Relationships Between Serum Adiponectin and Bone Density, Adiposity and Calcified Atherosclerotic Plaque in the African American-Diabetes Heart Study

Abstract

Context:

Adiposity, bone mineral density (BMD), and calcified atherosclerotic plaque (CP) exhibit complex interrelationships that are not well understood. Adipokines vary in relation to changes in body composition and may play roles in regulation of BMD and risk of cardiovascular disease.

Objective:

Our objective was to examine the relationship between serum adiponectin and quantitative computed tomography-derived measures of volumetric BMD (vBMD) in thoracic and lumbar vertebrae, adipose tissue volumes, and CP in coronary, carotid, and infrarenal aortoiliac arteries. Generalized linear models were fitted to test for associations between adiponectin and measured phenotypes.

Participants:

A total of 479 unrelated African Americans with type 2 diabetes, 57% female with a mean ± SD (median) age of 55.6 ± 9.5 (55.0) years and diabetes duration of 10.3 ± 8.2 (8.0) years.

Results:

Serum adiponectin was 8.26 ± 7.41 (6.10) μg/mL, coronary artery CP mass score was 280 ± 634 (14), carotid artery CP was 47 ± 133 (0), and aortoiliac CP was 1616 ± 2864 (319). Women had significantly higher body mass index and serum adiponectin and lower coronary and carotid artery calcium than males (all P < .05). Before and after adjusting for age, sex, body mass index, mean arterial pressure, smoking status, hemoglobin A_1c, thiazolidinedione use, and low-density lipoprotein-cholesterol, adiponectin was inversely associated with thoracic and lumbar vertebral vBMD [parameter estimates (PEs) of −0.06 and −0.021, respectively; both P < .0005], visceral adipose tissue (PE −0.02; P < 0.0001), and C-reactive protein (PE −0.07; P < .0001) and positively associated with intermuscular adipose tissue (PE 0.01; P = .03). After covariate adjustment, significant associations were not observed between adiponectin and CP in any vascular bed (P > .1).

Conclusion:

Serum adiponectin levels were inversely associated with cross-sectional measures of thoracic and lumbar vertebral vBMD, inflammation, and visceral adiposity in African Americans but not with vascular CP after adjustment for covariates. The data support a regulatory/signaling role for adiponectin in the modulation of bone density.

Adiponectin is an adipose-specific cytokine (adipocytokine), which has effects on glucose metabolism and insulin sensitivity, lipid metabolism, inflammation, and other processes (1). Despite its adipose tissue origin, circulating adiponectin concentrations are inversely associated with adiposity and body mass index (BMI), decreasing with weight gain and increasing with weight loss (2–5).

Our group was the first to describe inverse relationships between serum adiponectin and bone mineral density (BMD), which was assessed by both dual-energy X-ray absorptiometry and quantitative computed tomography (QCT) in a sample of 80 subjects enriched for type 2 diabetes (T2D), 61 European Americans (EAs) and 19 African Americans (AAs) (6). In that report, serum adiponectin was inversely associated with areal BMD measured by dual-energy X-ray absorptiometry in the spine, radius, and hip, with volumetric BMD (vBMD) measured by QCT in the thoracic and lumbar spine and with visceral fat volume measured by QCT, before and after adjustment for age, sex, race, smoking, and diabetes status. These associations remained significant after adjusting for whole-body fat mass, and relationships with vBMD were significant after adjusting for visceral fat volume. No significant associations of adiponectin were seen with sc fat volume, whole-body fat mass, and serum leptin levels (P > .1). Subsequent studies demonstrated these inverse associations with vBMD in predominantly European-derived individuals (7–9).

Many of the properties of adiponectin suggest the potential for cardiovascular benefit, including insulin sensitization, induction of nitric oxide production to limit obesity-induced endothelial dysfunction, reduction of endothelial cell activation, inhibition of generation of reactive oxygen species and apoptosis, and promotion of endothelial cell repair (10). An in vitro study suggested that adiponectin can inhibit the transformation of vascular smooth muscle cells into an osteoblastogenic phenotype, implying that adiponectin may reduce vascular calcification (11).

