|Home | About | Journals | Submit | Contact Us | Français|
African-Americans with hypertension are susceptible to left ventricular hypertrophy (LVH). Serum osteoprotegerin level has been reported to be associated with LVH. We investigated the association of osteoprotegerin with LV mass (LVM) in 898 African-Americans with hypertension (mean age 65 years, 71% women).
Osteoprotegerin levels were measured in serum by an immunoassay and log-transformed for analyses. LVM index (LVMi; LVM/height2.7) was estimated using M-mode echocardiography. Linear regression analyses using generalized estimating equations were used to assess the association of osteoprotegerin with LVMi.
Serum osteoprotegerin was correlated with LVMi (r = 0.21; P < 0.0001), an estimated increase in LVMi of 5.05 (95% confidence interval 2.93, 7.17) g/m2.7 in the highest compared to the lowest osteoprotegerin quartile. This association remained statistically significant after adjustment for conventional cardiovascular risk factors (age, sex, body mass index (BMI), history of smoking, diabetes, systolic blood pressure (BP), total and high-density lipoprotein cholesterol), estimated renal function, history of myocardial infarction and stroke, lifestyle factors (physical activity score, years of education, amount of alcohol consumption), medications (aspirin, antihypertensives, statins, estrogens), and C-reactive protein (CRP) (P = 0.02). Additionally, osteoprotegerin was correlated with early/atrial (E/A) ratio (r = −0.16; P < 0.0001), LV mean wall thickness (r = 0.17; P < 0.0001) and relative wall thickness (r = 0.14; P < 0.0001) but not ejection fraction (r = 0.04; P = 0.24) or internal end-diastolic dimension (r = 0.02; P = 0.60).
In African-Americans with hypertension, a higher serum osteoprotegerin level is weakly but independently associated with a higher LVM.
Osteoprotegerin is a glycoprotein belonging to the tumor necrosis factor superfamily of signaling proteins. It is a decoy receptor for receptor activator of nuclear factor-κB (RANK) ligand (RANKL), first described in 1997 as an inhibitor of RANK-mediated activation of osteoclasts in mice.1,2 In addition to bone remodeling, osteoprotegerin has been associated with dyslipidemia, insulin resistance, endothelial damage, atherosclerosis, cerebrovascular disease, coronary artery disease, and adverse cardiovascular outcomes.3–10 In experimental animals, osteoprotegerin has been shown to be an endothelial protective factor as it inhibits progression of arterial calcification11,12 and is a survival factor for growth factor deprived endothelial cells in vitro.13
Hypertension is a major risk factor for left ventricular (LV) hypertrophy (LVH).14 African-Americans have a higher incidence and, in general, greater severity of hypertension than whites.15 They also have a higher prevalence of LVH,16 even after accounting for clinical and hemodynamic factors.17 Traditionally, neurohumoral activation due to hypertension and upregulation of the renin–angiotensin–aldosterone system has been implicated in the pathogenesis of LVH.18 A role for systemic inflammation has also been proposed.19,20 Recently, the osteoprotegerin/RANK/RANKL axis has been hypothesized as a potentially important mediator of LVH.21,22 In a population-based cohort of over 2,700 adults, osteoprotegerin was independently associated with LVH in men, but not in women.23 We sought to evaluate the association of serum osteoprotegerin levels with LV mass (LVM) in a community-based sample of African-American men and women with hypertension.
The study population comprised African-Americans with hypertension who were enrolled in the Genetic Epidemiology Network of Arteriopathy (GENOA) cohort at Jackson, MS. This multicentric, community-based, study aimed at identifying genetic variants influencing blood pressure (BP) by recruiting African-Americans from Jackson, MS, Mexican-Americans from Starr County, TX, and non-Hispanic whites from Olmsted County, MN.24 In Jackson, the hypertensive probands in the Atherosclerosis Risk in Communities (ARIC) cohort and their siblings were recruited, if at least two individuals in the sibship were diagnosed with hypertension before the age of 60 years. The ARIC cohort in Jackson was a probabilistic sample of 45–64-year-old African-American residents ascertained through driver’s licenses between 1987 and 1989.25 The hypertensive proband and all available full-biologic siblings irrespective of age or hypertensive status were invited for interviews, physical examinations, and blood sampling after an 8-h fast between July 1997 and August 1999. Data on serum markers were collected as part of the Proteomic Markers of Arteriosclerosis Study investigating the association of multiple serum markers with phenotypes of arteriosclerosis.26,27 Written informed consent was obtained from each participant. The institutional review board of the University of Mississippi Medical Center (Jackson, MS), approved these studies. This study was restricted to participants with hypertension.
