Physical Activity and Osteoporotic Fracture Risk in Older Women

Methods

Patients

The study population consisted of 9704 nonblack women 65 years of age and older who were participating in the Study of Osteoporotic Fractures [9]. Participants were recruited from population-based lists (health maintenance organizations, voter registration, and motor vehicle tapes) in Baltimore, Maryland; Minneapolis, Minnesota; Portland, Oregon; and the Monongahela Valley, Pennsylvania, from 1986 to 1988. Women were excluded from the Study of Osteoporotic Fractures if they had had bilateral hip replacement, were unable to walk without the assistance of another person, or were institutionalized. All participants gave written informed consent and attended a clinic visit during which questionnaires were given, interviews were conducted, and measurements were taken.

Assessment of Physical Activity and Inactivity

Physical activity was assessed by using a modified version of the Harvard Alumni Questionnaire [27, 28]. Participants were asked to report the frequency and duration of their participation in 33 physical activities during the past year. They were also asked to give the number of city blocks or the equivalent walked each day for exercise or as part of daily activities and how many flights of stairs they climbed. A summary estimate of total energy expenditure was calculated according to methods described elsewhere [27, 28]. Leisure-time physical activities were attributed an intensity weighting of 5 kcal/min for low intensity, 7.5 kcal/min for medium intensity, or 10 kcal/min for high intensity; each city block walked was assigned 8 kcal/min, and each stair climbed was assigned 4 kcal/min. The total physical activity index, expressed in kilocalories per week, is the sum of kilocalories expended in sport and recreational activities, blocks walked, and stairs climbed.

To estimate physical activity related to household chores, participants were asked, “About how many hours per week do you usually spend doing heavy household chores, such as scrubbing floors, vacuuming, sweeping, yard work, gardening, or shoveling snow?” To estimate inactivity, women were asked how many hours per day they spend sitting upright.

Other Measurements

Body weight and height were measured by using a balance-beam scale and a stadiometer. Bone mineral density (g/cm²) was measured by using single-photon absorptiometry (OsteoAnalyzer, Siemens-Osteon, Wahiawa, Hawaii) at the distal radius and calcaneus. Hip abduction strength was measured by having the participant lay supine and exert a lateral force against a dynamometer (Sparks Instruments and Academics, Coralville, Iowa) placed 3 cm above the lateral malleolus and held by an observer. Details of these measurements have been reported elsewhere [9, 29].

Additional questions evaluated self-rated health; calcium intake from food; alcohol intake; and use of medications, including hormone replacement therapy, benzodiazepines, and anti-anxiety agents. To assess function, women were asked whether they had difficulty walking two to three blocks on level ground, walking up or down 10 steps, preparing meals, doing heavy housework, or shopping for groceries or clothing. They were also asked about their history of physician-diagnosed medical conditions, including osteoporosis, hypertension, diabetes, arthritis, and stroke.

Assessment of Incident Fractures

Methods for identifying fractures have been published elsewhere [30]. Participants were contacted every 4 months to ask whether they had had a fracture. Follow-up for fracture ascertainment was more than 99.5% complete. Radiographic reports and films were obtained to confirm hip fractures. Duration of hip and wrist fracture follow-up was calculated as the time to first occurrence of a fracture. Follow-up ranged from 0.2 to 9.6 years (mean, 7.6 years).

To measure vertebral fracture incidence, we obtained lateral radiographs of the thoracic and lumbar spine from 7238 women at baseline and an average of 3.7 years later. This sample was reduced because of inadequate technical measurements at baseline (n = 129), failure or refusal to undergo radiography (n = 341), failure to attend the follow-up clinic visit (n = 1528), or death (n = 468). Incident vertebral fracture was defined by morphometry as a reduction in the height of the anterior, middle, or posterior dimension of a vertebral body of 20% and at least 4 mm [31].