We have also observed inverse relationships between vBMD and calcified atherosclerotic plaque (CP) in EAs (12, 13). Nevertheless, relationships between adiponectin, vBMD, and subclinical cardiovascular disease (CVD) are poorly defined in AAs. This report assessed adiponectin concentrations in a larger cohort of well-characterized AAs with T2D from the African American-Diabetes Heart Study (AA-DHS), determining cross-sectional relationships with CP, vBMD, adipose tissue volumes, renal parameters, and systemic inflammation (14).

Materials and Methods

Study population

Self-reported and unrelated AAs with T2D were recruited from internal medicine clinics and community advertising in the AA-DHS (13). Participant examinations were conducted in the Clinical Research Unit of the Wake Forest Baptist Medical Center and included interviews for medical history and health behaviors, anthropometric measures, resting blood pressure (BP), 12-lead electrocardiography, fasting blood sampling, and spot urine collection. T2D was defined as a diagnosis of diabetes after 30 years of age, in the absence of historical evidence of diabetic ketoacidosis. History of CVD was provided by participant report and medical record review. Although histories of myocardial infarction or stroke were not exclusion criteria, CP scores in the coronary arteries were excluded in participants who underwent prior coronary artery bypass grafting and in the carotid arteries in participants who underwent carotid endarterectomy. Hypertension was based on a physician diagnosis or if coded in medical records, BP greater than 140/90 mm Hg, or use of antihypertensive medications. Laboratory assays included spot urine albumin and creatinine to calculate the urine albumin to creatinine ratio (ACR), total plasma cholesterol, low-density lipoprotein (LDL) cholesterol, high-density lipoprotein cholesterol, triglycerides, hemoglobin A_1c (HbA_1c), fasting glucose, and blood chemistries. Renal function was assessed by serum creatinine concentration and estimation of glomerular filtration rate (GFR) using the 4-variable Modification of Diet in Renal Disease Study equation (15). Serum creatinine concentration was measured using a modified kinetic Jaffe method and corrected for interlaboratory differences and calibrated to the Cleveland Clinic. All medications were recorded, including hormone therapies, bisphosphonates, angiotensin-converting enzyme inhibitors and angiotensin receptor blockers, β-blockers, statins, fibric acid derivatives, thiazide diuretics, and thiazolidinediones (TZDs). The study was approved by the Institutional Review Board at the Wake Forest School of Medicine, and all participants provided written informed consent.

Adiponectin assay

Serum adiponectin was measured in freshly thawed serum samples that had been stored at −80°C since collection using RIA kits and protocols from Millipore (Linco Research based assay; St Charles, Missouri). Intra- and interassay coefficients of variation were less than 10%.

Vascular imaging

CP in the coronary arteries (Cor), carotid arteries (Car), and infrarenal aortoiliac (Aortoiliac) was determined using multidetector computed tomography (CT) with cardiac gating and capable of 500-millisecond temporal resolution using the segmented reconstruction algorithm (LightSpeed Qxi; General Electric Medical Systems, Waukesha, Wisconsin). Techniques for the coronary and carotid scans have been described in detail (13). In brief, participants were placed in the supine position on the CT couch over a quality control calibration phantom (Image Analysis, Inc, Columbia, Kentucky) for scans of the heart and abdomen. The abdomen scan series was used to measure aortoiliac CP. Technical factors for this series were as follows: 120 kV, 250 mA, 0.8-second gantry rotation helical mode (7.5 mm/s), 2.5-mm slice thickness, and standard reconstruction kernel. The display field of view was 35 cm, resulting in a pixel dimension of 0.68 × 0.68 mm. CT scans of the 3 vascular territories (Cor, Car, and Aortoiliac) were analyzed on a GE Advantage Windows Workstation with the SmartScores software package (General Electric Medical Systems) using a modified Agatston scoring method, which adjusts for slice thickness and uses the conventional threshold of 130 Hounsfield units.

BMD Measurement

The QCT for trabecular vBMD (milligrams per cubic centimeter) in the thoracic spine (T8-T11) and lumbar spine (T12-L3) was measured using QCT-5000 volumetric software (Image Analysis, Columbia, Kentucky) with an external calibration phantom from the same CT images used to measure CP in the coronary and aortoiliac arteries, respectively. Coefficients of variation for these measures have previously been reported and were less than 1% for thoracic and lumbar BMD, and in sequential studies performed in the same individual, the precision error was 2.3% (13).

