Adiponectin and its association with bone mass accrual in childhood

Authors

Errata

This article is corrected by:

  1. Errata: Erratum: Adiponectin and its association with bone mass accrual in childhood Volume 27, Issue 9, 2035, Article first published online: 17 August 2012

Abstract

Circulating adiponectin levels are inversely related to bone mineral density (BMD) in humans and animal models. Previous studies in humans have been confined largely to adult populations, and whether adiponectin influences bone mass accrual in childhood is unclear. We examined this question using the Avon Longitudinal Study of Parents and Children (ALSPAC) birth cohort by investigating relationships between circulating adiponectin levels at a mean age of 9.9 years, indices of bone mass as measured by total-body dual-energy X-ray absorptiometry (DXA) at ages 9.9 and 15.5 years, and cortical bone parameters as measured by peripheral quantitative computed tomography (pQCT) of the midtibia at age 15.5 years. A total of 4927 children were included at age 9.9 years, of whom 97% and 90% of boys and girls, respectively, were in prepuberty or early puberty, as defined by Tanner stage 1–2. A total of 2754 children were included at age 15.5 years, of whom 95% and 97% of boys and girls, respectively, were in late puberty, as defined by Tanner stage 4–5. Circulating adiponectin was found to be related to fat mass, lean mass, and, to a lesser extent, height, so analyses were adjusted for these three variables to identify possible independent effects of adiponectin on bone development. Adiponectin was inversely related to total-body-less-head bone mineral content (BMC; −3.0%), bone area (BA; −1.8%), BMC divided by BA (BMD; −4.8%), and BMC adjusted for BA by linear regression (aBMC; −5.6%), as measured at age 9.9 years (coefficients show change per doubling in adiponectin concentration, p < .001). Consistent with these results, inverse associations also were seen between adiponectin and cortical BMC (−4.8%) and cortical bone area (−4.7%), as measured by tibial pQCT at age 15.5 years (p < .001). Further pQCT results suggested that this inverse association of adiponectin with skeletal development predominantly involved a negative association with endosteal relative to periosteal expansion, as reflected by cortical thickness (−6.0%, p < .001). We conclude that, independent of fat mass, lean mass, and height, adiponectin is associated with lower bone mass in childhood predominantly owing to an influence on relative endosteal expansion. Since these associations were observed before and after puberty, this suggests that setting of adiponectin levels in midchildhood has the potential to exert long-term effects on bone strength and fracture risk. © 2010 American Society for Bone and Mineral Research.

Introduction

There is growing interest in the relationship between adiposity and bone mass, including possible influences on bone mass accrual in childhood. For example, in the Avon Longitudinal Study of Parents and Children (ALSPAC) cohort, we previously reported that fat mass at age 9.9 years is positively related both to bone mass measured concurrently and to subsequent gains in bone mass over the following 2 years.1 Equivalent associations were observed using genetic markers of obesity as instrumental variables, implying that the relationship we found between fat and bone represents a causal effect of fat mass as opposed to an effect of possible confounders such as socioeconomic status and physical activity.2 Although increased weight is expected to increase bone mass as a consequence of greater mechanical strain, fat mass also was related to bone mass of the upper limb, suggesting that systemic factors contribute to this association. Obesity is associated with insulin resistance and hyperinsulinemia, which might contribute to the associated increase in bone mass.3 Both insulin and insulin-like growth factor 1 (IGF-1) exert trophic effects on bone in laboratory studies,4, 5 and in clinical studies, insulin resistance is associated with both increased bone density6–9 and reduced fracture risk10 independent of body mass index (BMI).

Leptin and adiponectin, two adipocyte-derived hormones, also could influence skeletal metabolism directly. In addition to regulating appetite, leptin has been reported to directly stimulate osteoblast function, which might contribute to the positive effect of fat mass on bone mass accrual. Alternatively, leptin has been suggested to reduce bone mass via a hypothalamic relay involving the sympathetic nervous system.3 Adiponectin is associated with increased insulin sensitivity and improved glucose tolerance11 and, although produced exclusively by adipose tissue,12 is inversely related to fat mass.13 Several lines of evidence suggest that lower adiponectin levels also might contribute to higher bone mass in obese subjects because adiponectin is a negative regulator of bone mass. For example, transgenic mice with elevated levels of adiponectin but normal body weight were found to have impaired bone mass acquisition,14 whereas adiponectin knockout male mice have been reported to have increased trabecular bone.15 In clinical studies, adiponectin consistently has been found to be inversely related to bone mineral density (BMD) in men and women.16–18 For example, in 1735 nondiabetic women, a doubling of the adiponectin level was associated with 3.2% lower hip BMD, which largely attenuated after adjusting for obesity in pre- but not postmenopausal women.18

Adiponectin is thought to exert a similar influence on fat mass and insulin sensitivity in childhood as in adulthood.11 In a small study of 38 girls aged 12 to 18 years, an inverse association was observed between adiponectin and BMD.19 Taken together, these observations suggest that reduced adiponectin levels also might contribute to the positive association between fat mass and bone mass accrual in childhood. To examine this question further, in this study we investigated the relationship between adiponectin levels as measured in the ALSPAC cohort at age 9.9 years, fat mass, and indices of skeletal development. In particular, we wished to determine whether any inverse association between adiponectin and BMD reflects an influence on bone size or true bone density and whether any relationship between adiponectin and skeletal parameters is explained by coassociation with fat mass.