Statistical Analysis

Women were classified according to quintiles of total physical activity. For heavy chores and sitting, for which there was a limited range of responses and data were skewed, women were grouped into approximate tertiles. To evaluate fracture risk according to intensity of energy expenditure, women were grouped into three exclusive groups: inactivity, low-intensity activity, and moderate to vigorous activity. Inactive women did not participate in any sport or recreational activity and did not walk for exercise. Women who did any sport or recreational activity or walked for exercise were grouped according to the highest intensity activity in which they participated. For example, if a woman reported gardening (light) and aerobic dance (moderate to vigorous), she would be classified in the moderate-to-vigorous intensity group. Additional analyses further stratified women within exercise intensity groups according to whether they did less than or more than 2 hours per week of total sport and recreational activity.

Analysis of covariance adjusted for age and chi-square tests of homogeneity were used to evaluate potential confounders across physical activity levels. Proportional hazards regression was done to calculate the relative risk (and 95% CIs) for hip and wrist fracture associated with physical activity level; the least active group for activity variables was the reference group. A multivariate model was also used that included age, weight, smoking status (current or never/past), use of estrogen therapy (current or never/past), self-rated health (five levels ranging from excellent to very poor), dietary calcium intake (mg/d), alcohol intake (drinks/wk), falls in the year before baseline (two or fewer than two), and functional difficulty (yes or no). These variables were selected on the basis of the a priori hypothesis that they could be related to both physical activity and risk for fracture. To determine whether any relations between physical activity and risk for fracture were explained by differences in bone density or muscle strength, calcaneal bone mineral density and hip abductor strength were added to the multivariate models in separate steps. We used logistic regression for vertebral fractures by applying a similar modeling strategy.

Additional analyses using different exclusion criteria were done to determine whether any relations among physical activity and fracture risk were explained by differences in health or functional status at baseline. To determine whether the relation between physical activity and hip fracture was explained by poor acute health among inactive women at baseline, analyses excluded 1) women who reported any difficulty walking two to three blocks on level ground, had a severe gait abnormality, or required a walking aid; 2) women with fair or poor self-rated health or history of diabetes, stroke, falls, or hip fracture before baseline measurement; and 3) women who had a fracture or died in the first 3 years after baseline measurement. All analyses were done using Statistical Analysis Software (SAS Institute, Inc., Cary, North Carolina).

Results

Characteristics of the overall study population, stratified by quintile of total kilocalories of physical activity, are listed in Table 1. Women in lower quintiles of total physical activity were older, were heavier, were more likely to smoke, were less likely to use estrogen and alcohol, and had lower calcium intakes than women in higher quintiles (P < 0.05). Women who were more active were less likely to report functional difficulties, had higher self-rated health, and had higher calcaneal and distal radius bone mineral density (P < 0.001) than did less active women. The proportion of women who fell twice or more per year was greater among the least active (12%) and most active quintiles (11.9%) than among the middle quintiles (9.6 to 10.2%). After an average follow-up period of 7.6 years (range, 0.2 to 9.6 years), 426 women had hip fractures and 523 had wrist fractures, for a yearly incidence of 5.8 hip fractures and 7.2 wrist fractures per 1000 person-years. Incident vertebral fractures occurred in 389 of the 7238 women who had spine films at baseline and 3.7 years later (cumulative incidence, 5.4%).

Hip Fractures

Each increasing quintile of total physical activity was associated with a reduced relative risk for hip fracture. Women in the highest quintile had a 42% lower age-adjusted risk (relative risk, 0.58 [95% CI, 0.42 to 0.80]) compared with the least active quintile (Table 2). Increasing intensity of energy expenditure was also associated with a progressive decrease in hip fracture incidence: Women who reported participation in low-intensity and moderate-to-vigorous physical activity had 27% and 45% reductions in age-adjusted risk for hip fracture, respectively, compared with inactive women (Table 2).

When women were stratified within intensity groups according to whether they did more than 2 hours of physical activity per week (results not shown), the greatest reductions in risk were among women who did moderate-to-vigorous activities and at least 2 hours of sport and recreational activities per week (relative risk, 0.47 [CI, 0.32 to 0.68]). Women in the moderate-to-vigorous exercise group who reported fewer than 2 hours of activity per week had a relative risk of 0.76 (CI, 0.53 to 1.10), whereas those who reported only low-intensity activities for less than 2 hours and more than 2 hours had relative risks of 0.77 (CI, 0.60 to 0.99) and 0.74 (CI, 0.56 to 0.98), respectively.

