Adipokines and the risk of fracture in older adults

Authors


Abstract

Adiponectin and leptin are adipokines that influence bone metabolism in vitro and in animal models. However, less is known about the longitudinal association of leptin and adiponectin with fracture. We tested the hypothesis that low leptin and high adiponectin levels are each individually associated with fracture risk in a prospective cohort study in Memphis and Pittsburgh among 3075 women and men aged 70 to 79 years from the Health Aging and Body Composition (Health ABC) study. There were 406 incident fractures (334 nonvertebral and 72 vertebral) over a mean of 6.5 ± 1.9 years. Cox regression was used to estimate the hazard ratios for fracture. Sex modified the association between adiponectin and fracture (p = .025 for interaction). Men with the highest adiponectin level (tertile 3) had a 94% higher risk of fracture [hazard ratio (HR) = 1.94; 95% confidence interval (CI) 1.20–3.16] compared with the lowest tertile (tertile 1; p = .007 for trend) after adjusting age, race, body mass index (BMI), education, diabetes, weight change, and hip bone mineral density (BMD). Among women, after adjusting for age and race, this association was no longer significant (p = .369 for trend). Leptin did not predict fracture risk in women (p = .544 for trend) or men (p = .118 for trend) in the multivariate models. Our results suggest that adiponectin, but not leptin, may be a novel risk factor for increased fracture risk independent of body composition and BMD and that these relationships may be influenced by sex. More research is needed to understand the physiologic basis underlying these sex differences. © 2011 American Society for Bone and Mineral Research.

Introduction

Leptin is secreted by adipose cells and has been shown to be highly correlated with body fat mass1 and a regulator of fat metabolism and appetite.2 Leptin also may modulate bone metabolism by enhancing differentiation of bone marrow stroma cells into mature osteoblasts and by inhibiting the differentiation of osteoclasts.3, 4 Studies in mice have shown that leptin administration increases bone mineral density (BMD).5–7 Some population-based cross-sectional studies report that leptin is positively8–14 associated with BMD, whereas others have not confirmed this association15–17 or find a negative relationship.18–23

Adiponectin, a hormone secreted exclusively by adipocytes, influences insulin sensitivity and has anti-inflammatory properties.24, 25 Adiponectin and its receptors are also expressed in human osteoblasts, suggesting that adiponectin may be a hormone linking bone and fat metabolism.26 However, adiponectin may have negative effects on bone metabolism by stimulating the receptor activator of nuclear factor-κB ligand (RANKL) pathway and inhibiting production of the decoy receptor for RANKL, osteoprotegerin.27 Several epidemiologic studies have found that lower levels of adiponectin are associated with higher BMD,9, 12, 28–30 whereas others have failed to find an association between adiponectin and BMD.17, 19, 21, 24, 31–33

There are limited data available on the relationship of leptin14, 34 and adiponectin30, 35, 36 to fracture risk, and the results are conflicting. However, several of these studies were cross-sectional, had short follow-up times and a low prevalence or incidence of fracture, and failed to adjust for changes in weight. The aim of this analysis was to test the hypothesis that lower serum leptin levels and higher adiponectin levels are associated with a higher risk of a fracture in older adults.

Methods

Study population

The Health Aging and Body Composition (Health ABC) study consists of 3075 women and men aged 70 to 79 years from two field centers, Pittsburgh, PA, and Memphis, TN. Among the women and men enrolled, 46% and 37% were blacks, respectively. To be eligible to participate in Health ABC, subjects had to report no difficulty walking at least 1/4 mile and/or climbing a flight of stairs. Participants were identified from a random sample of white Medicare beneficiaries and all age-eligible black community residents in designated ZIP code areas surrounding Pittsburgh and Memphis. Exclusion criteria included reported difficulty performing basic activities of daily living, obvious cognitive impairment, inability to communicate with the interviewer, intention of moving within 3 years, or participation in a trial involving a lifestyle intervention. The institutional review board (IRB) at each center approved the study protocol, and written informed consent was obtained from all the participants. Baseline data were collected from 1997 to 1998.