Methods

Study population

ALSPAC is a birth cohort investigating factors influencing the health, growth, and development of children. All pregnant women resident within a defined part of the former county of Avon in southwest England with an expected date of delivery between April 1991 and December 1992 were eligible for recruitment, of whom approximately14,000 women were enrolled20 (www.alspac.bristol.ac.uk). Written informed consent was provided by the mothers, and informed assent was obtained from the children at the time of each assessment. Ethical approval was obtained from the ALSPAC Law and Ethics Committee and relevant local ethics committees. Data in the ALSPAC are collected by self-completion postal questionnaires sent to parents, by linkage to computerized records, by abstraction from medical records, and by examination of the children at research clinics. All children with available data were included in the analyses.

Blood measurements

The main exposure for this study was circulating level of nonfasting adiponectin measured in blood samples collected at the age 9.9 years research clinic visit. Nonfasting blood samples were taken using standard procedures, with samples spun immediately and frozen at −80°C. The measurements were assayed in 2008 after a median of 7.5 years in storage with no previous freeze-thaw cycles during this period. Total plasma adiponectin concentrations were determined using ELISA (R&D Systems, Abingdon, UK), with the interassay coefficient of variation (CV) being 7%.21 Other circulating factors measured on blood samples collected at age 9.9 years, examined as possible confounding factors, included plasma lipids (ie, total cholesterol, triglycerides, and high-density lipoprotein cholesterol) measured by a modification of the standard Lipid Research Clinics Protocol using enzymatic reagents for lipid determinations,22 apolipoprotein (apo) A1 and apoB measured by immunoturbidimetric assays (Roche UK, Welwyn Garden City, UK), leptin measured by an in-house ELISA validated against commercial methods,23 high-sensitivity interleukin 6 (IL-6) measured by ELISA (R&D Systems, Abingdon, UK), and C-reactive protein (CRP) measured by automated particle-enhanced immunoturbidimetric assay (Roche UK). All assay CVs were less than 5%.

Outcomes: DXA and pQCT Variables

At the age 9.9 and 15.5 years research clinic visits, whole-body dual-energy X-ray absorptiometry (DXA) scans were performed using a Lunar Prodigy scanner (Madison, WI, USA) with pediatric scanning software. DXA measures comprised total-body-less-head (TBLH) bone mineral content (BMC, g), bone area (BA, cm2), and BMD (g/cm2). To provide a more complete adjustment of BMC for BA, area-adjusted BMC (aBMC, g) was derived by adjusting BMC for BA by linear regression. Further details of these measures, including reproducibility, are described elsewhere.24

At the age 15.5 years research clinic visits, peripheral quantitative computed tomography (pQCT) scans of the midtibia (ie, 50% site) also were performed using the XCT2000L (Stratec, Pforzheim, Germany), following which cortical bone area (CBA), cortical bone mineral content (BMCC), cortical bone mineral density (BMDC), periosteal circumference (PC), endosteal circumference (EC), and cortical thickness (CT) were recorded. A threshold routine was used for defining cortical bone, which specified a voxel with a density greater than 650 mg/cm3 as cortical bone. EC and PC were derived using a circular-ring model. Of the 4502 pQCT scans performed, 88 were excluded owing to major motion artifacts. CVs for pQCT scans, based on 139 subjects scanned a mean of 31 days apart, were 2.7%, 1.3%, and 2.9% for BMCC, BMDC, and BAC, respectively.

Other variables

Anthropometric parameters collected at the age of 9.9 and 15.5 years research clinic visits comprised standing height, measured to the last complete millimeter using the Harpenden stadiometer (Holtain Ltd., Crymych, UK), and weight, measured to the nearest 50 g using the Tanita body fat analyzer (Model TBF305, Arlington Heights, IL, USA). Whole-body DXA scans performed at each of these research clinic visits were used to provide measures of total fat and lean mass. Central and peripheral fat measures were derived based on trunk and upper and lower limb regional fat mass. Maternal socioeconomic status was recorded at 32 weeks of gestation by questionnaire. Pubertal stage was assessed using a Tanner stage questionnaire completed annually, based on pubic hair development in boys and pubic hair combined with breast development in girls.25 Vigorous physical activity assessed by questionnaire at age 9, since found previously to be related to BMD and fractures in ALSPAC,26 and extent of moderate and vigorous physical activity as assessed by actigraph accelerometer at age 11, since also found to be related to BMD in ALSPAC,27 were included as confounders in light of the report that physical activity is inversely related to adiponectin levels in children.28