Graded reductions in hip fracture risk were also seen for women who reported more hours per week of heavy chores. Women who reported at least 10 hours per week had a 22% reduction in risk (CI, 2% to 39%) compared with women who reported less than 5 hours per week. The more hours a woman spent sitting per day, the higher her risk for hip fracture: Women who sat for at least 9 hours per day had a 43% higher risk (CI, 12% to 82%) than those who sat for less than 6 hours per day.

Adjustment for smoking, estrogen therapy, body weight, number of falls, intake of calcium and alcohol, functional difficulties, and self-rated health at baseline only slightly weakened the relation between physical activity and risk for hip fracture (4% to 6%). After adjustment for these confounders, very active women (fourth and fifth quintiles) had a 36% reduction in risk for hip fracture (multivariate relative risk, 0.64 [CI, 0.45 to 0.89]) compared with inactive women (lowest quintile). With the exception of those for heavy chores, tests for trend remained significant (Table 2).

Wrist Fractures

Women with higher levels of total physical activity and those who reported more heavy chores per week tended to have a reduced risk for wrist fracture, but these relations were not statistically significant (P > 0.2 for total physical activity; multivariate P for trend for heavy chores, 0.09) (Table 3). Neither intensity of energy expenditure nor hours sitting per week were significantly associated with risk for wrist fracture. Addition of other covariates to the regression models did not alter these findings appreciably.

Vertebral Fractures

Total physical activity was not significantly associated with risk for vertebral fracture (Table 4). Although women in the middle quintile had a significantly reduced odds ratio for vertebral fracture (0.63 [CI, 0.44 to 0.88]), these findings were not consistent across quintiles, and there was no evidence of a linear trend (P > 0.2). Intensity of physical activity, however, was inversely associated with odds of vertebral fracture (Table 4). Compared with inactive women, women who did moderate-to-vigorous exercise had a 33% reduction (CI, 6% to 51%) in vertebral fracture odds after multivariate adjustment. This reduction was apparent in women who reported less than 2 hours of sport and recreational activity per week (odds ratio, 0.70 [CI, 0.45 to 1.07]) and those who reported more than 2 hours of sport and recreational activity per week (odds ratio, 0.67 [CI, 0.46 to 0.97]) (results not shown). Reported hours of heavy chores per week and sitting per day were not significantly associated with vertebral fracture (Table 4).

Additional Analyses

We evaluated models in which calcaneal bone mineral density and hip abduction strength were added to the multivariate model separately and in combination to determine whether bone muscle density or muscle strength intervened in the relation between physical activity and hip fracture (Table 5). Adding these variables weakened the associations between physical activity quintile and risk for fracture by 3% to 11%. Risk estimates were not appreciably altered in separate multivariate models that excluded women with mobility difficulties, comorbid conditions, poor self-rated health, and previous fractures and women who were censored (had hip fracture or died) in the first 3 years of follow-up.

Discussion

In this prospective, 7.6-year evaluation of 9704 women, higher levels of physical activity were associated with a reduced incidence of hip fracture. This dose-response relation amounted to a 36% reduction in risk for hip fracture for women in the highest two quintiles of total physical activity and a 42% reduction in women who did moderately intense or vigorous recreational activities. Multivariate adjustment and follow-up analyses done using several exclusion criteria indicated that the relation of physical activity to hip fracture was not confounded by health status, functional status, history of falling, or other health behaviors.

These findings are consistent with previous case–control studies and one prospective study in which moderately active women had statistically significant 40% to 50% reductions in risk for hip fracture compared with sedentary women [18, 21, 22, 24]. These results also corroborate the earlier finding from the Study of Osteoporotic Fractures that women who walked for exercise had a 30% reduced risk for fracture compared with those who did not walk for exercise after an average follow-up of 4.1 years [10]. Other studies have suggested that physical activity protects against osteoporotic fractures, but these studies failed to find statistical significance, lacked comprehensive physical activity assessment, or did not control for potential confounding behaviors or health status [25, 26].