Body mass index (BMI) was calculated from height measured by a stadiometer and weight measured by an electronic balance. Resting systolic and diastolic BPs were measured by trained study coordinators using random zero sphygmomanometer and appropriate size cuffs. The average of last two of three measurements in the right arm after at least 5 min of rest was used. The diagnosis of hypertension was established either with a BP reading of ≥140/90 mm Hg at the study visit or a prior diagnosis of hypertension and current treatment with antihypertensive medication(s). Similarly, diabetes was diagnosed if the subject had fasting plasma glucose (measured by oxidase method) ≥126 mg/dl or had a prior diagnosis and was on treatment with oral hypoglycemic agent(s) or insulin. Estimated glomerular filtration rate (eGFR) was calculated using the 4-variable Modification of Diet in Renal Disease Study (MDRD) equation.28 Smoking status was defined as ever-smoker if the participant reported having smoked >100 cigarettes in his/her lifetime. Information about medication use was obtained from participants and each prescription drug was assigned the Medi-Span Generic Product Identifier (Medi-Span, Indianapolis, IN) to identify and categorize the pharmacologic agent. Antihypertensive medications were classified as diuretics, β-blockers, calcium-channel blockers, or renin–angiotensin–aldosterone system inhibitors.
Physical activity score was obtained by eliciting h/day of heavy, moderate, light, and sedentary activity the participant engaged in. The score was calculated as 3 × (heavy) + 2 × (moderate) + (light). Amount of alcohol consumption was estimated from the reported frequency and amount per serving of beer, wine, hard liquor, wine cooler, and sake. The amount was then multiplied by 5, 12, 40, 8, and 15%, respectively, and summed to obtain ounces of alcohol per month.
Blood samples were obtained by venipuncture after an overnight fast. Serum total and high-density lipoprotein cholesterol and standardized creatinine were measured by standard enzymatic methods. Serum osteoprotegerin was measured by solid phase sandwich immunoassay on a Sector Imager 2400 platform (Meso-Scale Discoveries, Gaithersburg, MD). Intra-assay coefficients of variation were 4.0% at 133 pg/ml and 2.5% at 180 pg/ml, and interassay coefficients of variation were 13.3% at 149 pg/ml and 11.3% at 189 pg/ml, respectively. Plasma C-reactive protein (CRP) level was measured by a highly sensitive immunoturbidimetric assay.29 Interassay coefficients of variation were 14% at 0.33 mg/dl, 3.2% at 1.05 mg/dl, 3.4% at 9.07 mg/dl, and 3.6% at 23.8 mg/dl.
Participants underwent M-mode and spectral Doppler exam with the Acuson XP128/10c echo machine (Siemens Medical, Iselin, NJ) using 2.5, 3.5, and 5.0-MHz transducers. The parasternal acoustic window was used to record ≥10 consecutive beats of M-mode recordings of the LV internal diameter and wall thicknesses at, or just below, the tips of the anterior mitral leaflet in long- and short-axis views, as recommended by the American Society of Echocardiography.30 LVM was calculated using a necropsy-validated correction31 for the overestimation by the cube equation,32 and was indexed to height2.7 (LVM index (LVMi) = LVM/height2.7).33 LVH was defined as LVMi >51 g/m2.7 (ref. 34). Details of echocardiographic methods and calculations have been described previously.35
African-American participants with hypertension were identified and those with missing data for the primary predictor (serum osteoprotegerin) or outcome (LVMi) were excluded. Markov chain Monte Carlo algorithm was used to impute the missing CRP values from the distribution of the other available serum markers. Osteoprotegerin and CRP were transformed on the natural logarithm scale and standardized to obtain standard-normal distributions. Cochrane–Armitage (for dichotomous variables) and Jonckheere–Terpstra tests (for continuous variables) were used to test the variables for trend with quartiles of serum osteoprotegerin.
