Journal of the American College of Cardiology Vol. 47, No.

9, 2006
© 2006 by the American College of Cardiology Foundation ISSN 0735-1097/06/$32.00
Published by Elsevier Inc. doi:10.1016/j.jacc.2005.12.054

Cardiac Imaging

The Relationship Between Plasma

Osteoprotegerin Levels and Coronary Artery
Calcification in Uncomplicated Type 2 Diabetic Subjects
Dhakshinamurthy Vijay Anand, MBBS, MRCP,*†‡ Avijit Lahiri, MBBS, MSC, MRCP, FACC, FESC,*
Eric Lim, MBCHB, MA, MRCP,* David Hopkins, BSC, MBCHB, FRCP,§
Roger Corder, PHD, MRPHARMS†
London, United Kingdom
OBJECTIVES This study sought to prospectively evaluate the relationship between plasma osteoprotegerin
(OPG), inflammatory biomarkers (high-sensitivity C-reactive protein [hs-CRP],
interleukin-6 [IL-6], coronary artery calcification (CAC), and cardiovascular events in
patients with type 2 diabetes.
BACKGROUND Arterial calcification is a prominent feature of atherosclerosis and is associated with an
increased risk of cardiovascular events. Osteoprotegerin is a cytokine that has recently been
implicated in the regulation of vascular calcification.
METHODS A total of 510 type 2 diabetic patients (53 ⫾ 8 years; 61% male) free of symptoms of
cardiovascular disease were evaluated by CAC imaging. Risk factors, hs-CRP, IL-6, and
OPG levels were measured. Patients were followed up for cardiovascular events (cardiac
death, myocardial infarction, acute coronary syndrome, late revascularization, and nonhem-
orrhagic stroke).
RESULTS Significant CAC (⬎10 Agatston units) was seen in 236 patients (46.3%); OPG was
significantly elevated in patients with increased CAC. In multivariable analyses, OPG
retained a strong association with elevated CAC scores after adjustment for age, gender, and
other risk factors (odds ratio ⫽ 2.84, 95% confidence interval 2.2 to 3.67; p ⬍ 0.01). Sixteen
cardiovascular events occurred during a mean follow-up of 18 ⫾ 5 months. The waist-to-hip
ratio, United Kingdom Prospective Diabetes Study (UKPDS) risk score, OPG level, and
CAC score were significant predictors of time to cardiovascular events in a univariate Cox
proportional hazards model. In the multivariate model, the CAC score was the only
independent predictor of adverse events. Levels of hs-CRP and IL-6 were related to neither
the extent of CAC nor short-term events.
CONCLUSIONS A high proportion of asymptomatic diabetic patients have significant subclinical atheroscle-
rosis. Of the biomarkers studied, only OPG predicted both subclinical disease and near-term
cardiovascular events. Therefore, measurement of OPG merits further investigation as a simple
test for identifying high-risk type 2 diabetic patients. (J Am Coll Cardiol 2006;47:1850 –7)
© 2006 by the American College of Cardiology Foundation

Arterial calcification is a prominent feature of atherosclero- and/or duration of diabetes and typically affects arteries that
sis that can be readily quantified noninvasively by radio- are less prone to atherosclerosis (4). Although some reports
graphic imaging techniques such as electron beam com- have suggested that it may be associated with an increased
puted tomography (EBCT) (1). Clinical studies have risk of cardiovascular complications (5), this is not a
reproducibly shown that high amounts of coronary artery consistent finding (6). Notably, medial calcification is un-
calcification (CAC) predict an increased risk of myocardial usual in the coronary arteries, implying that when calcifica-
infarction and sudden coronary death (2). In patients with tion is observed in the coronary arteries, it is almost certainly
diabetes, arterial calcification can occur either as a compo- associated with intimal plaque.
nent of atheroma (intimal calcification) or in the absence of It has been shown that calcification in intimal atheroscle-
atherosclerotic plaque (medial calcification or Monckeberg rotic lesions is not merely a passive consequence of chronic
sclerosis) (3). Medial calcification is related to the severity vascular inflammation but an active process that may lead to
a positive feedback loop of calcification and inflammation,
From the *Cardiac Imaging and Research Centre, Wellington Hospital; †William driving atherosclerotic disease progression further (7). Con-
Harvey Research Institute, Barts and The London, Queen Mary’s School of Medicine temporary studies have shown similarities between vascular
and Dentistry; ‡Department of Cardiology, Royal Free Hospital; and the §Depart-
ment of Endocrinology, Kings College Hospital, London, United Kingdom. Sup- and skeletal calcification suggesting a regulatory role for
ported by research grants from the Michael Tabor Foundation, Harrow Cardiovas- osteogenic and calcitropic factors (8) in the development of
cular Research Trust, GE Healthcare Ltd., and the Derrick Smith research grant. Dr. cardiovascular disease. Very recently, osteoprotegerin
Lahiri is on the advisory board of GE Healthcare Ltd.
Manuscript received July 10, 2005; revised manuscript received November 24, (OPG), a key factor in bone remodeling, a member of the
2005, accepted December 5, 2005. tumor necrosis factor receptor superfamily (9), and a decoy
JACC Vol. 47, No. 9, 2006 Anand et al. 1851
May 2, 2006:1850–7 Osteoprotegerin and Coronary Calcium in Diabetes