Statistical analysis

Descriptive statistics are presented as means and SD and medians and lower and upper quartiles points. Linear regression analyses were performed; explanatory variables were loge transformed to ensure homoskedasticity of residual errors. Sex-specific minimally adjusted effects of exposures on outcomes are presented as per doubling (coefficients multiplied by log2) in adiponectin for a unit increase in the outcome and 95% confidence intervals (95% CI). p Values are presented with respect to the sex interaction. To explore the role of potential confounders, analyses were repeated following adjustment for a range of other variables, with a change in the beta coefficient of more than 10% following adjustment taken as a threshold indicating that confounding may be present. The association between change in bone outcomes between 9.9 and 15.5 years and adiponectin levels was assessed using mixed-effects regression, employing a random intercept at the individual level and an identity covariance matrix. The null hypothesis that there was no association between adiponectin levels at age 9 and change in bone parameters between 9.9 and 15.5 years was assessed by using a Wald test of the slope between age and adiponectin. Confounding variables were introduced as either time-invariant (adiponectin levels) or time-varying (height and fat and lean mass) fixed effects, and change in the magnitude of the slope was assessed. All missing data were treated as missing at random, and complete case analysis was performed in all analyses. Analyses were conducted using STATA 9.2 (Stata Corp., College Station, TX, USA).

Results

Adiponectin versus DXA bone measures at age 9.9 years

A total of 4927 children were identified at age 9.9 years with both DXA and adiponectin data, of whom the majority were prepubertal [81%, 16%, and 3% of boys in Tanner stages 1, 2, and 3 + , respectively; 57%, 33%, and 10% of girls in Tanner stages 1, 2, and 3 + , respectively (analyses based on a subset of children with available puberty information)]. Total-body fat mass was greater in girls, whereas lean mass was higher in boys (Table 1). Adiponectin levels were approximately 7% higher in girls. In univariate analyses at age 9 years, total-body fat mass, lean mass, and height were inversely associated with adiponectin concentration. Fat mass explained 3% of the variation in adiponectin (p < .0001), lean mass explained 3% of the variation in adiponectin (p < .0001), and height explained 1% of the variation (p < .0001). This relationship was stronger in girls than in boys for fat mass, lean mass, and height (p = .015, p < .0001, and p < .0001, respectively). Total-body BMC was slightly higher in boys, reflecting their greater bone size.

Table 1. Means, Standard Deviations, and 25th, 50th, and 75th Centiles for Variables Measured in Children Attending the Age 9.9 Years Research Clinic With Available Results for Adiponectin Concentration and Total-Body-Less-Head DXA (n = 2495 males and 2432 females)
 SexMean(SD)25th%50th%75th%
  1. Note: DXA results are shown for bone mineral content (BMC), bone area (BA), bone mineral density (BMD), and BMC adjusted for bone area by linear regression (aBMC).

Age (years)Male9.94(0.33)9.71,9.87,10.08
 Female9.94(0.32)9.73,9.87,10.08
Height (cm)Male140.01(6.20)136.00,140.00,144.00
 Female139.65(6.46)135.30,139.35,143.70
Weight (kg)Male34.26(6.92)29.40,32.80,37.40
 Female34.96(7.50)29.60,33.50,38.80
Fat mass (kg)Male7.12(4.68)3.89,5.55,8.96
 Female9.53(4.99)5.90,8.37,11.89
Lean mass (kg)Male25.59(2.98)23.54,25.43,27.48
 Female23.74(3.14)21.63,23.40,25.52
BMI (kg/m2)Male17.37(2.63)15.57,16.74,18.54
 Female17.81(2.92)15.74,17.23,19.29
Adiponectin (µg/mL)Male12676.4(5254.1)8973.611879.615465.6
 Female13510.7(5560.3)9462.812856.616794.5
BMC (g)Male907.67(173.24)784.42,888.89,1014.41
 Female883.85(187.10)758.03,861.06,989.40
BA (cm2)Male1150.59(153.44)1044.77,1138.68,1243.04
 Female1133.39(168.15)1019.50,1120.99,1229.31
BMD (g/cm2)Male0.784(0.053)0.747,0.781,0.820
 Female0.773(0.054)0.736,0.772,0.809
aBMC (g)Male896.35(40.51)869.63,895.40,921.78
 Female891.38(38.46)865.39,891.11,916.14

Associations between adiponectin and DXA bone measures at age 9.9 years were analyzed initially in regression models minimally adjusted for age and sex, which revealed that adiponectin was inversely associated with TBLH BMC, bone area, BMD, and aBMC (Table 2). For BMC, bone area, and BMD, effect sizes were approximately twofold greater in girls than in boys. For example, doubling of adiponectin was associated with a 7.6% and 14.6% lower BMC, a 6.3% and 13.1% lower bone area, and a 8.7% and 14.4% lower BMD in boys and girls, respectively [percentages represent coefficients in Table 2 divided by the expected range of the dependent variable (2 × IQR) in Table 1]. In contrast, there was no evidence of a sex interaction in the case of aBMC, which was associated with a 7.2% lower value per doubling of adiponectin in both boys and girls.