The maximum-likelihood solutions of linear regression models using generalized estimating equations accounting for covariance among sibships were used to obtain the association of osteoprotegerin with LVMi. We first modeled the unadjusted association of osteoprotegerin with LVMi. Covariates were added to this model to adjust for age, sex, other conventional cardiovascular risk factors (BMI, history of smoking, diabetes, systolic BP, total and high-density lipoprotein cholesterol), eGFR, history of myocardial infarction and stroke. Subsequently, variables for medication use (aspirin, diuretics, β-blockers, calcium-channel blockers, renin–angiotensin–aldosterone system inhibitors, statins, estrogens) and lifestyle factors (physical activity score, years of education, amount of alcohol consumption) were added. Finally, CRP was added to the model to evaluate the association independent of the inflammatory marker. Effect-modification by sex was evaluated by adding the sex × (log-osteoprotegerin) interaction term. Sensitivity analyses were done using the distribution means in place of missing values. Similarly, sensitivity analyses were done excluding observations with 3 s.d. extreme outlier values of serum osteoprotegerin or LVMi. Bivariate scatter-plots and smoothed cubic-splines were obtained to rule out any overt nonlinear associations. Linear regression methods as described above were used to obtain associations of serum osteoprotegerin with the other LV measures/indices, i.e., ejection fraction, internal end-diastolic dimension, ratio of early (E) to late atrial (A) mitral Doppler peak flow velocity (E/A ratio), mean wall thickness ((posterior wall thickness + inter-ventricular septal thickness)/2) and relative wall thickness (posterior wall thickness/internal end-diastolic dimension). Two-tailed P = 0.05 was considered the significance threshold for all statistical tests and results from regression models are reported as β-estimate (95% confidence interval; P value) unless otherwise specified. The analyses were done in SAS for Windows version 9.1.3 (SAS Institute, Cary, NC).
Of the 1,324 African-American adults enrolled in GENOA, 1,058 had hypertension. Of these, 160 were excluded due to missing values (29 missing LVM, 128 missing osteoprotegerin, and 3 missing both). The analyses thus included 898 individuals belonging to 497 sibships (mean age 65 years, 71% women). The median (interquartile range, IQR) for osteoprotegerin was 325 (249–417) pg/ml and for LVMi 39 (34–47) g/m2.7. Sex-specific median osteoprotegerin levels were statistically different: 330 (255–422) pg/ml for men and 306 (229–384) pg/ml for women (Wilcoxon–Mann–Whitney P = 0.005).
Age, sex, diabetes, systolic BP, total cholesterol, eGFR, physical activity score, years of education, amount of alcohol consumption, diuretic use, and CRP were significantly associated with serum osteoprotegerin. Comparing the highest with the lowest quartile of osteoprotegerin, the median (IQR) systolic BP increased from 137 (125–148) mm Hg to 146 (132–162) mm Hg; the median eGFR decreased from 79 (68–91) ml/min to 65 (51–78) ml/min; and the median CRP increased from 3.5 (1.8–7.4) mg/l to 4.5 (2.1–10.2) mg/l (Table 1).
Log-serum osteoprotegerin was correlated with E/A ratio (r = −0.16; P < 0.0001), LV mean wall thickness (r = 0.17; P < 0.0001), relative wall thickness (r = 0.14; P < 0.0001), LVM (r = 0.13; P = 0.0001), and LVMi (r = 0.21; P < 0.0001) but not ejection fraction (r = 0.04; P = 0.24) or internal end-diastolic dimension (r = 0.02; P = 0.60). The c-statistic for discriminating presence vs. absence of LVH with osteoprotegerin alone was 0.61. There was no effect modification by sex for any of these associations (Table 2).
