Diabetes Is Associated Independently of Body Composition With BMD and Bone Volume in Older White and Black Men and Women: The Health, Aging, and Body Composition Study

Authors

  • Elsa S Strotmeyer,

    1. Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, Pennsylvania, USA
    Search for more papers by this author
  • Jane A Cauley,

    1. Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, Pennsylvania, USA
    Search for more papers by this author
    • Dr Cauley has received research support from Eli Lilly and Company, Merck & Co., Inc., Novartis, and Pfizer Inc. In addition, she has received honorarium and served on the speaker's bureau from Eli Lilly and Company and Merck & Co., Inc. Dr Bauer received funding from Merck & Co., Inc., Procter & Gamble, and SKD. All other authors have no conflict of interest

  • Ann V Schwartz,

    1. Department of Epidemiology and Biostatistics, The University of California, San Francisco, California, USA
    Search for more papers by this author
  • Michael C Nevitt,

    1. Department of Epidemiology and Biostatistics, The University of California, San Francisco, California, USA
    Search for more papers by this author
  • Helaine E Resnick,

    1. Department of Epidemiology, MedStar Research Institute, Hyattsville, Maryland, USA
    Search for more papers by this author
  • Joseph M Zmuda,

    1. Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, Pennsylvania, USA
    Search for more papers by this author
  • Douglas C Bauer,

    1. Department of Medicine, The University of California, San Francisco, California, USA
    Search for more papers by this author
  • Frances A Tylavsky,

    1. Department of Preventive Medicine, University of Tennessee Health Science Center, Memphis, Tennessee, USA
    Search for more papers by this author
  • Nathalie de Rekeneire,

    1. Laboratory of Epidemiology, Demography and Biometry, National Institute on Aging, Bethesda, Maryland, USA
    Search for more papers by this author
  • Tamara B Harris,

    1. Laboratory of Epidemiology, Demography and Biometry, National Institute on Aging, Bethesda, Maryland, USA
    Search for more papers by this author
  • Anne B Newman

    Corresponding author
    1. Division of Geriatric Medicine and Department of Epidemiology, University of Pittsburgh School of Medicine, Graduate School of Public Health, Pittsburgh, Pennsylvania, USA
    • Address reprint requests to: Anne B Newman, MD, MPH, Health Aging Research Program, University of Pittsburgh, 130 North Bellefield Avenue, 5th Floor, Pittsburgh, PA 15213, USA
    Search for more papers by this author

Abstract

The association between type 2 diabetes, BMD, and bone volume was examined to determine the effect of lean and fat mass and fasting insulin in the Health, Aging, and Body Composition Study, which included white and black well-functioning men and women 70-79 years of age (N = 2979). Diabetes predicted higher hip, whole body, and volumetric spine BMD, and lower spine bone volume, independent of body composition and fasting insulin.

Introduction: The purpose of this study was to determine if the association between type 2 diabetes and higher BMD observed in older white women is seen in elderly white men and blacks and to evaluate if higher BMD in diabetic individuals is accounted for by lean mass, fat mass, or fasting insulin differences.

Materials and Methods: In the Health, Aging, and Body Composition Study, which included white and black well-functioning men and women 70-79 years of age (N = 2979), 19% of participants had diabetes at baseline. Of those with diabetes, 57% were men, and 62% were black. Multivariate linear regression models examined independent effects of diabetes, lean mass, fat mass, visceral fat, and fasting insulin on BMD and bone volume while adjusting for relevant covariates.

Results and Conclusions: Fasting insulin, visceral fat, and volumetric spine BMD, assessed by CT, and lean mass, fat mass, and total hip and whole body BMD, assessed by DXA, were higher (p ≤ 0.05 for all) for those with diabetes. Hip BMD was higher in white men (0.99 ± 0.14 versus 0.93 ± 0.14 g/cm2, p < 0.001), black men (1.06 ± 0.17 versus 1.00 ± 0.15 g/cm2, p < 0.001), white women (0.83 ± 0.13 versus 0.76 ± 0.13 g/cm2, p < 0.001), and black women (0.90 ± 0.15 versus 0.85 ± 0.15 g/cm2, p < 0.001) with diabetes compared with those without diabetes, although the relationship was attenuated by body composition. In multiple regression models, diabetes was an independent predictor of higher hip, whole body, and volumetric spine BMD in all participants (p ≤ 0.001), but lower spine volume (p = 0.01) and higher hip BMD for each race-gender group (p ≤ 0.01). Type 2 diabetes was associated with a 4-5% higher total hip BMD in all race-gender groups of elderly adults, independent of body composition and fasting insulin levels.