Leptin and adiponectin

Specimens were obtained by venipuncture in the morning after an overnight fast, processed, aliquoted into cryovials, frozen at –70°C, and subsequently shipped to the Health ABC core laboratory at the University of Vermont. Baseline leptin (N = 3020) concentrations were measured in duplicate using the Sensitive Human Leptin Radioimmunoassy (RIA) Kit (Product Number SHL-81K) from Linco Research, Inc. (St Charles, MO, USA).37 The assay is a competitive RIA in which the concentration of leptin is determined by competition with [125I]human leptin, with a maximum detectable leptin level of 50 ng/mL. The intraassay coefficient of variation (CV) is 3.7% to 7.5%, and the interassay CV is 3.2% to 8.9%. Adiponectin (N = 3044) was assayed using baseline serum specimens that were frozen approximately 6 years earlier.38 Total circulating levels of adiponectin were measured in duplicate by RIA (Linco Research) with an intraassay CV of 1.8% to 3.6%.

Fractures

Incident fractures (nonvertebral and vertebral) were assessed every 6 months by self-report and validated by radiology reports. Adjudication of fractures was complete through 2007 for the Pittsburgh clinic and through 2006 for the Memphis clinic. The mean follow-up time was 6.5 ± 1.9 years. Nontraumatic and traumatic fractures were included because both have been linked to low BMD.39 Pathologic fractures and fractures of unknown etiology were excluded.

Potential confounders

Demographic variables included self-report of age, race (black or white), sex, site, and education [less than high school (HS), HS graduate, or postsecondary]. Weight was measured on a standard balance-beam scale to the nearest 0.1 kg, and height was measured by a stadiometer to the nearest 0.1 cm. BMI (kg/m2) was calculated by using the formula weight (kg)/height2 (m2). Whole-body dual-energy X-ray absorptiometry (DXA; QDR 4500A, Software Version 9.03; Hologic, Bedford, MA, USA) also was used to measure total lean body mass (kg) and body fat (kg). A separate scan of the proximal femur was done to measure hip BMD. DXA quality-assurance procedures were conducted at both study sites and monitored by the study coordinating center, ensuring scanner reliability and identical scan protocols. An anthropometric spine phantom was scanned daily, and a hip phantom was scanned once per week to assess longitudinal performance of the scanners. Annual weight change (through visit 10) was estimated as the weight difference from baseline to the most recent reassessment divided by the respective time. Lifestyle factors included self-report of smoking (never, current, or former) and alcohol consumption (no consumption in last year, less than 1 drink per week, 1 to 7 drinks per week, or more than 1 drink per day). To assess supplementary intake for vitamin D and calcium and nonsteroidal anti-inflammatory drug (NSAID) use, participants were asked to bring all prescription and over-the-counter medications, which were coded based on the Iowa Drug Information System.40 To estimate dietary intake of calcium and vitamin D, participants completed a 108-item interviewer-administered food frequency questionnaire (FFQ; Block Dietary Data Systems, Berkeley, CA, USA). Physical activity (kcal/week) was determined using the caloric expenditure in the past week for self-reported walking, climbing stairs, and exercise.41 Diabetes was defined using fasting glucose (≥126 mg/dL), self-report, or hypoglycemic medication use. Similarly, subjects were classified as having hypertension through measurement of blood pressure (systolic ≥ 140 mmHg or diastolic ≥ 90 mmHg), self-report, or antihypertensive medication use. Other medical conditions were determined by asking respondents if they have ever been told by a doctor that they had a specific diagnosis of myocardial infarction (MI) and history of fracture after age 45. Handgrip strength (kg) was measured using a handheld dynamometer (Jamar; TEC, Clifton, NJ, USA). The dynamometer was adjusted for hand size for each participant, and two trials were performed on each hand with the maximum value recorded as the observation.

