Osteoprotegerin levels predict mortality in patients with symptomatic aortic stenosis


Thor Ueland, Section of Endocrinology, Medical Department National University Hospital, N-0027 Oslo, Norway.
(fax: +47 23073630; e-mail: thor.ueland@medisin.uio.no).


Abstract.  Ueland T, Aukrust P, Dahl CP, Husebye T, Solberg OG, Tønnessen T, Aakhus S, Gullestad L (Oslo University Hospital Rikshospitalet; University of Oslo; Oslo University Hospital Rikshospitalet; Oslo University Hospital Ullevål; Oslo University Hospital Rikshospitalet; and Oslo University Hospital Ullevål, Oslo, Norway). Osteoprotegerin levels predict mortality in patients with symptomatic aortic stenosis. J Intern Med 2011; 270: 452–460.

Objectives.  To examine the prognostic value of osteoprotegerin (OPG) levels in relation to all-cause mortality in patients with symptomatic severe aortic stenosis (AS).

Design.  We measured plasma OPG levels in 136 patients with symptomatic severe AS and investigated associations with transvalvular gradients, valve area, valve calcification (using ultrasonic backscatter analysis as an estimate) and measures of heart failure. Then, we assessed the prognostic value of elevated plasma OPG in determining all-cause mortality (= 29) in these patients.

Results.  Elevated OPG was poorly correlated with the degree of AS but was associated with increased backscatter measurements and impaired cardiac function. Furthermore, OPG was associated with all-cause mortality in patients with symptomatic AS, even after adjustment for conventional risk markers. The strongest association was obtained by using a combination of high levels of both OPG and N-terminal pro-brain natriuretic peptide (NT-proBNP), suggesting that these markers may reflect distinct pathways in the development and progression of AS.

Conclusion.  The level of circulating OPG is significantly associated with all-cause mortality alone and in combination with NT-proBNP in patients with severe symptomatic AS.


Calcific aortic stenosis (AS) is a progressive disease that shares many characteristics with atherosclerosis [1, 2]. Early in the disease process, active microscopic areas of calcification are seen co-localizing in areas of accumulation of lipoprotein and infiltration of inflammatory cells [1–3]. As the disease progresses, active bone formation is seen, involving cells with chondrocytic and/or osteoblastic characteristics capable of phenotypic differentiation and spontaneous calcification [3, 4]. Thus, there are similarities between vascular and skeletal calcification, suggesting a regulatory role for osteogenic and calcitropic factors in the development of AS. Indeed, major proteins involved in the regulation of tissue calcification have been detected in calcified valvular tissue [3].

One of the mediators with a role in tissue calcification that has been detected in AS is osteoprotegerin (OPG), a soluble member of the tumour necrosis factor (TNF) receptor superfamily, with pleiotropic effects on bone metabolism, endocrine function and the immune system [5, 6]. OPG acts as a decoy receptor for receptor activator of NFκB ligand (RANKL), inhibiting the interaction between RANKL and its receptor, RANK. It has been reported that RANKL and OPG expressions are related to the degree of calcification in AS [7–9]. In vitro, RANKL promotes transition towards an osteogenic phenotype in cultured human aortic valve myofibroblasts, and it has been suggested that the OPG/RANKL/RANK pathway may be involved in valvular calcification in AS [9].

Patients with symptomatic AS develop increased left ventricular (LV) outflow obstruction and LV hypertrophy to compensate and maintain ejection performance, and there is an increased mortality rate after the onset of symptoms [10]. Recent studies have shown that plasma levels of N-terminal pro-brain natriuretic peptide (Nt-proBNP) and C-reactive protein (CRP), markers of LV dysfunction and systemic inflammation, respectively, are related to disease severity, progression and survival in patients with AS [11–13]. Increased circulating levels of OPG have also been demonstrated in a range of cardiovascular (CV) disorders including atherosclerosis [14] and heart failure (HF) [15], and the results of several cohort studies in selected patient groups at high CV risk suggest that OPG can provide independent prognostic information on all-cause and CV mortality [14, 16, 17]. However, data on OPG levels in patients with AS with and without accompanying HF are lacking.

