Bone Mineral Density and Progression of Subclinical Atherosclerosis in African-Americans With Type 2 Diabetes

Abstract

Context:

Relative to European Americans, calcified atherosclerotic plaque (CP) is less prevalent and severe in African-Americans (AAs).

Objective:

Predictors of progression of CP in the aorta, carotid, and coronary arteries were examined in AAs over a mean 5.3 ± 1.4-year interval.

Design:

This is the African American-Diabetes Heart Study.

Setting:

A type 2 diabetes (T2D)-affected cohort was included.

Participants:

A total of 300 unrelated AAs with T2D; 50% female, mean age 55 ± 9 years, baseline hemoglobin A1c 8.1 ± 1.8% was included.

Main outcome measures:

Glycemic control, renal parameters, vitamin D, and computed tomography-derived measures of adiposity, vascular CP, and volumetric bone mineral density (vBMD) in lumbar and thoracic vertebrae were obtained at baseline and follow-up.

Results:

CP increased in incidence and quantity/mass in all three vascular beds over the 5-year study (P < .0001). Lower baseline lumbar and thoracic vBMD were associated with progression of abdominal aorta CP (P < .008), but not progression of carotid or coronary artery CP. Lower baseline estimated glomerular filtration rate was associated with progression of carotid artery CP (P = .0004), and higher baseline pericardial adipose volume was associated with progression of coronary artery (P = .001) and aorta (P = .0006) CP independent of body mass index. There was a trend for an inverse relationship between change in thoracic vBMD and change in aortic CP (P = .05).

Conclusions:

In this longitudinal study, lower baseline thoracic and lumbar vBMD and estimated glomerular filtration rate and higher pericardial adipose volumes were associated with increases in CP in AAs with T2D. Changes in these variables and baseline levels and/or changes in glycemic control, albuminuria, and vitamin D were not significantly associated with progression of CP.

We studied change in atherosclerosis in African Americans and found that lower baseline bone density and kidney function, and higher pericardial adiposity associated with increasing atherosclerosis.

African-Americans (AAs) are known to be at lower risk for developing calcified atherosclerotic plaque (CP) than European Americans (EAs) (1). Coronary artery CP (CAC) predicts future cardiovascular disease (CVD) and all-cause mortality in AAs as well as in other ethnic groups (2), yet few studies have investigated the progression of CP in AAs. In the Multi-Ethnic Study of Atherosclerosis (MESA), incident CP and progression of CP were lower in AAs than EAs. However, AAs with type 2 diabetes (T2D) were at significantly greater risk for progression of CP compared to EAs with T2D (P = .0002) (3).

Apart from MESA, studies investigating progression of CAC in AAs are extremely limited, of small sample size, and with wide variation in progression rates (4, 5). Studies of the progression of CP in persons with T2D are also limited, as are studies examining predictors of progression. In the Veterans Affairs Diabetes Trial, progression of CAC and abdominal aorta CP was observed in more than 75% of the cohort; however, progression was not influenced by conventional risk factors (6). Others have found that hemoglobin (Hb) A1c is an important predictor of progression in persons with T2D (7).

Numerous studies have demonstrated age-independent inverse relationships between bone mineral density (BMD) and CP (8, 9), which may be causally related by a variety of mechanisms (10). Thus, BMD and changes in BMD may be uniquely associated with change in CP. However, few studies have addressed this possibility, and none in AAs or populations with diabetes. The present study from the longitudinal AA-Diabetes Heart Study cohort attempts to fill gaps in this literature. AAs with T2D may be at uniquely high risk for progression of CAC compared to other racial/ethnic groups (3). This may translate into greater mortality risk in AAs (2). Studies of the progression of CP in other vascular beds and its predictors are limited. This longitudinal study in AAs with T2D provides an opportunity to evaluate a panel of potential predictors of the progression of CP across multiple vascular beds.

Materials and Methods

Participants

AAs with T2D were recruited from internal medicine clinics and community advertising (11). The initial examination was attended by 691 participants between 2007 and 2010. Forty-three percent (N = 300) returned for a follow-up examination at an approximate 5-year interval. This was an epidemiologic study; no interventions were provided between visits. Examination components included interviews for medical history and health behaviors, anthropometric measures, resting blood pressure, electrocardiography, fasting blood draw, spot urine collection, and vascular and abdominal imaging by computed tomography (CT). Protocols were identical at the baseline and follow-up visits. T2D was defined as a diagnosis of diabetes after age 30 years in the absence of historical evidence of diabetic ketoacidosis. History of CVD was provided by participant report and medical record review and included myocardial infarction, stroke and coronary artery interventions. Chronic kidney disease was defined as urine albumin:creatinine ratio greater than 30 mg/g and/or estimated glomerular filtration rate (eGFR) less than 60 ml/min/1.73 m² (Chronic kidney disease-Epidemiology Collaboration equation). Hypertension was based on a physician diagnosis, clinic blood pressure higher than 140/90 mm Hg, or use of antihypertensive medications.