Table 2. Regression Coefficients for Associations Between Adiponectin Concentrations at Age 9.9 Years and Total-Body-Less-Head Bone Mineral Content (BMC), Bone Area (BA), Bone Mineral Density (BMD), and BMC Adjusted for Bone Area by Linear Regression (aBMC) at Age 9.9 Years in 2495 Boys and 2432 Girls
 SexUnadjustedFat adjustedFat, lean, and height adjusted
Beta95% CIp ValueSex int.Beta95% CIp ValueSex int.Beta95% CIp ValueSex int.
  1. Note: Results are shown for minimally adjusted model (adjusted for age of scan) and following additional adjustment for total-body fat mass and total-body fat mass, lean mass, and height. Beta represents change in each parameter per doubling in adiponectin concentration. p Values are shown for individual regressions and for sex interactions.

BMCM−34.92(−45.82, −24.02)≤.0010.001−30.12(−38.71, −21.53)≤.0010.488−15.16(−20.24, −10.07)≤.0010.467
F−67.63(−78.42, −56.84)≤.001 −34.43(−43.07, −25.78)≤.001 −12.48(−17.61, −7.35)≤.001 
M/F−49.50(−57.41, −41.59)≤.001 −32.26(−38.36, −26.17)≤.001 −13.83(−17.44, −10.21)≤.001 
BAM−25.12(−34.89, −15.35)≤.0010.001−20.68(−28.23, −13.13)≤.0010.523−9.20(−13.10, −5.30)≤.0010.193
F−54.83(−64.49, −45.16)≤.001 −24.17(−31.77, −16.58)≤.001 −5.52(−9.45, −1.58)≤.006 
M/F−38.39(−45.48, −31.30)≤.001 −22.42(−27.77, −17.06)≤.001 −7.37(−10.15, −4.60)≤.001 
BMDM−0.013(−0.016, −0.009)≤.0010.001−0.012(−0.014, −0.009)≤.0010.483−0.007(−0.009, −0.004)≤.0010.793
F−0.021(−0.024, −0.018)≤.001 −0.013(−0.016, −0.010)≤.001 −0.007(−0.010, −0.005)≤.001 
M/F−0.016(−0.019, −0.014)≤.001 −0.012(−0.014, −0.010)≤.001 −0.007(−0.009, −0.005)≤.001 
aBMCM−7.41(−9.89, −4.92)≤.0010.926−7.47(−9.95, −4.98)≤.0010.789−5.08(−7.53, −2.62)≤.0010.446
F−7.58(−10.04, −5.12)≤.001 −7.95(−10.45, −5.45)≤.001 −6.43(−8.91, −3.96)≤.001 
M/F−7.45(−9.20, −5.70)≤.001 −7.71(−9.47, −5.94)≤.001 −5.75(−7.50, −4.00)≤.001 

In light of the association between adiponectin and fat mass, analyses were repeated subsequently following additional adjustment for fat mass, following which the relationship between adiponectin and bone measures attenuated particularly in girls in the case of BMC, bone area, and BMD, with the result that regression coefficients for these associations now were similar in both sexes with no evidence of a sex interaction (Table 2). For example, a doubling of adiponectin now was associated with a 6.6% and 7.4% lower BMC, a 5.2% and 5.7% lower bone area, and 7.9% and 9.0% lower BMD in boys and girls, respectively. In contrast, aBMC showed no evidence of attenuation by adjustment for fat mass (7.5% decrease in aBMC for boys and girls combined). In our fully adjusted model, in which height and lean mass also were included, further attenuation was observed particularly for variables related to bone size (Table 2). For example, a doubling of adiponectin now was associated with a 3.0% lower BMC, a 1.8% lower bone area, a 4.8% lower BMD, and 5.6% lower aBMC in boys and girls combined.