weight, body mass index, waist-to-hip ratio, blood pressure,

Abbreviations and Acronyms and a 12-lead electrocardiogram were also obtained.
AU ⫽ Agatston units Biochemical measurements. Blood and urine samples
CAC ⫽ coronary artery calcification were obtained for laboratory testing (Diabetes Control
CAD ⫽ coronary artery disease
and Complications Trial [DCCT]-aligned hemoglobin
EBCT ⫽ electron beam computed tomography
hs-CRP ⫽ high-sensitivity C-reactive protein A1c, lipid profile, urea, creatinine, and urine albumin/
IL ⫽ interleukin creatinine ratio). Serum hs-CRP (ICN Pharmaceuticals,
IQR ⫽ interquartile range Orangeburg, New York), IL-6 (R&D systems Inc., Min-
OPG ⫽ osteoprotegerin neapolis, Minnesota), and plasma OPG levels (Biomedica,
OR ⫽ odds ratio
UKPDS ⫽ United Kingdom Prospective Diabetes Study
Vienna, Austria) (16) were measured by enzyme-linked
immunosorbent assay. Intra-assay and inter-assay coeffi-
cients of variation for hs-CRP, IL-6, and OPG measure-
receptor for the receptor activator of nuclear factor-␬␤ ments were 3.3% and 3.7%, 6.4% and 7.8%, and 5% and
ligand (RANKL), was implicated in human atherogenesis 7.5%, respectively.
(10). Clinical studies have shown that elevated OPG levels CAC imaging. The CAC imaging was performed using an
are associated with the progression of vascular calcification EBCT scanner (GE Imatron C-150, San Francisco, Cali-
in patients on long-term hemodialysis (11), growth of fornia) equipped with high-resolution detectors. Forty con-
human abdominal aortic aneurysms (12), the presence and tiguous 3-mm slices were obtained at 100-ms scan time per
severity of coronary artery disease (CAD) (13), and cardio- slice in a single breath-hold, starting at the carina and
vascular risk in the general population (14). proceeding to the level of the diaphragm. Scan time was
Because type 2 diabetes increases the risk of coronary 100-ms per slice, synchronized to 40% of the R-R interval.
atherosclerosis and cardiovascular events, the objectives of The CAC scores were calculated by an experienced inves-
our study were: 1) to prospectively evaluate the relationship tigator blinded to clinical risk factor information on an
between plasma OPG levels, the extent of coronary artery Aquarius workstation (TeraRecon Inc., San Mateo, Cali-
calcification (measured by EBCT), and short-term cardio- fornia). Scores were classified into five categories, based on
vascular events, and 2) to determine the relationship be- cut-offs that have been widely used in the literature: ⱕ10
tween OPG and established cardiovascular risk factors, (minimal or insignificant CAC), 11 to 100 (mild CAC),
biomarkers of vascular inflammation (high sensitivity 101 to 400 (moderate CAC), 401 to 1000 (severe CAC),
C-reactive protein [hs-CRP] and interleukin [IL]-6). and ⬎1,000 (extensive CAC) (2).
Patient management and follow-up. Attending physi-
METHODS cians were informed of study results; management was left
Patients. Asymptomatic type 2 diabetic patients without to their discretion. Procedures for follow-up were defined a
prior evidence of cardiovascular disease were prospectively priori. Telephone interviews and review of medical notes
recruited from four diabetes clinics in secondary care were used to determine the occurrence of cardiovascular
(Northwest London). Inclusion criteria were presence of events (cardiac death, myocardial infarction, acute coronary
type 2 diabetes of ⬎1 year duration and age at onset ⱖ35 syndrome [defined by electrocardiographic changes and/or
years. Exclusion criteria were age ⬍35 years or ⬎65 years, myocardial markers], late coronary revascularization [⬎60
documented CAD, typical angina pectoris, abnormal rest- days after CAC scan], or nonhemorrhagic stroke). Subjects
ing electrocardiogram (e.g., Q waves, left bundle branch who underwent coronary revascularization within 60 days
block), cerebrovascular or peripheral arterial disease, renal after the CAC scan were excluded from survival analysis to
impairment (creatinine ⬎1.4 mg/dl), or serious life- eliminate events that were driven by the results of CAC
threatening illness. Diabetes was diagnosed according to imaging. In cases in which a study participant experienced
American Diabetes Association criteria. Participants were more than one end point, only the first end point was
considered to have type 2 diabetes if they had no history of counted.
ketoacidosis and if they did not start insulin treatment either Statistical analysis. Statistical analyses were carried out
in the 2 years after diagnosis or before the age of 40 years. using the Stata V7 statistical package (Statacorp LP, Col-
The study was approved by the local research ethics com- lege Station, Texas) or SPSS version 12.0 (SPSS Inc.,
mittees of participating hospitals. Detailed medical history Chicago, Illinois). Continuous variables were summarized
including the presence of cardiovascular risk factors, dura- as mean ⫾ standard deviation or median ⫾ interquartile
tion of diabetes, presence of microvascular complications, range (IQR). Bivariate correlations between OPG levels,
treatment history, and the predicted 10-year absolute coro- established cardiovascular risk factors, and inflammatory
nary heart disease risk based on the Framingham risk markers (hs-CRP and IL-6) were performed using the
function and United Kingdom Prospective Diabetes Study Pearson correlation for normally distributed variables and the
(UKPDS) risk engine (15) were recorded at baseline. A Spearman rank correlation for variables distributed otherwise.
physical examination was performed, and the height, The distribution of variables was assessed examining the
1852 Anand et al. JACC Vol. 47, No. 9, 2006
Osteoprotegerin and Coronary Calcium in Diabetes May 2, 2006:1850–7