To further understand the relationship between OPG levels and disease severity in AS, we analysed plasma levels of this soluble receptor in patients with AS and examined the relationship between OPG and transvalvular gradients, valve area and measures of HF, as well as the prognostic value of OPG in relation to all-cause mortality in these patients.


Cross-sectional study

A total of 136 patients with symptomatic AS, evaluated for aortic valve replacement (AVR) surgery between May 2005 and January 2007 at the Department of Cardiology, Oslo University Hospital Rikshospitalet, were consecutively enroled in the study (Table 1). Only patients with confirmed AS were included. Echocardiographic parameters and blood samples were obtained from each patient. Coronary angiography was performed in all patients to diagnose the presence of concomitant coronary artery disease [CAD; i.e. disease in at least one vessel (>50% narrowing of luminal diameter)]. The exclusion criteria were severe (grade III) aortic or mitral regurgitation or serum creatinine >150 μmol L−1. All investigations were made within a period of a few days. Of the 136 patients, 108 were scheduled for AVR, whereas surgical intervention was not scheduled in the remaining 28 because of comorbidity/high risk for operation (= 19), unwillingness of the patient to undergo surgery (= 4), or because clinical benefit was uncertain owing to less severe symptoms (low gradient, = 5). Two patients amongst those scheduled for AVR died pending surgery, thus 106 patients underwent surgical intervention.

Table 1. Patient characteristics with relation to OPG levels in 136 patients with symptomatic aortic stenosis
 Total populationOPG ≤ 6.69 ng mL−1 (= 102)OPG > 6.69 ng mL−1 (= 34)P-value
  1. Values given as percentage, mean ± SD, amedian and interquartile range. OPG, osteoprotegerin; BMI, body mass index; NYHA, New York Heart Association; Functional Classification eGFR, estimated glomerular filtration rate; DM2, type 2 diabetes mellitus; LVEF, left ventricular ejection fraction; HDL, High-density lipoprotein; Ch, cholesterol; LDL, Low-density lipoprotein; CRP, C-reactive protein; ACE, angiotensin-converting enzyme; ARB, angiotensin receptor blocker. Full datasets were not available for all measures.

Age (years)74 ± 1071 ± 1081 ± 5<0.001
Male (%)5558500.34
BMI (kg m−2)26.3 ± 4.326.9 ± 4.525.2 ± 3.80.035
NYHA functional class I/II/III/IV (%)5/33/61/14/34/62/06/32/58/40.25
Coronary artery disease (%)4337560.033
Current smokers (%)3334320.96
DM2 (%)1111130.72
Hypertension %2524280.63
Atrial fibrillation (%)3425500.002
 HDL-Ch (mmol L−1)a1.6 (1.3, 1.9)1.6 (1.3, 1.9)1.5 (1.2, 1.9)0.43
 LDL-Ch (mmol L−1)a3.0 (2.4, 3.9)3.1 (2.5, 3.9)2.8 (2.3, 4.0)0.47
 eGFRa66 (52, 86)74 (58, 95)52 (42, 62)<0.001
 CRP (mg L−1)a1.9 (0.9, 4.4)1.6 (0.9, 3.9)3.0 (0.9, 5.6)0.069
Medication (%)
 ACE inhibitor1415130.69

Longitudinal study

Serial measurements of OPG were taken in a cohort of 20 patients (76 ± 1 years, eight men) with severe AS (mean aortic gradient >50 mmHg or aortic area <0.7 cm2) who underwent AVR surgery at the Department of Cardiothoracic Surgery, Oslo University Hospital Ullevål. Echocardiographic measurements were taken preoperatively, on the second postoperative day and at 6 and 12 months after AVR. A detailed description of this cohort has previously been provided [18].


Informed consent was obtained from each study subject, and the study protocol was approved by the regional committee for ethics in medicine.


Continuous wave Doppler ultrasound from multiple positions was used to obtain the maximum aortic annular blood flow velocities and to calculate aortic valve area using the continuity equation [19]. Doppler echocardiographic calculations of stroke volume and cardiac output (CO) were made on the basis of the aortic annular cross-sectional and flow velocity profile. Left ventricular ejection fraction (LVEF) was obtained by the biplane Simpson method [20]. Peak systolic right ventricular pressure was estimated from the maximal tricuspid regurgitative blood flow velocity. Dimensional and velocity parameters were averaged from at least three (at least five in atrial fibrillation) cardiac cycles. To obtain a semiquantitative measure of the morphology of the stenotic aortic valve, ultrasound backscatter data analysis was performed as described by Ngo et al. [21]. Observers were blinded to the clinical status of the patient and the standard echo findings.