When trunkal fat mass was used in place of total-body fat mass in regression models, if anything, attenuation of the association between adiponectin and TBLH BMC and bone area was slightly greater, whereas attenuation was slightly weaker for peripheral fat mass. For example, for associations between adiponectin and BMC, regression coefficients were −49.5, −32.3, −36.3, and −29.0 in models adjusted for minimal, total-body fat, peripheral fat, and trunk fat, respectively (beta coefficients for boys and girls combined). Regression coefficients were unchanged by further adjustment for maternal socioeconomic status, physical activity as assessed by questionnaire or accelerometry, leptin, CRP, IL-6, and lipid levels (results not shown). Results also were unaffected by confining analyses to the subgroup of prepubertal and early pubertal children (ie, Tanner stages 1 and 2; results not shown).

Adiponectin versus change in DXA bone measures between 9.9 and 15.5 years

We also analyzed the relationship between adiponectin levels at age 9.9 years and subsequent change in bone mass based on comparison of DXA results between 9.9 and 15.5 years of age in 2765 children with repeated measurements at these two time points. Adiponectin was inversely associated with change in BMC, bone area, BMD, and aBMC (Table 3). However, in contrast to cross-sectional analyses at age 9.9 years, these associations showed relatively little attenuation following adjustment for fat mass, lean mass, and height, and associations generally were stronger in boys than in girls across all models, with the exception of aBMC (Table 3). For example, in fully adjusted analyses, a doubling in adiponectin was associated with a 1.9% and 1.0% annual decrease in BMC, a 0.6% and 0.2% annual decrease in bone area, a 1.2% and 0.7% annual decrease in BMD, and a 1.9% and 2.0% annual decrease in aBMC in boys and girls, respectively.

Table 3. Mixed-Effects Regression Coefficients for Associations Between Adiponectin Concentrations at Age 9.9 Years and the Associated Change in Total-Body-Less-Head Bone Mineral Content (BMC), Bone Area (BA), Bone Mineral Density (BMD), and BMC Adjusted for Bone Area by Linear Regression (aBMC) Between the Ages 9.9 and 15.5 Years in 1295 Boys and 1380 Girls
 SexUnadjustedFat adjustedFat, lean, and height adjusted
Beta95% CIp ValueSex int.Beta95% CIp ValueSex int.Beta95% CIp ValueSex int.
  1. Note: Analyses are adjusted for fat mass, lean mass, and height as time-varying covariates. Beta represents change in bone parameter per year for a doubling in adiponectin concentrations at age 9.9 years. p Values are shown for individual regressions and for sex interactions.

BMCM−10.70(−15.27, −6.13)≤.0010.109−10.58(−15.12, −6.05)≤.0010.104−8.86(−11.76, −5.97)≤.0000.029
F−5.54(−9.87, −1.21)≤.012 −5.38(−9.71, −1.06)≤.015 −4.41(−7.16, −1.66)≤.002 
M/F−7.98(−11.12, −4.84)≤.001 −7.87(−11.00, −4.74)≤.001 −6.53(−8.52, −4.53)≤.000 
BAM−2.87(−5.45, −0.28)≤.0300.074−2.83(−5.37, −0.29)≤.0290.112−2.42(−3.90, −0.93)≤.0010.145
F0.38(−2.07, 2.83)≤.761 0.02(−2.40, 2.45)≤.986 −0.89(−2.31, 0.52)≤.214 
M/F−1.16(−2.93, 0.62)≤.202 −1.34(−3.10, 0.41)≤.134 −1.62(−2.64, −0.59)≤.002 
BMDM−0.0018(−0.0026, −0.0010)≤.0010.084−0.0017(−0.0026, −0.0009)≤.0010.160−0.0017(−0.0025, −0.001)≤.0000.159
F−0.0008(−0.0016, 0.00001)≤.053 −0.0009(−0.0017, −0.00009)≤.030 −0.0010(−0.0017, 0.000)≤.003 
M/F−0.0013(−0.0018, −0.0007)≤.001 −0.0013(−0.0019, −0.0007)≤.001 −0.0014(−0.0019, −0.001)≤.000 
aBMCM−3.07(−4.91, −1.24)≤.0010.832−2.84(−4.65, −1.03)≤.0020.970−1.98(−3.70, −0.26)≤.0240.949
F−2.80(−4.53, −1.06)≤.002 −2.89(−4.61, −1.17)≤.001 −2.05(−3.69, −0.42)≤.014 
M/F−2.93(−4.19, −1.67)≤.001 −2.87(−4.11, −1.62)≤.001 −2.02(−3.20, −0.83)≤.001 

Adiponectin versus pQCT bone measures at age 15.5 years

A total of 2769 children were identified with adiponectin levels at age 9.9 years and DXA and pQCT scans at the age 15.5 years research clinic visits, when the great majority of subjects were in late puberty (approximately 95% of boys and 97% of girls were in Tanner stage 4 or 5 based on analysis of the subset with available puberty information). As expected, at age 15.5 years, boys were taller than girls and had greater lean mass, whereas fat mass was considerably higher in girls (Table 4). BMCC, CBA, PC, and CT of the midtibia, as measured by pQCT, were higher in boys, whereas BMDC was higher in girls. In univariate analyses at age 15.5 years, the association between fat mass, lean mass, and adiponectin were much weaker, explaining 0.8% (p = .004) and 1.5% (p < .0001) of the variation, respectively. There was no evidence of a univariate association between height and adiponectin (p = .64). There also was no evidence of any difference between associations between boys and girls at age 15.5 years.