The estimated LVMi was higher in the highest compared to the lowest quartile of osteoprotegerin by 5.05 (95% confidence interval 2.93, 7.17; P < 0.0001) g/m2.7. The unadjusted linear association of log-osteoprotegerin with LVMi was statistically significant (unadjusted β = 2.32 (1.52, 3.12; P < 0.0001)) (Figure 1). The association remained significant on sequential additional adjustments for (i) age and sex (P = 0.0002), (ii) conventional cardiovascular risk factors, eGFR, history of myocardial infarction and stroke (P = 0.01), (iii) lifestyle factors (P = 0.02), (iv) medications (P = 0.02), and (v) CRP (multivariable-adjusted β = 0.99 (0.19, 1.80; P = 0.02) g/m2.7) (Table 3). Apart from osteoprotegerin, other significant independent predictors of LVMi were age, BMI, systolic BP, physical activity score, β-blocker, calcium-channel blocker, and estrogen use (Table 4). There was no effect-modification by sex (interaction term P = 0.95). The results were robust to sensitivity analyses using means of osteoprotegerin and LVMi to account for missing values (unadjusted β = 2.07 (1.34, 2.80; P < 0.0001) g/m2.7; multivariable-adjusted β = 0.86 (0.14, 1.59; P = 0.02) g/m2.7). On removing 14 observations with 3 s.d. extreme outlier values for osteoprotegerin or LVMi, the unadjusted association remained statistically significant (β = 1.89 (1.14, 2.64; P < 0.0001) g/m2.7), however, the multivariable-adjusted association was attenuated (β = 0.60 (−0.14, 1.35; P = 0.12) g/m2.7). On the other hand, the association between serum osteoprotegerin and E/A ratio became nonsignificant on adjustment for age and sex (P = 0.27).
Several cardiovascular risk factors were associated with serum osteoprotegerin levels in our study. This was true for systolic BP, an established risk factor for LVH; eGFR, a proxy for vascular damage; and, CRP, a marker of systemic inflammation. There was a statistically significant association of osteoprotegerin with LVMi, independent of cardiovascular risk factors, lifestyle factors, and medications.
As expected, age, BMI, and systolic BP were the strongest independent predictors of LVMi. Estrogens inhibit and reverse LVH via their direct effect on calcineurin signaling in cardiomyocytes36,37 and estrogen use had a strong inverse association with LVMi. The magnitudes and strengths of the associations of age, BMI, systolic BP, and estrogen use with LVMi remained robust to multivariable adjustment. However, the statistically significant univariate associations of diabetes, eGFR, physical activity, aspirin, and antihypertensive medication use (especially diuretics and renin–angiotensin–aldosterone system inhibitors) with LVMi were either attenuated or nullified in the multivariable-adjusted model; and similarly, the predictive value of CRP was preempted by osteoprotegerin and the other covariates. It is noteworthy, that these second group of variables are to some extent surrogates for neurohumoral activation or systemic inflammation pathways implicated in the development of LVH, and the association of LVMi with osteoprotegerin was independent of these variables. These results might be indicative of upregulated osteoprotegerin/RANK/RANKL system being involved in the established association of endothelial damage and systemic inflammation with LVH.20 However, with the modest correlation between osteoprotegerin and LVMi (r = 0.21) and the absence of a well-defined pathobiology implicating osteoprotegerin, its role in LVH remains a speculation at best.21 Also, we did not find any association of serum osteoprotegerin with measures of LV systolic (ejection fraction) and diastolic (E/A ratio) function that was independent of age and sex. Certain criteria would have to be established for serum osteoprotegerin levels to be considered as a biomarker for LVH, including a relationship with myocardial expression of the osteoprotegerin/RANK/RANKL system, an epidemiologic association with LVH, variation of serum levels in parallel with LVH, as well as good sensitivity and specificity.38
Atherosclerosis, including coronary, aortic, and peripheral arterial disease, is associated with higher levels of serum osteoprotegerin.6,10,22,39–42 Unstable coronary plaques43 and growing abdominal aortic aneurysms7 are associated with increased expression of osteoprotegerin. Higher levels of osteoprotegerin with vascular disease may be a mechanism to regulate endothelial function and limit atherosclerosis progression,3,9 a hypothesis supported by experimental animal models in which osteoprotegerin inhibits arterial calcification and plaque progression.11,12,44 Osteoprotegerin has been shown to protect endothelial cells by binding tumor necrosis factor-related apoptosis-inducing ligand and thereby inhibiting endothelial cell apoptosis.13,45 Notably, osteoprotegerin is also a survival factor for smooth muscle cells and induces matrix metalloproteinase-9 activity in smooth muscle.12 A mechanism similar to this causing LVH is plausible, where osteoprotegerin promotes cardiomyocyte hyperplasia and regulates matrix turnover. Serum osteoprotegerin has been shown to be associated with development of heart failure in patients who present with acute coronary syndrome or aortic stenosis.46 There is upregulated systemic expression of osteoprotegerin and its increased uptake by cardiac tissue in chronic heart failure. The abnormal pattern of osteoprotegerin/RANK/RANKL expression and matrix metalloproteinase activity in left ventricle remodeling has been studied.21,46 Clinically, higher serum osteoprotegerin levels seem to be associated with higher heart failure related hospitalizations and higher mortality.22 The circulating serum levels of osteoprotegerin are easily measurable, in contrast to RANK and RANKL that are expressed locally in the cardiac tissue. This easy measurability of serum osteoprotegerin imparts it potential as a biomarker.