INTRODUCTION

OLDER WHITE WOMEN with type 2 diabetes have higher areal BMD than women without diabetes, even after adjusting for body size,(1–5) but not much evidence exists for older white men and black adults or for volumetric BMD. Some large studies have found higher BMD among white men with type 2 diabetes,(2,4) but others reported nonsignificant differences.(1,6) Although type 2 diabetes is more prevalent among older black adults in the United States,(7) whether the effect of diabetes on BMD is similar to older white adults is unknown. The Rotterdam Study found higher BMD among white men and women with type 2 diabetes as well as a decreased risk of nonvertebral fractures in diabetic women.(2) However, recent prospective studies have shown that, despite higher BMD, women with type 2 diabetes may have an increased risk of fractures, which may reflect their higher risk of falls and diabetes complications.(5,8,9)

Higher BMD in those with type 2 diabetes may be due in part to greater obesity associated with the disease.(10,11) Specific components of body weight, that is, lean or fat mass, have not been considered previously and may account for BMD differences among those with and without diabetes. Other possible mechanisms for the relationship between type 2 diabetes and BMD include greater androgenicity,(1) hyperinsulinemia,(10,12–15) poor glycemic control,(14,16,17) or an effect on parathyroid function and bone metabolism.(16,18) The effect of diabetes on bone volume is also unknown.

The objectives of this study were to determine (1) whether type 2 diabetes was associated with a higher BMD or bone volume in white and black men and women, 2) whether body composition, as assessed by DXA and CT, accounted for these differences, or (3) whether fasting insulin levels accounted for higher BMD among those with diabetes.

MATERIALS AND METHODS

Subjects

Participants were 3075 well-functioning older white and black adults (48.4% male; 41.6% black), 70-79 years of age, participating in the Health, Aging, and Body Composition (Health ABC) Study, which is a prospective cohort investigating changes in body composition as a common pathway by which multiple diseases contribute to disability. Recruitment was from a random sample of white Medicare beneficiaries and all age-eligible black community residents in Pittsburgh, PA and Memphis, TN. Eligible participants reported no difficulty walking one-quarter mile (400 m), climbing 10 steps, or performing activities of daily living; they were free of life-threatening cancers, with no active treatment within the past 3 years; and they planned to remain within the study area for at least 3 years. Participants provided informed consent, approved by the institutional review boards at the University of Pittsburgh and the University of Tennessee, Memphis. Data were from the baseline examination during 1997-1998. For the current analyses, we excluded participants with missing diabetes status (n = 22), diabetes onset in childhood (≤20 years old; n = 5), and oral steroid use (n = 69). Complete data were available for 2979 participants (914 white men, 542 black men, 816 white women, and 707 black women), with the exception of volumetric BMD measured in Pittsburgh for only 1446 participants (441 white men, 267 black men, 367 white women, and 371 black women).

Diabetes mellitus

Diabetes was defined as self-report of diabetes previously diagnosed by a physician, use of hypoglycemic medications, or fasting glucose ≥126 mg/dl (≥7.0 mM) in accordance with the American Diabetes Association criteria.(19) The baseline clinic visit measures included fasting glucose, 75-g oral glucose tolerance test (OGTT; only for those not using diabetes medications), fasting insulin (only for those not using exogenous insulin), and glycosylated hemoglobin (HbA1c) as previously reported.(20) Pre-existing diabetes, diabetes duration, and use hypoglycemic medications were determined by an interviewer-administered questionnaire. Of those with diabetes by ADA criteria at baseline (N = 566), 27% of white men, 21% of black men, 17% of white women, and 14% of black women did not report a prior diagnosis. Including additional individuals with ≥200 mg/dl 2-h plasma glucose from the 75-g OGTT (N = 136) did not change results in final multivariate models.