Table 4. Means, Standard Deviations, and 25th, 50th, and 75th Centiles for Variables Measured at Age 15.5 Years in Subjects With Available Results for Adiponectin Concentration and DXA and Tibial pQCT Scans (1332 Boys and 1422 Girls)
 SexMean(SD)25th%50th%75th%
  1. Note: Tibial pQCT results comprise cortical bone mineral content (BMCC), cortical bone area (CBA), cortical bone mineral density (BMDC), periosteal circumference (PC), and cortical thickness (CT).

Age (years)Male15.46(0.25)15.3015.4115.54
 Female15.47(0.28)15.3015.4115.56
Height (cm)Male174.4(7.53)170.0174.5179.3
 Female164.8(6.13)160.8164.8168.8
Weight (kg)Male63.30(11.24)55.9561.7068.90
 Female58.79(10.15)52.0057.1563.90
Fat mass (kg)Male10.66(7.66)5.668.1912.60
 Female18.49(7.75)13.1217.0822.02
Lean mass (kg)Male49.72(6.54)45.7549.8654.01
 Female21.61(3.38)19.3621.0023.15
BMI (kg/m2)Male2218.25(447.14)1904.642211.662488.75
 Female1942.15(354.18)1705.291908.892131.15
BMCC (mg)Male352.9(52.73)316.7352.0387.0
 Female310.0(40.36)283.0309.5335.1
CBA (mm2)Male328.2(46.59)295.6328.6357.5
 Female275.8(35.96)251.8275.1297.8
BMDC (mg/cm3)Male1074.6(34.56)1054.11077.21099.9
 Female1124.3(22.85)1110.61125.91139.7
PC (mm)Male76.18(5.45)72.776.079.5
 Female69.59(4.80)66.569.372.6
CT (mm)Male5.62(0.70)5.185.636.08
 Female5.18(0.58)4.805.195.56

An inverse association was observed between adiponectin and BMCC, BMDC, CBA, PC, and CT as measured by tibial pQCT at age 15.5 years in minimally adjusted analyses (Table 5). For example, a doubling of adiponectin was associated with a 9.9%, 9.8%, 2.6%, 5.8%, and 9.8% lower BMCC, CBA, BMDC, PC, and CT, respectively (boys and girls combined). Whereas adjustment for fat had relatively little effect on the results, attenuation was observed in the fully adjusted model, in which a doubling in adiponectin was associated with a 4.8%, 4.7%, 1.7%, 1.6%, and 6.0% lower BMCC, CBA, BMDC, PC, and CT, respectively (boys and girls combined). There also was some evidence of a gender interaction in that inverse associations between adiponectin and BMDC appeared to be stronger in girls (0.8% higher and 5.9% lower per doubling of adiponectin), whereas the opposite held for associations with PC (3.0% and 0.2% lower per doubling of adiponectin) (results for boys and girls, respectively; fully adjusted model). Equivalent results were obtained when analyses were restricted to children in late puberty (ie, Tanner stages 4 and 5; results not shown).

Table 5. Regression Coefficients for Associations Between Adiponectin Concentrations at Age 9.9 Years and Cortical Bone Mineral Content (BMCC), Cortical Bone Area (CBA), Cortical Bone Mineral Density (BMDC), Periosteal Circumference (PC), and Cortical Thickness (CT) as Measured at Age 15.5 Years in 1322 Boys and 1422 Girls
 SexUnadjustedFat adjustedFat, lean, and height adjusted
Beta95% CIp ValueSex int.Beta95% CIp ValueSex int.Beta95% CIp ValueSex int.
  1. Note: Results are shown for minimally adjusted model (adjusted for age of scan) and following additional adjustment for total-body fat mass and total-body fat mass, lean mass, and height. Beta represents change in each parameter per doubling in adiponectin concentration. p Values are shown for individual regressions and for sex interactions.

