
number of blocks walked daily and the number of flights of stairs
climbed daily and to list all sports or recreation in which they had
actively participated during the past year.26 For each sport or
recreation listed, we asked for details regarding frequency (weeks
per year) and duration (time per week when active). This assessment
of physical activity has been shown to be reliable and valid.2733
We analyzed physical activity in 3 different ways. First, we
examined the relation between total energy expended on physical
activity at baseline and stroke incidence. We estimated total energy
expenditure thus: walking 1 block daily rated 56 kcal/wk; climbing
1 flight of stairs daily required 28 kcal/wk. To each sport or
recreation, we assigned a multiple of resting metabolic rate (MET
score).34 Since resting metabolic rate is '1 kcal/kg body wt per hour,
we estimated the average weekly energy expenditure for that activity
by multiplying its MET score by body weight and hours per year of
participation, then dividing by 52.35 We summed kilocalories per
week from walking, stair climbing, and sports or recreation to
estimate total energy expenditure. We then defined 5 categories:
,1000 (31.0% of men), 1000 to 1999 (28.7%), 2000 to 2999
(18.2%), 3000 to 3999 (10.2%), and $4000 kcal/wk (11.9%).
Second, to understand the effects of specific components of
physical activity, we examined separately walking, stair climbing,
and sports or recreation. We categorized men into approximate
fourths of distance walked (1 block50.13 km): ,5 (31.8% of men),
5 to ,10 (21.3%), 10 to ,20 (26.2%), and $20 km/wk (20.7%).
Similarly, we grouped men into approximate fourths of flights
climbed (2 flights51 story): ,10 (22.9% of men), 10 to ,20
(21.2%), 20 to ,35 (20.8%), and $35 stories/wk (35.1%). Because
of interest in whether nonvigorous and vigorous activities have
equivalent benefits,36 we calculated separately energy expenditure
from nonvigorous (,6 METs) and vigorous ($6 METs) sports or
recreation.35 (Examples of vigorous activities reported by alumni
include jogging or running, swimming laps, playing tennis, and
shoveling snow.) For nonvigorous energy expenditure, we defined 5
categories: none (51.1% of men), 1 to ,250 (13.8%), 250 to ,600
(11.7%), 600 to ,1400 (12.0%), and $1400 kcal/wk (11.4%). For
vigorous energy expenditure, we did likewise (41.0%, 16.9%,
12.1%, 14.7%, and 15.3%, respectively). These cut points were
chosen so that among those who did report some sport or recreation,
men would be approximately evenly distributed among the remaining 4 categories of energy expenditure. Furthermore, the choice of
identical cut points would enable direct comparison of stroke rates
between, for example, men who expended 250 to ,600 kcal/wk in
nonvigorous activities and those who expended the same amount of
energy but in vigorous activities.
With the growing recognition that it is difficult to persuade the many
sedentary individuals25 to take up vigorous activities, research interest
lately has focused on moderate intensity physical activity.37 Therefore,
we analyzed physical activity in a third way. We calculated the energy
expended on sports or recreation separately for light activities (,4.5
METs) and for activities of at least moderate intensity ($4.5 METs).
(Examples of moderate activities reported by alumni include dancing,
bicycling, snorkeling, and digging or filling in garden.) We used the

same cut points as before in analyses. For light energy expenditure, men
distributed themselves thus: none, 63.9%; 1 to ,250, 10.3%; 250 to
,600, 8.2%; 600 to ,1400, 8.7%; and $1400 kcal/wk, 8.9%. For
energy expended on activities of $4.5 METs, the distribution was
36.4%, 15.8%, 12.7%, 16.7%, and 18.4%, respectively.
On the 1977 questionnaire, we also asked about cigarette smoking
(categorized in analyses as nonsmoker, smoker of #20, or .20
cigarettes per day), alcohol consumption (,50, 50 to 199, or $200
g/wk), ages of parental death (neither, 1, or both parents died before
age 65 years), weight and height (combined as body mass index;
,22.5, 22.5 to ,23.5, 23.5 to ,24.5, 24.5 to ,26.0, or $26.0
kg/m2), physician-diagnosed hypertension (no, yes, or unknown),
and diabetes mellitus (no, yes, or unknown).