Statistical analysis

Two-sample t tests (Wilcoxon rank-sum test for nonparametric measures) and chi-square tests were used to evaluate mean and proportion differences by sex, respectively. Leptin and adiponectin means differed in women and men, and this difference could not be explained by BMI. Therefore, sex-specific tertiles were used in the analysis. For normally distributed variables, a test of linear trend was performed by treating leptin or adiponectin tertiles as continuous variables. The Cochran-Armitage test for trend was used for dichotomous variables. For categorical variables with more than two groups, or non-normal continuous variables, the Jonckheere-Terpstra test of trend was performed.

Cox proportional-hazards models were used to compare the time to fracture by tertiles of leptin and adiponectin and estimate the hazards ratio while controlling for potential confounders. Secondary analyses were performed separately for nonvertebral and vertebral fractures. Schoenfeld residuals were used to the test the assumption of proportionality.

To further test whether there was a linear relationship between adiponectin and fracture risk, we performed spline analysis to test for possible inflection points. Restricted cubic spline linear regression was used with knots for adiponectin at the 5th, 25th, and 75th percentiles, and a reference group at the 95th percentile was set to create the spline plot. Threshold effects were evaluated by identifying potential inflection points on the spline and performing a test of equality to determine if the slopes above and below the cut point are equal. We did not adjust for any covariates to preserve the shape and smoothness of the spline plot.

A backward elimination procedure was used with age, race, BMI, and the exposure of interest forced in all the multivariate models. Diabetes, weight change, and hip BMD were added in the final model to determine whether these factors further explained our associations. Multicollinearity was assessed using the variance inflation factor (VIF). Interactions between sex and leptin or adiponectin also were evaluated in the multivariate models. All statistical analyses were performed using the Statistical Analysis System (SAS, Version 9.1; SAS Institute, Cary, NC, USA).

Results

Table 1 shows the baseline characteristics by sex. Women had significantly greater serum leptin (21.4 versus 7.9 ng/mL) and adiponectin levels (13.3 versus 9.5 µg/mL) than men (p < .001). Women also were somewhat younger, had higher BMIs and total body fat, lower weight loss and total lean mass, were more likely to never smoke or consume alcohol, had lower dietary vitamin D or calcium intake, had a lower prevalence of diabetes and MI, and had a lower grip strength than men. Men had significantly higher hip aBMD than women (0.97 ± 0.15 g/cm2 versus 0.91 ± 0.15 g/cm2, p < .001) at baseline.

Table 1. Characteristics Among Older Women and Men
 Women (N = 1478)Men (N = 1568)p Value
Leptin (ng/mL), mean (SD)21.4 ± 14.77.9 ± 6.9<.001
Adiponectin (µg/mL), mean (SD)13.3 ± 7.49.5 ± 5.6<.001
Age (years), mean (SD)73.5 ± 2.973.8 ± 2.9.013
BMI (kg/m2), mean (SD)27.7 ± 5.527.1 ± 4.0<.001
Weight change (kg), mean (SD)–0.37 ± 1.74–0.40 ± 1.94<.001
Total fat mass (kg), mean (SD)29.2 ± 9.324.2 ± 7.2<.001
Total lean mass (kg), mean (SD)41.3 ± 6.257.1 ± 7.3<.001
Race, %  <.001
 Whites53.963.2 
 Blacks46.136.8 
Education, %  <.001
 <HS23.227.3 
 HS graduate39.525.5 
 Postsecondary37.347.2 
Site, %  .885
 Memphis50.150.4 
 Pittsburgh49.949.6 
Smoking, %  <.001
 Never57.229.4 
 Current9.910.8 
 Former32.959.8 
Alcohol, %  <.001
 No consumption in last year57.742.6 
 <1 drink per week21.919.6 
 1 to 7 drinks per week16.926.2 
 >1 drink per day3.511.6 
Dietary calcium (mg/d), mean (SD)753.2 ± 363.6817.3 ± 428.9<.001
Supplementary calcium, %28.47.2<.001
Dietary vitamin D (IU/d), mean (SD)192.6 ± 131.1225.9 ± 147.5<.001
Supplementary vitamin D, %13.43.4<.001
Physical activity (kcal/week), median IQR)298.5 (26.7–920.4)688.0 (165.1–1832.1)<.001
History of fracture after age 45, %27.416.1<.001
Diabetes, %16.322.0<.001
Hypertension, %77.674.0.020
MI, %17.922.4.051
NSAID use, %25.218.4<.001
Grip strength (kg), mean (SD)25.1 ± 6.340.8 ± 8.6<.001
Hip aBMD (g/cm2)0.81 ± 0.150.97 ± 0.15<.001