Biochemistry and blood sampling

Peripheral venous blood was drawn into pyrogen-free tubes with EDTA as anticoagulant. The tubes were immediately immersed in melting ice and centrifuged within 30 min at 2000 g for 20 min to obtain platelet-poor plasma. All samples were stored at −80 °C and had been thawed once prior to assay. NT-proBNP and CRP were assayed on a MODULAR platform (Roche Diagnostics, Basel, Switzerland). Plasma levels of low-density lipoprotein cholesterol, high-density lipoprotein cholesterol and creatinine were measured enzymatically using a Roche/Hitachi 917 analyser (Roche Diagnostics, Mannheim, Germany). Plasma OPG levels were measured by enzyme immunoassay (R&D Systems, Stillwater, MN, USA) as previously described [22]. Briefly, the mean recovery of two samples spiked with different concentrations of recombinant OPG was 93%, range 78 ± 101%. The intra- and inter-assay coefficients of variation were 3.6% and 10.6%, respectively. The sensitivity, defined as the mean ± 3SDs of the zero standard, was calculated to be 15 pg mL−1. Serial dilution of two samples 1 : 1 ± 1 : 8 gave a recovery of 126% and 127% at a dilution of 1 : 8. The CV% between fresh samples and aliquots that were thawed once, twice and three and four times was 2.4 ± 1.1 (mean ± SD), 2.8 ± 0.9, 2.7 ± 0.6 and 2.7 ± 1.0, respectively. The CV% between the immediately frozen sample and the samples exposed to room temperature for 1, 4 and 24 h was 2.8 ± 1.8, 4.1 ± 2.0 and 4.1 ± 2.0, respectively.


We chose to use parametric statistics on all measures because ultimately they would be included in a multiple regression model. Variables not normally distributed as evaluated by the Kolmogorov–Smirnov test were log-transformed for statistical analysis but may be presented as nontransformed data. Differences between groups were analysed using the Student’s t-test. Paired samples with more than two time-points were analysed with repeated measures anova followed by paired samples t-test if significant. Relationships between variables were tested by simple linear (bivariate) regression analysis (Pearson correlation). The importance of OPG as a risk factor for all-cause mortality was investigated by multivariable analyses including, besides OPG, variables significantly associated with mortality in Table 1. Cox proportional hazard analysis was performed to estimate hazard ratios (HRs) using forward stepwise conditional and forced-entry methods. Follow-up time for all-cause mortality was calculated from time of inclusion to death from any cause. Receiver-operating characteristics (ROC) curves were generated to evaluate the accuracy of prediction of all-cause mortality. Kaplan–Meier analysis with log-rank test was performed to compare the number of events in different groups (comparisons pooled over strata). Multicollinearity was evaluated by examining the tolerance and variance inflation factor (VIF) in linear regression. The tolerance was >0.4 and the VIF <2.5 for OPG, as well as the other independent variables. P-values are two-sided and considered significant at <0.05.


OPG levels in patients with AS

The characteristics of patients according to OPG levels comparing the fourth quartile to the lower three are presented in Tables 1 and 2. Amongst patients with AS, those with the highest OPG levels were older and had a lower body mass index, a higher incidence of CAD and atrial fibrillation and higher creatinine levels (Table 1). Further investigation into OPG levels (as a continuous variable) showed significantly increased levels in patients with CAD [combined (P<0.001) and assessed separately in the left anterior descending (= 0.001), left circumflex (P = 0.007) and right coronary arteries (P = 0.004)]. By contrast, there was no association between OPG levels and traditional (i.e. smoking, type 2 diabetes mellitus, hypertension, lipid parameters) risk markers for CAD. Moreover, although patients with high OPG levels were more frequently using warfarin and aspirin, the use of statins, β-blockers and medications that interfere with the angiotensin system showed no association with OPG levels (Table 1).