Our results are in general consistent with those from the Dallas Heart Study, which was a general-population based cohort, and used magnetic resonance imaging instead of echocardiography.23 Consistent with this study, our analyses showed serum osteoprotegerin to be associated with indicators of LVH such as mean wall thickness, relative wall thickness, LVM, and LVMi. In contrast, we did not find an inverse association of serum osteoprotegerin with LV ejection fraction. Nor did we find any effect modification by sex, though both study populations had lower median osteoprotegerin levels in men compared to women. These differences might reflect differences in the study populations, including racial and age composition, hemodynamic risk factors, study methodology, and random effects. The median serum osteoprotegerin level in our study was significantly lower than that in the Dallas Heart Study, likely due to differences in the assay.
In summary, established cardiovascular risk factors are associated with serum osteoprotegerin level. Osteoprotegerin is also associated with echocardiographic evidence of LVH. We found a statistically significant independent association between serum osteoprotegerin and LVMi in hypertensive African-Americans. Whether osteoprotegerin is a clinically useful biomarker for LVH in hypertensive patients needs further investigation.
Osteoprotegerin level was measured using a well-standardized assay with low coefficients of variation. Indexing of the LVM to height (LVM/height2.7) rather than the body-surface area is a more valid method and free from a reflection of the underlying anthropometric variables.33 We were able to adjust for several potential confounding factors. Our sample is not representative of the general population; however, our study included a large sample of hypertensive African-Americans from the community, the population at the highest risk of LVH and its adverse sequelae. Although the association of serum osteoprotegerin level and LVMi has independent of confounders and statistically significant, the effect size was small. We were unable to evaluate temporal effects of serum osteoprotegerin levels on LVM in our analysis as echocardiographic measurements for LV measurements were made in advance of recording the clinical variables and serum markers. With limited understanding of the pathophysiological mechanisms, it is difficult to parse out a causal relationship linking osteoprotegerin with LVH from an epiphenomenon. Other limitations included missing data (15%) and potential sampling bias. Our study was composed mostly of women (71%) and thus lacked statistical power for associations of echocardiographic measures with osteoprotegerin among men, although the effect sizes in both groups were similar in general (Table 2). Moreover, the technique to measure LV indices using M-mode echocardio graphy is imprecise and outdated and now has been replaced by 2-D and 3-D echocardiography and magnetic resonance imaging.
Our results show that serum osteoprotegerin is independently but weakly associated with LVM in hypertensive African-American men and women. Only a longitudinal study with serial measurements of serum osteoprotegerin and LVM can elucidate the potential causal effect of osteoprotegerin/RANK/RANKL system on LVM. Our results lend preliminary support to such a potential effect; however, basic science research is needed to delineate out the pathophysiology of osteoprotegerin/RANK/RANKL-mediated effects on LVH. With epidemiologic data suggesting worse outcomes associated with higher serum osteoprotegerin levels in various subgroups, it is exciting to consider that osteoprotegerin could potentially serve as a screening biomarker in hypertensive patients to stratify risk of LVH and initiate preventative interventions. However, due to the limitations of this study and deficiencies in our understanding of the osteoprotegerin/RANK/RANKL pathophysiology, osteoprotegerin is as yet far from being a serious contender as a clinically useful biomarker. On the other hand, a host of other markers involved in the modulation of extracellular matrix, cardiomyocyte injury, inflammation, oxidative stress, and neurohormonal activation are being evaluated as biomarkers for hypertensive heart disease.38
This work was supported by the NIH grant HL-81331.
Disclosure: The authors declared no conflict of interest.