Greater serum leptin levels were significantly associated with higher BMI, no alcohol use in the past year, diabetes, hypertension, and higher aBMD in both women and men (Table 2). Among women only, increasing leptin tertiles were significantly associated with lower age, greater weight loss, less than HS education, supplementary calcium intake, lower dietary and supplementary vitamin D intake, lower physical activity, lower history of fracture prevalence, higher NSAID use, and greater grip strength. The direction of these associations was reversed for adiponectin with the exception of NSAID use in women.

Table 2. Characteristics by Leptin and Adiponectin Tertiles in Older Women and Men
 LeptinAdiponectin
 LowMiddleHighp TrendLowMiddleHighp Trend
WomenN = 518N = 517N = 517 N = 562N = 501N = 505 
 Leptin (ng/mL)0.3–12.412.5–24.124.2–99.3<.001
 Adiponectin (µg/mL) 1.0–9.010.0–15.016.0–49.0<.001
 Age (years), mean (SD)73.8 ± 3.073.5 ± 2.873.2 ± 2.8.00173.1 ± 2.873.4 ± 2.874.1 ± 2.9<.001
 BMI (kg/m2), mean (SD)23.7 ± 3.927.4 ± 4.031.8 ± 5.0<.00130.1 ± 5.427.5 ± 5.125.3 ± 4.8<.001
 Weight change (kg), mean (SD)–0.22 ± 1.39–0.30 ± 1.67–0.56 ± 2.07.002–0.40 ± 1.50–0.31 ± 1.35–0.38 ± 2.25.780
 Total fat mass (kg), mean (SD)22.1 ± 6.728.6 ± 6.136.7 ± 8.4<.00132.5 ± 9.329.3 ± 8.825.4 ± 8.3<.001
 Total lean mass (kg), mean (SD)38.3 ± 5.040.9 ± 5.744.6 ± 6.0<.00144.1 ± 6.340.8 ± 5.838.7 ± 5.1<.001
 Race, %   <.001   <.001
  Whites68.055.339.1 29.961.972.9 
  Blacks32.044.760.9 70.138.127.1 
 Education, %   <.001   <.001
  <HS18.021.829.6 33.117.717.8 
  HS graduate40.142.237.4 37.242.838.8 
  Postsecondary41.937.033.0 29.739.643.4 
 Site, %   .007   .117
  Memphis47.746.856.1 48.049.752.9 
  Pittsburgh52.353.243.9 52.050.347.1 
 Smoking, %   .136   .007
  Never58.754.957.7 53.058.061.1 
  Current13.511.05.2 10.98.89.9 
  Former27.834.037.1 36.133.229.0 
 Alcohol, %   .002   <.001
  No consumption in last year54.355.962.6 65.154.252.8 
  <1 drink per week21.922.821.1 20.822.822.2 
  1 to 7 drinks per week19.716.614.5 11.220.020.2 
  >1 drink per day4.14.71.8 2.93.04.8 
 Dietary calcium (mg/d), mean (SD)788.1 ± 375.0732.7 ± 357.2743.9 ± 358.7.065764.5 ± 388.6729.2 ± 345.2764.5 ± 352.9.970
 Supplementary calcium, %37.726.421.1<.00120.430.535.1<.001
 Dietary vitamin D (IU/d), mean (SD)212.8 ± 140.2178.3 ± 123.5186.7 ± 127.4.002192.9 ± 134.4184.9 ± 122.7199.7 ± 135.2.450
 Supplementary vitamin D, %18.712.08.7<.0019.812.617.3<.001
 Physical activity (kcal/week), median (IQR)425 (55–1040)289 (47–895)213 (5–788)<.001207 (8–793)343 (42–1025)371 (70–898).001
 History of fracture after age 45, %30.927.324.0.01320.727.734.5<.001