Laboratory assays included high sensitivity C-reactive protein, HbA1c, 25 hydroxyvitamin D and 1,25 di-hydroxyvitamin D, intact PTH, serum calcium, and serum phosphate. Markers of kidney disease included urine albumin:creatinine ratio and eGFR. Medications were recorded; those relevant for these analyses included hormone replacement therapies, calcium and vitamin D supplements, bisphosphonates, steroids, angiotensin-converting enzyme inhibitors and angiotensin receptor blockers. The study was approved by the Institutional Review Board at the Wake Forest School of Medicine and all participants provided written informed consent.

Imaging

CP was measured in the coronary arteries, infrarenal aorta, and carotid arteries using single and multidetector CT systems incorporating a standardized scanning protocol (12). This report uses the calcium mass score (milligrams of calcium), derived from the volume score but also accounting for the density of CP on a pixel-by-pixel basis (13). Additional scoring parameters included a 90 Hounsfield unit threshold and two adjacent pixels to define the maximum calcified lesion size; the program accounted for slice thickness. Quantitative coronary CP mass scores were excluded from analyses in participants who had undergone coronary artery bypass grafting, stenting, or angioplasty before the baseline visit or between visits (N = 59). Repeat CP measures were available in N = 235 for coronary, 294 for carotid, and 296 for aorta. As previously reported, correlation coefficients for CP measurements were 0.96 for measures on two sequential CT scans (8) and 0.99 for readings obtained by two different CT image analysts (12).

Quantitative CT for trabecular volumetric BMD (vBMD; mg/cm³) in the thoracic (T₈–T₁₁) and lumbar vertebrae (T₁₂–L₃) were measured using QCT-5000 volumetric software (Image Analysis) with an external calibration phantom from the CT images measuring CP in coronary arteries and aorta, respectively. Precision of CT-derived vBMD is high with coefficients of variation of less than 1% (14). Vertebral fractures were read as present/absent from the anteroposterior and lateral scout images of the CT scans. Visceral adipose tissue, SC adipose tissue, intermuscular adipose tissue, and hepatic steatosis were measured on abdominal CT scans. Pericardial adipose tissue was measured on the thoracic CT scan. A threshold of –190 to –30 Hounsfield units was used to define adipose-containing tissue.

Statistical analysis

Sample means and SDs were computed for continuous traits and proportions were calculated for discrete traits. For variables with highly skewed distributions, median values and interquartile ranges were reported to reflect the central tendency and dispersion. Three classes of predictor variables were considered: vBMD (primary hypothesis), clinical chemistry biomarkers, and adipose tissue volumes. Variables are listed in Table 1.

Table 1.

Baseline Characteristics of the African American-Diabetes Heart Study Cohort (n = 300), Mean (sd) or percent

Age, y	55.1 (9.0)
Duration of diabetes, y	10.2 (7.2)
BMI, kg/m2	34.7 (7.7)
Smoking (% past, % current)	35.3, 22.3
Chronic kidney disease, %	38.7
Cardiovascular disease, %	29.3
Hypertension, %	80.7
Medications
Insulin, %	42.0
Oral hypoglycemic, %	76.3
Statins, %	49.0
ACEi/ARB, %	52.3
Steroids, %	7.3
Hormone replacement therapy, women only, %	13.7
Bisphosphonates, %	0.3
Calcium supplements, %	9.3
Vitamin D supplements, %	3.0
Clinical Chemistry Biomarkers
High-sensitivity CRP, mg/dl	1.03 (1.58)
HbA1c, %	8.1 (1.8)
25OHD, ng/ml	20.1 (12.3)
1,25(OH)2D3, pg/ml	47.3 (18.0)
Intact PTH, pg/ml	56.2 (30.3)
Serum calcium, mg/dl	9.57 (0.42)
Serum phosphorus, mg/dl	3.56 (0.58)
Urine albumin:creatinine, mg/g	12.0 (4.2, 48.0)a
Estimated GFR, ml/min/1.73 m2	92.1 (21.1)
Bone and Adiposity Measures
Lumbar vBMD, mg/cm3	179 (148, 209)a
Thoracic vBMD, mg/cm3	205 (172, 232)a
Vertebral facture, %	28.3
Pericardial adipose, cm3	90.6 (42.9)
Visceral adipose, cm3	181.7 (78.2)
IM adipose, cm3	10.5 (7.3)
SC adipose, cm3	433.9 (189.0)
Hepatic steatosis, cm3	52.6 (10.7)
CP Measures
Aorta CP mass, mg Ca+ (n = 296)
Baseline	997 (41, 5472)a
Follow-up	2767 (411, 10 194)a
Carotid artery CP mass, mg Ca+ (n = 294)
Baseline	1 (0, 72)a
Follow-up	27 (0, 238)a
Coronary artery CP mass, mg Ca+ (n = 235)
Baseline	20 (0, 312)a
Follow-up	89 (11, 688)a
Aorta CP mass = 0 (N, %)
Baseline	56 (18.9%)
Follow-up	2 (0.7%)
Carotid artery CP mass = 0 (N, %)
Baseline	142 (48.3%)
Follow-up	87 (29.6%)
Coronary artery CP mass = 0 (N, %)
Baseline	94 (40.0%)
Follow-up	38 (16.2%)