Table 1. Patient Characteristics A univariate and multivariate Cox proportional hazards

Variable Number Percentage model was devised to estimate the time to cardiovascular
events. From the univariate Cox models, survival curves
Male 309 60.6
Ethnicity were calculated based on CAC scores and OPG levels. For
Caucasian 110 21.6 this analysis, CAC scores were classified into four catego-
South Asian 273 53.5 ries, namely ⱕ100, 101 to 400, 401 to 1,000, and ⬎1,000
Black 115 22.5 Agatston units (AU), and biochemical marker values (hs-
Other (Chinese/Middle Eastern) 12 2.4
CRP, IL-6, and OPG) were classified into tertiles. Addi-
Blood pressure ⬎140/90 mm Hg or 380 74.5
antihypertensive medication tionally, a receiver operating characteristics curve was plot-
Treated with blood pressure ⬍140/90 mm Hg 104 20.4 ted for plasma OPG, CAC, and UKPDS and Framingham
Treated with blood pressure ⬎140/90 mm Hg 140 27.4 risk scores with the end point of cardiovascular events, thus
Untreated (blood pressure ⬎140/90 mm Hg) 136 26.7 evaluating the ability of each variable to classify the primary
Hyperlipidemia or lipid-lowering medication 318 62.4
end point of cardiovascular outcomes. The area under the
Treated with total cholesterol ⬍200 mg/dl 133 26.1
Treated with total cholesterol ⬎200 mg/dl 64 12.6 curve and 95% confidence intervals were calculated for this
Untreated (total cholesterol ⬎200 mg/dl) 121 23.7 receiver operating characteristics curve.
Family history of premature CAD 158 31
Smoking RESULTS
Past 110 21.6
Current 97 19 The first 510 consecutive patients who fulfilled the eligibil-
Insulin only 57 11.2 ity criteria were enrolled (mean age 53 ⫾ 8 years, 61%
Insulin ⫹ oral agent 51 10
male). Patient characteristics and biochemical variables are
Oral agent 434 85.1
Diet only 19 3.7 shown in Table 1.
Microalbuminuria 74 14.5 Distribution of CAC. A total of 273 patients (53.5%) had
Proteinuria 40 7.8 minimal or insignificant CAC (ⱕ10 AU). The remaining
Peripheral neuropathy 90 17.6 236 patients (46.3%) had varying degrees of coronary
Retinopathy 116 22.7
calcification. The CAC was mild (11 to 100) in 100 patients
Mean SD (19.6%), moderate (101 to 400) in 77 (15.1%), severe (401
Age (yrs) 52.7 8.4 to 1,000) in 31 (6.1%), and extensive (⬎1,000) in 28 (5.5%).
Duration of diabetes (yrs) 8 6 Image quality was poor in one patient because of significant
BMI 28.54 5.04 respiratory motion artifacts.
Waist-to-hip ratio 0.94 0.08 Ethnic variation in CAC. There was a high proportion of
CAC 6.2* 0–115.8*; 0–5725†
DCCT-aligned 8.2 1.7 South Asians compared with white Caucasians and black
hemoglobin A1c (%) African/Caribbeans (Table 1) because of the geographic
Total cholesterol (mg/dl) 187 36 location of the study population and the high prevalence of
LDL cholesterol (mg/dl) 107 31
HDL cholesterol (mg/dl) 50 16
Triglycerides (mg/dl) 167 104
hs-CRP (mg/l) 4.28* 1.72–8.01*
IL-6 (pg/ml) 1.93* 1.2–3.1*
Osteoprotegerin (pmol/l) 6.82* 4.68–11.55*
*Median values and interquartile ranges of CAC score, hs-CRP, IL-6, and osteo-
protegrin. †The overall range of CAC score.
BMI ⫽ body mass index; CAC ⫽ coronary artery calcification; CAD ⫽ coronary
artery disease; DCCT ⫽ Diabetes Control and Complications Trial; HDL ⫽
high-density lipoprotein; hs-CRP ⫽ high-sensitivity C-reactive protein; IL ⫽
interleukin; LDL ⫽ low-density lipoprotein.

frequency histograms and the normal plots, and by using the

Shapiro-Wilk test. Log transformation was applied to
normalize the distribution of biomarker values. Ordered
logistic regression was performed to identify clinical predic-
tors of coronary calcification and evaluate the association
between OPG and CAC. A backward selection procedure
was applied to identify candidate variables for the multivar-
iate model. Candidate variables included, from the pool of Figure 1. Box and whisker plots illustrate the median and interquartile
historical variables, risk factors, and laboratory variables, ranges of osteoprotegerin (OPG) in patients with minimal, mild, moder-
those univariate predictors with p ⱕ 0.2. For the multivar- ate, severe, and extensive coronary calcification respectively. Open circles
show outliers (values between 1.5 and 3 box lengths from the upper edge
iate model, only those variables with a p ⬍ 0.05 were of the box) and asterisks show extreme outliers (values ⬎3 box lengths
retained for the final variable selection. from the upper edge of the box).
JACC Vol. 47, No. 9, 2006 Anand et al. 1853
May 2, 2006:1850–7 Osteoprotegerin and Coronary Calcium in Diabetes

