Association of BMD and FRAX Score With Risk of Fracture in Older Adults With Type 2 Diabetes

Context Type 2 diabetes mellitus (DM) is associated with higher bone mineral density (BMD) and paradoxically with increased fracture risk. It is not known if low BMD, central to fracture prediction in older adults, identifies fracture risk in patients with DM.

Objective To determine if femoral neck BMD T score and the World Health Organization Fracture Risk Algorithm (FRAX) score are associated with hip and nonspine fracture risk in older adults with type 2 DM.

Design, Setting, and Participants Data from 3 prospective observational studies with adjudicated fracture outcomes (Study of Osteoporotic Fractures [December 1998-July 2008]; Osteoporotic Fractures in Men Study [March 2000-March 2009]; and Health, Aging, and Body Composition study [April 1997-June 2007]) were analyzed in older community-dwelling adults (9449 women and 7436 men) in the United States.

Main Outcome Measure Self-reported incident fractures, which were verified by radiology reports.

Results Of 770 women with DM, 84 experienced a hip fracture and 262 a nonspine fracture during a mean (SD) follow-up of 12.6 (5.3) years. Of 1199 men with DM, 32 experienced a hip fracture and 133 a nonspine fracture during a mean (SD) follow-up of 7.5 (2.0) years. Age-adjusted hazard ratios (HRs) for 1-unit decrease in femoral neck BMD T score in women with DM were 1.88 (95% confidence interval [CI], 1.43-2.48) for hip fracture and 1.52 (95% CI, 1.31-1.75) for nonspine fracture, and in men with DM were 5.71 (95% CI, 3.42-9.53) for hip fracture and 2.17 (95% CI, 1.75-2.69) for nonspine fracture. The FRAX score was also associated with fracture risk in participants with DM (HRs for 1-unit increase in FRAX hip fracture score, 1.05; 95% CI, 1.03-1.07, for women with DM and 1.16; 95% CI, 1.07-1.27, for men with DM; HRs for 1-unit increase in FRAX osteoporotic fracture score, 1.04; 95% CI, 1.02-1.05, for women with DM and 1.09; 95% CI, 1.04-1.14, for men with DM). However, for a given T score and age or for a given FRAX score, participants with DM had a higher fracture risk than those without DM. For a similar fracture risk, participants with DM had a higher T score than participants without DM. For hip fracture, the estimated mean difference in T score for women was 0.59 (95% CI, 0.31-0.87) and for men was 0.38 (95% CI, 0.09-0.66).

Conclusions Among older adults with type 2 DM, femoral neck BMD T score and FRAX score were associated with hip and nonspine fracture risk; however, in these patients compared with participants without DM, the fracture risk was higher for a given T score and age or for a given FRAX score.

Prevention of fractures is an important goal in older adults. It is increasingly recognized that adults with type 2 diabetes mellitus (DM), an estimated 17% of older adults in the United States,¹ have a higher fracture rate.^2-6 Preventive identification of adults at higher fracture risk is based on bone mineral density (BMD) T scores, used alone or in the World Health Organization (WHO) Fracture Risk Algorithm (FRAX) score.⁷ However, because type 2 DM is paradoxically associated with higher BMD and increased fracture risk,² there is concern that these established methods for predicting fractures may not perform adequately in patients with type 2 DM.^8,9 There is a need to clarify the use of standard methods for assessing fracture risk in this expanding population of older adults.

No prospective studies are available on prediction of fracture in older adults with type 2 DM using BMD T scores or FRAX score. We used data from 3 prospective observational studies with adjudicated fracture outcomes, the Study of Osteoporotic Fractures (SOF), the Osteoporotic Fractures in Men Study (MrOS), and the Health, Aging, and Body Composition (Health ABC) study, to assess the associations of BMD T score and FRAX score with hip and nonspine fracture risk in older adults with type 2 DM. Those older adults using insulin are reported to have a higher fracture risk,^3,6 so our results are stratified by insulin use.