 Diabetes, %10.716.021.9<.00130.110.76.3<.001
 Hypertension, %72.078.182.2<.00184.376.571.3<.001
 MI, %8.38.36.6.2997.89.06.7.529
 NSAID use, %21.026.027.9.01123.624.527.6.140
 Grip strength (kg), mean (SD)24.4 ± 6.325.2 ± 6.125.5 ± 6.4.00326.2 ± 5.724.9 ± 6.324.0 ± 6.7<.001
 Hip aBMD (g/cm2)0.75 ± 0.140.80 ± 0.130.88 ± 0.14<.0010.87 ± 0.140.80 ± 0.140.75 ± 0.14<.001
MenN = 488N = 491N = 489 N = 519N = 452N = 505 
 Leptin (ng/mL)0.1–4.34.4–8.48.5–60.3<.001 
 Adiponectin (µg/mL)1.0–6.07.0–10.011.0-53.0<.001
 Age (years), mean (SD)73.8 ± 2.973.7 ± 2.873.7 ± 2.9.30773.4 ± 2.873.7 ± 2.974.1 ± 2.9<.001
 BMI (kg/m2), mean (SD)24.2 ± 2.827.1 ± 3.029.9 ± 3.8<.00128.1 ± 3.727.1 ± 3.926.0 ± 4.0<.001
 Weight change (kg), mean (SD)–0.33 ± 2.18–0.36 ± 1.67–0.51 ± 1.94.163–0.43 ± 1.81-0.38 ± 2.18–0.37 ± 1.84.615
 Total fat mass (kg), mean (SD)18.2 ± 4.424.4 ± 4.829.8 ± 6.9<.00125.2 ± 6.824.6 ± 7.022.7 ± 7.5<.001
 Total lean mass (kg), mean (SD)54.2 ± 6.357.3 ± 6.959.7 ± 7.7<.00158.8 ± 7.457.3 ± 7.255.1 ± 6.9<.001
 Race, %   .078   <.001
  Whites63.368.057.9 49.364.876.2 
  Blacks36.732.042.1 50.735.223.8 
 Education, %   .248   .033
  <HS30.026.126.0 30.124.626.7 
  HS graduate24.625.926.0 27.824.824.0 
  Postsecondary45.448.048.0 42.150.649.3 
 Site, %   .049   .117
  Memphis46.552.352.8 48.948.053.9 
  Pittsburgh53.547.747.2 51.152.046.1 
 Smoking, %   .023   .174
  Never31.127.829.9 29.733.025.8 
  Current16.08.67.5 10.411.710.3 
  Former52.963.662.6 58.955.363.8 
 Alcohol, %   .051   <.001
  No consumption in last year39.043.445.4 47.141.139.3 
  <1 drink per week19.720.119.2 21.720.716.4 
  1 to 7 drinks per week29.125.823.5 22.325.330.9 
  >1 drink per day12.210.711.9 8.912.913.4 
 Dietary calcium (mg/d), mean (SD)824.4 ± 431.3807.5 ± 416.9820.0 ± 440.5.878802.1 ± 441.3822.4 ± 445.1828.0 ± 401.8.355
 Supplementary calcium, %7.66.77.4.8875.67.88.3.089
 Dietary vitamin D (IU/d), mean (SD)225.8 ± 151.6221.7 ± 138.7230.8 ± 152.9.630222.7 ± 153.4221.6 ± 146.2232.2 ± 142.2.330
 Supplementary vitamin D, %2.53.34.5.0802.53.64.2.142
 Physical activity (kcal/week), median (IQR)766 (189–1977)657 (121–1840)647 (178–1715).279579 (148–1539)794 (172–2003)761 (178–1944).070
 History of fracture after age 45, %17.214.916.0.59116.614.916.6.983
 Diabetes, %15.222.028.9<.00135.019.411.2<.001
 Hypertension, %72.370.578.9.01976.575.070.9.041
 MI, %16.812.517.6.71516.616.613.7.211
 NSAID use, %17.718.319.1.58316.717.820.8.094
 Grip strength (kg), mean (SD)40.9 ± 9.140.9 ± 8.440.6 ± 8.4.59541.9 ± 8.741.0 ± 8.939.5 ± 8.2<.001
 Hip aBMD (g/cm2)0.93 ± 0.150.97 ± 0.141.02 ± 0.15<.0011.02 ± 0.150.98 ± 0.140.92 ± 0.15<.001