Table 2. Univariate Predictors of Increased Coronary confidence interval [CI] 2.42 to 3.92; p ⬍ 0.001) (Fig. 1).
Calcification: Ordered Logistic Regression Analysis Univariate predictors of CAC are listed in Table 2. In
Odds Ratio p multivariable analyses, OPG levels retained a strong asso-
Variable (95% Confidence Interval) Value ciation with elevated CAC scores after stepwise adjustment
Age* 2.18 (1.76–2.69) ⬍0.001 for age, gender, and other variables such as hypertension,
Gender ethnicity, duration of diabetes, and statin use (OR [age- and
Female 1 gender-adjusted] 2.96 [95% CI 2.34 to 3.78], p ⬍ 0.01; OR
Male 2.64 (1.88–3.71) ⬍0.001
[multivariate adjustment] 2.84 [95% CI 2.2 to 3.67], p ⬍
BMI 0.96 (0.70–1.32) 0.80
Waist-to-hip ratio† 1.67 (1.34–2.07) ⬍0.001 0.01). In contrast, neither hs-CRP nor IL-6 measurements
Systolic blood pressure 1.20 (1.09–1.32) ⬍0.001 correlated with coronary atherosclerotic plaque burden (r ⫽
Diastolic blood pressure 1.18 (1.04–1.35) 0.01 0.02, p ⫽ not significant for hs-CRP; r ⫽ 0.09, p ⫽ 0.03 for
Hyperlipidemia 1.71 (1.21–2.41) 0.002 IL-6) (Fig. 2).
Family history of premature CAD 1.19 (0.84–1.67) 0.33
Smoking (pack-yrs)
Association between OPG and cardiovascular risk fac-
1–10 yrs 1.23 (0.71–1.82) tors. In agreement with previous studies, OPG levels were
⬎10 yrs 2.23 (1.45–3.44) 0.001 positively correlated with age as well as waist-to-hip ratio,
Duration of diabetes* 1.79 (1.38–2.32) ⬍0.001 systolic blood pressure, duration of diabetes, and Framing-
Microalbuminuria 1.04 (0.67–1.62) 0.84 ham and UKPDS risk scores (Table 3), but not inflamma-
Retinopathy 1.47 (1.01–2.15) 0.04
Peripheral neuropathy 1.43 (0.95–2.15) 0.08
tory biomarkers (hs-CRP, IL-6). Patients with evidence of
Hemoglobin A1c 1.00 (0.91–1.10) 0.96 microvascular disease, i.e., retinopathy, peripheral neuropa-
hs-CRP (log scale) 0.95 (0.83–1.09) 0.44 thy, or microalbuminuria, had higher OPG levels in com-
IL-6 (log scale) 1.02 (0.98–1.06) 0.45 parison with those without (median OPG ⫽ 7.8 [IQR 5.3
Osteoprotegerin (log scale) 3.08 (2.42–3.92) ⬍0.001 to 13.9] vs. 6.1 [IQR 4.3 to 10.5], p ⫽ 0.05).
Statin therapy 1.72 (1.25–2.38) 0.001
Framingham risk score* 1.2 (0.72–1.68) ⬍0.001
Follow-up. Follow-up was completed in 99.6% of patients
UKPDS risk score* 1.8 (1.35–2.25) ⬍0.001 (average length of follow up, 18 ⫾ 5 months). A total of 16
*Odds ratios are given for a 10-U increase in the explanatory variable. †Odds ratios are
cardiovascular events occurred (1 cardiac death, 8 nonfatal
given for a 0.1-U increase in the explanatory variable. myocardial infarctions, 2 acute coronary syndromes, 2 late
UKPDS ⫽ United Kingdom Prospective Diabetes Study; other abbreviations as in revascularizations, and 3 nonhemorrhagic strokes). No car-
Table 1.
diovascular events occurred in subjects with a CAC ⬍10
type 2 diabetes in South Asians. Median (interquartile AU. The majority of events (n ⫽ 14) occurred in subjects
range) CAC scores in black African/Caribbeans, South with severe CAC (⬎400 AU). The OPG levels were higher
Asians, and Caucasians were 0 (0 to 42.5) AU, 7.2 (0 to in subjects with events than in those without: 32 (IQR 13.4
124.3) AU, and 18.8 (0 to 268.4) AU, respectively (p ⬍ to 74.5) vs. 6.8 (IQR 4.7 to 11) (p ⬍ 0.0001).
0.001). The waist-to-hip ratio, UKPDS risk score, OPG level,
Relationship between OPG and CAC scores. The me- and CAC score were significant predictors of time to
dian level of OPG was 6.82 pmol/l (IQR 4.68 to 11.55). cardiovascular events in a univariate Cox proportional haz-
The OPG levels were significantly elevated in patients with ard model, whereas age and male gender were of borderline
increased CAC (unadjusted odds ratio [OR] ⫽ 3.08, 95% significance (Table 4). Inflammatory biomarkers such as

Figure 2. The correlation between markers of inflammation (high-sensitivity C-reactive protein [hs-CRP, in mg/l], interleukin [IL]-6 [in pg/ml]) and the
coronary calcium (CAC) score. Log transformation was applied to normalize the distribution of hs-CRP and IL-6 and to reduce the skew of CAC scores.
1854 Anand et al. JACC Vol. 47, No. 9, 2006
Osteoprotegerin and Coronary Calcium in Diabetes May 2, 2006:1850–7