Ascertainment of Stroke Occurrence

In 1988 we sent another health survey to surviving alumni. Included on
the 1988 survey was a question on whether a physician had ever
diagnosed stroke and if so, the year of first diagnosis. Men who did not
provide this information were considered lost to follow-up and excluded
from study. Self-reported, physician-diagnosed chronic diseases among
alumni were believed to be valid.38 40 We used death certificates to
ascertain additional strokes (underlying or contributing cause of death)
not reported on the 1988 questionnaire. We traced deaths occurring
through 1990 using information from the alumni office to obtain death
certificates. Mortality follow-up in this cohort is .99% complete.38

Statistical Analyses
We used proportional hazards regression to analyze time to first
stroke occurrence (ascertained from the 1988 questionnaire or death
certificate) or censoring (1988 questionnaire return or death, whichever occurred later).41 There was no evidence that proportional
hazards assumptions were violated. For strokes ascertained from
death certificates alone, we took the date of diagnosis to be the date
of death. Since this could create a potential bias, we first examined
the association between physical activity and stroke risk, excluding
alumni with strokes determined from death certificates. We counted
only cases of stroke ascertained from the 1988 questionnaire in
which date of diagnosis was known. The findings from these
analyses were similar to those from analyses in which all strokes (ie,
determined from the 1988 questionnaire or death certificate) were
included. Therefore, we present our findings only from analyses that
combined strokes ascertained from both sources.
We first modeled rate ratios (relative risks) of stroke as a function
of total energy expenditure and age. We then proceeded to adjust
additionally for other predictors of stroke in these data: cigarette
smoking, alcohol consumption, and early parental death. Since
age-adjusted and multivariate analyses yielded similar findings, only
results from the latter are presented. We calculated 95% CIs for
estimated relative risks and used 2-tailed tests of significance.
In secondary analyses, we examined the association between total
energy expenditure and stroke risk separately for 2 time periods
(chosen so that there would be approximately the same number of
events in each period): 1977 to 1984 and 1985 to 1990, and for 3 age
groups: ,65, 65 to 74, and $75 years at study entry. We also
separately analyzed a subgroup of men at lower risk of stroke: men
who did not smoke, consumed ,200 g/wk of alcohol, and whose
parents did not die before age 65 years.
When analyzing the components of physical activity, we included
indicator terms for categories of walking, stair climbing, nonvigorous
(,6 METs) energy expenditure, and vigorous ($6 METs) energy
expenditure in multivariate models. To assess the effects of light activity
and activity of at least moderate intensity, we conducted parallel
analyses with energy expenditure dichotomized at 4.5 instead of 6.0
METs. Finally, we investigated the effects of walking and stair climbing
among only men reporting no sports or recreation in 1977.

Results
At study entry, alumni were aged 58 years on average (range,
43 to 88 years) (Table 1). Their prevalence of smoking,
16.9%, was much lower than in the general population of that

Downloaded from http://stroke.ahajournals.org/ by guest on November 4, 2014

Lee and Paffenbarger

TABLE 1.

October 1998

2051

Baseline Characteristics of Harvard Alumni in 1977

Characteristic
Mean age, years (6SD)

58.0 (68.95)

Cigarette habit, %
Nonsmoker

83.1

Smoker, 120 cigarettes per day

8.5

Smoker, $20 cigarettes per day

8.4

Alcohol consumption, %
,50 g/wk

27.2

50,200 g/wk

37.1

$200 g/wk

35.7

Early (,65 years) parental death, %

Neither parent died early

65.8

1 parent died early

29.8

Both parents died early

4.4

Body mass index (kg/m2), %

,22.5

23.0

22.5,23.5

14.4

23.5,24.5

19.1

24.5,26.0

20.4

$26.0

23.1

Physician-diagnosed hypertension, %
Physician-diagnosed diabetes mellitus, %

20.1
3.9

Energy expenditure (kcal/wk),* %

,1000

31.0

10001999

28.7

20002999

18.2

30003999

10.2

$4000

11.9

*Estimated from blocks walked, stairs climbed, and sports or recreational

activities performed.