BMD and body composition

Height was measured using a stadiometer, and weight was measured with a calibrated balance beam scale. Body mass index (BMI) was calculated as weight divided by square height (kg/m2). Total hip, femoral neck, and whole body areal BMD (g/cm2) was assessed at both field centers by DXA (Hologic 4500A, software version 9.03; Hologic, Bedford, MA, USA). Total bone mineral-free lean mass (LM) and total fat mass (FM) were derived from the whole body scan. DXA quality assurance measurements were performed at both study sites to ensure scanner reliability, and identical patient scan protocols were used. CT in Pittsburgh (9800 Advantage; General Electric, Milwaukee, WI, USA) and Memphis (Somatron Plus; Siemens, Iselin, NJ, USA, or PQ2000S; Picker, Cleveland, OH, USA) of the abdomen was used to measure abdominal visceral fat area (cm2) at the L4-L5 disk space as previously described.(21) CT of the spine L3 region was conducted, only in Pittsburgh participants, to obtain volumetric trabecular BMD (mg/ml) and volume of the integral region (ml). CT data were analyzed with a standardized protocol at the University of Colorado Health Sciences Center for soft tissue and at the University of California, San Francisco, for volumetric BMD.(22,23)

Other covariates

Additional covariates were included in models as possible confounders of the association of diabetes and BMD. Smoking history, alcohol consumption history, previous diagnosis of osteoporosis, weekly physical activity from walking and exercise, and weight changes in the prior 12 months were determined by an interviewer-administered questionnaire. Medications used in the prior week were brought to the clinic. Thiazide diuretic, statin, estrogen, osteoporosis medications (bisphosphonates, calcitonin, raloxifene, fluoride), and calcium and vitamin D supplement use was coded using the Iowa Drug Information System (IDIS) ingredient codes.(24) The Health ABC performance battery was a supplemented version of the lower-extremity performance test used in the Established Populations for the Epidemiologic Studies of the Elderly (EPESE; chair stands, standing balance, 6-m walk for gait speed),(25) with increased test duration, a single foot stand, and a narrow walk test of balance as previously described (score range 0-12).(26)

Statistical analyses

Data were from the baseline study visit. Differences in prevalence and univariate associations between diabetic and nondiabetic participants were tested separately by race and gender using Pearson χ2 methods. Fishers exact methods were used when the expected value of any cell was <5. For continuous variables, nonparametric one-way Mann-Whitney tests were performed for comparisons between diabetic and nondiabetic participants. Among diabetic participants, partial correlations coefficients were computed for BMD and diabetes duration as well as BMD and HbA1C levels, adjusting for age, race, gender, and study site. Nonparametric Kruskal Wallis tests of means were performed to test trends in BMD and bone volume means with diabetes duration, classified as newly diagnosed at baseline (N = 122) or previously diagnosed <5 years (N = 126), 5-10 years (N = 70), >10 to ≤20 years (N = 111), and >20 years (N = 100). Total hip, whole body, and volumetric spine BMD and bone volume means for diabetic and nondiabetic participants were calculated with analysis of covariance (ANCOVA), adjusted first for age, race, and gender, and then additionally for LM, FM, and visceral fat. Stepwise multiple linear regression models were done to determine the independent effects of diabetes, LM, FM, visceral fat, and fasting insulin on BMD (total hip, femoral neck, whole body, or spine BMD) and also spine volume, while adjusting for age, race, gender, site (Memphis or Pittsburgh), smoking status, drinking status, Health ABC performance battery categorical score (0-12), gain or loss of ≥5 lbs in the past 12 months, previously diagnosed osteoporosis and/or osteoporosis medication use, calcium and vitamin D supplement use, thiazide diuretic use, and statin use. (The Health ABC performance battery was chosen for multivariate analyses instead of physical activity because it had a stronger association with BMD. The addition of physical activity to regression models did not change the results.) Regression models were also performed separately by race and gender and adjusted for oral estrogen use in women. Multicollinearity for variables was assessed using the variance inflation factor (VIF), the inverse of the proportion of variance not accounted for by other independent variables; no VIF was >10, and the mean VIF for each regression model was ≤2.(27) Percentage change in BMD or bone volume because of diabetes in linear regression models was calculated using the following formula: [(unstandardized β for diabetes)(unit change in diabetes)/BMD or bone volume mean for entire sample] × 100. 95% CIs for percentage change in BMD or bone volume caused by diabetes were calculated using the following formula: {[(unstandardized β for diabetes)(unit change in diabetes) ± (SE of β for diabetes)(1.96)]/BMD or bone volume mean for entire sample} × 100. Data were analyzed using the SPSS (SPSS, Chicago, IL, USA) statistical software package.