Adipose tissue imaging

Pericardial adipose tissue (PAT) and visceral adipose tissue (VAT) were measured from volumetric CT acquisitions to reduce variability related to slice location using the Volume Analysis software (Advantage Windows Workstation, GE Healthcare, Waukesha, Wisconsin) and a threshold of −190 to −30 as the definitions of adipose tissue containing tissue. PAT is the combined adipose tissue superficial (paracardial) and deep (epicardial) to the pericardium (16); however, the pericardium extends superiorly to encase the great vessels and inferiorly borders the diaphragm. Our methods for measuring PAT segments a volume for measurement that covers 45 mm in length along the z-axis (cephalocaudad) of the individual, based on origin of the left main coronary such that it extends 15 mm above and 30 mm below. This PAT volume includes most the coronary arteries and myocardium and excludes PAT located superiorly around the aorta and pulmonary arteries and adjacent to the abdomen (13).

In the abdomen, VAT, sc adipose tissue (SAT), and intermuscular adipose tissue (IMAT) were measured on abdominal CT scans with the following technical factors: helical mode, 120 kVp, 250 mA, 4 × 2.5 mm collimation, a standard reconstruction kernel, and a display field of view of 500 mm. The landmark for the analysis was the first lumbar disk above the lumbar-sacrum junction, most commonly designated as L4-L5. A volume of 15 mm in the z-axis length of the abdomen was segmented for the sc, abdominal wall and intraabdominal compartments. VAT was defined as the adipose tissue containing pixels located within the abdominal cavity, SAT was defined as the adipose tissue containing pixels between the skin surface and lean tissue of the abdominal wall, and IMAT was measured within the abdominal wall and paraspinal muscles.

Statistical methods

Generalized linear models were fitted to test for associations between serum adiponectin concentration and BMD, adipose tissue volumes, renal parameters, Cor CP, Car CP, and Aortoiliac CP. Adiponectin values greater than 21.1 μg/mL (corresponding to the 95th percentile in the distribution) were Winsorized to 21.1. The Box-Cox method was applied to identify the appropriate transformation best approximating the distributional assumptions of conditional normality and homogeneity of variance of the residuals (17). This method suggested taking the natural logarithm of (Cor CP + 1), (Car CP + 1), (Aortoiliac CP + 1), PAT, VAT, SAT, and IMAT and the square root of lumbar and thoracic vBMD. Generalized linear models were fitted using the Winsorized values of adiponectin as the predictor. Preliminary analyses assessing association between β-blockers, fibric acid derivatives, statins, thiazide diuretics, and TZDs on vBMD revealed that only TZDs were associated with reduced lumbar vBMD (P = .02), with a trend toward reduced thoracic vBMD (P = .07). Therefore, after an unadjusted analysis and an age-, sex-, and BMI-adjusted analysis, the final analysis adjusted for age, sex, BMI, LDL, mean arterial pressure (MAP), HbA_1c, smoking status, and TZD status. Adjustment for medication status was achieved by including an indicator variable if the participant was taking a TZD.

Results

The study sample consisted of 479 unrelated AAs with T2D. Mean ± SD (median) age was 55.6 ± 9.5 (55.0) years and diabetes duration 10.3 ± 8.2 (8.0) years, and 57% were female and 50.7% hypertensive (Table 1). The mean ± SD (median) levels of serum adiponectin were 8.26 ± 7.4 (6.1) μg/mL, estimated GFR was 95.2 ± 27.2 (93.3) mL/min per 1.73m², and the urine ACR was 151 ± 568 (14) mg/g. Cor CP, Car CP, and Aortoiliac CP was present in 62.7%, 48.5%, and 77.9% of participants, respectively (Table 2). Because the distributions of CP were skewed, mean values should be interpreted with caution, and values were log transformed for analysis; median CP scores were 13.5 for Cor CP, 0 for Car CP, and 319 for Aortoiliac CP. Sixty-four percent (n = 308) of participants met criteria for metabolic syndrome defined in the third report of the National Cholesterol Education Program Expert Panel on Detection, Evaluation, and Treatment in Adults, although applying this definition in AAs may be problematic (18). Medications taken by these participants included β-blockers (n = 112; 23.4%), fibrates (n = 7; 1.5%), statins (n = 245; 51.2%), thiazide diuretics (n = 130; 27.1%), and TZDs (n = 64; 13.4%). The mean (SD) median number of alcoholic drinks per month were 16.3 (27.5) 4.33; however, alcohol intake did not have a significant impact on the association results (data not shown).