Table 2. Echocardiographic and neurohormonal characteristics with relation to OPG levels in 136 patients with symptomatic aortic stenosis
 Total populationOPG ≤ 6.69 ng mL−1 (= 102)OPG > 6.69 ng mL−1 (= 34)P-value
  1. Data are given as mean ± SD, amedian and interquartile range. OPG, osteoprotegerin; LVEF, left ventricular ejection fraction; CO, cardiac output.; NT-proBNP, N-terminal pro-brain natriuretic peptide. To convert NT-proBNP values from pmol L−1 to pg mL−1, multiply by 8.47.

 LVEF (%)62 ± 1263 ± 1261 ± 130.39
 CO (mmHg)a4.8 (4.2, 5.6)4.9 (4.4, 5.8)4.4 (4.1, 5.2)0.015
 Aortic valve area (cm2)a0.62 (0.50, 0.80)0.69 (0.50, 0.80)0.60 (0.50, 0.70)0.13
 Mean aortic gradient (mmHg)53.5 ± 20.253.5 ± 19.254.0 ± 22.00.90
 Backscatter (dB)18.8 ± 5.017.9 ± 4.520.4 ± 5.60.007
 NT-proBNP (pmol L−1)a107 (42, 291)76 (27, 163)208 (59, 611)<0.001

OPG levels in relation to echocardiographic and neurohormonal features

Evaluation of the association between high OPG levels (fourth vs. lower three quartiles) and echocardiographic and neurohormonal characteristics of AS patients revealed two significant findings (Table 2). First, although there was no association between high plasma OPG levels and LVEF, aortic valve area or mean pressure gradient across the aortic valve, elevated OPG levels were associated with decreased CO and increased NT-proBNP levels. Secondly, measurement of backscatter has been shown to correlate well with the degree of calcification in atherosclerotic human aorta both in vivo and in vitro [23–25], and notably, those with high OPG levels had the highest backscatter measurements (Table 2).

OPG levels following AVR

To further characterize OPG in relation to AS, plasma levels were measured before and after AVR in 20 patients with severe AS (Fig. 1). We first evaluated the temporal course of OPG and found a significant increase postoperatively (= 0.011), with a slight increase also seen at both 6 (= 0.065) and 12 months (= 0.042). When investigating associations between OPG and echocardiographic parameters, we found no association between OPG levels, LVEF or the mean aortic gradient at 6 or 12 months. No associations were found between the changes in OPG and the changes in these variables (LVEF or the mean aortic gradient) either. However, the change in OPG at 6 months was strongly correlated with both the aortic valve area and the change in aortic valve area at 6 months (Fig. 1b). Similar but more modest associations were observed at 12 months (change in OPG and aortic valve area at 12 months, = −0.45, = 0.059; change in OPG and change in aortic valve area at 12 months, = −0.44, = 0.069). Thus, patients with the largest increase in OPG have the lowest increase in valve area following surgery, suggesting that higher OPG levels are associated with residual LV stress after AVR.

Figure 1.

Osteoprotegerin (OPG) levels following aortic valve replacement (AVR). (a) Temporal course of OPG and aortic valve area over 1 year following AVR. (b) Correlations between change in OPG and change in aortic valve area from preoperative levels to 6 months (left graph) and between change in OPG levels and actual levels of aortic valve area at 6 months after AVR (right graph). *P<0.05, ***P<0.001 vs. before AVR.

