Customize your JAMA Network experience by selecting one or more topics from the list below.
Feskanich D, Willett W, Colditz G. Walking and Leisure-Time Activity and Risk of Hip Fracture in Postmenopausal Women. JAMA. 2002;288(18):2300–2306. doi:10.1001/jama.288.18.2300
Author Affiliations: Channing Laboratory, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School (Drs Feskanich, Willett, and Colditz) and Departments of Nutrition (Dr Willett) and Epidemiology (Drs Willett and Colditz), Harvard School of Public Health, Boston, Mass.
Context Physical activity can reduce the risk of hip fractures in older women,
although the required type and duration of activity have not been determined.
Walking is the most common activity among older adults, and evidence suggests
that it can increase femoral bone density and reduce fracture risk.
Objective To assess the relationship of walking, leisure-time activity, and risk
of hip fracture among postmenopausal women.
Design, Setting, and Participants Prospective analysis begun in 1986 with 12 years of follow-up in the
Nurses' Health Study cohort of registered nurses within 11 US states. A total
of 61 200 postmenopausal women (aged 40-77 years and 98% white) without
diagnosis of cancer, heart disease, stroke, or osteoporosis at baseline.
Main Outcome Measures Incident hip fracture resulting from low or moderate trauma, analyzed
by intensity and duration of leisure-time activity and by time spent walking,
sitting, and standing, measured at baseline and updated throughout follow-up.
Results From 1986 to 1998, 415 incident hip fracture cases were identified.
After controlling for age, body mass index, use of postmenopausal hormones,
smoking, and dietary intakes in proportional hazards models, risk of hip fracture
was lowered by 6% (95% confidence interval [CI], 4%-9%; P<.001) for each increase of 3 metabolic equivalent (MET)–hours
per week of activity (equivalent to 1 h/wk of walking at an average pace).
Active women with at least 24 MET-h/wk had a 55% lower risk of hip fracture
(relative risk [RR], 0.45; 95% CI, 0.32-0.63) compared with sedentary women
with less than 3 MET-h/wk. Even women with a lower risk of hip fracture due
to higher body weight experienced a further reduction in risk with higher
levels of activity. Risk of hip fracture decreased linearly with increasing
level of activity among women not taking postmenopausal hormones (P<.001), but not among women taking hormones (P = .24). Among women who did no other exercise, walking for at least
4 h/wk was associated with a 41% lower risk of hip fracture (RR, 0.59; 95%
CI, 0.37-0.94) compared with less than 1 h/wk. More time spent standing was
also independently associated with lower risks.
Conclusion Moderate levels of activity, including walking, are associated with
substantially lower risk of hip fracture in postmenopausal women.
Despite varying populations and diversity in methods of assessing physical
activity, evidence from epidemiological studies suggests that the risk of
hip fracture can be reduced by 20% to 50% for active compared with sedentary
adults.1,2 Most hip fractures
result from a fall,3 and several clinical trials
have demonstrated that regular activity can reduce fall occurrence4-6 through improvements
in muscle strength7-11 and
activity can also reduce fracture risk by increasing the mechanical load on
bone, which promotes remodeling. Clinical trials have demonstrated that femoral
bone density can be increased with weight-bearing exercise or resistance training.15-17
Although physical activity has definite benefits for bone health, its
relative contributions to fracture reduction by type, frequency, intensity,
and duration of activity have been difficult to define. In this analysis,
we examined associations between exercise and leisure-time activities and
the risk of hip fracture among postmenopausal women in the Nurses' Health
Study, considering type, intensity, and duration of activity. We also assessed
the concurrent influences of body mass index, postmenopausal hormone use,
smoking, and diet.
The Nurses' Health Study is an ongoing cohort of 121 700 women
who in 1976 (time of initial mail questionnaire) were registered nurses between
the ages of 30 and 55 years and resided in 1 of 11 US states. Approximately
98% of the cohort is white. Follow-up questionnaires are sent every 2 years
and the response rate is at least 90% in each cycle. Deaths are confirmed
through the National Death Index.18 On the
initial questionnaire, participants provided a medical history and information
on lifestyle and other risk factors related to cancer and heart disease. Subsequent
questionnaires updated these data and were expanded to include other diseases
and relevant risk factors. Time spent in specific exercise or leisure-time
activities was added to the questionnaire in 1986.