Table 1.

Demographic Characteristics of the Study Population

Variable	Females (n = 272)			Males (n = 207)			All (n = 479)			P Value
	Mean	SD	Median	Mean	SD	Median	Mean	SD	Median
Age, y	55.5	9.2	55	55.7	10.0	55	55.6	9.5	55	.8393
Age diabetes onset, y	45.3	9.5	46	45.9	10.1	46	45.5	9.8	46	.4672
Diabetes duration, y	10.2	7.6	8	10.4	8.9	8	10.3	8.2	8	.5751
BMI, kg/m2	37.5	9.0	36.51	32.9	7.5	31.61	35.5	8.7	34	<.0001
Systolic BP, mm Hg	133.5	19.6	131	132.8	18.0	132	133.2	18.9	132	.8171
Diastolic BP, mm Hg	76.6	11.4	76.75	79.0	11.0	79	77.6	11.3	77	.0216
Hypertension, %	50.7			50.7			50.7			.9982
Lipid medications, %	52.1			51.0			51.6			.8206
Current smoker, %	20.6			27.1			23.8			.0024
Past smoker, %	31.3			40.6			35.3
ACE inhibitor use, %	37.5			43.0			39.9			.2237
Angiotensin receptor blocker use, %	17.9			12.1			14.8			.1402
Insulin use, %	38.6			42.5			40.3			.3875
Hormone replacement therapy, %	27.2			NA			27.2			NA

Abbreviation: ACE, angiotensin-converting enzyme; NA, not available.

Table 2.

Laboratory and Imaging Variables Used in Analysis

Variable	Females (n = 272)			Males (n = 207)			All (n = 479)			P Value
	Mean	SD	Median	Mean	SD	Median	Mean	SD	Median
Adiponectin, μg/mL	8.6	7.8	6.5	7.9	6.9	5.8	8.3	7.4	6.1	.0148
Coronary calcium mass, mg	218.1	601.6	4.5	362.5	666.7	42.75	280.1	633.8	13.5	.0015
Coronary calcium mass > 0, %	59.4			67.2			62.7			.0839
Carotid calcium mass, mg	36.5	103.9	0	61.2	162.8	1.5	47.1	132.9	0	.0277
Carotid calcium mass > 0, %	45.2			52.9			48.5			.0535
Aortoiliac calcium mass, mg	1461.6	2472.0	233	1821.9	3311.0	368.5	1616.1	2864.0	319	.0883
Aortoiliac calcium mass > 0, %	74.7			82.2			77.9			.0535
Serum creatinine, mg/dL	0.9	0.3	0.8	1.1	0.3	1.1	1.0	0.3	0.9	<.0001
Urine ACR, mg/g	156.3	630.0	11	144.0	479.6	21	151.0	568.9	14	.0235
GFR, ml/min	94.0	28.0	93.71	96.8	26.0	92.59	95.2	27.2	93.34	.2425
Fasting glucose, mg/dL	145.2	62.3	129	158.8	70.9	139	151.1	66.4	134	.0283
HbA1c, %	8.0	2.0	7.5	8.2	1.9	7.8	8.1	2.0	7.7	.291
Total cholesterol, mg/dL	185.0	45.4	176.5	175.7	50.0	169	181.0	47.6	174	.0081
HDL cholesterol, mg/dL	50.2	14.1	48	44.6	12.3	43	47.8	13.6	46	<.0001
LDL cholesterol, mg/dL	110.4	39.0	101	103.5	38.0	101	107.4	38.7	101	.0404
Triglycerides, mg/dL	124.0	101.0	99.5	141.4	175.1	107	131.5	138.1	102	.2677
Intermuscular adipose, cm3 per 15 mm	12.2	9.4	9.69	9.2	6.3	7.89	10.9	8.3	8.87	<.0001
Pericardial adipose, cm3 per 45 mm	84.8	33.7	81.11	95.5	46.9	86.9	89.4	40.2	82.677	.0484
Subcutaneous adipose, cm3 per 15 mm	510.1	166.7	486.99	352.7	162.2	326.25	442.6	182.1	427.63	<.0001
Visceral adipose, cm3 per 15 mm	177.8	65.0	168.52	180.1	82.2	169.81	178.8	72.8	169.28	.9178
Thoracic vBMD, mg/cm3	208.5	54.9	206.01	202.7	49.5	200.03	206.0	52.7	204.96	.3144
Lumbar vBMD, mg/cm3	180.6	48.9	177.06	179.6	44.1	180	180.1	46.9	179	.8675

Abbreviation: HDL, high-density lipoprotein.