Abbreviations: ACEi, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; BMI, body mass index; CRP, C-reactive protein; 1,25(OH)₂D, 1,25 di-hydroxyvitamin D; 25OHD, 25 hydroxyvitamin D.

^aMedian and interquartile range.

The primary outcome variable was 5-year progression of CP in each of the three vascular beds (coronary, aorta, and carotid). Progression was defined as the absolute difference between follow-up (FU) and baseline (BL) CP in each bed; it was treated as a continuous endpoint. These differences were modeled on the log-scale to best approximate the normality assumption. The outcome used for each vascular bed can be written as Y = log (FU – BL + c). The constant c was added to ensure that logarithm returns a real value in the few cases where the difference between FU and BL values was nonpositive. We used c = 1001, 4, and 37 for coronary, carotid, and aorta, respectively.

Two models were fit for each predictor and outcome combination. Model 1 adjusted for the baseline value of the outcome and time between measurements. Model 2 additionally included age, gender, HbA1c, and a panel of biomarkers, including vitamin D metabolites and PTH, disease status, and medications. Detailed covariate lists are found as footnotes in the tables. Baseline predictors and 5-year change in predictor variables were evaluated. A sensitivity analysis was conducted by repeating the vBMD models after excluding 23 people on steroids or bisphosphonates. To account for multiple comparisons for the primary hypothesis (2 vBMD measures × 3 outcome measures), we set alpha at 0.05/6 or 0.008.

Results

The cohort was evaluated at a mean 5.3 ± 1.4-year interval, 50% (N = 150) were female, mean age was 55 ± 9 years, with baseline HbA1c 8.1 ± 1.8% (Table 1). CP increased over time in all three vascular beds (all P < .0001), both in terms of incidence of any CP (mass > 0) and quantity of CP (change in calcium mass over time). Baseline lumbar vBMD and thoracic vBMD were both inversely associated with change in aorta CP, but not with change in coronary or carotid artery CP (Table 2). In a fully adjusted model, lower baseline lumbar vBMD was associated with an increase in aorta CP (P = .006). Similarly, lower baseline thoracic vBMD associated with an increase in aorta CP (P = .005). Effect sizes were consistent across men and women but were modestly reduced in a sensitivity analysis that excluded 23 people on steroids or bisphosphonates (data not shown). In contrast, changes in vBMD were not significantly associated with changes in CP in any vascular bed (Supplemental Table 2), although two observations are noteworthy. Trends were observed for both decreasing lumbar vBMD (P = .13) and decreasing thoracic vBMD (P = .05) with increasing aorta CP, neither of which reached the study-wide alpha threshold of 0.008. Vertebral fractures were not associated with change in CP in any vascular bed (data not shown).

Table 2.

Effect sizes (β coefficients, SE, and P Value) for Baseline Vertebral Bone Mineral Density and 5-Year Change in Calcified Atherosclerotic Plaque in the Aortic, Carotid, and Coronary Arteries Minimally Adjusted Model and Fully Adjusted Model, Adjustments in Footnotes

	β Coefficient (SE) P Value
	Abdominal Aorta	Carotid	Coronary
Model 1a
Lumbar vBMD	−0.016 (0.005)	0.001 (0.001)	0.022 (0.03)
	.0004	.39	.47
Thoracic vBMD	−0.015 (0.004)	0.002 (0.001)	−0.013 (0.03)
	.0002	.12	.62
Model 2b
Lumbar vBMD	−0.013 (0.005)	0.0002 (0.001)	0.045 (0.031)
	.006	.89	.15
Thoracic vBMD	−0.012 (0.004)	0.001 (0.001)	0.007 (0.027)
	.005	.23	.80

^aModel 1: adjusted for baseline outcome, time between measurements.

