Association of Physical Activity and Fracture Risk Among Postmenopausal Women

Key Points Question Is the amount and intensity of physical activity associated with total and site-specific fracture among postmenopausal women who participated in the Women’s Health Initiative study? Findings In this cohort study of 77 206 postmenopausal women with a mean follow-up of 14 years, higher amounts of total, mild, moderate to vigorous, and walking physical activity were significantly associated with lower risk of hip fracture. Positive associations existed for moderate to vigorous physical activity with wrist or forearm fractures and for sedentary behavior with total fractures. Meaning Regular physical activity, including lighter-intensity activities, and less sedentary time is associated with reduced risk of fracture in older women.


Introduction
Approximately 1.5 million fractures occur annually in women who live in the United States, accounting for $12.7 billion in health care costs. 1 Approximately 14% of fractures occur in the hip 1 ; mortality after hip fracture is as high as 20%. 2 Fracture has been associated with low bone mineral density (BMD), propensity to fall, and declines in muscle strength, balance, mobility, and physical functioning. [3][4][5] The 2008 Physical Activity Guidelines Advisory Committee evaluated quality and quantity of evidence from 21 studies and concluded that people with higher total physical activity (PA) levels have 36% to 68% lower risk of hip fracture. 6 The Advisory Committee Report for the 2018 revision of the PA guidelines did not include an explicit update on fracture outcomes but did indicate that evidence supports the conclusion that higher amounts of total PA are associated with lower risk of falls and fall-related injuries, including bone fracture. 7 In both the 2008 and 2018 PA guidelines, consensus was lacking regarding fracture risk at sites beside the hip. The majority of published studies assessed PA as a composite measure; thus, the role of PA types and intensities in fracture is unclear. Sedentary behavior (eg, sitting time) is becoming an established modifiable risk factor for major forms of morbidity and mortality, independent of PA habits 7 ; however, its contribution to fracture has not been systematically evaluated. 6,7 The Women's Health Initiative (WHI) is a prospective cohort study among postmenopausal women with ongoing assessment of fractures. We examined recreational PA, household activities, walking, and sedentary behavior in association with incident fracture and the extent to which age, race/ethnicity, or fall frequency modified this association in older, community-dwelling, ambulatory women.

Study Population
The WHI observational study design has been published. 8 Recruitment of participants was conducted at 40 US clinic centers from October 1993 through December 1998, enrolling 93 676 postmenopausal women aged 50 to 79 years. 9 Women with predicted survival of less than 3 years or with conditions that might compromise retention were ineligible. Study protocols were approved by institutional review boards at participating institutions. Written informed consent was obtained from participants. The initial WHI observational study concluded in 2005. Additional follow-up was obtained from women who consented to participate in 2 WHI Extension Studies (2005-2010 and jogging, tennis, or swimming laps; moderate PA, less exhausting activities, such as biking outdoors, using an exercise machine, calisthenics, easy swimming, or popular or folk dancing; and mild PA, slow dancing, bowling, or golf. Walking was assessed separately from these activities with the following questions: "how often do you walk outside the home for more than 10 minutes without stopping," "when you walk outside the home for more than 10 minutes without stopping, for how many minutes do you usually walk," and "what is your usual speed." Nonrecreational activities, including the time (hours per week) spent on heavy indoor household chores (ie, scrubbing floors, sweeping, or vacuuming) and yard work (ie, mowing, raking, gardening, or shoveling snow) were queried using questions specific to these constructs.
Physical activity was summarized as energy expenditure, calculated as the product of metabolic equivalent (MET) intensity values for each activity multiplied by the hours per week of reported participation (MET hours per week). Standard MET values were assigned to mild (3.

Ascertainment of Incident Fracture
Participants were observed from enrollment through September 2015 using annual mailed health questionnaires. As hip fracture was a primary outcome in the WHI, all hip fractures were adjudicated in the main study (1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005) and Extension Study I (2005-2010). Trained physicians reviewed radiology reports, with hospital discharge summaries, operative reports, and clinic and progress notes as additional sources. Self-reported fractures at sites other than the hip were not adjudicated. From 2010 to 2015, self-reported fractures at all sites including hip were not adjudicated. Selfreported fracture in WHI has good agreement (a mean of 76%) with criterion medical records. 15 In the present study, we evaluated total PA in association with total and site-specific fracture end points. We then conducted additional analysis on specific PA types in association with a reduced set of fracture end points (ie, hip, wrist or forearm, and clinical vertebral), which are designated major osteoporotic fracture events in the WHI. 16 This approach was used to reduce the total number of statistical comparisons involving secondary exposures and in subgroups where power may be limited for site-specific end points. and more than 7.5 MET h/wk; (3) mild activity, categorized as none, more than 0 to 3.5 MET h/wk, and more than 3.5 MET h/wk; and (4) moderate to vigorous activity, categorized as less than 9 MET h/wk and at least 9 MET h/wk, to be comparable with guideline-recommended levels. Time to first fracture was analyzed, with censoring at death, loss to follow-up, or the end of follow-up on September 30, 2015. Death was not an event in the models. Owing to differences in hip fracture ascertainment between WHI Extension Study II (ie, self-reported) and the rest of the study (ie, adjudicated), models evaluating hip and total fracture outcomes included a time-dependent stratification by hip ascertainment type within the Cox models.
Multivariable-adjusted models included the following covariates: age, race/ethnicity, education, smoking status, alcohol use, height, weight, history of fracture after age 55 years, bone drug use, corticosteroid use, calcium intake, vitamin D intake, lifetime hormone therapy use, falls in the past year, physical function, thiazide use, diabetes, age at menopause, and osteoporosis history.
Additional models controlled mutually for sedentary time or PA to assess their independent association with fracture risk. The joint association of total PA and sedentary time with risk of total fracture was also examined. Multiplicative interactions for total PA with enrollment age, race/ ethnicity, and fall frequency were explored using cross-product terms. Sensitivity analysis was

