High serum adiponectin predicts incident fractures in elderly men: Osteoporotic fractures in men (MrOS) Sweden



Adipocytes and osteoblasts share a common progenitor, and there is, therefore, potential for both autocrine and endocrine effects of adiponectin on skeletal metabolism. The aim of the present study was to determine whether high serum adiponectin was associated with an increased risk of fracture in elderly men. We studied the relationship between serum adiponectin and the risk of fracture in 999 elderly men drawn from the general population and recruited to the Osteoporotic Fractures in Men (MrOS) study in Gothenburg, Sweden. Baseline data included general health questionnaires, lifestyle questionnaires, body mass index (BMI), bone mineral density (BMD), serum adiponectin, osteocalcin, and leptin. Men were followed for up to 7.4 years (average, 5.2 years). Poisson regression was used to investigate the relationship between serum adiponectin, other risk variables and the time-to-event hazard function of fracture. Median levels of serum adiponectin at baseline were 10.4 µg/mL (interquartile range, 7.7–14.3). During follow-up, 150 men sustained one or more fractures. The risk of fracture increased in parallel with increasing serum adiponectin (hazard ratio [HR]/SD, 1.46; 95% confidence interval [CI], 1.23–1.72) and persisted after multivariate-adjusted analysis (HR/SD, 1.30; 95% CI, 1.09–1.55). Serum adiponectin shows graded stepwise association with a significant excess risk of fracture in elderly men that was independent of several other risk factors for fracture. Its measurement holds promise as a risk factor for fracture in men. © 2012 American Society for Bone and Mineral Research.


Adiponectin is a hormone produced mainly by adipocytes; low serum concentrations are associated with central adiposity and insulin resistant diabetes mellitus.1–4 Receptors for adiponectin and the expression of adiponectin were subsequently demonstrated in a variety of other tissues including osteoblasts5, 6 and osteoclasts.6 Culture of osteoblast-like cells with adiponectin stimulates osteoblastogenesis and the osteoclast receptor activator of NF-κB ligand (RANKL) pathway while inhibiting its decoy receptor, osteoprotegerin.7 Adipocytes and osteoblasts share a common progenitor, and there is, therefore, potential for both autocrine and endocrine effects of adiponectin on skeletal metabolism.

Several clinical studies have investigated the association between serum adiponectin and skeletal metabolism. In cross-sectional studies, higher serum adiponectin has been generally associated with lower bone mineral density (BMD) values in premenopausal women,8–11 postmenopausal women,10, 12 and in women with the metabolic syndrome.13, 14 In one study, the association between serum adiponectin and BMD observed in postmenopausal women was not evident in premenopausal women.15 Correlations have usually been weak, suggesting that adiponectin explained only a small component of the variance in BMD or bone mineral content (BMC) (generally 5% to 15%).11 In a small prospective study, serum adiponectin concentration predicted bone loss at the lumbar spine in 35 elderly women followed for 1 year.14 In some but not all studies15 the association disappeared after adjustment for body mass8 or lean body mass.14 Such findings have not been invariant16, 17 and a small case-control study reported no difference in total- or high molecular weight-adiponectin in postmenopausal women with or without osteoporosis.18

Fewer data are available for men but, as for women, several studies report an inverse relationship between serum adiponectin and BMD at the femoral neck or lumbar spine in healthy young men,19, 20 in elderly men,17, 21 and in men with the metabolic syndrome.13 In some instances, the correlation disappeared after adjustment for body mass index (BMI).21 Significant negative correlations between adiponectin and BMD were found, which remained significant after adjustment of age and body fat.20 As is the case for women, other studies found no association between adiponectin and central BMD.12, 22 In one of these studies, the lack of an independent association of adiponectin with BMD assessed using pQCT at the tibia in 320 adult men contrasted to opposite findings in 271 women reported in the same study.12

Thus, although there is some biological and epidemiologic evidence suggesting that adiponectin is a risk factor for low bone mass, results are inconsistent, and evidence that serum adiponectin values might be a risk factor for future fracture are limited to two cross-sectional23, 24 and three prospective studies17, 25, 26 with divergent results. The aim of the present study was to prospectively investigate the relationship between serum adiponectin and fracture risk in elderly men in Sweden.