^bModel 2 = model 1 + age, sex and baseline HbA1c, eGFR, serum phosphorus, smoking (never/ever), steroids (yes/no), hormone replacement therapy in women (yes/no), calcium supplements (yes/no), vitamin D supplements (yes/no).

Relationships between clinical chemistry biomarkers and adiposity measures with change in the three vascular CP outcomes are provided in Supplemental Table 1. Lower baseline eGFR was associated with increased progression of carotid artery CP (P = .0004). Higher baseline pericardial fat was associated with increased progression of CAC (P = .001) and aorta CP (P = .0006). In contrast, changes in predictors (pericardial adipose volume, eGFR), or baseline level or change in the levels of HbA1c, vitamin D and mineral metabolites, and C-reactive protein were not associated with changes in CP in any vascular bed (Supplemental Table 2).

Discussion

The novel finding in this longitudinal study of AAs with T2D is the observation that lower baseline levels of lumbar and thoracic vBMD are significantly associated with increases in aorta CP over a 5-year period. This finding is novel because AAs and EAs have markedly different vitamin D axes, levels of bone mineralization, and urinary calcium handling (10). Additionally, longitudinal results extend previous work, including our own cross-sectional reports of age-independent inverse associations between vBMD and CP (8, 9). This finding supports a growing body of evidence that the inverse relationship between BMD and vascular calcification may have a causal, physiologic basis (10, 15).

A variety of complex mechanisms may be involved in age-independent inverse relationships between calcium accretion in bones and arteries (10). The exact nature of the initiation of calcification in atherosclerotic plaques is not fully understood, but this generally occurs in relatively advanced lesions when cell death is observed. Preservation of arterial health (inhibition of atherogenesis/atherosclerosis) would delay the plaque calcification process. Factors originating outside of bone and arteries likely to simultaneously inhibit bone loss and vascular calcification include estrogen, exercise, and vitamin D, although the relationship with vitamin D in AAs is complex (10, 16, 17). Conversely, factors simultaneously promoting bone loss and CVD/vascular calcification include systemic inflammation, autoimmune and connective tissue diseases, kidney dysfunction, and low levels of estradiol, exercise, and vitamin D. Inflammation in particular may act through pathways involving receptor activator of nuclear factor κB (RANK), the RANK-ligand, and osteoprotegerin, the decoy receptor for the RANK-ligand, to simultaneously increase bone resorption (via osteoclast activation). In vitro studies suggest it may be involved in promotion of vascular calcification (via upregulation of inflammation associated genes and altered smooth muscle programming); however, its involvement in CP initiation and progression is unclear (18). Advanced glycosylation end-products are also proinflammatory, acting through the advanced glycosylation end-product receptor (RAGE) to promote inflammation, and may interact with the RANK pathways to influence bone and vascular calcification in opposing directions, with new evidence suggesting that a RAGE binding protein (S100A12, aka calgranulin C or extracellular newly associated with RAGE protein, EN-RAGE), can exacerbate atherosclerosis and vascular calcification in a transgenic mouse model (19, 20). The importance of these pathways merits further exploration.

Statin use is also a potential confounder of the relationships between vascular and skeletal calcium content. Statins reduce plasma lipids and atherosclerosis progression, which may inhibit the initiation of calcification in early lesions. In more advanced disease, statins promote removal of lipid from atherosclerotic lesions and produce other changes in lesion composition and the associated inflammatory response. Recent longitudinal studies evaluating the effects of statin dose on plaque composition evaluated by intravascular ultrasound techniques suggest that high-intensity statin treatment increases the amount of calcified plaque compared to regular-intensity statins (21, 22). This possibility is supported by previous studies of atherosclerosis regression by dietary cholesterol lowering in cynomolgus monkeys, which showed that, along with loss of lipid, there were increases in the collagen, proteoglycan, and calcium content of the arteries (23). Our analyses considered statins as a possible confounder by covarying for it in the multivariate models.

More direct relationships between bone and vascular calcification may also play a role. Atherosclerosis and arterial calcification influence perfusion, which in turn alters physiology in bone and other tissues. Hence, greater vascular calcium indicative of more advanced atheroma may impair blood flow to the skeleton. The source of spinal blood flow is the aorta, and aortic calcium was inversely correlated with vBMD phenotypes in this study. Alternatively, perturbations in bone metabolism, such as elevations in bone resorption, may help to promote ectopic (and arterial) calcification by transiently increasing serum phosphate and calcium, the building blocks of the eventual calcium hydroxyapatite crystals in bones and arteries. Even modest elevations in these components could have long-term consequences. Better understanding of the mechanism(s) underlying this inverse relationship could lead to new approaches for prevention and/or treatment of CVD and bone loss.