Table 3. Bivariate Correlation Between Plasma Osteoprotegerin OPG and cardiovascular disease. These findings are con-
Levels and Cardiovascular Risk Factors sistent with those from reports showing a close correlation
Correlation p between OPG and several major cardiovascular risk factors.
Variable Coefficient Value Plasma levels of OPG increase with age in men and women
Age 0.30* ⬍0.0001 (16 –18). In a study of 286 healthy Korean women age 37 to
BMI ⫺0.04* 0.35 73 years (19), OPG was correlated with age, waist-to-hip
Waist-to-hip ratio 0.11* 0.01 ratio, and total and low-density lipoprotein cholesterol and
Systolic blood pressure 0.11* 0.01
follicle-stimulating hormone levels. In another study of 490
Diastolic blood pressure 0.03* 0.57
Smoking (pack-years) 0.07† 0.11 Caucasian women age ⬎65 years, OPG was shown to be
Total cholesterol 0.02* 0.70 30% higher in diabetic compared with nondiabetic subjects
LDL cholesterol 0.01* 0.93 (20).
HDL cholesterol 0.06* 0.23 Circulating levels of OPG have also been correlated with
Triglycerides ⫺0.06* 0.17
surrogate measures of cardiovascular risk including: 1)
Hemoglobin A1c 0.01* 0.94
Log hs-CRP 0.01* 0.81 carotid intima-media thickness in healthy men and women
Log IL-6 0.01* 0.87 (14), 2) left ventricular hypertrophy (21), and 3) microvas-
Duration of diabetes 0.20* ⬍0.0001 cular disease in diabetic subjects (22). Three cross-sectional
Framingham risk score 0.16† 0.0002 studies have shown a relationship between OPG levels
UKPDS risk score 0.26† ⬍0.0001
and the severity of coronary atherosclerosis in symptom-
*Pearson product moment correlation coefficient is calculated for normally distributed atic CAD patients undergoing coronary angiography
variables. †Spearman rank correlation coefficient is calculated for variables distributed
otherwise. (13,21,23). More recently, a small case-controlled study (n
Abbreviations as in Table 2.
Table 4. Univariate Predictors of Cardiovascular Event-Free
hs-CRP and IL-6 were not predictive of short-term cardio- Survival, Cox Proportional Hazards Model
vascular events. Compared with patients in the lowest OPG Relative Risk p
tertile (⬍5.44 pmol/l), the relative risk of a cardiac event Variable (95% Confidence Interval) Value
was 5.76 (p ⫽ 0.01) for patients with OPG levels in the
Age* 1.89 (0.92–3.88) 0.08
highest tertile (⬎9.52 pmol/l) (Fig. 3). In the multivariate Gender
model, the CAC score was the only independent predictor Female 1
of adverse cardiovascular events (Fig. 3). Male 4.30 (0.97–19.1) 0.06