In unadjusted analyses, serum adiponectin levels showed positive association with Log(Cor CP + 1), Log(Aortoiliac CP + 1), Log(Car CP + 1) and Log(IMAT) and an inverse association with the Log(VAT), and the square root of the thoracic vertebral vBMD and the lumbar vertebral vBMD (Table 3). Adjusted models including demographic characteristics (age, sex, BMI, mean arterial BP), HbA_1c, smoking, TZDs, and LDL cholesterol maintained significant evidence of an inverse association between adiponectin and the square root of the thoracic vertebral vBMD [parameter estimate (PE) = −0.06, P = .0005], the square root of the lumbar vertebral vBMD (PE = −0.02, P < .0001), Log(VAT) (PE = −0.02, P < .0001), Log[C-reactive protein (CRP)] (PE = −0.07, P < .0001) and a positive association with Log(IMAT) (PE = 0.01, P = .03). Significant associations were not observed between adiponectin and Log(Cor CP + 1), Log(Aortoiliac CP + 1), Log(Car CP + 1), or Log(GFR) in adjusted models (Table 3).

Table 3.

Relationships Between Serum Adiponectin and Indices of Bone and Vascular Calcium, Fat Distribution, and Inflammation: Unadjusted and Fully Adjusted Adiponectin Association Analyses in AA-DHS

Outcome	Unadjusted Model			Fully Adjusted Modela
	Estimate	SE	P Value	Estimate	SE	P Value
Log(Cor CP + 1)	0.0752	0.0258	.0037	0.03	0.03	.3
Log(Car CP + 1)	0.059	0.0195	.0026	0.03	0.02	.1
Log(Aortoiliac CP + 1)	0.0829	0.0299	.0058	0.03	0.03	.3
Log(VAT)	−0.0174	0.0042	4.50E-05	−0.02	0.00	<.0001
Log(PAT)	−0.0059	0.0041	.2	−0.01	0.00	.07
Log(SAT)	−0.0075	0.0046	.1	0.00	0.00	.3
Log(IMAT)	0.0159	0.0059	.0069	0.01	0.01	.03
Sqrt(LvBMD)	−0.1005	0.0162	<.0001	−0.021	0.004	<.0001
Sqrt(TvBMD)	−0.1012	0.0167	<.0001	−0.06	0.02	.0005
Log(CRP)	−0.0376	0.0128	.0035	−0.07	0.02	<.0001

Abbreviation: Sqrt(LvBMD), square root of lumbar vertebral vBMD; Sqrt(TvBMD), square root of thoracic vertebral vBMD.

^aAdjusted for age, sex, BMI, LDL, MAP, smoking status, HBA_1c, and TZD medication status.

Tables 4 and 5 shows the relationships between adiponectin and variables in sex-stratified analyses. After adjusting for age, BMI, smoking status, LDL cholesterol, MAP, and HbA1c, inverse relationships were present in both men and women between adiponectin and the square root of the lumbar vertebral vBMD (men: PE = −0.071, P = .0012; women: PE = −0.053, P = .012), the square root of the thoracic vertebral vBMD (men: PE = −0.086, P = .0003; women: PE = −0.057, P = .011), and the Log(VAT) (men: PE = −0.019, P = .0021; women: PE = −0.012, P = .0119), whereas the Log(CRP) was associated in men only (men: PE = −0.039, P = .0455) (Table 5). Inclusion of TZDs in the model (Table 4) did not alter the relationships between adiponectin and Log(VAT) (men: PE = −0.02, P = .0009; women: PE = −0.02, P = .004), although the relationship between adiponectin and bone density remained significant only in the men (lumbar vBMD; men: PE = −0.08, P = .0014; women: PE = −0.03; P = .15; and thoracic vBMD, men: PE = −0.1, P = .0001; women: PE = −0.04, P = .12).