OPG and all-cause mortality

During a mean follow-up of 37 months (range 1–54 months), 29 patients died, 12 in the nonsurgical group and 17 in the group of patients who underwent surgery. Figure 2a shows a strong association between the concentration of OPG and all-cause mortality, especially in the fourth quartile. This corresponds well to the ROC curve analysis (Fig. 2b). With alpha = 0.05 and considering mortality proportions in the top quartile (proportion = 0.5) versus the three lower quartiles (proportion = 0.13), a power estimate of 0.94 is obtained. Unadjusted Cox regression demonstrated that age, diabetes mellitus, atrial fibrillation, estimated glomerular filtration rate, LVEF, aortic valve area and CRP and NT-proBNP levels were significantly associated with all-cause mortality (Table 3). When these together with OPG were included in a multivariable forced Cox regression, diabetes, NT-proBNP and OPG were found to be significantly associated with all-cause mortality with the highest Wald value for OPG (Table 3). A stepwise regression identified the same factors. Presence of CAD was not associated with all-cause mortality and, when included in the models, had no impact on the association between OPG and survival. Considering patients with and without AVR separately, a significant association between OPG and mortality was observed regardless of AVR [AVR 2.78 (1.53–5.05) = 0.001; non-AVR 6.55 (2.09–20.55) = 0.001]. Stepwise multivariable analysis as conducted earlier showed that OPG remained a significant predictor of all-cause mortality in both subgroups [AVR 2.75 (1.43–5.27) = 0.002; non-AVR 4.86 (1.11–21.27) = 0.036]. The two patients who died during surgery had OPG levels in the second and third quartile, and a significant association between OPG and all-cause mortality was seen even when these two patients were excluded from the analyses (data not shown).

Figure 2.

Association between circulating osteoprotegerin (OPG) and all-cause mortality in patients with symptomatic aortic stenosis. (a) Kaplan–Meier curves showing the cumulative incidence of all-cause mortality (n  = 29) during the whole study period (mean follow-up 37 months, range 1–54), according to quartiles of OPG at enrolment. (b) Receiver-operating characteristics curve analysis for the predictive value of OPG, N-terminal pro-brain natriuretic peptide and CRP for all-cause mortality. Values show area under the curve (and 95% confidence interval).

Table 3. Multivariable analyses: OPG as an independent predictor of all-cause mortality in patients with symptomatic aortic stenosis
VariableUnadjustedMultivariable (forced)Multivariable (stepwise)
β (SD)HR (95% CI)Pβ (SD)HR (95% CI)WaldPβ (SD)HR (95% CI)WaldP
  1. CRP, C-reactive protein; eGFR, estimated glomerular filtration rate; NT-proBNP, N-terminal pro-brain natriuretic peptide; OPG, osteoprotegerin.

Age/100.83 (0.29)2.29 (1.30–4.01)0.004−0.06 (0.48)0.94 (0.37–2.41)0.00.901    
Diabetes mellitus1.20 (0.44)3.31 (1.41–7.76)0.0061.05 (0.63)2.87 (0.83–9.90)2.80.0961.25 (0.47)3.50 (1.40–8.74)7.20.007
Atrial fibrillation0.80 (0.37)2.22 (1.07–4.61)0.0320.07 (0.45)1.07 (0.44–2.60)0.00.885    
[Loge] eGFR−0.50 (0.17)0.61 (0.44–0.85)0.0040.06 (0.26)1.06 (0.64–1.75)0.00.829    
[Loge] ejection fraction−0.39 (0.13)0.67 (0.53–0.87)0.002−0.03 (0.20)0.97 (0.65–1.44)0.00.869    
[Loge] aortic valve area−0.55 (0.24)0.58 (0.36–0.92)0.022−0.18 (0.27)0.84 (0.50–1.41)0.40.508    
[Loge] CRP0.58 (0.19)1.78 (1.24–2.56)0.0020.11 (0.20)1.12 (0.75–1.64)0.30.579    
[Loge] NT-proBNP1.25 (0.25)3.50 (2.17–5.66)<0.0010.74 (0.31)2.11 (1.15–3.86)5.80.0160.85 (0.41)2.33 (1.41–3.86)10.80.001
[Loge] OPG1.26 (0.25)3.53 (2.15–5.80)<0.0011.16 (0.38)3.20 (1.52–6.68)9.40.0021.11 (0.29)3.04 (1.71–5.40)14.2<0.001

Combinations of OPG and NT-proBNP in the prediction of all-cause mortality

Figure 3 shows HRs for different combinations of OPG and NT-proBNP quartiles in relation to their association with all-cause mortality. HRs increased with a combination of higher quartiles (HR 21.0 for patients with both OPG and NT-proBNP in the fourth quartile). Finally, stepwise multivariable regression including the above-mentioned predictors from Table 3 as well as the interaction term of OPG and NT-proBNP identified that the interaction term showed the strongest association with all-cause mortality [Beta 0.08, SE 0.01, Wald 30.7, HR 1.08 (1.05–1.11)].

Figure 3.