era (36.5% among US white men, 1979).1 With regard to

body weight, again, alumni possessed a healthier profile,
being of lower body mass index than the general population
of that time.42 Table 1 also describes the distributions of
alcohol consumption, early parental death, and physiciandiagnosed hypertension and diabetes mellitus among
subjects.
Between 1977 and 1990, 378 strokes occurred in 119 880
person-years. For total energy expenditure of ,1000, 1000 to
1999, 2000 to 2999, 3000 to 3999, and $4000 kcal/wk, the
relative risks of stroke, adjusted for age, smoking, alcohol
intake, and early parental death, were 1.00 (referent), 0.76
(95% CI, 0.59 to 0.98), 0.54 (0.38 to 0.76), 0.78 (0.53 to
1.15), and 0.82 (0.58 to 1.14), respectively; P50.05 for linear
trend (Figure ). When we added a quadratic term for energy
expenditure to examine how well the data fitted a U-shaped
curve, this was significant at P50.004. In assessing the roles
that body mass index, hypertension, and diabetes mellitus
might play in mediating the inverse association between
physical activity and stroke risk, we conducted an analysis
that additionally adjusted for these 3 variables. Relative risks
were slightly attenuated from before: 1.00, 0.79, 0.56, 0.81,
and 0.85, respectively; P50.09 for linear trend.

Relative risks of stroke among Harvard alumni, 1977 to 1990,

according to energy expenditure estimated in kilocalories per
week from blocks walked, stairs climbed, and sports or recreational activities performed in 1977. Relative risks are adjusted
for age, cigarette smoking, alcohol consumption, and early
parental death. Vertical bars indicate 95% CIs.

In secondary analyses, we examined separately 2 time

intervals: 1977 to 1984 (176 strokes) and 1985 to 1990 (202
strokes). Relative risks of stroke for the 5 categories of total
energy expenditure during the earlier period were 1.00
(referent), 0.65, 0.55, 0.76, and 0.86, respectively; for the
later period, they were 1.00, 0.86, 0.53, 0.80, and 0.78,
respectively. We also investigated separately men aged ,65
(n58344), 65 to 74 (n52259), and $75 years (n5527) at
study entry; 149, 164, and 65 strokes, respectively, occurred.
Among men aged ,65 years, the relative risks for the 5
categories of total energy expenditure were 1.00 (referent),
0.75, 0.54, 1.02, and 0.87, respectively. For men aged 65 to
74 years, corresponding relative risks were 1.00, 0.72, 0.52,
0.44, and 0.68, respectively. Among men aged $75 years,
they were 1.00, 0.89, 0.60, 1.10, and 1.28, respectively. These
findings were adjusted for differences in age, smoking,
alcohol consumption, and early parental death.
Next, we analyzed a subgroup of alumni at lower risk for
stroke (men who did not smoke, consumed ,200 g/wk of
alcohol, and whose parents did not die before age 65;
n54164; 111 strokes). After adjustment for age and alcohol
intake, the relative risks of stroke for the 5 categories of total
energy expenditure were 1.00 (referent), 0.74, 0.56, 0.63, and
0.84, respectively.
We proceeded to explore the associations between specific
components of physical activity and stroke incidence (Table
2). With walking, attaining a distance of $20 km/wk was
associated with reduced risk. With stair climbing, activities at
,6 METs, or activities at $6 METs, we observed a U-shaped
relation of stroke rates to each physical activity component,
similar to that seen for total energy expenditure. However, the
quadratic terms here were not significant (P50.12, P50.22,
and P50.06, respectively). Noteworthy was the similarity of
age-adjusted incidence rates for each level of energy expen-

Downloaded from http://stroke.ahajournals.org/ by guest on November 4, 2014

2052

Physical Activity and Stroke Incidence

TABLE 2. Incidence Rates* and Relative Risks of Stroke

Among Harvard Alumni, 19771990, According to Different
Components of Physical Activity in 1977

Physical Activity
Component

Incidence
Rate
No. of
(per
Events 10 000)

Multivariate Relative
Risk (95% CI)

Incidence
Rate
No. of
(per
Events 10 000)

Physical Activity
Component

Multivariate Relative
Risk (95% CI)

Activities at ,4.5 METs, kcal/wk

Distance walked, km/wk

,5

TABLE 3. Incidence Rates* and Relative Risks of Stroke

Among Harvard Alumni, 19771990, According to Activities of
<4.5 and >4.5 METs in 1977