Fractures

A total of 406 incident fractures (nonvertebral = 334 and vertebral = 72) occurred over a mean follow-up of 6.5 ± 1.9 years. For women and men, the fracture rates per 1000 person-years were 27.5 and 14.0, respectively. Based on the unadjusted model, greater leptin levels were significantly (p = .009 for trend) associated with lower fracture rates, and higher adiponectin concentration predicted (p = .003 for trend) greater fracture risk in women (Table 3). In men, higher adiponectin was a predictor (p = .001 for trend) of increased fracture risk in the unadjusted models. However, the association of leptin and adiponectin with fracture in women was attenuated after adjusting for age, race, and BMI (p > .658 for trend). Men with the highest adiponectin level (tertile 3) had a 94% higher risk of fracture [hazard ratio (HR) = 1.94, 95% confidence interval (CI) 1.09–2.77] compared with the lowest tertile (tertile 1; p = .007 for trend) after adjusting for age, race, BMI, education, diabetes, weight change, and hip BMD. Sex modified the association between adiponectin tertiles and fracture risk in the full multivariate model (p = .025 for interaction). However, there was no evidence of a threshold for serum adiponectin and risk of a fracture in men (p > .05 for test of threshold; Fig. 1).

Table 3. Cox Proportional Hazards Modela for Fractures According to Sex-Specific Tertiles of Leptin (ng/mL) and Adiponectin (µg/mL)
 LeptinAdiponectin
 MiddleHighp TrendMiddleHighp Trend
  • a

    Values are hazard ratios (95% CI).

  • b

    Adjusted for age, race, BMI, education, history of fracture, and grip strength.

  • c

    Adjusted for age, race, BMI, and education.

  • d

    Adjusted for age, race, BMI, education, history of fracture, grip strength, weight change, and hip BMD.

  • e

    Adjusted for age, race, BMI, education, weight change, and hip BMD.