Our study expanded on previous research by evaluating the characteristics of physical activity and their relation to risk for fracture. The association of total physical activity with risk for hip fracture seemed to be related to both the amount and the intensity of physical activity. Women who did moderately intense or vigorous activities, such as aerobic and other forms of dance, tennis, and weight training, had greater reductions in hip and vertebral fracture risk than did women who did lighter activities, such as walking and gardening, particularly if they were active for at least 2 hours per week. This decrease in risk among moderately and vigorously active women compared with women who did low-intensity activities for a similar duration suggests that vigorous activity was not simply a marker of greater total duration.

Physical activity may prevent hip fractures in several ways. Exercise may reduce the likelihood of falling or may enable a protective response in the event of a fall through enhanced balance, reaction time, coordination, mobility, and muscle strength [15-17, 32]. Exercise may also enhance bone mineral density or the structural integrity of bone, reducing the likelihood of fracture in the event of a fall [11-14]. Although our analyses did not attempt to explain how physical activity prevents fractures, the fact that physical activity was related to fracture risk after we controlled for calcaneal bone mineral density, hip strength, and number of falls suggests that the link between physical activity and hip fracture is multifactorial and is not completely explained by its effects on bone mass and muscle strength.

We found that total physical activity was not significantly related to the incidence of wrist or vertebral fractures. For wrist fractures, moderate-to-vigorous activities were related to a nonsignificantly increased risk for fracture. These findings may reflect the fact that osteoporotic fractures have heterogeneous causes [32]. It has been suggested that more agile women are more likely to extend their hands during a fall to absorb the impact, causing wrist fractures but preventing other types of fractures [32]. Thus, physical activity may have competing effects on the risk for wrist fracture, enhancing bone density but putting some women at greater risk for falling. Vertebral fractures, on the other hand, may depend more on deficits in bone density and may be less influenced by the tendency to fall than are hip or wrist fractures. Physical activity may not influence bone density sufficiently to have an appreciable effect on vertebral fractures.

Our study had several limitations. The primarily white, independently living, female study sample may limit generalization to other populations. Most physical activity assessment tools, including the Paffenbarger questionnaire used in this study, were originally developed for younger men and may not be sensitive to differences at the low end of the physical activity spectrum. However, Cauley and colleagues [33] found this questionnaire to be reliable and valid in older women. We did not comprehensively assess neuroleptic medications, which could affect the risk for falls, but multivariate models adjusted for falls at baseline and additional analyses that controlled for use of benzodiazepines and anxiety medications (results not shown) did not alter these findings.

The sample size was reduced for vertebral fractures because these fractures were assessed by using vertebral radiography at specific time points (baseline and 3.7 years later). Therefore, women who died or did not attend the follow-up visit were not included in this analysis. Follow-up analyses indicated that women who did not attend the follow-up visit were less active (median kilocalories expended, 719 compared with 1071 for those who attended the visits) and were more likely to have hip fractures (age-adjusted odds ratio, 1.29 [CI, 1.05 to 1.60]). If these women also had higher vertebral fracture rates, our findings may have underestimated the relation between physical activity and risk for vertebral fracture. Despite this limitation, this is the first prospective evaluation of physical activity and vertebral fractures, perhaps because it is difficult to measure vertebral fracture incidence [31].

In the absence of a randomized design, we cannot rule out the possibility of selection bias or residual confounding whereby women with higher bone density, muscle strength, and physical function choose a more active lifestyle. However, multivariable control for factors related to health and functional status and separate evaluation of models excluding women on the basis of mobility factors, health status, and early fracture or death did not alter these findings. It is also possible that higher-intensity activities are more easily recalled than lower-intensity activities; this could explain in part the observed gradient in fracture risk across exercise intensity groups.