BMCCM−12.91(−16.90, −8.92)≤.0010.647−12.12(−15.85, −8.39)≤.0010.256−5.87(−8.64, −3.10)≤.0010.972
F−11.62(−15.39, −7.86)≤.001 −9.15(−12.68, −5.62)≤.001 −5.94(−8.55, −3.33)≤.001 
M/F−12.23(−14.97, −9.49)≤.001 −10.55(−13.12, −7.99)≤.001 −5.91(−7.81, −4.01)≤.001 
CBAM−11.63(−15.18, −8.09)≤.0010.426−10.93(−14.23, −7.62)≤.0010.133−5.69(−8.19, −3.19)≤.0010.486
F−9.65(−13.00, −6.31)≤.001 −7.43(−10.56, −4.30)≤.001 −4.47(−6.83, −2.11)≤.001 
M/F−10.59(−13.02, −8.15)≤.001 −9.08(−11.36, −6.81)≤.001 −5.04(−6.76, −3.33)≤.001 
BMDCM−0.85(−3.36, 1.65)≤.5040.227−0.76(−3.27, 1.74)≤.5500.2021.14(−1.33, 3.61)≤.3640.116
F−2.97(−5.33, −0.61)≤.014 −3.01(−5.38, −0.64)≤.013 −3.41(−5.74, −1.08)≤.004 
M/F−1.97(−3.69, −0.26)≤.024 −1.95(−3.67, −0.23)≤.027 −1.27(−2.96, 0.43)≤.144 
PCM−0.92(−1.36, −0.48)≤.0010.310−0.83(−1.25, −0.42)≤.0010.089−0.41(−0.75, −0.07)≤.0190.116
F−0.60(−1.02, −0.19)≤.005 −0.34(−0.73, 0.06)≤.092 −0.03(−0.36, 0.29)≤.841 
M/F−0.75(−1.06, −0.45)≤.001 −0.57(−0.86, −0.29)≤.001 −0.21(−0.45, 0.02)≤.079 
CTM−0.15(−0.21, −0.10)≤.0010.557−0.15(−0.20, −0.09)≤.0010.894−0.07(−0.12, −0.03)≤.0030.256
F−0.17(−0.22, −0.12)≤.001 −0.15(−0.20, −0.10)≤.001 −0.11(−0.16, −0.07)≤.001 
M/F−0.16(−0.20, −0.13)≤.001 −0.15(−0.18, −0.11)≤.001 −0.10(−0.13, −0.06)≤.001 

Discussion

We have found that circulating adiponectin levels are inversely related to total-body-less-head bone mass in a large population-based cohort of children at a mean age of 9.9 years. Bone mass as measured by total-body DXA reflects “volumetric” bone density as well as bone size. While true volumetric density cannot be measured directly on DXA scans, aBMC represents a reasonable estimate because, unlike conventional “areal” BMD, aBMC is fully adjusted for bone size. Therefore, the inverse association that we found between adiponectin and total-body aBMC suggests that effects on volumetric density contribute to the inverse association between adiponectin and bone mass accrual that we observed. This was particularly the case for adjusted models, in which associations with aBMC showed less evidence of attenuation in comparison with variables that predominantly reflect overall bone size, such as total-body bone area.

pQCT scans performed at age 15.5 years were able to examine the basis for this apparent influence of adiponectin on volumetric bone density as estimated by DXA by virtue of the more detailed information provided about cortical bone. A relatively strong inverse association was observed between adiponectin and cortical thickness that we assume contributes to the relationship with aBMC as measured by DXA. Since cortical thickness represents the net effect of endosteal relative to periosteal expansion, our findings imply that adiponectin affects bone mass accrual in childhood by influencing activity at the endosteum. Consistent with this conclusion, in further analyses, adiponectin was found to be positively related to endosteal circumference after adjustment for periosteal circumference, which relationship almost exactly mirrored that with cortical thickness (results not shown).

Adiponectin also was inversely related to periosteal circumference in unadjusted models, suggesting that inhibition of periosteal growth contributes to this relationship between adiponectin and cortical thickness. However, in contrast to the association with cortical thickness, that between adiponectin and periosteal circumference was largely attenuated in our fully adjusted model, implying that independent effects of adiopnectin are predominantly mediated through those on endosteal expansion. In girls, association between adiponectin and BMDc also may have contributed to that with aBMC because adiponectin was inversely related to BMDC in all models, although an equivalent relationship was not seen in boys. An effect of adiponectin on trabecular bone also may have contributed to the relationship with aBMC, as suggested by our finding that adiponectin was inversely related to aBMC as measured at the spine subregion, which has a relatively high content of trabecular bone (results not shown).

Our observation that adiponectin levels are inversely related to gain in BMC, bone area, and areal BMD between the ages of 9.9 and 15.5 years more strongly in boys than in girls raises the possibility that puberty partly attenuates the influence of adiponectin on bone development, particularly in girls. The report that the relationship between adiponectin and areal BMD in 1735 women was considerably weaker in premenopausal versus postmenopausal subjects is consistent with the suggestion that some of the effects of adiponectin on the skeleton are partly reduced by estrogen.18 Rising estrogen levels are known to influence cortical bone development in pubertal girls,29 although the basis for any interaction with adiponectin is currently unknown. However, since puberty did not entirely abolish the relationship between adiponectin and gain in bone mass in this study, the lack of association between adiponectin and areal BMD in premenopausal women observed by Richards and colleagues18 seems inconsistent with our results.