128

34.5

1.00 (referent)

None

248

33.2

1.00 (referent)

5,10

81

34.1

1.06 (0.801.40)

1,250

33

30.2

0.92 (0.641.34)

10,20

105

32.9

1.01 (0.771.31)

250,600

28

28.2

0.83 (0.561.23)

64

23.6

0.71 (0.520.96)

600,1400

30

27.7

0.87 (0.591.27)

$1400

39

29.1

$20

P50.05 for linear trend

,10

0.89 (0.631.25)

P50.53 for linear trend

Stairs climbed, stories/wk

Activities at $4.5 METs, kcal/wk

116

34.8

1.00 (referent)

10,20

78

33.3

0.95 (0.711.27)

None

206

35.1

1.00 (referent)

20,35

59

25.0

0.71 (0.520.98)

1,250

56

32.0

0.88 (0.651.19)

125

31.6

0.97 (0.741.25)

250,600

27

21.6

0.60 (0.400.90)

P50.53 for linear trend

600,1400

40

25.7

0.72 (0.511.03)

$1400

49

31.3

$35
Activities at ,6 METs, kcal/wk
None

221

34.3

1.00 (referent)

1,250

41

30.7

0.90 (0.641.26)

250,600

30

24.0

0.69 (0.471.01)

600,1400

40

29.1

0.86 (0.611.21)

$1400

46

29.2

0.86 (0.631.19)

P50.16 for linear trend

0.89 (0.641.23)

P50.09 for linear trend

METs indicates multiples of resting metabolic rate.
*Age-adjusted.
Adjusted for age (single years), cigarette smoking (nonsmoker, smoker of
#20, or .20 cigarettes per day), alcohol consumption (,50, 50 199, or
$200 g/wk), early parental death (neither, 1, or both parents died before age
65), walking, and stair climbing, as well as activities of the other intensity.

Activities at $6 METs, kcal/wk

None

217

33.8

1.00 (referent)

1,250

55

30.6

0.90 (0.671.22)

250,600

27

22.9

0.67 (0.451.02)

600,1400

35

27.0

0.82 (0.561.18)

$1400

44

33.9

1.03 (0.731.44)

P50.48 for linear trend

METs indicates multiples of resting metabolic rate.
*Age-adjusted.
Adjusted for age (single years), cigarette smoking (nonsmoker, smoker of
#20, or .20 cigarettes per day), alcohol consumption (,50, 50 199, or
$200 g/wk), early parental death (neither, 1, or both parents died before age
65), and the other 3 components of physical activity.

diture at ,6 METs and the corresponding level of energy

expenditure at $6 METs (eg, for 250 to ,600 kcal/wk, 24.0
and 22.9 per 10 000, respectively).
When we analyzed energy expenditure for sports or recreation using a 4.5 instead of a 6.0 MET cut point, we observed
that activities of ,4.5 METs were unrelated to stroke risk
(Table 3). For activities of $4.5 METs, a U-shaped relation
was seen with a significant quadratic term (P50.04).
Finally, we investigated another subgroup: men reporting
no sports or recreation in 1977 (n52892; 148 strokes). The
event rate was higher in this than the larger group (data not
shown). The results in this subgroup were similar to those
based on the larger group; however, the findings in this
smaller group were no longer significant. The relative risks of
stroke associated with walking ,5, 5 to ,10, 10 to ,20, and
$20 km/wk in this subgroup were 1.00 (referent), 1.15, 0.77,

and 0.73, respectively; P50.09 for linear trend. For climbing

,10, 10 to ,20, 20 to ,35, and $35 stories/wk, they were
1.00 (referent), 1.00, 0.60, and 0.94, respectively; P50.50 for
linear trend.