In all 3 studies, race/ethnicity was determined based on self-report. Protocols were approved by the institutional review boards at all institutions involved and all participants signed a written informed consent.

The SOF is a longitudinal study of older white women (N = 9704) at 4 clinical centers in the United States designed to identify risk factors for fracture.¹⁰ At baseline (1986-1987), DM was ascertained based on a self-reported diagnosis, and insulin use was ascertained in those women reporting DM. Women were queried regarding use of bone-active medications, including oral glucocorticoids. Approximately 2 years after baseline, femoral neck hip BMD was measured at visit 2 by dual-energy x-ray absorptiometry (DXA) on 7959 women. Women were excluded due to missing data on DM status (n=20) and due to fractures adjudicated as uncertain (n=13), resulting in 7926 women in these analyses. Women were queried about fractures every 4 months by postcard and at clinic visits. Reported fractures were centrally adjudicated and confirmed based on a radiology report or radiograph. In these analyses, only fractures occurring between December 1988 and July 2008, after the hip BMD measurement at visit 2, were included.

The MrOS study is a longitudinal study of older men (N = 5995) who were recruited at 6 clinical centers in the United States and was designed to assess risk factors for fracture.^11,12 Diabetes was ascertained at baseline based on a self-reported diagnosis, self-reported use of a DM medication, or an increased fasting glucose (≥126 mg/dL; to convert to mmol/L, multiply by 0.0555). Participants were asked to identify all prescription medications used in the previous 30 days. Each medication was matched to its ingredients based on the Iowa Drug Information Service Drug Vocabulary (College of Pharmacy, University of Iowa, Iowa City, Iowa). Hip BMD was measured at baseline using DXA on 5994 men who are included in these analyses. Fractures were ascertained and adjudicated as described in the SOF. These analyses included fractures that occurred between March 2000 and March 2009.

The Health ABC study is a longitudinal study of older adults (N = 3075), approximately 50% white and 50% black, recruited at 2 clinical centers in the United States, and designed to assess body composition and physical functioning changes in 70- to 79-year-old adults.¹³ Diabetes was ascertained at baseline using the same criteria as in the MrOS study. Participants were asked to identify all prescription medications used in the previous 2 weeks. Medications were coded as described in the MrOS study. Hip BMD was measured at baseline using DXA on 3043 participants. Of these, 78 participants were excluded because of missing data on DM status. These analyses included 1523 women and 1442 men. Fractures were adjudicated at the clinical centers based on radiograph confirmation. These analyses included fractures that occurred between April 1997 and June 2007.

Femoral neck BMD T scores included in our models were calculated using sex- and race-specific young adult reference values from the third National Health and Nutrition Examination Survey (NHANES III), the method currently used in the output from clinical densitometers.¹⁴ We grouped the race/ethnicity data from the 3 cohorts into the following categories in comparison with NHANES categories (white, Asian, and other categories for NHANES non-Hispanic white; black for NHANES non-Hispanic black; and Hispanic for NHANES Mexican-American). T scores included in the WHO FRAX score were calculated using reference values for white women as required for this algorithm.¹⁵

In the SOF and MrOS studies, the WHO 10-year absolute risks of hip and osteoporotic fracture (FRAX scores) were calculated by the WHO Collaborating Center for Metabolic Bone Disease, using the FRAX algorithm (version 3).^15-19 The FRAX algorithm includes femoral neck BMD T score, age, sex, body mass index, previous history of fracture, parental history of hip fracture, current smoking, recent use of corticosteroids, presence of rheumatoid arthritis, and at least 3 alcoholic beverages per day. The FRAX scores have not been calculated for Health ABC study participants.