Receiver-operating characteristic curve analysis (Fig. 4) Waist-to-hip ratio† 2.25 (1.23–4.13) 0.009
Hyperlipidemia 0.99 (0.34–2.91) 0.99
showed that the CAC score and plasma OPG level were
Family history of premature CAD 0.77 (0.25–2.43) 0.66
better predictors of short-term cardiovascular risk in com- Smoking (pack-yrs)
parison to the clinically derived UKPDS/Framingham risk 1–10 yrs 1.03 (0.32–3.29)
scores (area under the curve values were 0.93, 0.8, 0.75, and ⬎10 yrs 0.33 (0.04–2.61) 0.56
0.62 for CAC, OPG, UKPDS, and Framingham risk Duration of diabetes* 1.66 (0.87–3.17) 0.13
Microalbuminuria 0.86 (0.20–3.83) 0.85
scores, respectively [p ⬍ 0.0001]). The sum of sensitivity
Retinopathy 2.12 (0.76–5.96) 0.15
and specificity for prediction of short-term cardiovascular Peripheral neuropathy 1.53 (0.49–4.82) 0.47
events was maximal at an OPG level of ⱖ19 pmol/l Hemoglobin A1c 0.98 (0.72–1.34) 0.90
(sensitivity ⫽ 74% [95% CI 48% to 92%], specificity ⫽ 91% Framingham risk score* 1.39 (0.83–2.32) 0.21
[95% CI 87% to 93%]). UKPDS risk score* 1.43 (1.12–1.82) 0.004
Osteoprotegerin
⬍5.44 1
DISCUSSION 5.45–9.51 1.03 (0.14–7.34)
⬎9.52 5.76 (1.28–26.0) 0.01
Key findings. In this prospective study involving patients
hs-CRP
with uncomplicated type 2 diabetes, we evaluated the ⬍2.32 1
relationship between established risk factors, biochemical 2.35–6.62 0.28 (0.06–1.34)
markers (hs-CRP, IL-6, and OPG), and coronary calcifi- ⬎6.62 0.87 (0.29–2.59) 0.27
cation, together with their ability to predict cardiovascular IL-6
⬍1.4 1
events. Key findings were that OPG is correlated with
1.4–2.7 1.09 (0.35–3.37)
cardiovascular risk factors (age, waist-to-hip ratio, and ⬎2.7 0.55 (0.14–2.22) 0.61
systolic blood pressure), duration of diabetes, microvascular Calcium score
disease (retinopathy and peripheral neuropathy), UKPDS/ ⱕ100 1
Framingham risk scores, and extent of calcified coronary 101–400 9.67 (0.88–105) 0.064
401–1,000 69.22 (8.08–585) ⬍0.0001
plaque. Increasing CAC scores and OPG levels were
⬎1,000 115.37 (14.17–936) ⬍0.0001
associated with worsening event-free survival. In contrast,
*Relative risk is given for a 10-U increase in the explanatory variable. †Relative risk is
biomarkers of inflammation (hs-CRP and IL-6) predicted given for a 0.1-U increase in the explanatory variable.
neither CAC nor near-term cardiovascular events. Abbreviations as in Table 2.
JACC Vol. 47, No. 9, 2006 Anand et al. 1855
May 2, 2006:1850–7 Osteoprotegerin and Coronary Calcium in Diabetes