RESULTS

Overall, 19.0% (N = 566) of 2979 participants had diabetes, similar to the prevalence among black and white men and women in NHANES III.(7) Black men had the highest prevalence of diabetes (27.9%), followed by black women (24.5%), white men (18.8%), and white women (8.6%). Participants with diabetes were more likely to be men (57% versus 47%; p < 0.001) and less likely to be white (43% versus 62%; p < 0.001) than those without diabetes. White men, black men, white women, and black women with diabetes had significantly higher total hip and whole body areal BMD than the nondiabetic counterparts of the same race and gender. QCT spine volumetric BMD was higher in white men and black women with diabetes compared those without diabetes, but this relationship was not statistically significant for black men or white women (Table 1). No unadjusted differences in spine bone volume existed between diabetic and nondiabetic participants for any race or gender. Participants with diabetes had higher fasting insulin than those without diabetes, but this difference was not statistically significant for blacks.

Table Table 1.. Descriptive Characteristics by Diabetes Status for Gender- and Race-Specific Groups
original image

Diabetic participants had significantly higher weight, BMI, LM, FM, and visceral fat than the participants without diabetes in each of the race and gender groups. Adjusting for LM, FM, and abdominal visceral fat area reduced differences in BMD at the total hip, whole body, and spine between participants with and without diabetes (Table 2), although diabetic participants still had higher BMD. Adjusting for these body composition variables also qualitatively changed the direction of the association between spine bone volume and diabetes, with diabetic participants having significantly higher bone volume before adjustments and significantly lower bone volume after adjustments. Adjustment for body composition lowered BMD by 3% for whole body, 5% for the total hip, and 5% for the volumetric spine, lowered spine bone volume by 5% for diabetic participants, and raised BMD and bone volume by 1% at all sites for nondiabetic participants.

Table Table 2.. Adjusted BMD and Volume Means ± SE for Diabetic and Nondiabetic Participants*
original image

Traditional risk factors associated with BMD are also shown in Table 1 by diabetes status for each race and gender group. Diabetic participants from each race and gender group had lower Health ABC performance battery scores and less physical activity from walking or exercise than nondiabetic participants. Differences in self-reported fractures were significant only for white women, with diabetic white women reporting fewer fractures than nondiabetic white women after 45 years of age. No differences between diabetic and nondiabetic participants were found for current smoking, osteoporosis medication use, or oral estrogen use.

Multivariate linear regression analyses for all participants

Diabetes was a significant independent correlate of higher BMD at the hip, whole body, and spine and of lower spine volume (Table 3). Type 2 diabetes was associated with 4.7% (95% CI: 3.4-6.1%) higher total hip BMD, 1.9% (95% CI: 1.0-2.8%) higher whole body BMD, 8.2% (95% CI: 3.8-12.7%) higher spine volumetric BMD, and 2.4% lower spine volume (95% CI: −0.47 to −4.3%). This relationship between diabetes and BMD persisted with adjustment for abdominal visceral fat area, total FM, and total LM. The relationship between diabetes and lower bone volume became significant only after adjustment for these body composition variables. Male gender and black race were significantly associated with higher BMD in all models. Black race was significantly associated with lower spine bone volume. In all models, total LM was independently associated with higher BMD, but the relationship of fat to BMD was not uniform across BMD sites. Total LM was positively associated with bone volume, but FM and abdominal visceral fat were negatively associated with bone volume.