Discussion
In these prospective data, physical activity was associated
with decreased risk of stroke in men. With higher levels of
energy expenditure up to 3000 kcal/wk, risk declined
steadily; beyond this, however, the association weakened. We
observed similar findings even in older individuals. What
kinds of activities were beneficial? The data indicated that
walking $20 km/wk was associated with significantly lower
risk, independent of other physical activity components.
Climbing stairs and activities of at least moderate intensity
($4.5 METs) each showed U-shaped relations to stroke risk,
with the risk being significantly lower at the nadir of the
curve. Light-intensity activities (,4.5 METs), however, were
unrelated to stroke risk. Furthermore, men who participated
in recreational activities experienced lower stroke rates than
those who only walked or climbed stairs but did not engage
in recreational activities.
It is unclear why stroke incidence rates exhibited a
U-shaped relation to physical activity in this study group.
Two of 3 randomized trials specifically testing exercise of
different intensities suggest that higher-intensity physical
activity is less effective in decreasing blood pressure than
lower intensity activity.43,44 This could plausibly explain our
observations. We did, however, explore whether the

Downloaded from http://stroke.ahajournals.org/ by guest on November 4, 2014

Lee and Paffenbarger

U-shaped association might be artifactual. For example, as
alumni aged, perhaps those most active scaled down their
activities, while the less active maintained theirs. We observed some evidence for this: of alumni falling in the least
active category in 1977, 69% remained in the 2 least active
categories in 1988, while 57% of alumni in the most active
category in 1977 remained in the 2 most active categories in
1988. This might attenuate relative risks among the most
active, the phenomenon becoming more pronounced as
alumni aged. However, we observed U-shaped associations
during early and late follow-up and among younger and older
men. Thus, physical activity misclassification may not totally
account for the U-shaped association.
Next, we wondered whether men at higher risk for stroke
exercised more assiduously to ameliorate their risk. Although
we did adjust for potential confounders, residual confounding
might still exist. We therefore examined separately men at
lower risk (nonsmokers who drank ,200 g/wk of alcohol and
whose parents did not die early). Findings were little
changed; thus, residual confounding is less likely to have
caused the U-shaped relation.
Finally, we examined the association between physical
activity and coronary heart disease. The consensus is that
coronary heart disease risk declines steadily with increasing
activity25; thus, if the observed U-shaped curve for stroke
resulted from bias in the design or analysis of the present
study, a U-shaped curve also might be observed for coronary
heart disease in this study group. Among alumni, 1433
developed coronary heart disease (ascertained in similar
fashion as for strokes) during follow-up. After adjustment for
the same potential confounders, the relative risks of this
disease associated with the 5 categories of total energy
expenditure were 1.00 (referent), 0.84, 0.74, 0.83, and 0.71,
respectively, with a highly significant linear trend
(P,0.0001). Therefore, the U-shaped association between
stroke rates and physical activity is less likely to be an artifact
of the data. Confirmation of this finding requires similar
observations in other populations.
Several investigators have examined dose-response in the
relation between physical activity and stroke.6,8 10,1317,20,23
Two studies also observed U-shaped relations between stroke
rates and physical activity. In these 2 studies, investigators
categorized physical activity into 3 levels among Italian
railroad workers20 and Seventh-Day Adventist men.23 In 2
other studies in which physical activity was divided into 3
levels, stroke risk decreased at the second level but did not
decline with further activity.14,15 Of the remaining 7 studies
that assessed dose response, stroke risk declined steadily with
increasing activity.6,8 10,13,16,17 These differences may have
reflected the nature of the different populations studied;
furthermore, physical activity assessments in the various
studies are not directly comparable.
Strengths of the present study include a well-characterized
study group, with detailed and valid physical activity assessment at baseline. The latter enabled investigation of specific
kinds of physical activity, providing useful information for
formulating public health recommendations.
However, several limitations exist. We successfully followed only 78% of eligible men. Mortality follow-up in this