Our study was designed to assess calcified atherosclerotic plaque through CT imaging of coronary arteries, carotid bifurcation, and the abdominal aorto-iliac regions known for clinically significant atherosclerotic disease and plaque formation. Medial arterial calcification or Monkenberg's sclerosis, observed particularly in diabetes, is classically a diffuse calcification along the elastic fibers of the medial layer of arteries, although recent evidence suggests that the intima is also involved (24). Medial arterial calcifications are typically identified by radiographs and are most frequently observed in the muscular arteries of the extremities, head, and neck. The spatial resolution of CT and minimum CP lesion size of more than 1 mm² used in this study can identify calcification only after aggregation of calcium occurs in fibroatheromas, which is observed near the intimal medial border involving large areas of adjoining necrotic core and collagen, thus making detection of the small calcifications of medial calcinosis unlikely, even if this condition were present in the coronary, carotid, or abdominal aortas of our participants (25). In addition, medial calcification is generally associated with renal dysfunction, and our subject population was selected to have relatively preserved renal function.

The present study extends several previous reports of bone and vascular calcification. The Framingham Offspring Study reported cross-sectional associations between vBMD and arterial and cardiac valvular calcification obtained by CT in 1317 participants (26). They found that men and women with low spine vBMD had more calcification in the abdominal aorta. Cross-sectional associations have also been reported by the MESA (27). An earlier report from the original Framingham Offspring cohort reported bone loss in the lumbar spine and hand on radiographs performed at 25-year intervals were associated with increases in aortic calcification index in women but not men (28). Our data extend these results by reporting an association between vBMD and change in CP, giving greater credence to a causal relationship when using a more accurate CT measurement. In addition, our analysis was conducted in an AA population in which bone/vascular calcification relationships are less well-defined. The value of this cohort lies in its extensive phenotyping and the longitudinal design, which provide temporal relationships between these factors, previously reported by very few studies, especially in AAs (3). These results extend cross-sectional associations between pericardial adipose volumes and CAC (29, 30) and with a binary measure of progression (31). The current report provides evidence that pericardial adipose volumes predict quantitative measures of progression of CP in both the coronary and aorta vasculature in AAs.

Association was also observed between baseline eGFR and progression of carotid CP. This is striking because of the relatively preserved kidney function in this cohort at baseline. Of particular importance was the lack of association between a number of biomarkers for diabetes control, inflammation, albuminuria, vitamin D metabolites, and adiposity (other than pericardial) on the progression of CP in three vascular beds. In addition, change in these predictors over 5 years did not associate with progression of CP in any vascular bed; baseline vBMD, eGFR, and pericardial adipose were the only predictors significantly impacting the outcome of change in CP. Many of these biomarkers were previously shown to be associated with the incidence and/or quantity of vascular calcification in cross-sectional studies, particularly in EA cohorts. The lack of association in this study may relate to limited variability in the measures of change or outcome, modest power based on sample size (particularly for CAC), variable influences of polypharmacy on targeted phenotypes in long-term diabetes (noted previously), and/or lack of true causal association between biomarkers and progression of CP in AAs. Our extensive database allowed us to provide rigorous adjustment for multiple potential confounders, limiting type 1 error. In this case, it is all the more impressive that significant consistent associations were observed between vBMD measures and progression of abdominal aorta CP. Finally, it is important to note that our measures of vBMD were limited to the trabecular bone compartment of the lumbar and thoracic spine, theoretically more metabolically active than the cortical compartment, but not representative of the entire skeleton which is generally measured with whole body dual x-ray absorptiometry.

In conclusion, lower baseline vBMD in the lumbar and thoracic vertebrae is significantly associated with the progression of CP in the abdominal aorta in AA men and women with T2D. Given the increased risk of CVD among those with T2D and the known risk of vascular morbidity and mortality that is associated with calcification of the abdominal aorta (32), it seems prudent to carefully evaluate AAs who manifest bone loss for CVD risk factors and initiate early treatment of risk factors. Despite marked differences in propensities to develop CP and osteoporosis between AAs and EAs, along with different vitamin D axes, longitudinal analyses in both ethnic groups demonstrate significant inverse associations between bone mineralization and CP, supporting a causal relationship beyond conventional CVD risk factors.