Table Table 3.. Final Multiple Linear Regression Models for Total Hip (N = 2883), Whole Body (N = 2765) and Spine BMD (N = 1425), and Spine Volume (N = 1345) Including All Participants*
original image

When diabetic participants using insulin as a medication were removed from multivariate analyses for BMD of the total hip and whole body (N = 122) and for spine bone volume, final models were unchanged, and fasting insulin level was not related to BMD (data not shown). However, when the individuals using insulin were removed from the model for spine volumetric BMD (N = 50), higher fasting insulin levels were associated with higher volumetric BMD (β = 0.056; p = 0.04), in addition to the variables shown in Table 3. Removal of participants using insulin did not change associations between diabetes and BMD or spine bone volume in any final multivariate models. Separate multivariate analyses including only nondiabetic participants showed no relationship of fasting insulin levels and BMD at any site or spine bone volume.

Multivariate linear regression analyses of total hip BMD for race- and gender-specific groups

Race- and gender-specific regression analyses were performed for total hip BMD (Table 4). Diabetes remained a significant independent factor in each race-gender model. Total hip BMD was 3.8% (95% CI: 1.5-6.1%) higher in white men with diabetes, 4.2% higher (95% CI: 1.5-6.9%) in black men with diabetes, 4.7% higher (95% CI: 1.1-8.3%) in white women with diabetes, and 5.1% (95% CI: 2.6-7.6%) higher in black women with diabetes. Body composition measures of total FM and LM were also significant independent predictors of hip BMD. Abdominal visceral fat and fasting insulin measures were not related to total hip BMD in the multivariate race-gender models. Additionally, multivariate analyses including only nondiabetic participants showed no relationship of fasting insulin levels and total hip BMD for any race-gender group. For race- and gender-specific regression analyses with covariates included, diabetes was not a significant predictor of whole body or volumetric spine BMD in white men, black men, or white women. In black women, diabetes was independently associated with 1.9% (95% CI: 0.04-3.8%) higher whole body BMD (standardized β = 0.132; p = 0.01) and 10.8% (95% CI: 2.6-19.0%) higher volumetric spine BMD (standardized β = 0.077; p = 0.03).

Table Table 4.. Final Multiple Linear Regression Models for Total Hip BMD (g/cm2) in White Men (N = 894), Black Men (N = 521), White Women (N = 799), and Black Women (N = 681)*
original image

Participants with diabetes

Overall, 46% of participants with diabetes used oral hypoglycemic agents, and 22% used insulin. Blacks used insulin more frequently than whites (27% versus 15%; p < 0.001), but there were no race or gender differences for use of oral hypoglycemic medications. Blacks had slightly longer diabetes duration than whites (10.9 ± 10.9 versus 9.6 ± 11.0 years; p = 0.07), but women and men had a similar duration (10.8 ± 10.8 versus 10.0 ± 11.0 years; p = 0.12).

No differences in BMD were seen for diabetic participants with HbA1C ≥8% compared with <8% (mean, 7.9 ± 1.6%), although diabetic participants with HbA1C ≥8% had lower bone volume (19.6 versus 20.7 ml; p = 0.007). HbA1C levels in diabetic participants were not correlated with BMD or bone volume after controlling for age, race, gender, and study site. However, longer diabetes duration was negatively correlated with total hip BMD (r = −0.14; p = 0.002) but not with whole body BMD or spine volumetric BMD or bone volume, after controlling for age, race, gender, and study site. A significant trend of decreasing total hip BMD but not whole body BMD or spine volumetric BMD or bone volume was present with increasing diabetes duration (0.99 ± 0.16 g/cm2 for newly diagnosed diabetes, 0.98 ± 0.16 g/cm2 for <5 years, 0.96 ± 0.17 g/cm2 for 5-10 years, 0.97 ± 0.17 g/cm2 for >10-20 years, and 0.91 ± 0.18 g/cm2 for >20 years; p = 0.003). This trend remained significant after insulin users were excluded. Diabetic participants using insulin, thiazide diuretics, or statins had similar BMD at each site and spine bone volume compared with nonusers.