One possible explanation for this apparent discrepancy is that adiponectin levels as measured at age 9.9 years are not themselves directly responsible for the effect on skeletal development but are a marker of a separate causative factor. We attempted to examine this question by adjusting for other factors potentially related to both adiponectin and skeletal development, such as physical activity levels and levels of leptin, IL-6, and CRP, but this had little effect on the relationship that we observed with adiponectin. Adiponectin is also related to other factors involved in peripheral insulin sensitivity that may exert trophic effects on bone, such as IGF-1 and insulin, which were not analyzed in this study. Another explanation for this inconsistency is that although no association was observed between adiponectin and BMD in premenopausal women in adjusted analyses, Richards and colleagues found a similar relationship in unadjusted analyses to that reported here. Adiponectin is inversely related to fat mass, and fat mass is associated with similar parameters to adiponectin, such as bone size and periosteal circumference.1, 30 However, although it is therefore important to adjust analyses for fat mass, the strategy used in these two studies differed because whereas we adjusted for total-body fat mass, lean mass, and height, Richards and colleagues adjusted for a range of different factors, including fasting insulin level, BMI, and central fat, which may have led to overadjustment because central fat is a component of the denominator of BMI.

Since fat mass is higher in girls than in boys at age 9.9 years, and since adiponectin is related to fat mass more strongly in girls than in boys, this may explain the sex differences observed in unadjusted analyses between adiponectin and total-body bone area and BMC at this age. Our finding that adjustment for total-body fat mass attenuated these sex differences suggests that this strategy was largely successful in removing any influence of coassociation with fat mass on the relationship between adiponectin and bone measures. Although adiponectin has been reported previously to be associated with visceral as opposed to subcutaneous fat mass,17 in sensitivity analyses, where we compared the use of total, peripheral, and trunk fat as adjustment factors, broadly similar results were obtained. In terms of other approaches to evaluating the contribution of adiponectin independent of fat mass, preliminary analyses suggested that Mendelian randomization techniques,31 as applied recently to examine fat-bone relationships,2 may lack sufficient power owing to the smaller sample size and the weaker association between bone variables and adiponectin compared with fat mass.

One unexpected observation was our finding that adiponectin levels are inversely related to lean mass as well as fat mass and that adjustment for lean mass led to attenuation of some of the associations that we found. Whereas adiponectin levels are known to be inversely related to weight and BMI, this is generally assumed to reflect a relationship between adiponectin and fat mass,11 and we are not aware of any previous reports of a direct association between adiponectin and lean mass. It is well recognized that lean mass and bone mass are closely related, which, based on our recent study, appears to reflect a combined influence of lean mass on both periosteal growth and relative endosteal expansion.30 Therefore, our observation that associations between adiponectin levels and pQCT parameters such as periosteal circumference and cortical thickness were attenuated by adjustment for lean mass is consistent with current understanding of the influence of lean mass on bone development.

In terms of limitations to this study, adiponectin exists as high-, medium-, and low-molecular-weight isoforms, which may have distinct functions. For example, mutations that inhibit the formation of high-molecular-weight adiponectin lead to type 2 diabetes, suggesting that this isoform may be a better predictor of insulin resistance and diabetes than total adiponectin, as reported recently.32 Since the assay used in this study measured total adiponectin levels, we were unable to determine the relative contribution of different isoforms to the associations with the parameters related to skeletal development that we observed. A further limitation to this study was the use of self-assessment by questionnaire to evaluate pubertal stage as opposed to evaluation by a trained examiner.

In conclusion, we have found that adiponectin levels are inversely related to bone mass accrual in childhood, which pQCT analyses suggest is likely to reflect effects on periosteal growth, endosteal expansion, and BMDC. Coassociation of adiponectin with fat mass and lean mass contributed to this inverse relationship, particularly with respect to parameters related to overall bone size, such that in our fully adjusted model adiponectin was most strongly related to aBMC as measured by DXA at age 9.9 years and to cortical thickness as measured by pQCT at age 15.5 years. Taken together, these findings suggest that setting of the adiponectin axis in midchildhood may relate to subsequent bone development, particularly such aspects of skeletal architecture as cortical thickness, and may have long-term implications for skeletal strength and fracture risk.

Disclosures

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

Acknowledgements

We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists, and nurses. The UK Medical Research Council, the Wellcome Trust, and the University of Bristol provide core support for ALSPAC. Salary support for AS was provided by Wellcome Trust Grant 079960, which also funded the pQCT scans. ADH is supported by a British Heart Foundation Senior Research Fellowship (FS05/125). This publication is the work of the authors, who serve as guarantors for the contents of this article.

Ancillary