This analysis began in 1986 with the postmenopausal women who responded
to the specific activity questions and had not reported a previous hip fracture
or a diagnosis of cancer, heart disease, stroke, or osteoporosis. Eligible
women entered the analysis after menopause. A total of 61 200 women,
aged 40 to 77 years, contributed to this analysis with follow-up through 1998.
In 1982, participants reported all previous hip fractures with the date
and circumstances leading to fracture. Incident fractures were reported on
subsequent biennial questionnaires. Only fractures due to low or moderate
trauma (eg, slipping on ice, falling from the height of a chair) were included
as cases in this study. Those associated with high trauma (eg, skiing, falling
off a ladder) were excluded from analysis (about 15% of reported hip fractures).
During the 12 years of follow-up, 415 cases were identified among the women
in this study. The median age at fracture was 67 years (range, 46-75 years).
Although we relied on self-reports of hip fractures, we expected reliable
information in a cohort of registered nurses. Specificity was demonstrated
in a small validation study in which all 30 reported hip fractures were confirmed
by medical records.19
In 1986, participants were asked to report the average amount of time
spent per week during the previous year in each of 7 activities: walking or
hiking outdoors, jogging (>10 min/mile), running, bicycling (including stationary
machine), racquet sports, lap swimming, and other aerobic activity (eg, aerobic
dance, rowing machine). These activities were the most common ones reported
by women in the University of Pennsylvania Alumni Health Study. For each activity,
women chose one of 11 duration categories that ranged from zero to 11 h/wk
or more. Walking pace was also reported as either easy (<2 mph), average
(2-2.9 mph), brisk (3-3.9 mph), very brisk (≥4 mph), or unable to walk.
Activity was reassessed in 1988, 1992, 1994, and 1996. The last 3 activity
questionnaires included 2 additional items: other vigorous activities (eg,
lawn mowing) and lower intensity exercise (eg, yoga, stretching).
Each activity on the questionnaire was assigned a metabolic equivalent
(MET) score based on the classification by Ainsworth et al.20 One
MET is the energy expenditure for sitting quietly. MET scores for specific
activities are defined as the ratio of the metabolic rate associated with
that activity divided by the resting metabolic rate. For example, walking
at an average pace was assigned a MET score of 3; jogging, 7; and running,
12. MET scores for walking were assigned based on walking pace; for other
activities, a leisurely to moderate intensity score was selected. The scores
for MET-hours per week for each activity were calculated from the reported
hours per week engaged in that activity multiplied by the assigned MET score,
and the values from the individual activities were summed for a total MET-hours
per week score. To obtain the best long-term measure of physical activity,
total values were cumulatively averaged in analyses. That is, at the beginning
of each 2-year follow-up cycle, the MET-hours per week is the mean of all
MET-hours per week calculated from responses to the questionnaires up to that
We also assessed inactivity with hours per week spent sitting and standing
(at home, at work, and other time away from home). These items were on the
questionnaires in 1988, 1990, and 1992, and hours of standing were cumulatively
averaged over follow-up in this analysis. For sitting, the data were collected
with one general question in 1988, which was later expanded to 2 (in 1990)
and 3 (in 1992) more specific questions. Predictably, the total reported hours
per week of sitting in the cohort increased as the number of questions increased.
Therefore, separate category cut points were created for each year of data
collection and hours of sitting were updated, but not cumulatively averaged,
The ability of the activity questionnaire to assess total activity and
inactivity over the previous year was tested in a sample of 151 white women.21 Compared with four 7-day activity diaries, the questionnaire
underascertained total activity by approximately 20% and inactivity by 35%.
However, the correlations for total MET-hours per week of activity (r = 0.62; 95% confidence interval [CI], 0.44-0.75) and
total hours of inactivity (r = 0.41; 95% CI, 0.25-0.54)
suggest that the questionnaire is a reasonably valid tool for categorical
ranking of respondents. The activity questionnaire was also compared with
4 past-week questionnaires collected seasonally during the year. For walking,
the primary activity among postmenopausal women, the correlation was 0.70
(95% CI, 0.49-0.84).
In 1980, participants were asked to report the number of hours per week
spent in moderate and vigorous activity as well as the frequency in which
they engaged in any regular activity long enough to work up a sweat. From
the responses to these questions, we estimated the number of hours per week
that participants engaged in leisure-time activities in 1980. This was used
with the 1986 hours per week from the activity questionnaire to determine
a 6-year change in activity level.
Weight was requested on all biennial questionnaires and body mass index
(BMI) was calculated using the height reported on the initial 1976 questionnaire.