All regression analyses were conducted using Stata version 11.2 (StataCorp, College Station, Texas). We used Cox proportional hazards regression models to estimate the effect of T score, adjusted for age, or the effect of FRAX score on the hazard of hip and nonspine fracture in those older adults with and without DM. We checked for interactions of T score with age, race, and insulin use in those older adults with DM, and verified the Cox proportional hazards regression assumption. The C index was calculated using the estat concordance command in Stata.

Cox proportional hazards regression models were also used to estimate the effects of DM with and without insulin use on the hazard of hip and nonspine fracture, controlling for age and femoral neck BMD T score. We did not take incident DM into account in these analyses because our focus was on prediction in older adults with prevalent DM. To minimize residual confounding, the effects of age and T score were flexibly modeled using restricted cubic splines. We checked for interaction between DM and both covariates, and verified the Cox proportional hazards regression assumption. Ten-year cumulative risks were estimated using the baseline survival function, evaluated at 10 years, raised to the power of the relative hazard for each combination of DM group, age, and T score.

Similarly, we used Cox proportional hazards regression models to estimate the 10-year risk of fracture for women in the SOF with and without DM, stratified by insulin use, in models controlling for the FRAX 10-year fracture risk score rather than T score and age. For men, because none of the MrOS study participants had 10 years of follow-up, we calculated 8-year risks instead, adjusting the FRAX scores under the assumption that the hazard for fracture is approximately constant.

To estimate the reduction in T scores equivalent in terms of added fracture risk to having DM, we equated the log hazards of fracture for 2 same-age participants with and without DM, then solved for the difference in T scores, given by the ratio of the coefficient for DM to the coefficient for T score. Confidence intervals (CIs) for the differences were obtained by the δ method, using the Stata nlcom command.

Power to detect associations between T score or FRAX score and fracture can be evaluated based on the lower limits of the 95% CIs for the hazard ratios (HRs). For interactions between DM and T score, the combined samples provided 80% power in 2-sided tests, with a type 1 error rate of 5% to detect the following ratios of the age-adjusted T score HRs in adults with and without DM (in women, 1.37 for hip and 1.21 for nonspine fracture; and in men, 1.72 for hip and 1.30 for nonspine fracture). For interactions between DM and FRAX score, the minimally detectable ratios of the HRs were slightly higher.

Femoral neck BMD T score was higher in women with type 2 DM than in women without type 2 DM (Table 1). In the SOF and Health ABC studies, 1117 women without DM and 84 women with type 2 DM had at least 1 hip fracture, and 3231 women without DM and 262 women with type 2 DM had at least 1 nonspine fracture during a mean (SD) follow-up of 12.6 (5.3) years (Table 2).

Femoral neck BMD T score was associated with hip and nonspine fracture in women with DM (Table 3). Because fracture risk varies with age, race/ethnicity, and insulin use, we assessed interactions of these variables with T score. T score had a greater effect on the hazard of nonspine fracture with older age (P for interaction = .04). Otherwise, we found no statistically significant interactions (P = .10) between T score and age, race/ethnicity, or insulin use. Femoral neck BMD T score was also associated with fracture risk in women without DM, as observed previously in older women,²⁰ with no evidence of interaction of T score with DM status. The ability of femoral neck BMD T score to predict fracture, based on the C index, was similar in older adults with and without DM (Table 3). For example, in women with DM, the HR for hip fracture for each 1-unit decrease in T score was 1.88 (95% CI, 1.43-2.48). The corresponding C indexes were 0.72 in women with DM and 0.74 in women without DM.

However, for a given T score, women with DM had a higher risk of hip or nonspine fracture than women without DM of similar age. The increased fracture risk at a given T score was found for women with DM using and not using insulin, which is shown in Figure 1 in the hip and nonspine fracture risk vs T score for women aged 75 years, based on our Cox proportional hazards regression models.