DISCUSSION

The association of type 2 diabetes and BMD has not been rigorously examined in the multiracial elderly populations to determine if other confounding factors, such as insulin levels, combined with accurate information on quantity of FM, LM, and visceral fat mass possibly explain this association. In this cohort, type 2 diabetes was associated cross-sectionally with 2-8% higher regional and whole body BMD, both areal and volumetric measures, even with adjustment for body composition variables of LM, FM, and abdominal visceral fat and other confounding factors. This relationship was consistent across race and gender groups for total hip BMD, which is the BMD site often used in fracture prediction. Our results indicate that, before adjustment for body composition measures, these factors were contributing to an additional 3-5% higher unadjusted BMD for diabetic participants.

The association of diabetes with lower spine bone volume is novel and intriguing. Our results suggest that higher BMD in diabetic older adults could be related to smaller bone volume, with more mineral in a smaller bone space. We are not aware of previous reports comparing bone volume in those with and without diabetes. Our finding of lower bone volume in diabetic participants after adjustment for body composition and size may have identified a factor contributing to bone strength in type 2 diabetes. The lower bone volume might account for the apparent paradox of increased fracture risk, reported in other studies,(5,8,9) in type 2 diabetes even with higher BMD.

Previous studies have not documented the association of diabetes and BMD in black populations, despite the higher prevalence of diabetes in this group.(7) Other large epidemiology studies, such as the Rancho Bernardo Study, the Rotterdam Study, the European Vertebral Osteoporosis Study (EVOS), and the Study of Osteoporotic Fractures (SOF), had similar findings of 3-10% higher BMD for white women(1,2,4,5) and white men(2) with type 2 diabetes compared with nondiabetic individuals. Our results differ from those of the Rancho Bernardo Study, which found no difference in BMD at the femoral neck and lumbar spine between 39 diabetic and 138 nondiabetic white men,(1) and the EVOS Study,(4) which found higher lumbar spine, but not femoral neck or trochanter, BMD among diabetic white men. These large cohort studies of diabetes and bone measured areal but not volumetric BMD.

We found the relationship of diabetes and spine volumetric BMD, which measures trabecular BMD, to be similar to the relationship between diabetes and areal BMD at largely cortical bone sites. This was in contrast to Sosa et al.,(28) who measured QCT at the spine, comparing a small number of peri- and postmenopausal women with non-insulin-dependent diabetes to controls, and found no differences in BMD. Volumetric BMD measures, which remove differences in bone size, may be important in our population because race and gender may influence bone size, and areal BMD measures are not independent of bone size differences between groups. Our results indicate that controlling for bone size does not eliminate the association between diabetes and higher BMD. Overall, our results support an association of diabetes with higher areal and volumetric BMD at cortical and trabecular bone sites in men and women, regardless of white or black race.

Previous studies reporting higher BMD in older white women with diabetes have adjusted for body size using BMI or waist circumference(1,2,5) rather than DXA and CT measures, which consider the quantity of LM, FM, and abdominal visceral fat. Each of these components of body composition is affected differently by diabetes and is associated separately with BMD. Additionally, surrogate measures of body composition may be problematic for comparisons across races because differences in abdominal visceral fat, FM, and LM may exist that are not adequately reflected in weight or BMI measures.(29,30) Thus, we hypothesized that adjustment for specific components of body composition might account for the higher BMD found in type 2 diabetes in this multiracial population. We have shown that these confounding factors, while contributing to higher BMD in diabetic persons, did not completely account for the association of diabetes and high BMD.