Postmenopausal hormone use (never, past, or current) and smoking (never, past,
or current, with time since quitting for past smokers and number of cigarettes
per day for current smokers) were also assessed every 2 years. Diet was measured
at least every 4 years beginning in 1980 with a semiquantitative food frequency
questionnaire, and intakes of calcium, vitamin D, retinol, protein, vitamin
K, alcohol, and caffeine were calculated from the reported consumption of
foods and use of multivitamins and specific vitamin or mineral supplements.
The BMI and nutrient intakes were cumulatively averaged over follow-up in
Study participants contributed person-time from the return date of their
1986 questionnaire or the questionnaire on which they first became postmenopausal
until the occurrence of a hip fracture, death, or the end of follow-up on
June 1, 1998. A total of 576 518 person-years was accrued from the 61 200
women in this analysis. Median follow-up time per woman was 11.6 years.
Person-time was allocated to the appropriate category for each exposure
and covariate variable at the beginning of every 2-year follow-up cycle. Age-adjusted
incidence rates were calculated within exposure categories and relative risks
(RRs) are the ratio of the rate in each upper category compared with the rate
in the lowest category. Cox proportional hazards models were used to calculate
multivariate RRs adjusted for other risk factors for hip fracture. P values for linear trend and for interaction in stratified analyses
were determined using continuous exposure variables in the models. Statistical
analysis was conducted using SAS statistical software (Version 6.12; SAS Institute
Inc, Cary, NC) and P<.05 was used as the level
The postmenopausal women in this analysis were fairly sedentary. From
the 7 activity questions in 1986, the median total activity was 7 MET-h/wk
(equivalent to 2.3 h/wk of walking at an average pace), while 19% of the women
reported zero or minimal leisure-time activity (ie, <15 min/wk). In the
general US population, 29% of adults engage in no leisure-time activity.22 Walking was by far the most popular activity in this
cohort, contributing 66% of the total MET-hours per week. The median duration
among walkers was 1.25 h/wk. Biking (14%) and other aerobic activity (11%)
were contributors toward total activity.
Table 1 outlines the characteristics
of the study population by level of activity. Active women spent more time
walking and standing, but sitting was unrelated to activity. Active women
also had a lower BMI, were less likely to smoke, were more likely to take
postmenopausal hormones, and were more likely to take a calcium supplement
and a multivitamin, although diet in general was not strongly related to activity.
Although thiazide diuretic use was somewhat higher among less active women,
this factor was not included in multivariate models because it did not confound
results. Hip fracture incidence rates for this cohort are also presented.
Among the postmenopausal women in this study, both activity and BMI
exhibited significant inverse associations with risk of hip fracture (Table 2). These associations were independent
of one another and showed little confounding by the other measured risk factors.
Compared with the women in the lowest category of less than 3 MET-h/wk, those
with 24 MET-h/wk or higher had a significantly lower (55%) hip fracture risk
(RR, 0.45; 95% CI, 0.32-0.63) in the multivariate analysis. Risk declined
in a dose-dependent manner with a 6% decrease in risk (95% CI, 4%-9%) for
every 3 MET-h/wk increase in activity (equivalent to 1 h/wk of walking at
an average pace). Risks of hip fracture among women with BMIs between 25.0
and 29.9 were not different from that of the reference group with BMIs between
23 and 24.9. A BMI of 30 or higher was associated with 50% the hip fracture
risk of women in the reference group, and women with a BMI of less than 23
had significantly higher risk (45%-83%; Table 2). These inverse associations between activity, BMI, and
risk of hip fracture were unchanged when women were excluded during follow-up
because of diagnosis of cancer, heart disease, stroke, or diabetes. We also
found that fracture risks were unchanged when women who reported balance problems
in 1990 (5% of the study population) were excluded from analysis.
Based on our risk estimates, we calculated the percentage of hip fractures
in the Nurses' Health Study cohort that could have been prevented if all participants
had exercised at a higher level. If all had exercised at 9 MET-h/wk or higher,
23% (95% CI, 15%-34%) of the hip fractures could have been prevented; at 15
MET-h/wk or higher, 32% (95% CI, 21%-44%) could have been prevented; and if
all exercised at 24 MET-h/wk or higher, 42% (95% CI, 27%-59%) of the hip fractures
could have been prevented.
Higher levels of physical activity were significantly protective against
hip fractures among both the leaner (BMI <25; P for
trend <.003) and heavier women (BMI ≥25; P for
trend <.001). However, the heavier women had a lower fracture risk in every
activity category (Figure 1). Even
among the leanest (BMI <21) and heaviest (BMI ≥30) women, we observed
significant inverse linear associations between activity and risk of hip fracture
(P for trend = .04 in both BMI strata).