In this prospective study of physical activity and risk for fracture in older women, physical activity was related to a reduced risk for hip fracture but was less related to wrist and vertebral fractures. The findings of a dose-response relation between physical activity and hip fracture could have important implications for fracture prevention and underscores the need to enhance physical activity among older women. Although we found the greatest risk reductions among women who participated in moderately intense or vigorous activities, such as tennis or aerobic dance, women who did lower-intensity activities, such as walking, gardening, or social dancing, for at least 1 hour per week also had significant reductions in risk for hip fracture. The latter activities were also the activities of choice for most women in the higher quintiles of physical activity, who simply tended to do more of them (that is, ≥ 2 hours per week). Thus, low-intensity activities may be the most prudent recommendation for sedentary older women. Future research should evaluate whether different types or patterns of physical activity affect other types of osteoporotic fractures and whether they do so primarily through effects on the skeleton, muscular fitness, or balance or through other mechanisms.

Appendix: Investigators in the Study of Osteoporotic Fractures Research Group

University of California, San Francisco (Coordinating Center): S.R. Cummings (Principal Investigator), M.C. Nevitt (Co-Investigator), D.G. Seeley (Project Director), D.M. Black (Study Statistician), H.K. Genant (Director, Central Radiology Laboratory), C. Arnaud, D. Bauer, W. Browner, L. Christianson, M. Dockrell, C. Fox, R. Gore, S. Harvey, M. Jaime-Chavez, L. Laidlaw, R. Lipschutz, L. Lui, G. Milani, L. Palermo, R. San Valentin, K. Stone, H. Tabor, D. Tanaka, and C. Yeung.

University of Maryland: J.C. Scott (Principal Investigator), R. Sherwin (Co-Investigator), M.C. Hochberg (Co-Investigator), J. Lewis (Project Director), E. Peddicord (Clinic Coordinator), A. Bauer, C. Boehm, G. Cullum, L. Finazzo, M.E. Flaks, T. Ford, D. Harris, B. Hohman, E. Oliner, T. Page, J. Schlossberg, C. Shaffer, A. Trimble, and S. Trusty.

University of Minnesota: K. Ensrud (Principal Investigator), P. Schreiner (Co-Investigator), C. Bell (Project Director), E. Mitson (Clinic Coordinator), C. Bird, D. Blanks, S. Estill, S. Fillhouer, S. Fincham, J. Griffith, J. Hansen, F. Imker-Witte, K. Jacobson, K. Kiel, K. Knauth, N. Nelson, E. Penland-Miller, and M. Riley-Alves.

University of Pittsburgh: J.A. Cauley (Principal Investigator), L.H. Kuller (Co-Principal Investigator), M. Vogt (Co-Investigator), L. Harper (Project Director), L. Buck (Clinic Coordinator), C. Bashada, D. Cusick, G. Engleka, A. Githens, M. Gorecki, K. McCune, D. Medve, M. Nasim, C. Newman, S. Rudovsky, and N. Watson.

The Kaiser Permanente Center for Health Research, Portland, Oregon: E. Harris (Principal Investigator and Project Director), W.M. Vollmer (Co-Investigator), E. Orwoll (Co-Investigator), H. Nelson (Co-Investigator), K. Crannell (Project Administrator and Clinic Coordinator), J. Bender, A. Doherty, K. Easter, M. Erwin, F. Heinith, J. Kann, K. Redden, C. Romero, K. Snider, and C. Souvanlasy.

From University of Vermont College of Medicine, Burlington, Vermont; University of Pittsburgh, Pittsburgh, Pennsylvania; University of California at San Francisco, San Francisco, California; and Veterans Affairs Medical Center, University of Minnesota, Minneapolis, Minnesota.

Drs. Seeley and Bauer: Departments of Medicine, Epidemiology, and Biostatistics, 74 New Montgomery Street, Suite 520, University of California at San Francisco, San Francisco, CA 94105.

Dr. Ensrud: Section of General Medicine, Veterans Affairs Medical Center, One Veterans Drive, Minneapolis, MN 55454.