October 1998

2053

study group is known to be virtually complete38; hence, those

lost to follow-up were most likely alive as of 1990. Because
they did not return a questionnaire in 1988 (or did so but did
not answer the stroke question), we could not ascertain their
stroke status. Men lost to follow-up (n53169) were younger
at baseline (mean age, 47.2 versus 58.0 years among those in
the present study), somewhat more likely to be smoking
(18.1% versus 16.9%), but less likely to be drinking $200
g/wk of alcohol (32.0% versus 35.7%) and to have both
parents dying early (3.3.% versus 4.4%). The age differential
is not unexpected, since virtually all those lost to follow-up
were alive as of 1990, while those in the present study
included deceased (therefore, older) men as of 1990. However, the energy expenditure of those lost to follow-up was
quite similar to that of men in the present study (mean, 2212
versus 2104 kcal/wk). This was supported by similar distributions of body mass index in the 2 groups. Since physical
activity levels were similar in the 2 groups of men, bias from
incomplete follow-up is less likely.
Additionally, strokes were ascertained through self-report
and death certificates. Some misclassification was likely,
even though alumni self-report of physician-diagnosed
chronic diseases has been shown to be valid.38 40 The misclassification is likely to be random, since physical activity
information was collected before asking about stroke. Hence,
the misclassification would likely have weakened findings
but not created spurious inverse associations. We also were
unable to examine different types of stroke since we did not
pursue medical records to determine stroke subtype. Furthermore, we lacked dietary data at baseline and thus could not
adjust for this. However, in about two thirds of men, we did
collect dietary information in 1988. We did not see clear
differences in percent fat or saturated fat intake among men
with different physical activity levels, decreasing the likelihood that diet was a strong confounder. Finally, men in this
study were healthier than the US general population. While
this may limit the generalizability of findings, it does not
preclude their internal validity.
Biologically, it is plausible for physical activity to decrease
stroke risk by curtailing obesity, decreasing blood pressure,
maintaining normal glucose tolerance, and improving insulin
sensitivity.45,46 However, when we additionally adjusted in
analyses for body mass index, hypertension, and diabetes
mellitus, these variables explained only a very modest portion
of the benefit of physical activity. Other pathways through
which physical activity may decrease stroke risk include
improving lipid profile, decreasing fibrinogen levels, increasing fibrinolytic activity, and reducing the tendency of platelets to aggregate.3
In conclusion, physical activity is associated with decreased stroke risk in men, including older men. A decreased
risk was observed at energy expenditures of 1000 to 1999
kcal/wk, with further risk decrement seen at 2000 to 2999
kcal/wk but not beyond. Confirmation of the U-shaped
relation observed in these data requires similar observations
in other populations.

Acknowledgments
This study was supported by grants from the National Heart, Lung,
and Blood Institute (HL 34174) and the National Cancer Institute

Downloaded from http://stroke.ahajournals.org/ by guest on November 4, 2014

2054

Physical Activity and Stroke Incidence

(CA 44854), US Public Health Service. We are grateful to Stacey

DeCaro, Sarah E. Freeman, Tina Y. Ha, James B. Kampert, Rita W.
Leung, Doris C. Rosoff, Howard D. Sesso, and Alvin L. Wing for
their help with the College Alumni Health Study.