Women      
 Unadjusted0.74 (0.56–0.98)0.68 (0.51–0.92)0.0091.55 (1.14–2.11)1.61 (1.19–2.19)0.003
 Age adjusted0.75 (0.56–0.99)0.70 (0.52–0.94)0.0151.52 (1.12–2.08)1.56 (1.14–2.12)0.006
 Age and race adjusted0.80 (0.60–1.06)0.84 (0.62–1.44)0.2171.24 (0.90–1.72)1.19 (0.86–1.66)0.369
 Age race and BMI adjusted0.86 (0.63–1.16)0.98 (0.67–1.41)0.7941.19 (0.86–1.66)1.11 (0.79–1.56)0.658
 Base modelb0.86 (0.63–1.16)0.96 (0.66–1.39)0.7321.17 (0.84–1.64)1.08 (0.76–1.53)0.796
 Base modelb + diabetes0.85 (0.62–1.16)0.96 (0.66–1.41)0.7541.32 (0.94–1.85)1.21 (0.84–1.74)0.412
 Base modelb + weight change0.85 (0.62–1.17)0.99 (0.68–1.45)0.8731.12 (0.79–1.57)0.96 (0.70–1.42)0.869
 Base modelb + hip BMD0.92 (0.67–1.25)1.11 (0.76–1.63)0.6781.16 (0.83–1.62)0.94 (0.66–1.34)0.565
 Full multivariate modelc0.93 (0.68–1.28)1.16 (0.78–1.72)0.5441.21 (0.85–1.73)0.98 (0.67–1.43)0.729
 Full multivariate model,c nonvertebral0.94 (0.66–1.34)1.18 (0.77–1.81)0.5081.37 (0.94–2.01)0.99 (0.65–1.51)0.698
 Full multivariate model,c vertebral0.91 (0.40–2.06)1.07 (0.37–3.12)0.9560.56 (0.22–1.44)0.86 (0.35–2.09)0.959
Men      
 Unadjusted0.89 (0.59–1.33)0.70 (0.45–1.09)0.1181.44 (0.88–2.35)2.30 (1.48–3.58)<0.001
 Age adjusted0.90 (0.60–1.35)0.72 (0.46–1.12)0.1461.41 (0.86–2.31)2.22 (1.43–3.47)<0.001
 Age and race adjusted0.90 (0.60–1.35)0.75 (0.49–1.17)0.2091.35 (0.82–2.21)2.00 (1.27–3.14)0.002
 Age race and BMI adjusted0.94 (0.61–1.46)0.82 (0.48–1.39)0.4671.34 (0.82–2.19)1.95 (1.23–3.08)0.004
 Base modeld0.87 (0.56–1.36)0.75 (0.44–1.28)0.7311.35 (0.82–2.21)2.04 (1.28–3.23)0.002
 Base modeld + diabetes0.86 (0.55–1.34)0.69 (0.40–1.20)0.1891.42 (0.87–2.34)2.23 (1.38–3.61)0.001
 Base modeld + weight change0.85 (0.54–1.33)0.72 (0.42–1.25)0.2421.29 (0.78–2.13)2.07 (1.30–3.30)0.002
 Base modeld + hip BMD0.85 (0.55–1.33)0.75 (0.43–1.29)0.2961.30 (0.79–2.14)1.67 (1.05–2.66)0.029
 Full multivariate modele0.80 (0.51–1.25)0.64 (0.37–1.21)0.1181.39 (0.84–2.32)1.94 (1.20–3.16)0.007
 Full multivariate model,e nonvertebral0.79 (0.47–1.32)0.72 (0.39–1.34)0.2941.70 (0.94–3.06)2.33 (1.32–4.10)0.004
 Full multivariate model,e vertebral0.86 (0.34–2.14)0.42 (0.11–1.57)0.2200.68 (0.23–1.98)1.03 (0.39–2.70)0.888
Figure 1.

Plot showing the unadjusted hazard ratios and 95% confidence limits of incident fractures by baseline serum adiponectin in older men.

Secondary analyses for nonvertebral and vertebral fractures

Results of the secondary analyses for serum leptin/adiponectin and risk of nonvertebral and vertebral fractures are shown in Table 3. In the full multivariate model, men in tertile 3 for adiponectin had a 133% higher risk of nonvertebral fracture (HR = 2.33, 95% CI 1.32–4.10) compared with tertile 1 (p = .004 for trend). However, there was very limited power to detect any significant associations for incident vertebral fractures by serum leptin or adiponectin.

Discussion

We conducted the largest and most comprehensive analysis of the association of leptin and adiponectin with risk of fracture among older individuals. We found that among older men, higher adiponectin levels were associated with an increased risk of fracture independent of potential confounders, including BMI, diabetes, weight change, and BMD. This association was not observed among older women, and the interaction between adiponectin and sex was statistically significant. On the other hand, the impact of leptin on fracture risk was explained largely by BMI in both groups.