Subjects and Methods

Osteoporotic Fractures in Men (MrOS) is a multicenter, prospective cohort study of elderly men in Sweden, Hong Kong, and the United States.27 The present study is based on data from the Swedish MrOS cohort of older men recruited at medical centers in Gothenburg (n = 1010). Details have been described previously.28, 29 In brief, men aged 70 to 81 years were randomly identified using national population registers. To be eligible for the study, men had to be able to walk without aids, provide self-reported data, and give written informed consent. There were no other exclusion criteria. The participation rate in MrOS Sweden was 45%. In the present report, the baseline data in MrOS Sweden (Gothenburg) was used together with 7 years of follow-up for fracture.

The international MrOS questionnaire27 was administered at baseline to collect information about current smoking, physical activity, number and type of medications, self-estimated general health, fracture history, family history of hip fracture, history of diseases (has a doctor told you that you have rheumatoid arthritis, hypertension, cancer, stroke, myocardial infarction, diabetes, or angina?) and the use of alcohol. Physical activity was quantified using some of the questions in the Physical Activity Scale for Elderly (PASE).30 The participants were asked to estimate the amount of time spent in the past week heavy training (weight training, pushups). The activity was categorized as: never (0); seldom, 1 or 2 days (1); sometimes, 3 to 4 days (2); and often 5 to 7 days (3). General health was self-reported as “compared to other people of your own age, how would you describe your own health? Very good (1), good (2), fair (3), bad (4), or very bad (5).” Use of alcohol was expressed as three or more glasses of alcohol-containing drinks per day, calculated from the reported frequency and amount of alcohol use.

Areal BMD, total body fat mass, and total body lean mass was measured using the Hologic QDR 4500/A-Delphi (Hologic, Bedford, MA, USA). Height (in cm) and weight (in kg) were measured and BMI was calculated as kg/m2. Plasma and serum samples were collected at 8 a.m. after at least 10 hours of fasting and abstinence from smoking. Samples were frozen immediately and stored at −80°C. Serum adiponectin was measured at baseline for 999 men and analyzed with an ELISA-kit (Linco Research, St. Charles, MO, USA); interassay coefficient of variation (CV) 8% at the Department of Clinical Chemistry, Sahlgrenska University Hospital. Serum 25-hydroxyvitamin D [25(OH)D] was measured with a competitive RIA (Diasorin, Stillwater, MN, USA; intraassay CV 6%, interassay CV 15% to 16%). The plasma levels of total osteocalcin (carboxylated + uncarboxylated) were analyzed at the first thaw, using monoclonal antibodies against human osteocalcin, the N-terminal amino acids 1–43 and 1–49, and detected by electrochemiluminescence (Elecsys N-MID Osteocalcin Cal-Set; Roche Diagnostics, Indianapolis, IN, USA). Leptin was measured in serum samples using a commercially available kit (Diagnostic Systems Laboratories, Webster, TX, USA; interassay CV 5.3%). Fasting serum insulin was measured with an immunoelectric method based on chemiluminescence with an Advia Centaur (Bayer Diagnostics, Tarrytown, NY; interassay CV <10%). Fasting plasma glucose was determined by an enzymatic method on a Modular (Roche; interassay CV <4%). Parathyroid hormone (PTH) levels were analyzed using the Immulite 2000 Intact PTH Assay (Diagnostic Products Corporation, Los Angeles, CA, USA; intraassay CV 2% to 3%, interassay CV 9% to 10%).

Fracture outcomes were collected up to September 1, 2009, and the site of fracture and time to fracture were retrieved from computerized X-ray archives in Gothenburg using a unique personal registration number held by all Swedish citizens. Fractures reported by the study subject, but not independently confirmed, were not included in the present study. Deaths were documented from the National Cause of Death Register up to the end of 2009. This register comprises records of all deaths in Sweden and is more than 99% complete. Emigrants were followed up to the day of emigration. Participants were followed for a mean of 5.2 years (range, 0.0–7.4) years after the baseline examination. Any fracture was studied as the primary endpoint regardless of site. Sites considered to be characteristic of osteoporosis31 and all nonvertebral fractures were studied separately as secondary endpoints.

Statistical methods

Correlations between serum adiponectin and other variables were analyzed using nonparametric Pitman's permutation test.32 Linear regression was used to adjust for other variables.