Increased risk prediction for all-cause mortality in symptomatic aortic stenosis using combinations of osteoprotegerin and N-terminal pro-brain natriuretic peptide.


Previously, Helske et al. [26] reported an increase in circulating OPG levels in AS patients with HF, and in the present study, we have extended this finding in several ways. First, we have shown that raised OPG levels are not only associated with markers of LV dysfunction and HF, but also with backscatter as a marker of valve calcification. Secondly, plasma OPG is associated with all-cause mortality in patients with symptomatic AS, also after adjustment for conventional risk markers. Thirdly, NT-proBNP was correlated with OPG and was itself associated with all-cause mortality, but the strongest association was obtained for a combination of these markers (i.e. high levels of both NT-proBNP and OPG). Our findings suggest that OPG should be further investigated as a risk marker in patients with AS.

OPG in symptomatic AS

Since its initial discovery as a key regulator in bone metabolism, OPG has become the subject of intense interest for its role in vascular disease and calcification. Thus, several recent studies have demonstrated the OPG/RANKL/RANK pathway in human aortic valves and provided data to suggest that OPG may act as a calcification inhibitor by limiting RANKL-induced matrix calcification [7–9, 26]. Our findings showing an association between plasma OPG and valvular calcification as estimated by backscatter analysis in symptomatic AS patients may seem at odds with these previous reports. However, although OPG at the cellular level within a sclerotic aortic valve may inhibit RANKL activity, circulating OPG levels have been suggested to be a stable, reliable and overall measure of the activity in the OPG/RANKL/RANK axis, and even better than RANKL that circulates at low levels. Moreover, circulating OPG appears to be a reliable marker not only of vascular calcification, but also of inflammation, potentially mirroring two interacting pathogenic processes in the development of AS. The lack of a significant correlation between CRP and OPG does not exclude the possibility that OPG is a marker of inflammation but rather underscores the fact that inflammation involves a multitude of pathways and is unlikely to be reflected by one marker.

OPG is associated with indices of HF and all-cause mortality in symptomatic AS

We have recently demonstrated increased myocardial and circulating OPG levels in experimental and human HF, significantly correlated with neuroendocrine activation and LV dysfunction in clinical HF [15]. Thus, a deteriorating LV function could potentially contribute to the increased plasma OPG in our patients, as also suggested by Helske et al. [26] who demonstrated increased systemic OPG levels in patients with severe AS and HF because of LV overload. Indeed, we found a significant correlation between OPG and both neuroendocrine (i.e. NT-proBNP) and haemodynamic (i.e. CO) measures of HF in patients with symptomatic AS. However, the association between OPG and all-cause mortality was seen after adjustment for NT-proBNP, suggesting that these markers may reflect overlapping, but also distinct pathways in the development and progression of AS. Additionally, the combination of high levels of both OPG and NT-proBNP was strongly associated with all-cause mortality, thus providing more information together than when either of these markers was used alone.

In the current study, we showed that OPG was associated with mortality in patients with symptomatic AS, also after adjustment for conventional risk markers. Notably, this association between high OPG levels and all-cause mortality was observed in patients who underwent AVR as well as in those who did not. Thus, OPG could be of interest as a risk marker not only in untreated AS patients, but also in patients who undergo AVR, as a preoperative marker for long-term risk evaluation. However, the total mortality in those who underwent AVR was relatively low, and these associations should be interpreted with caution. Also, although we examined OPG levels during longitudinal testing after AVR, this study population was small, and OPG levels following surgery do not necessarily reflect levels during spontaneous progression of AS. A larger study population, including both asymptomatic and symptomatic patients, assessed longitudinally, would reveal whether plasma OPG assessment may benefit patient management and elucidate the role of OPG in the spontaneous progression of AS.


Circulating OPG is strongly associated with all-cause mortality alone and in combination with NT-proBNP in symptomatic AS patients. Future studies should evaluate the predictive usefulness of this marker in larger patient populations, including in subjects with asymptomatic disease.

Conflict of interest statement

None of the authors has any conflicts of interest to declare.

Source of funding

This work was supported by grants from the Norwegian Council of Cardiovascular Research, Helse Sør-øst and Norwegian Research Council. The authors also thank an anonymous benefactor.