Table 4.

Relationships Between Serum Adiponectin and Indices of Bone and Vascular Calcium and Fat Distribution Stratified by Sex, Adjusted for Age, Gender, BMI, Smoking Status, LDL Cholesterol, MAP, HbA_1c, and TZD

Outcome	Males			Females
	Estimate	SE	P Value	Estimate	SE	P Value
Log(Cor CP + 1)	0.03	0.04	.3866	0.01	0.04	.8818
Log(Car CP + 1)	0.03	0.03	.2816	0.01	0.03	.6216
Log(Aortoiliac CP + 1)	0.04	0.04	.3412	0	0.04	.9341
Log(VAT)	−0.02	0.01	.0009	−0.02	0.01	.004
Log(PAT)	−0.01	0.01	.114	−0.005	0.01	.4097
Log(SAT)	−0.004	0.006	.5265	−0.004	0.004	.383
Log(IMAT)	0.01	0.01	.108	0.01	0.01	.1184
Sqrt(LvBMD)	−0.08	0.02	.0014	−0.03	0.02	.1541
Sqrt(TvBMD)	−0.1	0.03	.0001	−0.04	0.03	.1232
Log(CRP)	−0.04	0.02	.0594	0	0.02	.8334

Abbreviations: Sqrt(LvBMD), square root of lumbar vertebral vBMD; Sqrt(TvBMD), square root of thoracic vertebral vBMD.

Table 5.

Relationships Between Serum Adiponectin and Indices of Bone and Vascular Calcium and Fat Distribution Stratified by Sex, Adjusted for Age, BMI, Smoking Status, LDL Cholesterol, MAP, and HbA_1c

Dependent	Males			Females
	Estimate	SE	P Value	Estimate	SE	P Value
Log(Cor CP + 1)	0.031	0.035	.3723	0.050	0.031	.1112
Log(Car CP + 1)	0.022	0.027	.4191	0.021	0.025	.4085
Log(Aortoiliac CP + 1)	0.026	0.035	.457	0.026	0.033	.4299
Log(VAT)	−0.019	0.006	.0021	−0.012	0.005	.0119
Log(PAT)	−0.005	0.006	.3569	−0.003	0.005	.584
Log(SAT)	−0.004	0.005	.3892	−0.001	0.004	.75
Log(IMAT)	0.016	0.008	.0533	0.010	0.007	.1847
Sqrt(LvBMD)	−0.071	0.022	.0012	−0.053	0.021	.0114
Sqrt(TvBMD)	−0.086	0.023	.0003	−0.057	0.022	.0109
Log(CRP)	−0.039	0.019	.0455	−0.018	0.016	.2656

Abbreviations: Sqrt(LvBMD), square root of lumbar vertebral vBMD; Sqrt(TvBMD), square root of thoracic vertebral vBMD.

Discussion

This report characterized the relationships between serum adiponectin, adiposity, BMD, inflammation, and subclinical CVD in the understudied AA population with T2D. Adjusting for potential confounders, adiponectin levels had significant inverse associations with thoracic and lumbar vertebral vBMD, CRP, and the volume of visceral adipose tissue. Adiponectin was positively correlated with intermuscular adipose tissue volume in the trunk but not with CP in the coronary, carotid, or aortoiliac vascular beds or with the sc fat volume in AAs.

Several characteristics of adiponectin provide plausible avenues for modulation of bone metabolism (6). Adiponectin bears a marked structural similarity to TNFα family members; a notable example is receptor activator of nuclear factor-κB ligand, which is involved in the regulation of osteoclastogenesis. Adiponectin activates (19), as well as inhibits, nuclear factor-κB (20), a ubiquitous transcription factor critical for osteoclastogenesis, providing a mechanism by which adiponectin may affect bone. Adiponectin receptors are expressed in both osteoblasts and osteoclasts, and there is in vitro evidence suggesting that adiponectin may directly suppress osteogenesis (21). Recombinant adiponectin prevents adipogenesis in bone marrow-derived preadipocytes, suggesting that it can influence the marrow environment (22). Adiponectin is also structurally similar to collagenous repeat-containing sequence of 26-kDa protein, a novel protein discovered in a mouse embryonic fibroblast cell line, which is likely to be involved in osteogenesis (23). Another similar protein is type X collagen, and mutations in the COL10A1 gene are known to cause metaphyseal chondrodysplasia (type Schmid) (24, 25). Thus, although the precise effects of adiponectin on the skeleton are not known, several potential molecular mechanisms and pathways may be involved.