The association between activity and hip fractures appeared dissimilar
for users and nonusers of postmenopausal hormones (Figure 2). However, a test for interaction was not statistically
significant (P = .12). Among the nonusers, there
was a steep decline in risk (P for trend <.001)
with higher levels of activity. For the postmenopausal hormone users, risk
was significantly lower compared with nonusers in the lowest activity category
of less than 3 MET-h/wk (RR, 0.45; 95% CI, 0.26-0.78) and there was little
further risk reduction with higher activity levels (P for
trend = .24). In the highest activity category of 24 MET-h/wk or higher, the
reduced risk of hip fracture was essentially the same for the hormone users
(RR, 0.29; 95% CI, 0.16-0.51) and nonusers (RR, 0.33; 95% CI, 0.22-0.50) when
both were compared with nonusers in the lowest activity category. We also
examined the association between physical activity and hip fracture stratified
by median years of age and by median intakes of calcium, vitamin D, and retinol,
but did not find any evidence that the association differed in the upper and
lower strata of these variables.
We explored the risk of hip fracture among women who increased or decreased
their level of activity based on the differences in hours per week reported
on the 1980 and 1986 questionnaires (Table
3). Consistent with our primary analyses, risk was assessed from
1986-1998 and women with a diagnosis of cancer, heart disease, stroke, or
osteoporosis were excluded at baseline. Among women who reported a low activity
of less than 1 h/wk in 1980, those who increased their activity to 4 h/wk
or more by 1986 had an RR of 0.53 (95% CI, 0.27-1.04) compared with those
who remained in the low-activity category. Risk appeared to decrease as the
1986 activity level increased (P for trend = .07).
Among women who reported a high activity level of 4 h/wk or more in 1980,
risk of hip fracture was doubled among those who decreased to less than 1
h/wk in 1986 (RR, 2.08; 95% CI, 1.20-3.61) compared with those who remained
in the high activity category. Risk increased linearly with increasing reduction
of activity (P for trend = .004). Similar results
were found when comparing change in activity between 1986 and 1992. For those
who increased activity from less than 3 to 15 MET-h/wk or higher from 1986
to 1992, the adjusted RR for hip fracture was 0.34 (95% CI, 0.13-0.88) and
for those who decreased activity from 15 MET-h/wk or more to less than 3 MET-h/wk,
the RR was 1.84 (95% CI, 0.86-3.92).
Since walking was the primary activity for the postmenopausal women
in this cohort, we examined whether walking was associated with a lower risk
of hip fracture. No other activity was reported with sufficient frequency
for an individual analysis. To focus only on walking, we excluded women at
baseline and during follow-up when they reported engaging in any other activity
for 20 min/wk or more. Compared with women who reported no activity or who
walked for less than 1 h/wk, those who walked 4 h/wk or more had a significantly
lower risk of hip fracture (RR, 0.59; 95% CI, 0.37-0.94) and there was a significant
dose-response of lower risk with longer duration of walking (P for trend = .02; Table 4).
Walking pace was also a significant predictor of hip fracture. Compared with
an easy pace, women reporting an average pace had 49% lower risk and women
reporting a brisk to very brisk pace had 65% lower risk. When both duration
and pace were analyzed in the same multivariate model, the RRs for pace did
not change while those for duration were attenuated (RR, 0.72; 95% CI, 0.45-1.16
for ≥4 h/wk).
Sitting and standing were assessed as measures of inactivity in this
cohort. Sitting was not significantly associated with risk of hip fracture
(Table 5), although a nonsignificant
increase in risk was observed among the women sitting 55 h/wk or more (RR,
1.29; 95% CI, 0.85-1.96) compared with those sitting for less than 10 h/wk
after controlling for hours of standing, total MET-hours per week, BMI, and
the other measured risk factors. In contrast to sitting, we observed a significant
dose-response relationship between standing and risk of hip fracture (P for trend = .01). Compared with women who stood for less
than 10 h/wk, women standing for 55 h/wk or more had a significantly lower
(46%) risk. Standing for any duration of 10 h/wk or more was associated with
a significantly lower (28%) fracture risk (RR, 0.72; 95% CI, 0.53-0.97).