An extension of the Poisson regression model33 was used to study the association between age, the time since baseline, serum adiponectin, other covariates on the one hand, and on the other hand, the risk of fracture. In contrast to logistic regression, the Poisson regression uses the length of each individual's follow-up period and the hazard function is assumed to be exp(β0 + β1 [current time from baseline] + β2 · [current age] + β3 · [current variable of interest]). The observation period of each participant was divided in intervals of 1 month. One fracture per person was counted. A similar approach was used to examine other predictors of fracture and a final multivariable model was constructed to determine which predictors had an independent contribution to risk. A forward stepwise approach was used, choosing all variables that had a value of p < 0.05 in the univariate analysis for the association with fracture as candidates for the multivariable analysis. The distribution of serum adiponectin was not normal and was normalized using a piecewise linear function. The association between predictive factors and risk of fracture was described as a gradient of risk (GR), expressed as the hazard ratio per 1 SD change.

In order to study the linearity of the association between serum adiponectin and fracture risk in more detail, a spline Poisson regression model was fitted using knots at the 10th, 50th, and 90th percentiles of serum adiponectin. The splines were second-order functions between the breakpoints and linear functions at the tails resulting in a smooth curve.

Two-sided values of p were used for all analyses and p < 0.05 was considered to be significant.


The mean ± SD serum adiponectin level in the 999 men was 11.9 ± 6.4 µg/mL. The 10th percentile was at 5.6 µg/mL and the 90th percentile was at 20.1 µg/mL. Serum values of adiponectin were not normally distributed but were skewed to the right with a median value 10.4 µg/mL. At baseline univariate analyses showed a significant positive correlation between serum adiponectin and age, cancer, physical activity, chronic obstructive airway disease, and osteocalcin. Serum adiponectin was also inversely associated with weight, BMI, BMD, hypertension, myocardial infarction, angina, diabetes, total fat mass, total lean mass, plasma glucose, fasting insulin, and serum leptin. Serum adiponectin was significantly negatively correlated to all BMD sites (total hip, lumbar spine, femoral neck, and trochanteric). BMD at total hip had the highest estimated correlation coefficient (r = −0.22; 95% confidence interval [CI], −0.28 to −0.16) and BMD at lumbar spine the lowest (r = −0.13; 95% CI, −0.19 to −0.06). In all further analyses, total hip BMD was used. When the association between serum adiponectin and all variables with significant relationship above was investigated in a multivariable linear analysis, BMI, fasting insulin, BMD, plasma glucose, chronic obstructive airway disease, myocardial infarction, physical activity, age, and cancer remained significant covariates (Table 1). There was a significant positive correlation between serum adiponectin and osteocalcin in a univariate setting (r = 0.15; 95% CI, 0.09–0.21). In the multivariate setting there was no significant correlation.

Table 1. Multivariable Linear Regression of the Correlations Between Serum Adiponectin (µg/mL) and Other Variables
  1. BMI = body mass index; BMD = bone mineral density.

Age (year)0.1430.05920.016
BMI (kg/m2)−0.2260.0647<0.001
Total hip BMD (g/cm2)−4.951.50<0.001
Fasting insulin (µU/mL)−0.07900.0189<0.001
Plasma glucose (mM)−0.5120.137<0.001
Chronic obstructive airways disease2.340.685<0.001
Myocardial infarction−1.670.5340.0017
Physical activity training (0–3)0.7170.2270.0016

During a follow-up of up to 7.4 years (average, 5.2 years) 150 men (15%) sustained one or more confirmed fractures. Forty-seven were spine fractures (31%), 21 were hip fractures (14%), 19 were forearm fractures (13%), 15 were hand fractures (10%), 12 were humeral fractures (8%), 9 were rib fractures (6%), 6 were foot fractures (4%), 5 were pelvic fractures (3%), and the rest (11%) were fractures of the scapula, clavicle, skull, face, patella, ankle, and other parts of the leg. The risk of fracture was significantly associated with serum adiponectin adjusted for age and time since baseline (hazard ratio per SD [HR/SD], 1.46; 95% CI, 1.23–1.72).