The skeleton is a dynamic structure. In response to changes in stress and strain, local remodeling of the skeleton occurs to build bone in appropriate overloaded microdomains to increase support and relieve the physical strain and associated biochemical signals from nearby strain-sensitive cells such as osteocytes and bone-lining cells. Appropriate proportional increases in BMD in response to load also accommodate those changes. In the case in which adiposity is suddenly increasing, there may be a need for more rapid responses of the skeleton to future increases in load, which may be advantageous in the prevention of overload-based fracture. Adipose tissue adds load without adding support to the body and thus requires both the skeleton and the skeletal muscle to increase in their ability to support the structural demands. We hypothesize that normal adiponectin levels during times of normal body weight may appropriately keep the skeleton in balance with related loads. Furthermore, we hypothesize that during times of weight gain, associated decreases in circulating adiponectin may poise the skeleton toward an osteogenic phenotype, which would be important in preparing the skeleton for the increased load to come in the future. Additional signals from muscle may play a role in the impact of gains in fat and muscle mass in response to load demands.

Although adiponectin was inversely associated with BMD, and BMD and CP were inversely associated with one another in previous studies (12, 13), there was no significant association of adiponectin with CP at any site after adjustment for covariates, suggesting that serum adiponectin may not influence the process of vascular calcification.

The AA-DHS has several important strengths and some limitations. This relatively large and well-phenotyped AA sample enabled the simultaneous evaluation of circulating inflammation-related proteins and adipokines, which may play roles in the communication between adipose tissue, the skeleton, and arterial calcification. The AA-DHS allowed us to provide sensitive molecular characterization in AAs, a population group known to display different patterns of bone disease and subclinical CVD, relative to EAs (13). An important strength in this report is the relatively preserved renal function in participants, removing the potential consequences of altered serum adiponectin levels due to advanced kidney disease, whether mediated by poor renal excretion or inflammation-driven overproduction.

In addition, adjustment was performed for TZDs because this medication class was associated with reduced vBMD, whereas β-blockers, thiazides, statins, and fibrates did not impact vBMD (data not shown). TZDs act as agonists for peroxisomal proliferator-activated receptor-γ, an important transcription factor in adipocyte development and metabolism. TZDs are known to promote differentiation of bone marrow mesenchymal stem cells toward adipogenesis and away from osteoblastogenesis and to have adverse effects on the skeleton, reducing bone density and increasing fracture risk (26). TZDs also are known to significantly increase serum adiponectin concentrations (27). The mechanistic explanation for the finding that inclusion of TZD use in the analytical models diminished relationships between adiponectin and BMD in women but not men is not clear, although this could relate to the impact of TZDs on both variables and complex interactions between estrogens and TZDs. More research into the underlying mechanisms is indicated.

Limitations of the current study include the cross-sectional nature of study measurements, which provide only an instantaneous view of the adiponectin relationships with other phenotypes. We believe that longitudinal studies characterizing the relationship between adiponectin levels and the progression of vBMD, adiposity, and CP are necessary to better understand the nature of the relationship. We were unable to assess insulin resistance due to the lack of insulin or C-peptide levels. Analyses adjusted for BMI, not waist circumference; however, the correlation between these variables was 0.86. Substitution of waist circumference for BMI did not alter significance of any results reported (data not shown). In addition, adjustment for alcohol ingestion and estimated GFR did not impact results.

We conclude that serum adiponectin concentrations manifest inverse associations with vBMD, visceral adiposity, and CRP (and direct associations with IMAT) in AAs with T2D; without significant evidence of association with subclinical atherosclerosis in the coronary, carotid, or aortoiliac vascular beds or sc adipose tissue volume. The inverse relationship between adiponectin and bone density observed here is consistent with findings in the adiponectin knockout mouse, which exhibited increased bone mass and strength relative to wild-type control mice (28). Adiponectin may provide a unique signaling mechanism between adipose tissues and the skeleton.