Sex modified the association between adiponectin and fracture. Men in the highest adiponectin tertile had a 94% higher risk of fracture than men in lowest tertile. This association was independent of BMI, weight change, and other potentially important covariates. In contrast, adiponectin levels were unrelated to fracture risk in women. Our results are consistent with those from the Rancho Bernardo Study, which recently reported an association between adiponectin and fractures [adjusted odds ratio (OR) = 1.13, 95% CI 1.08-1.23] in men not women.36 Participants of the Rancho Bernardo Study were similar to our subjects in age and smoking prevalence but had lower BMI, fat mass, adiponectin, and prevalence of diabetes. Also, consistent with our findings, a cross-sectional study in Japan among 231 men and 170 postmenopausal women reported that a 1 SD difference in adiponectin was associated with higher odds of vertebral fracture in men but not women.35 In contrast, a recent prospective study (15-year follow-up) in Sweden did not find an association between adiponectin levels and fracture among 507 men aged 70 years and older.30 The results of our spline analysis among men suggest that this association was linear with no evidence of an adiponectin threshold for fracture risk.

Recombinant adiponectin induces RANKL and inhibits OPG mRNA expression in human osteoblasts in a dose- and time-dependent manner, leading to osteoclast formation.27 Our findings are consistent with these data showing that adiponectin at higher levels may have a direct influence on bone turnover and remodeling. The risk of fracture with greater levels of adiponectin may reflect greater osteoclast activation and bone resorption. However, population-based longitudinal studies evaluating the association of adiponectin and markers of bone turnover are needed.

Our results suggest that adiponectin is a unique risk factor for fracture in men. It is unclear why there are gender differences for the association of adiponectin with incident fractures. Adiponectin was 40% higher in women than in men. However, it is uncertain if sex differences in adiponectin account for these differential associations. BMI and weight change explained fracture risk in women, but in men these variables did not predict fractures. Prior studies suggest that the association between adiponectin and bone may be influenced by sex hormones.36 Additional prospective studies that include both adiponectin and sex hormones are needed to further our understanding of adiponectin and fracture.

In contrast to adiponectin, we found no evidence of an association between serum leptin levels and the risk of fracture in this cohort of older adults. A 10-year cohort study among 250 Italian women and men with comparatively lower leptin levels reported that subjects in the highest leptin tertile had a significantly reduced risk of a nontraumatic fracture compared with the lowest tertile group after adjusting for weight and other confounders.34 However, this study included relatively younger participants (mean age 58.5 years) and also included few fracture cases (N = 31). Higher leptin levels also were associated with a lower prevalence of vertebral fractures independent of percent body fat among 139 Japanese postmenopausal women aged 48 to 78 years.14 These two studies included relatively younger subjects, and older age is associated with a greater number of comorbidities. Therefore, age and other potentially confounding variables such as BMI, weight change, and grip strength in our study were able to explain fracture risk better than leptin levels.

Our study had several potential limitations. We measured leptin and adiponectin at baseline only. In addition, survivors may have greater BMD than nonsurvivors,42 and study participants are likely to be healthier than the general population. Also, this assay cannot distinguish high- and low-molecular-weight forms of adiponectin, which may have varying biologic activities.43 Finally, we adjusted for many covariates but lacked sufficient data to adjust for insulin (a known correlate of adiponectin) and several bone markers, including osteocalcin, cross-links, N-terminal type 1 procollagen propeptide (P1NP), and serum cross-linked C-telopeptide (CTX). Our study also had notable strengths, including its large sample size, reliable measurement of leptin and adiponectin, and validation of fractures, adjustment for many potential confounders, and long follow-up period.

In summary, in a large cohort of racially diverse older women and men, higher adiponectin levels were associated with an increased risk of fracture in men but not women after controlling for multiple confounders, including measures of adiposity and BMD. Serum leptin levels were not associated with fracture risk in either men or women. Our findings suggest that adiponectin may be a novel risk factor for fractures among older men. Future studies are needed to evaluate and understand the physiologic basis for the potential sex differences in the relationship between adiponectin and fracture.

Dsiclosures

All the authors state that they have no conflicts of interest.

Acknowledgements

The Health Aging and Body Composition Study (Health ABC) includes the contract numbers N01-AG-6-2101, N01-AG-6-2103, N01-AG-6-2106, and NIH/NIA R01-AG028050. This research was supported in part by the Intramural Research Program of the NIH, National Institute on Aging.

Ancillary