Table 2 shows baseline characteristics of the study population divided into men with and without an incident fracture. Serum adiponectin was significantly higher in men who sustained a fracture than those that did not. Men who sustained a fracture were also older, had lower total lean mass, lower BMD, worse general health, and a higher prevalence of previous fracture than men without a fracture. There was no difference in 25(OH)D, PTH, serum osteocalcin, plasma glucose, serum leptin, and indices of physical activity by fracture status. There was also no difference between men with fractures and those without in any of the reported comorbidities (diabetes, hypertension, cancer stroke, cardiovascular disease, and chronic obstructive airways disease).

Table 2. Characteristics of 999 Men Aged 70 to 81 Years at Baseline According to Fracture Status During Follow-Up
VariableNo fracture during follow-up (n = 849) (mean ± SD)Fracture during follow-up (n = 150) (mean ± SD)Significance (two-sided value of p)a
Age (years)75.2 ± 3.275.5 ± 3.20.039
Height (cm)175.9 ± 6.3174.8 ± 6.80.051
Weight (kg)81.2 ± 12.279.5 ± 12.30.16
BMI (kg/m2)26.2 ± 3.526.0 ± 3.7>0.30
Total fat mass (kg)18.6 ± 5.618.3 ± 6.1>0.30
Total lean mass (kg)59.5 ± 6.757.9 ± 6.70.0095
Total hip BMD (g/cm2)0.97 ± 0.140.90 ± 0.13<0.001
Adiponectin (µg/mL)11.5 ± 5.814.2 ± 8.9<0.001
Fasting insulin (µU/mL)10.7 ± 11.29.6 ± 8.5>0.30
Plasma glucose (mM)5.8 ± 1.45.9 ± 1.7>0.30
Osteocalcin (µg/L)26.7 ± 13.627.9 ± 11.50.28
Leptin (ng/mL)22.2 ± 19.824.2 ± 26.40.22
PTH (pmol/L)6.1 ± 4.15.8 ± 2.5>0.30
25(OH)D (nmol/L)66.9 ± 18.865.7 ± 20.5>0.30
General health (1–5)2.0 ± 0.82.2 ± 0.80.0012
Physical activity walking (0–3)2.4 ± 0.82.3 ± 0.80.28
Physical activity training (0–3)0.4 ± 0.80.5 ± 0.9>0.30
 % (n)b% (n)c 
  • BMI = body mass index; BMD = bone mineral density; PTH = parathyroid hormone; 25(OH)D = 25-hydroxyvitamin D.

  • a

    Poisson regression adjusted for age and current time since baseline.

  • b

    Percent of total number of men without a fracture (n) with a positive response.

  • c

    Percent of total number of men with a fracture (n) with a positive response.

Glucocorticoids (current)1 (849)2 (150)>0.30
Current smoking8 (832)10 (148)>0.30
Previous fracture of any kind30 (833)40 (148)0.017
Parental history of hip fracture10 (680)9 (127)>0.30
Rheumatoid arthritis2 (830)1 (144)>0.30
Diabetes11 (841)13 (150)0.23
Hypertension38 (841)32 (150)0.17
Cancer15 (840)21 (149)0.076
Stroke6 (841)9 (150)0.083
Myocardial infarction15 (840)16 (150)>0.30
Chronic obstructive airways disease8 (841)7 (150)>0.30
Alcohol 3 or more glasses per day4 (730)5 (126)>0.30

In a multivariate analysis of the risk of any fracture during follow-up, serum adiponectin, total hip BMD, and general health were significantly and independently associated with the risk of any fracture. None of the other variables had any significant independent predictive role, including previous fracture, although this association variable fell just short of our definition of statistical significance (p = 0.053). Higher adiponectin was a predictor of increased fracture risk in all models, unadjusted and adjusted (p < 0.001) (Table 3). The gradient of fracture risk (GR) (HR/SD change) for adiponectin was 1.30 (95% CI, 1.09–1.55), multivariable adjusted. When BMD was dropped from the model the GR was 1.42 (95% CI, 1.20–1.68). There was no significant interaction between BMD and serum adiponectin (p = 0.14); ie, the association between adiponectin and risk of fracture is not significantly different for different BMD values.

Table 3. Poisson Analysis of the HRs for any Fracture According to Tertiles of Adiponectin (µg/mL) and Hazard Function per SD
 HR (95% CI) (middle versus low)HR (95% CI) (high versus low)GR (95% CI)p*
  • HR = hazard ratio; CI = confidence interval; GR = gradient of risk; BMD = bone mineral density.