References
1. National Center for Health Statistics. Health, United States, 1995.
Hyattsville, Md: Public Health Service; 1996.
2. Kistler JP, Ropper AH, Martin JB. Cerebrovascular diseases. In:
Isselbacher KJ, Braunwald E, Wilson JD, Martin JB, Fauci AS, Kasper
DL, eds. Harrisons Principles of Internal Medicine. New York, NY:
McGraw-Hill, Inc; 1994:22332256.
3. Moore S. Physical activity, fitness, and atherosclerosis. In: Bouchard C,
Shephard RJ, Stephens T, eds. Physical Activity, Fitness, and Health:
International Proceedings and Consensus Statement. Champaign, Ill:
Human Kinetics Publishers; 1994:570 578.
4. Fagard RH, Tipton CM. Physical activity, fitness, and hypertension. In:
Bouchard C, Shephard RJ, Stephens T, eds. Physical Activity, Fitness,
and Health: International Proceedings and Consensus Statement.
Champaign, Ill: Human Kinetics Publishers; 1994:633 655.
5. Paffenbarger RS Jr, Williams JL. Chronic disease in former college
students: early precursors of fatal stroke. Am J Public Health. 1967;57:
1290 1299.
6. Paffenbarger RS Jr, Hyde RT, Wing AL, Steinmetz CH. A natural history
of athleticism and cardiovascular health. JAMA. 1984;252:491 495.
7. Salonen JT, Puska P, Tuomilehto. Physical activity and risk of myocardial
infarction, cerebral stroke and death. Am J Epidemiol. 1982;115:526537.
8. Herman B, Schmitz PIM, Leyten ACM, van Luijk JH, Frenken CWGM, Op
de Coul AAW, Schulte BPM. Multivariate logistic analysis of risk factors for
stroke in Tilburg, the Netherlands. Am J Epidemiol. 1983;118:514525.
9. Lapidus L, Bengtsson C. Socioeconomic factors and physical activity in
relation to cardiovascular disease and death: a 12 year follow up of
participants in a population study of women in Gothenburg, Sweden. Br
Heart J. 1986;55:295301.
10. Wannamethee G, Shaper AG. Physical activity and stroke in British
middle aged men. BMJ. 1992;304:597 601.
11. Shinton R, Sagar G. Lifelong exercise and stroke. BMJ. 1993;307:
231234.
12. Lindenstrm E, Boysen G, Nyboe J. Lifestyle factors and risk of cerebrovascular disease in women: the Copenhagen City Heart Study. Stroke.
1993;24:1468 1472.
13. Hheim LL, Holme I, Hjermann I, Leren P. Risk factors of stroke
incidence and mortality: a 12-year follow-up of the Oslo Study. Stroke.
1993;24:1484 1489.
14. Abbott RD, Rodriguez BL, Burchfiel CM, Curb JD. Physical activity in
older middle-aged men and reduced risk of stroke: the Honolulu Heart
Program. Am J Epidemiol. 1994;139:881 893.
15. Kiely DK, Wolf PA, Cupples LA, Beiser AS, Kannel WB. Physical
activity and stroke risk: the Framingham Study. Am J Epidemiol. 1994;
140:608 620.
16. Gillum RF, Mussolino ME, Ingram DD. Physical activity and stroke
incidence in women and men: the NHANES I Epidemiologic Follow-up
Study. Am J Epidemiol. 1996;143:860 869.
17. Sacco RL, Gan R, Boden-Albala B, Lin I-F, Kargman DE, Hauser WA,
Shea S, Paik MC. Leisure-time physical activity and ischemic stroke risk:
the Northern Manhattan Stroke Study. Stroke. 1998;29:380 387.
18. Paffenbarger RS Jr. Factors predisposing to fatal stroke in longshoremen.
Prev Med. 1972;1:522527.
19. Kannel WB, Sorlie P. Some health benefits of physical activity: the
Framingham Study. Arch Intern Med. 1979;139:857 861.
20. Menotti A, Seccareccia F. Physical activity at work and job responsibility
as risk factors for fatal coronary heart disease and other causes of death.
J Epidemiol Community Health. 1985;39:325329.
21. Menotti A, Keys A, Blackburn H, Aravanis C, Dontas A, Fidanza F,
Giampaoli S, Karvonen M, Kroumhout D, Nedeljkovic S, Nissinen A,
Pekkanen J, Punsar S, Seccareccia F, Toshima H. Twenty-year stroke
mortality and prediction in twelve cohorts of the Seven Countries Study.
Int J Epidemiol. 1990;19:309 315.
22. Harmsen P, Rosengren A, Tsipogianni A, Wilhelmsen L. Risk factors for
stroke in middle-aged men in Goteborg, Sweden. Stroke. 1990;21:
223229.
23. Lindsted KD, Tonstad S, Kuzma JW. Self-report of physical activity and
patterns of mortality in Seventh-Day Adventist men. J Clin Epidemiol.
1991;44:355364.