Results
After exclusions (n = 3165) and missing covariates (n = 13 305) (eTable 8 in the Supplement), 77 206 women were included in this study. Baseline characteristics of the cohort appear in Table 1. size, most baseline characteristics were significantly associated with total PA, although, in some instances, absolute differences were modest (Table 1).
During a mean (SD) follow-up period of 14.0 (5.2) years, 25 516 women (33.1%) reported experiencing at least 1 fracture.    We next examined nonrecreational activity, including yard work and heavy household chores.
In models adjusted for total recreational PA and sedentary time, more than 6 MET h/wk of yard work was associated with lower risk of total fracture (HR, 0.95; 95% CI, 0.92-0.98; P for trend = .002) and hip fracture (HR, 0.90; 95% CI, 0.82-0.99; P for trend = .04) compared with no yard work ( Table 3).
Yard work was not associated with risks of clinical vertebral or wrist and forearm fractures. Energy expenditure from heavy chores was not associated with total or site-specific fractures (eTable 5 in the Supplement). Table 4 presents associations between sedentary behavior and fracture risks. In age-adjusted models, 9.5 hours of daily sitting or lying down was associated with higher risks of hip fracture (HR, 1.11; 95% CI, 1.01-1.21; P for trend = .03), clinical vertebral fracture (HR, 1.09; 95% CI, 1.01-1.17; P for trend = .02), wrist or forearm fracture (HR, 1.07; 95% CI, 1.01-1.14; P for trend = .11), and total fracture (HR, 1.10; 95% CI, 1.07-1.13; P for trend < .001). Associations were attenuated and not statistically significant in the multivariable-adjusted models, with the exception of total fracture risk, which was attenuated but remained statistically significant, including when further adjusted for total PA (HR, 1.04; 95% CI, 1.01-1.07; P for trend = .01).
We next evaluated risk of total fracture according to jointly classified total PA and sedentary time exposures (Figure). Fracture was inversely associated with total PA, regardless of time spent sedentary.
We examined whether the associations of total PA with hip, wrist or forearm, clinical vertebral, and total fracture differed after stratifying on categories of age, race/ethnicity, and fall frequency history at baseline. There were no significant interactions observed (data not shown).
Finally, to evaluate the robustness of results from our primary analysis that used baseline PA and sedentary exposures, we repeated the analysis using time-varying exposures based on updated information collected after baseline (eTable 6 and eTable 7 in the Supplement). Adjusting for CI, 0.59-0.86; P for trend = .03). Total PA was positively associated with wrist or forearm fracture  Sedentary h/d, median (range) 5.0 (0-6.5) 8.0 (7.0-9.5) 12.0 (10.0-24.0) NA

Discussion
This large cohort study among older, community-dwelling, ambulatory women found that recreational and nonrecreational PA was inversely associated with risks of hip, clinical vertebral, and total fractures. Total PA was positively associated with knee and elbow fracture. Mild-intensity PA was associated with lower risks of hip, vertebral, and total fracture, and MVPA was associated with lower risk of hip fracture but higher risk of wrist or forearm fracture. Yard work was inversely

Strengths and Limitations
Strengths of our study include its prospective design, large sample size, and long follow-up period with low loss to follow-up. We were able to distinguish between various types and intensities of PA; sedentary behavior was also assessed. Mutual adjustment for PA and sedentary behavior when analyzing each exposure provides an approach for teasing out these interrelated factors of fracture incidence. Repeated assessments of PA and sedentary time permitted time-varying exposure analysis, the results of which were similar to those for baseline sedentary exposure and somewhat stronger than those for baseline PA. Limitations include the use of self-report questionnaires to assess PA; misclassification on exposure is inevitable. 49 Because the PA (and sedentary behavior) assessment preceded fracture occurrence, exposure misclassification might be expected to be nondifferential and any resulting bias of associations toward the null, 50 but this may not be the case in all such circumstance. 51 The questionnaire reliability, 10,13 similarity of results for baseline and time-varying exposure analyses, and previously observed associations with major disease end points affecting older women 12,52,53 enhances confidence in the study results. Fracture outcome ascertainment was based only on selfreport after 2010 in the WHI. High validity for self-reported hip (78%) and wrist or forearm (81%)

JAMA Network Open | Public Health
fractures, but lower validity (51%) for vertebral fractures has been reported in WHI. 15 Physical activity can be influenced by multilevel sociocultural and ecological forces, 54 which are challenging to quantify and account for statistically in studies such as ours. Further, we conducted many analyses, and some associations could be owing to chance alone; results should be interpreted accordingly.

Conclusions
In this cohort study, greater amounts of total PA were associated with lower risk of total fracture, but associations varied by fracture site. Greater MVPA was associated with lower risk of hip fracture but higher risk of wrist or forearm fracture. Mild activity was inversely associated with risks of hip, clinical vertebral, and total fracture, independent of other PA and sedentary behavior. The current results suggest that lower-intensity activities, including walking and nonrecreational activities, could have benefit on fracture risk at older ages. If confirmed, future recommendations on fracture prevention in postmenopausal women should promote light PA, especially in those who are frail and unable to safely engage in more intense activities. Sedentary behavior as an independent factor predisposing individuals to fracture requires further investigation.