Femoral neck BMD T score is widely used in the diagnosis of osteoporosis.²¹ In women with DM, interpretation of a T score has to consider the higher risk of fracture associated with DM (Figure 1). To assist with this interpretation, Table 4 provides the mean difference in T score, comparing women with and without DM with a similar fracture risk. For hip fracture, the mean difference in T scores estimated from the Cox proportional hazards regression model, comparing women with and without DM who have the same fracture risk, was 0.59 (95% CI, 0.31-0.87). Thus, a woman with DM with a T score of −1.9 would have an estimated 10-year hip fracture risk similar to a woman without DM with a T score of −2.5, the threshold generally used for a diagnosis of osteoporosis.

The FRAX score (available for the SOF but not for the Health ABC study) is an estimate of the 10-year risk of fracture that takes into account age, BMD, and additional risk factors for fracture (Table 1), but not DM status. A higher FRAX score was associated with a higher fracture risk during follow-up for women with and without DM (Table 3). However, in the SOF, the FRAX score in women with DM underestimated the observed long-term risk of fracture (Figure 2).

Similar to women, the femoral neck BMD T score was higher in men with DM than in men without DM in the MrOS and Health ABC studies (Table 1). In the MrOS and Health ABC studies, 158 men without DM and 32 men with DM experienced at least 1 hip fracture and 690 men without DM and 133 men with DM had at least 1 nonspine fracture during a mean (SD) follow-up of 7.5 (2.0) years (Table 2).

The femoral neck BMD T score was associated with hip and nonspine fracture in men with DM (Table 3), with no evidence of interaction with age, race/ethnicity, or insulin use. The femoral neck BMD T score was also strongly associated with fracture risk in men without DM, as observed in previous studies,²⁰ with no evidence of interaction by DM status. The C indexes were similar for men with and without DM (Table 3).

Similar to the results for women, for a given femoral neck BMD T score, men with DM generally had a higher risk of fracture than men without DM of similar age. However, men with DM who were not using insulin did not have an increased risk of nonspine fracture at a given BMD T score compared with men without DM, which is shown in Figure 1 in the hip and nonspine fracture risk vs femoral neck BMD T score for men aged 75 years, based on our Cox proportional hazards regression models.

Table 4 provides mean differences in T scores, comparing men with and without DM at a similar fracture risk, to assist with the interpretation of T scores in men with DM. For hip fracture, the mean difference in T scores comparing men with and without DM who have the same fracture risk was 0.38 (95% CI, 0.09-0.66). Thus, a man with DM with a T score of −2.1 would have an estimated 10-year hip fracture risk similar to a man without DM with a T score of −2.5.

A higher FRAX score was associated with fracture risk for men in the MrOS study with and without DM (Table 3). However, as with women, the FRAX score in men with DM underestimated the long-term risk of fracture that was observed in the MrOS study (Figure 2).

Because the SOF relied solely on self-report to identify DM while fasting glucose assays were available in the MrOS and Health ABC studies, we performed a sensitivity analysis defining DM solely based on self-report in all cohorts. Because thiazolidinedione use is associated with increased fracture risk in women and possibly in men,²² we performed a sensitivity analysis excluding participants who reported thiazolidinedione use during the study (63 participants in the SOF, 204 in the MrOS, and in the Health ABC study, 70 women and 79 men). These sensitivity analyses supported our findings in the main analyses.

In the first study to our knowledge to prospectively examine the relationship between BMD and fracture in older adults with type 2 DM, we found that lower femoral neck BMD T score and higher FRAX score are associated with hip and nonspine fracture risk. The ability of femoral neck BMD T score or FRAX score to predict fracture is similar in older adults with and without DM. However, for a given T score and age, those adults with DM had a higher risk of fracture than those without DM, consistent with previous studies.^2-4 Participants with DM also experienced higher fracture rates at a given FRAX score than participants without DM.