24. Ellekjr EF, Wyller TB, Sverre JM, Holmen J. Lifestyle factors and risk
of cerebral infarction. Stroke. 1992;23:829 834.
25. US Department of Health and Human Services. Physical Activity and
Health: A Report of the Surgeon General. Atlanta, Ga: US Department of
Health and Human Services, Centers for Disease Control and Prevention,
National Center for Chronic Disease Prevention and Health Promotion;
1996:1 8, 85172, 175207.
26. Lee I-M, Paffenbarger RS Jr, Hsieh C-c. Time-trends in physical activity
among college alumni, 19621988. Am J Epidemiol. 1992;135:915925.
27. LaPorte RE, Cauley JA, Kinsey CM, Corbett W, Robertson R, BlackSandler R, Kuller LH, Falkel J. The epidemiology of physical activity in
children, college students, middle-aged men, menopausal females and
monkeys. J Chron Dis. 1982;35:787795.
28. LaPorte RE, Black-Sandler RB, Cauley JA, Link M, Bayles C, Marks B.
The assessment of physical activity in older women: analysis of the
interrelationship and reliability of activity monitoring, activity surveys,
and caloric intake/expenditure. J Gerontol. 1983;30:394 397.
29. Siconolfi SF, Lasater TM, Snow RCK, Carleton RA. Self-reported
physical activity compared with maximal oxygen uptake. Am J Epidemiol. 1985;122:101105.
30. Cauley JA, LaPorte RE, Sandler RB, Schramm MM, Kriska AM. Comparison of methods to measure physical activity in postmenopausal
women. Am J Clin Nutr. 1987;45:14 22.
31. Albanes D, Conway JM, Taylor PR, Moe PW, Judd J. Validation and
comparison of eight physical activity questionnaires. Epidemiology.
1990;1:6571.
32. Washburn RA, Smith KW, Goldfield SRW, McKinlay JB. Reliability and
physiologic correlates of the Harvard Alumni Activity Survey in a general
population. J Clin Epidemiol. 1991;44:1319 1326.
33. Ainsworth BE, Leon AS, Richardson MT, Jacobs DR, Paffenbarger RS
Jr. Accuracy of the College Alumnus Physical Activity Questionnaire.
J Clin Epidemiol. 1993;46:14031411.
34. Ainsworth BE, Haskell WL, Leon AS, Jacobs DR Jr, Montoye HJ, Sallis
JF, Paffenbarger RS Jr. Compendium of physical activities: classification
of energy costs of human physical activities. Med Sci Sports Exerc.
1993;25:71 80.
35. Taylor HL, Jacobs DR Jr, Schucker B, Knudsen J, Leon AS, Debacker G.
A questionnaire for the assessment of leisure time physical activities.
J Chron Dis. 1978;31:741755.
36. NIH Consensus Development Panel on Physical Activity and Cardiovascular Health. Physical activity and cardiovascular health. JAMA. 1996;
276:241246.
37. Pate RR, Pratt M, Blair SN, Haskell WL, Macera CA, Bouchard C,
Buchner D, Ettinger W, Heath GW, King AC, Kriska A, Leon AS,
Marcus BH, Morris J, Paffenbarger RS Jr, Patrick K, Pollock ML, Rippe
JM, Sallis J, Wilmore JH. Physical activity and public health: a recommendation from the Centers for Disease Control and Prevention and the
American College of Sports Medicine. JAMA. 1995;273:402 407.
38. Paffenbarger RS Jr, Wing AL, Hyde RT. Physical activity as an index of
heart attack risk in college alumni. Am J Epidemiol. 1978;108:161175.
39. Lee I-M, Paffenbarger RS Jr, Hsieh C-C. Physical activity and risk of
developing colorectal cancer among college alumni. J Natl Cancer Inst.
1991;83:1324 1329.
40. Lee I-M, Paffenbarger RS Jr, Hsieh C-C. Physical activity and risk of
prostatic cancer among college alumni. Am J Epidemiol. 1992;135:
169 179.
41. Cox DR. Regression models and life tables. J R Stat Soc (B). 1972;34:
187220.
42. Kuczmarski RJ, Flegal KM, Campbell SM, Johnson CL. Increasing
prevalence of overweight among US adults: the National Health and
Nutrition Examination Surveys, 1960 to 1991. JAMA. 1994;272:205211.
43. Hagberg JM, Montain SJ, Martin WH III, Ehsani AA. Effect of exercise
training in 60- to 69-year-old persons with essential hypertension. Am J
Cardiol. 1989;64:348 353.
44. Matsusaki M, Ikeda M, Tashiro E, Koga M, Miura S-i, Ideishi M, Tanaka
H, Shindo M, Arakawa K. Influence of workload on the antihypertensive
effect of exercise. Clin Exp Pharmacol Physiol. 1992;19:471 479.
45. Bronner LL, Kanter DS, Manson JE. Primary prevention of stroke.
N Engl J Med. 1995;333:13921400.
46. Whisnant JP, Wiebers DO, OFallon WM, Sicks JD, Frye RL. A
population-based model of risk factors for ischemic stroke: Rochester,
Minnesota. Neurology. 1996;47:1420 1428.

Downloaded from http://stroke.ahajournals.org/ by guest on November 4, 2014