Insulin and insulin-like growth factors have been shown to have an anabolic effect on bone.(31–33) However, fasting insulin levels were not independently associated with BMD and did not explain the relationship between diabetes and BMD, after adjustment for LM, FM, and abdominal visceral fat, in the entire group of participants or in the race-gender-specific groups. Furthermore, when only nondiabetic participants were included, fasting insulin levels were still not related to BMD at any site or in race-gender-specific analyses for the total hip, after controlling for LM, FM, and abdominal visceral fat. The San Antonio Heart Study also found that serum insulin levels in women without diabetes do not explain increased BMD once BMI was taken into account.(10) A significant positive association between fasting insulin and BMD, which was not explained by BMI and waist-to-hip ratio, was present among nondiabetic women, but not men, in the Rancho Bernardo Study.(15) Similar analyses from the Rotterdam Study indicate that, after adjustment for BMI, the association of higher postload insulin levels and higher BMD was not statistically significant in men and women, even when those with diabetes were excluded.(14) The relationship of fasting insulin and BMD may differ depending on areal or volumetric measure of BMD, because Reid et al.(12) have shown a positive relationship between fasting insulin and BMD/height and volumetric BMD of the spine controlling for FM and LM in normal postmenopausal women. Our results suggest an independent relationship of fasting insulin levels with higher spine volumetric BMD in Pittsburgh participants, but this relationship did not completely account for the positive association between non-insulin-dependent diabetes and volumetric spine BMD.

Diabetes control as measured by HbA1C and medication use among diabetic participants was not related to BMD or bone volume. However, longer diabetes duration was correlated with lower total hip BMD, suggesting that BMD differences exist within diabetic participants. Other macrovascular or microvascular complications of diabetes, often associated with a longer duration of disease, may also be related to lower BMD or fracture risk, and longer duration of disease has been associated with higher fracture risk.(5,8,9,17,34) We observed no change in multivariate results when diabetes was additionally detected with the OGTT rather than using the fasting glucose test and previous diagnosis and/or treatment to define diabetes. The Rancho-Bernardo and Rotterdam Studies(1,2) used both the OGTT and fasting glucose test to classify diabetes and compared BMD between diabetic and nondiabetic participants. Both of these studies excluded diabetic participants taking insulin, although we observed no difference in the association of diabetes and BMD. Excluding those using exogenous insulin may be a way to ensure that type 2, rather than type 1, diabetes is considered, although we excluded participants with diabetes onset in childhood, who would most likely have type 1 diabetes. Among type 2 diabetic individuals, the effect of exogenous insulin on the relationship between diabetes and BMD use may be important in understanding underlying etiology.

Baseline data from the Health ABC Study allow only suggestions of causal relationships. Longitudinal data on diabetes and changes in BMD will be available as follow-up of this cohort continues. Another limitation is that some type 2 diabetic individuals may not have participated in the Health ABC Study because of greater physical disability. We did adjust for performance battery scores in an attempt to address this potential bias. However, our association of diabetes with higher BMD may only be useful when considering individuals with type 2 diabetes who maintain a certain level of physical functioning, because it is possible that the severity or length of disease may affect BMD in different ways. For example, although type 2 diabetes was associated with higher BMD overall, longer diabetes duration was associated with a significant trend to lower hip BMD.

In summary, this unique multiethnic study of men and women showed a consistent association of type 2 diabetes and higher BMD, not completely accounted for by increased obesity, central adiposity, or fasting insulin levels among diabetic participants. Further research is needed to evaluate bone turnover markers, as well as sex steroids, insulin-like growth factor 1, and parathyroid hormone,(35) to understand the mechanisms underlying the relationship between type 2 diabetes and higher BMD. Because diabetes in this population was associated with a higher total hip BMD of ∼0.5 SD, a protective effect on fractures may be small.(36) Lower bone volume may also play a role in fractures, because bone size contributes to strength. An examination of whether higher BMD and lower bone volume among those with type 2 diabetes will translate into reduced fracture risk is needed.

Acknowledgements

This study was funded by National Institutes on Aging Grants N01-AG-6-2102, N01-AG-6-2103, N01-AG-6-2106, and 5-T32-AG00181.

Ancillary