Cross-sectional studies have not found an association between BMD and fracture risk in type 2 DM, perhaps due to small sample size. In a cross-sectional study of 150 older women with type 2 DM in England, those with a previous clinical fracture had lower age-matched lumbar spine and total hip BMD (Z scores), but the differences were not statistically significant.⁹ In older women with DM in Japan (n = 150), prevalent vertebral fractures were associated with lower, but not statistically different, lumbar spine BMD Z score.⁸ In contrast, our prospective study found that femoral neck BMD was associated with risk of hip and nonspine fracture in older men and women with type 2 DM.

Our results indicate that femoral neck BMD and the FRAX score are as useful for the assessment of fracture risk in older adults with DM as in those without DM. However, interpretation of T score or FRAX score in an older patient with DM must take into account the higher fracture risk associated with DM. For example, using the mean differences in T scores between older adults with and without DM estimated from these cohorts (Table 4), a T score in a woman with DM is associated with hip fracture risk equivalent to a woman without DM with a T score of approximately 0.5 units lower (mean difference, 0.59; 95% CI, 0.31-0.87).

The FRAX score was designed to provide an estimate of absolute fracture risk in older adults that could be used in combination with country-specific cost-effectiveness data to set intervention thresholds.²³ The FRAX score has been incorporated into US guidelines for prevention and treatment of osteoporosis.²¹ The FRAX algorithm does not currently include type 2 DM as a risk factor for fracture, and our results indicate that use of the FRAX score in patients with DM will likely underestimate risk. Our results were most consistent for women, but also indicate that the FRAX score tends to underestimate risk in men with DM, particularly in those using insulin. To be widely useful, the FRAX score must necessarily be as brief as possible. However, an adjustment of this algorithm for type 2 DM seems justified, given the prevalence of DM among older adults.

The reasons for increased fracture risk at a given BMD in older adults with DM are not clearly understood. Bone strength may be compromised through changes that are not captured with DXA, such as higher levels of advanced glycation end products in bone collagen.²⁴ More frequent falls in older adults with DM could also increase fracture risk for a given BMD.²⁵ We did not find an increased risk of nonspine fracture in men with DM who were not using insulin. Men may be relatively protected from the negative skeletal effects of DM due to less rapid bone loss with aging. Our results suggest that insulin therapy might be a useful marker of increased nonspine fracture risk at a given T score or FRAX score. However, we had limited ability to assess these associations due to the small number of fractures among patients using insulin.

Because of small numbers, we did not assess the ability of T score or FRAX score to predict fracture risk among older adults using a thiazolidinedione. There is growing evidence that thiazolidinedione use increases fracture risk in women, and possibly men.^22,26 Thus, the fracture risks presented herein for a given femoral neck BMD T score (Figure 1) or FRAX score (Figure 2) are likely an underestimate of risk in those adults using a thiazolidinedione.

Our study has several limitations. We did not have glucose measurements in the SOF. Diabetes was ascertained solely by self-report, which may have led to the inclusion of women with DM among women without DM in the SOF. However, this misclassification would tend to underestimate any real differences between study participants with and without DM. It is possible that some participants with DM had type 1 DM as we were not able to distinguish between type 1 and type 2 DM. However, given the age range of these cohorts, it is likely that very few participants had type 1 DM. Additionally, the analyses did not include vertebral fractures and did not adjust for DM duration, and the FRAX scores were not available in the Health ABC study. Strengths of our study include the use of data from 3 large and racially diverse cohorts of older men and women with extended follow-up and adjudication of fracture outcomes.

In conclusion, we found that femoral neck BMD T score and FRAX score are both associated with fracture risk in older adults with type 2 DM and appear to be useful for clinical evaluation of fracture risk, despite the paradox of higher BMD with increased fracture risk in this population. However, a given T score or FRAX score is associated with a higher risk of fracture in older adults with DM compared with those without DM. Refinements are needed in current treatment and diagnostic algorithms for use in older patients with type 2 DM.

Corresponding Author: Ann V. Schwartz, PhD, Department of Epidemiology and Biostatistics, University of California San Francisco, 185 Berry St, Lobby 5, Ste 5700, San Francisco, CA 94107-1762 (aschwartz@psg.ucsf.edu).