  • a

    Adjusted for age, time, BMD total hip, general health, and previous fracture.

  • *

    Two-sided value of p for the GR.

Unadjusted0.98 (0.63–1.50)1.78 (1.21–2.62)1.44 (1.27–1.63)<0.001
Age and time adjusted0.97 (0.63–1.49)1.72 (1.17–2.54)1.46 (1.23–1.72)<0.001
Adjusted for age, time, and BMD0.81 (0.52–1.26)1.44 (0.97–2.14)1.33 (1.12–1.58)<0.001
Multivariable modela0.81 (0.52–1.26)1.42 (0.95–2.11)1.30 (1.09–1.55)<0.001
Multivariable modela omitting total hip BMD0.87 (0.56–1.36)1.66 (1.12–2.45)1.42 (1.20–1.68)<0.001
Multivariable modela: osteoporotic fracture0.91 (0.56–1.48)1.50 (0.96–2.34)1.32 (1.13–1.54)<0.001
Multivariable modela: any nonvertebral fracture0.83 (0.49–1.40)1.39 (0.86–2.23)1.26 (1.08–1.47)0.0031

Men with the highest adiponectin level (tertile 3) had a 42% higher risk of fracture (HR, 1.42; 95% CI, 0.95–2.11) compared to the lowest tertile (tertile 1) after adjusting for age, time since baseline, BMD total hip, general health, and previous fracture. As an extra control, BMI, osteocalcin, total lean mass, insulin, plasma glucose, leptin, chronic obstructive airways disease, myocardial infarction, cancer, diabetes, and physical activity was forced into the multivariable model in Table 3 one at a time, and the association between serum adiponectin and the risk of fracture was still significant and the GR was stable.

The association between serum adiponectin and risk of any osteoporotic fracture or any nonvertebral fracture are shown in Table 3. There is a similar significant association between serum adiponectin and osteoporotic fracture risk (p < 0.001) with or without nonvertebral fracture (p = 0.0031).

Figure 1 shows the relationship between serum adiponectin and the risk of fracture when using spline functions and allowing the relationship to be nonlinear in the multivariable setting described in Table 3. Fracture risk increased with increasing values of serum adiponectin but the association was not significant at values below 18 µg/mL (the 86th percentile of the distribution).

Figure 1.

The hazard function of fracture (momentary risk) assessed and 95% confidence intervals according to baseline serum adiponectin for a man, age set to 75 years, and the time since baseline set to 2 years of follow-up. Previous fracture was set to “no” and BMD and general health is set to the average value of the cohort. The second axis to the right and the thin line shows the number of men at different values of adiponectin. The circles on the x axis represent the knots used in the spline model (10th, 50th, and 90th percentile).


In this prospective study of community-living older men, we describe a significant association between serum adiponectin and risk of any fracture independent of multiple potential confounders, including BMI, BMD, insulin, plasma glucose, general health, cardiovascular diseases, and history of previous fracture (GR, HR/SD change in serum adiponectin was 1.30 (95% CI, 1.09–1.55). Our findings are in accord with some previous studies that have reported the association between serum adiponectin and the risk of fracture in men. In a recent publication, Barbour and colleagues25 studied 3075 men and women aged 70 to 79 years followed for 6.5 years and found a significant association between serum adiponectin and all fractures in men but not in women. Two other studies have shown a significant association between vertebral fracture and adiponectin in men.17, 23 In contrast, Michaelsson and colleagues26 could not demonstrate a significant association after an extended follow-up of 15 years. Thus far no studies have shown a relationship between adiponectin and fracture risk in women.17, 23, 24

The mechanism to account for the association between serum adiponectin and fracture risk is unknown. The present study was not designed to address this but suggests that the association is independent of BMD. Several other important risk factors have been shown to be independent of BMD including age, a family history of hip fracture, and a prior fragility fracture,34, 35 and these findings reinforce the view that the measurement of BMD captures only a small component of fracture risk. The association appeared to be independent of comorbidities—at least those examined in this report. If the association is causal then any hypothesis would need to accommodate the general finding that the relationship is not observed in women. Irrespective of the mechanism, the findings of the present study suggest that serum adiponectin might add to fracture risk prediction in men.