In this 12-year prospective study among postmenopausal women, total
physical activity from exercise and leisure-time activities was associated
with a significantly lower risk of hip fracture. Our primary measure of activity
was a MET-hour, which combined an assessment of duration and intensity. Risk
of hip fracture declined 6% for every increase of 3 MET-h/wk (equivalent to
1 h/wk of walking at an average pace). Previous prospective studies using
differing measures of activity among older men and women have reported a 25%
to 39% lower risk of hip fracture in the active vs inactive participants.23-25
As observed in this and other studies,26,27 higher
BMI is also associated with a reduced risk of hip fracture, likely due to
its weight-bearing effect on bone, the protection supplied by padding around
the hips in the event of a fall, and the conversion of androgens to estrogen
in fatty tissues.28 However, we found that
heavier women could further reduce their fracture risk by engaging in more
physical activity. Though lean women also appeared to benefit from activity,
the very elderly or those with involuntary weight loss may be at higher risk
of fractures due to general frailty.29
Even during adult years, initiation of regular physical activity can
reduce fracture risk, but activity must be maintained to preserve the benefits.
We found that risk of hip fracture decreased among sedentary women who increased
their activity to 4 h/wk or more compared with those who remained sedentary.
Conversely, risk increased among those who were actively exercising but became
sedentary. Although women with a major chronic disease were removed from this
analysis, we cannot exclude the possibility that other medical conditions
or underlying disease contributed to both the reduced activity and increased
fracture risk. Similar to our finding, Hoidrup et al25 reported
that risk of hip fracture increased among moderately active men and women
who were sedentary 6 years later compared with those who remained in the moderately
Several studies have reported an interaction between activity and postmenopausal
hormone use. In clinical research, a combination of estrogen supplementation
plus exercise was more effective than exercise alone in increasing trabecular
bone mineral density in older women.30 Population
studies have observed a reduced risk of hip fracture with postmenopausal hormone
use among sedentary women, but not among physically active women.31,32 In our cohort, we found that active
women not taking supplemental estrogen had similar protection against hip
fractures as that provided by hormone use. Interactions reported between the
effects of calcium intake and physical activity on bone density33,34 were
not supported by our data.
Based on accumulated evidence for all health outcomes, at least 30 minutes
to 1 hour of moderate intensity exercise on most days of the week is recommended
for adults.35,36 However, recommendations
for bone health may be different from those focused on cardiovascular fitness
in which intensity of activity to raise heart rate is a critical factor. A
high peak load or impact may be more important than endurance.37,38 Also,
vigorous exercise is associated with a higher risk of fall-related fractures,2 particularly in the elderly and those with functional
limitations.39 For bone, activities that improve
balance or flexibility are important to reduce the risk of falling,40 while weight-bearing activities and resistance training
can increase muscle size and strength7,10 and
lead to higher bone mineral density at the muscle site.17,41
Walking may increase femoral bone density,42 and
it is a relatively safe and easy activity and already the most common exercise
among older adults.35 In our cohort, walking
for 4 h/wk or more was associated with a 41% lower risk of hip fracture. A
faster pace was also associated with lower risk, perhaps because of a greater
impact on the bone. Several cross-sectional studies have reported positive
correlations between walking and bone density.43-45 A
prospective study reported a 30% lower risk of hip fracture among women who
walked for exercise.46 In relatively short-term
clinical trials, brisk walking attenuated femoral bone loss, but increased
the risk of falling,47 while a walking program
increased spinal bone mineral density, but had no effect at the femoral site.48
Standing was also associated with a lower risk of hip fracture in our
cohort, independent of body weight and time spent in leisure-time activities.
As a weight-bearing activity, standing could confer benefits to balance and
muscle that may translate into improved bone strength and protection against
hip fracture. Although prior research is limited, the prospective Study of
Osteoporotic Fractures46 reported a 70% increased
risk of hip fracture among postmenopausal women who stood for less than 4
h/d, and a cross-sectional study49 found that
active nurses had higher femoral bone mineral densities compared with clerks
sitting at a desk.
The results of this study are applicable to white postmenopausal women
and may not be generalizable to men, to women of other racial or ethnic backgrounds,
or to a more elderly or frail population. Also, we lacked prospective data
on frequency of falling and it is possible that women who experienced a bad
fall but did not break a bone were more cautious and therefore limited their
In conclusion, more leisure-time activity is associated with a lower
risk of hip fractures in postmenopausal women. Walking is the most common
exercise and is a suitable activity for lowering fracture risk. Both lean
and heavy women can reduce their fracture risk by increasing their level of