Author Contributions: Dr Schwartz had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Schwartz, Black.

Acquisition of data: Ensrud, Cauley, Harris.

Analysis and interpretation of data: Schwartz, Vittinghoff, Bauer, Hillier, Strotmeyer, Donaldson, Cauley, Koster, Womack, Palermo, Black.

Drafting of the manuscript: Schwartz.

Critical revision of the manuscript for important intellectual content: Schwartz, Vittinghoff, Bauer, Hillier, Strotmeyer, Ensrud, Donaldson, Cauley, Harris, Koster, Womack, Palermo, Black.

Statistical analysis: Vittinghoff, Palermo.

Obtained funding: Schwartz, Strotmeyer, Ensrud, Harris.

Administrative, technical, or material support: Cauley, Harris.

Study supervision: Womack.

Funding/Support: This study was funded by an investigator-initiated grant from Amgen (Dr Schwartz). The Study of Osteoporotic Fractures is supported by National Institutes of Health funding. The National Institute on Aging provides support under the following grants: R01 AG005407, R01 AR35582, R01 AR35583, R01 AR35584, R01 AG005394, R01 AG027574, and R01 AG027576. The Osteoporotic Fractures in Men study is supported by National Institutes of Health funding and by grant 1-04-JF-46 from the American Diabetes Association (Dr Strotmeyer). The National Institute of Arthritis and Musculoskeletal and Skin Diseases, the National Institute on Aging, the National Center for Research Resources, and the National Institutes of Health Roadmap for Medical Research provided support under the following grants: U01 AR45580, U01 AR45614, U01 AR45632, U01 AR45647, U01 AR45654, U01 AR45583, U01 AG18197, U01-AG027810, and UL1 RR024140. The Health, Aging, and Body Composition study is supported by contracts N01-AG-6-2101, N01-AG-6-2103, and N01-AG-6-2106 from the National Institute on Aging; grant R01-AG028050 from the National Institute on Aging; grant R01-NR012459 from the National Institute of Nursing Research; and by the Intramural Research Program of the National Institutes of Health, National Institute on Aging.

Role of the Sponsors: The funding organizations were independent of the design and conduct of the study, the collection, management, analysis, and interpretation of the data, or the preparation of the manuscript. Before submission for publication, the manuscript was reviewed by Amgen and was reviewed and approved by the National Institute on Aging.

The SOF Research Group Investigators:San Francisco Coordinating Center (California Pacific Medical Center Research Institute and University of California San Francisco): S.R. Cummings (principal investigator), M.C. Nevitt (co-investigator), D.C. Bauer (co-investigator), D.M. Black (co-investigator), K.L. Stone (co-investigator), W. Browner (co-investigator), R. Benard, T. Blackwell, P.M. Cawthon, L. Concepcion, M. Dockrell, S. Ewing, M. Farrell, C. Fox, R. Fullman, S.L. Harrison, M. Jaime-Chavez, W. Liu, L. Lui, L. Palermo, N. Parimi, M. Rahorst, D. Kriesel, C. Schambach, R. Scott, J. Ziarno; University of Maryland: M.C. Hochberg (principal investigator), R. Nichols (clinic coordinator), S. Link; University of Minnesota: K.E. Ensrud (principal investigator), S. Diem (co-investigator), M. Homan (co-investigator), P. Van Coevering (program coordinator), S. Fillhouer (clinic director), N. Nelson (clinic coordinator), K. Moen (assistant program coordinator), F. Imker-Witte, K. Jacobson, M. Slindee, R. Gran, M. Forseth, R. Andrews, C. Bowie, N. Muehlbauer, S. Luthi, K. Atchison; University of Pittsburgh: J.A. Cauley (principal investigator), L.H. Kuller (co-principal investigator), J.M. Zmuda (co-investigator), L. Harper (project director), L. Buck (clinic coordinator), M. Danielson (project administrator), C. Bashada, D. Cusick, A. Flaugh, M. Gorecki, M. Nasim, C. Newman, N. Watson; The Kaiser Permanente Center for Health Research, Portland, Oregon: T. Hillier (principal investigator), K. Vesco (co-investigator), K. Pedula (co-investigator), J. Van Marter (project director), M. Summer (clinic coordinator), A. MacFarlane, J. Rizzo, K. Snider, J. Wallace.