The performance characteristics of serum adiponectin in the assessment of fracture risk are similar to that expected of BMD in men. In a meta-analysis, the risk of any fracture increased by 1.44 (range, 1.32–1.58) in men for each SD decrease in femoral neck BMD.36 In the present study, serum adiponectin had approximately the same predictive value (1.46; 95% CI, 1.23–1.72). In the present study, serum adiponectin was inversely correlated to BMD at any site as found in other studies.17, 23, 25, 26 After adjustment for total hip BMD, the association between serum adiponectin persisted although the GR was somewhat decreased (1.33; 95% CI, 1.12–1.58). These data suggest that adiponectin has almost the same predictive value as BMD, and is only in part dependent on BMD. Similar findings were noted when GRs were adjusted for other significant covariates (see Table 3). If the associations that we describe are reproduced in other male cohorts with a wider age range, serum adiponectin may have a role in the assessment of fracture risk in men.

In contrast, the present study found no association of serum osteocalcin with fracture risk. In a review of the utility of markers of bone turnover for fracture risk prediction, Vasikaran and colleagues37 reported heterogeneous results for osteocalcin in the prediction of fractures, depending on fracture site studied and the analyte used, and data for men were sparse. In a study of elderly men in northern Finland, a decrease in the ratio of carboxylated serum osteocalcin (s-OC) to total s-OC was associated with increased risk of subsequent fractures.38 In the present study, s-OC was correlated to serum adiponectin in a univariate setting and was no longer significantly correlated in a multivariate setting.

The work presented in this article has a number of strengths and weaknesses. A strength is the detail of the baseline assessment so that we were able to examine a large number of potential classical and more novel covariates (confounders or potential effect modifiers) and adjust for these—including insulin and plasma osteocalcin. There are also several limitations. The participation rate (45%) is likely to have caused a healthy participant selection bias. By the inclusion of time since baseline assessment in the models when estimating the association between serum adiponectin and the risk of fracture, we have in part adjusted for this effect. A significant limitation is the narrow age range of men that were recruited to the MrOS (70–81 years), so that our findings may not be extrapolated to younger ages. The assay that we used did not distinguish high and low molecular weight forms of adiponectin, which may have varying biological activities, but this limitation would be expected to weaken rather than strengthen any association of the analyte with fracture risk.

We conclude that there is a significant association between serum adiponectin levels and risk of fracture in elderly men. The findings suggest that the measurement of adiponectin might be of value in the assessment of fracture risk in men.


HJ: Supported by ESCEO-AMGEN Osteoporosis Fellowship Award. EM: Research funding and/or advisory board and/or speaker's fees from the following: Amgen, AstraZeneca, GSK, Hologic, Innovus, Lilly, MSD, Novartis, Pfizer, Roche, Servier, Tethys, and Warner Chilcott. ML: Grant support from The Swedish Research Council, the Avtal om läkarutbildningen och forskning research grant in Gothenburg research grant in Gothenburg, and the Lundberg Foundation. DM: Grant support from The Swedish Research Council, and the ALF/LUA research grant in Gothenburg. UL: Grant support from Swedish Research Council for Medicine, the Swedish Rheumatism Association, the Royal 80 Year Found of King Gustav V, Combine, the Lundberg Foundation, and ALF/LUA grants from the Sahlgrenska University Hospital; lecture fees from Amgen. All other authors state that they have no conflicts of interest.


This work was supported by an ESCEO-AMGEN Osteoporosis Fellowship Award. Amgen had no input into the analysis plan or in the writing of this report. The research was also supported by the ALF/LUA research grant in Gothenburg.

Authors' roles: Study design: HJ, AO, JAK, and DM. Data collection: MK, ÖL, CO, and DM. Data analysis: HJ. Data interpretation: HJ, AO, JAK, and DM. Drafting manuscript: HJ, JAK, and DM. Revising manuscript content: AO, UHL, HJ, ML, EBC, MK, ÖL, US, EM, and CO. Approving final version of manuscript: HJ, AO, UHL, HJ, ML, EBC, MK, ÖL, US, EM, JAK, CO, and DM. HJ takes responsibility for the integrity of the data analysis.