Methods
Study Population
Data from 5,544 white and black men (mean age 73) from the Osteoporotic Fractures in Men Study (MrOS) and 1,114 white and black women (mean age 77.6) from the Study of Osteoporotic Fractures (SOF) were examined. MrOS and SOF are multicenter prospective cohort studies designed to identify risk factors for osteoporosis and osteoporotic fracture. In MrOS, 5,994 older men were recruited from six sites (Birmingham, AL; Minneapolis, MN; Palo Alto, CA; Pittsburgh, PA; Portland, OR; and San Diego, CA) across the United States from March 2000 to April 2002. In SOF, 9,704 women were recruited from four U.S. sites (Baltimore, MD; Minneapolis, MN; Pittsburgh, PA; Portland, OR) between 1986 and 1988. The original SOF cohort was enhanced by the addition of 662 African-American women recruited between 1997 and 1998. To be eligible, men and women needed to be aged 65 and older, be able to walk without assistance from another person, and have reported no bilateral hip replacement. Human subjects approval was obtained at all sites, with written informed consent obtained from all participants.
Body composition determined using whole-body dual energy X-ray absorptiometry (DXA) was available for the 5,544 white and black men from MrOS at baseline and a subset of 1,114 white and black women from SOF (when recruited for the Year 10 examination). Women without whole-body DXA were not included because sarcopenia cannot be assessed without appendicular lean mass.
BMD Measurement
BMD (g/cm) was measured using DXA at the total hip and femoral neck in men (QDR 4500, Hologic, Bedford, MA) and in women (Hologic 1,000 and 2,000). Details of the measurement and densitometry procedures have been published elsewhere. Briefly, certified technicians performed the DXA scans following a strict protocol. To assess longitudinal performance of the scanners, an anthropometric spine phantom was scanned daily and a hip phantom weekly at each clinical center. In both sexes, the right hip was scanned unless there was a fracture, implant, hardware, or other problem, in which case the left hip was scanned. Individuals were classified as having low BMD if their femoral neck T-score was less than −1, according to the 1994 World Health Organization recommendation. Subjects were considered to have normal BMD if their T-score was −1 or greater. T-scores were calculated using the National Health and Nutrition Examination Survey III reference database. Young Caucasian women were used as the reference population in men and women, as the International Society for Clinical Densitometry recommends.
Sarcopenia Assessment
The definition of sarcopenia was based on the European Working Group on Sarcopenia in Older Persons. Participants were classified as sarcopenic if they had low lean mass plus slowness (classified according to gait speed) or weakness (assessed according to grip strength). Low lean mass was defined by correcting appendicular lean mass for height and fat mass. Linear regression was used to model the relationship between appendicular lean mass on height (meters) and fat mass (kg). The 20th percentile of the distribution of residuals was used as the cutpoint for low muscle mass. Separate models were fit for men and women. The residuals method was used because, in the Health, Aging and Body Composition (Health ABC) and Framingham studies, the residuals method was a better predictor of disability and mobility limitations in men and women than other definitions of sarcopenia. Walking speed was calculated as the average of two usual-walking-pace attempts over 6 m and expressed as m/s. Slowness was defined as gait speed slower than 0.8 m/s. Grip strength was measured using a dynamometer in men (Jamar, Jackson, MI) and in women (Sparks Instruments and Academics, Coralville, IA). Maximum grip strength from all attempts was used in the analysis. Weakness was assessed according to grip strength and characterized as less than 30 kg for men and less than 20 kg for women. Height was measured on a stadiometer. Lean mass of extremities and total body fat were obtained using the Hologic QDR 4,500 for men and 2,000 for women. Appendicular lean mass was calculated as the sum of lean mass in the arms and legs. Bone mineral content was removed from the lean mass calculation.
Subjects' Classification
Participants were classified into four groups based on their bone mass and sarcopenia status: normal BMD, no sarcopenia (3,367 men (61%), 308 women (28%)); normal BMD, sarcopenia (79 men (1%), 48 women (4%)); low BMD, no sarcopenia (1,986 men (36%), 626 women (56%)); low BMD, sarcopenia (112 men (2%), 132 women (12%)).
Other Measurements
Covariates were assessed at baseline in men and at Year 10 in women, at the time of the whole body DXA. Participants completed questionnaires and interviews that collected information on demographic characteristics, lifestyle, medical history, and a medication inventory. Participants were asked to bring all prescription and over-the-counter medications to the clinic for verification of use. Smoking status was categorized as current or not (former, never), and alcohol consumption was assessed as the average number of drinks per week. Participants were asked whether they walked for exercise. Self-rated health was categorized as excellent or good versus fair, poor, or very poor. Information on history of falls in the past year and previous fractures was obtained. Functional status was assessed by asking about difficulty with five instrumental activity of daily living (IADLs) (walking 2 or 3 blocks outside on level ground, climbing 10 steps without resting, preparing meals, doing heavy housework, shopping for groceries or clothes). Weight was measured on balance beam scales (except for one MrOS site, which used a digital scale). Body mass index (BMI) was calculated as weight in kilograms divided by square of height in meters.
Fracture Ascertainment
All nonspine fractures were identified through questionnaire follow-ups mailed every 4 months to ask whether participants had sustained a new fracture; these contacts were more than 95% complete. Participants who reported a fracture were asked about the circumstances of the fracture. The degree of trauma was categorized as fall from a standing height or less; fall on stairs, steps, or curb; fall from more than standing height; and traumatic. Traumatic fractures (minimal, moderate, severe) were included because they have been found to be associated with low BMD. Pathological fractures were excluded. All fractures were confirmed according to radiographic report and adjudicated centrally over a mean of 9 years for men and 8 years for women. These analyses included fractures that occurred between 2000 and 2012 in men and between 1997 and 2009 in women. The follow-up time ended at the date of the first fracture, date of death, date of last contact, or database lock. Traumatic fractures were excluded from sensitivity analyses (167 in men (19%), 23 in women (8%)).
Statistical Analysis
Baseline characteristics were compared across groups using analysis of variance for continuous variables and chi-square tests for categorical variable. Pairwise comparisons of the baseline characteristics were calculated; P-values are shown in Table 1 and Table 2.
For the primary outcome, age was initially adjusted for. The incidence rates of nonspine fractures for each of the four groups were estimated using a Poisson distribution. Using Cox proportional hazards models, age- and multivariable-adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated. Participants with normal BMD and no sarcopenia formed the reference group. The multivariable-adjusted model included established risk factors for fracture: age, race, fall history, previous fracture history, current smoking, glucocorticoids, rheumatoid arthritis, alcohol consumption, IADL impairments, and physical activity. Backward elimination was used to drop all variables that did not reach a statistically significant level of P < .10. The effect of the interaction term between low BMD and sarcopenia on fracture risk was assessed. Sensitivity analyses were conducted for different sarcopenia operational definitions. The associations between low BMD and fracture risk, adjusting for sarcopenia, and between sarcopenia and fracture risk, adjusting for low BMD, were also examined. Pairwise comparisons were made in participants who had a nonspine fracture to compare the circumstances of the fracture between the four groups. Separate analyses were done for men and women using SAS version 9.3 (SAS Institute, Inc., Cary, NC).