The MrOS Research Group Investigators:Coordinating Center (California Pacific Medical Center Research Institute and University of California, San Francisco): S.R. Cummings (principal investigator), D.C. Bauer (co-investigator), D.M. Black (co-investigator), P.M. Cawthon (co-investigator), M.C. Nevitt (co-investigator), K.L. Stone (co-investigator), R. Fullman (project director), R. Benard, T. Blackwell, A. Chau, L. Christianson, L. Concepcion, J. Diehl, S. Ewing, M. Farrell, C. Fox, S. Hoffland, J. Ireland, M. Jaime-Chavez, E. Kwan, S.L. Harrison, W. Liu, L.Y. Lui, A. Mills, L. Nusgarten, L. Palermo, N. Parimi, L. Perreault, J. Schneider, R. Scott, D. Tanaka, C. Yeung; Administrative Center (Oregon Health and Sciences University): E. Orwoll (principal investigator), K. Phipps (co-investigator), L. Marshall (co-investigator), J. Babich Blank (project director), L. Lambert, C. Nielson, Y. Wang, C. Petersen, M. Powell; University of Alabama, Birmingham: C.E. Lewis (principal investigator), J. Shikany (co-investigator), P. Johnson (project director), N. Webb, K. Hardy, S. Felder, J. Wilkoff, J. King, T. Johnsey, M. Young, C. Atkins, C. Collier, J. Smith, C. Sassaman; University of Minnesota: K. Ensrud (principal investigator), H. Fink (co-investigator), N. Nelson (clinic coordinator), P. Van Coevering (program director), S. Fillhouer (project director), R. Andrews, C. Bowie, M. Forseth, R. Gran, F. Imker-Witte, S. Luthi, K. Moen, N. Muehlbauer, M. Paudel, M. Slindee, S. Ziesche; Stanford University: M. Stefanick (principal investigator), A. Hoffman (co-investigator), K. Kent, N. Ellsworth, A. Krauss, R. Gupta, S. Hartley, M. Bowers; University of Pittsburgh: J. Cauley (principal investigator), J. Zmuda (co-investigator), M. Danielson (study administrator), L. Harper (project director), L. Buck (clinic coordinator), M. Nasim, D. Cusick, M. Gorecki, N. Watson, C. Bashada, C. Newman; University of California, San Diego: E. Barrett-Connor (principal investigator), T. Dam (co-investigator), M.L. Carrion-Petersen (project director), P. Miller, N. Kamantigue, K. Marksbury, M. Stephens, Z. Torres.

The Health ABC Research Group Investigators:University of Pittsburgh: A.B. Newman (principal investigator), P. Kost (study coordinator); University of Tennessee, Memphis: S. Satterfield (principal investigator), F.A. Tylavsky (principal investigator), S.B. Kritchevsky (former Memphis principal investigator, now at Wake Forest School of Medicine), S. Thomas (study coordinator); University of California, San Francisco (Health ABC Coodinating Center): S.R. Cummings (principal investigator), M.C. Nevitt (principal investigator), S.M. Rubin (project director); National Institute on Aging (Project Office): T.B. Harris (National Institute on Aging project officer), M.E. Garcia.

Additional Contributions: Helaine Resnick, PhD, Institute for the Future of Aging Services of the American Association of Homes and Services for the Aging, provided helpful comments on a draft of the manuscript. Dr Resnick did not receive any compensation for her contribution.