Figure 3. Event-free survival at 18 ⫾ 5 months by the Cox proportional hazards model according to the extent of coronary calcification level (A) and plasma
osteoprotegerin (OPG) level (B), respectively. RR ⫽ relative risk ratio.

⫽ 40) has confirmed that OPG is an independent predictor tions (26). In addition, we have also confirmed the prog-
of silent CAD in asymptomatic diabetic subjects (24). nostic value of elevated OPG levels in this population.
Recent reports have also established the prognostic value of OPG and vascular calcification. Several key regulators of
OPG measurements in healthy men and women (14), and bone formation and bone structural proteins have recently
in patients with heart failure after acute myocardial infarc- been detected in atherosclerotic plaques (8). The RANKL/
tion (25). RANK/OPG system is emerging as an important regulator
Our report is the first to show that OPG predicts the of vascular, immune, and skeletal biology (9,10). Genetically
extent of CAC (a surrogate measure of total atherosclerotic altered mice unable to express functional OPG show both
plaque burden) in asymptomatic diabetic patients. Prelimi- osteoporosis and arterial calcification in the aorta and large
nary data from the population based Multi-Ethnic Study of arteries (27). Interestingly, the intravenous injection of
Atherosclerosis (MESA) are consistent with our observa- recombinant OPG protein and the transgenic overexpres-
sion of OPG have been shown to reverse osteoporosis and
prevent vascular calcification (28). Furthermore, it has been
shown that administration of recombinant OPG to rats is
able to inhibit arterial calcification induced by warfarin and
vitamin D (29). Data from experimental studies indicate
that pro-inflammatory cytokines such as tumor necrosis
factor-alpha and IL-6 induce OPG expression in smooth
muscle and endothelial cells, and OPG in turn modulates
various inflammatory signaling pathways (10,12).
All of the above clinical and experimental data suggest
an active role for OPG in vascular pathophysiology, but
it is not immediately apparent whether OPG is beneficial
or injurious to the vasculature. When elevated in patients
with atherosclerosis, OPG levels clearly seem to indicate
the overall burden of disease (13,14,21,23). This may be
because OPG is protective and counterregulatory, as
proposed by Browner, et al. (20) and others (14), or it
Figure 4. Receiver-operating characteristic curve analysis showing the may simply reflect the level of vascular inflammatory
prognostic value of coronary calcium (CAC) scores, osteoprotegerin processes that underlie the development and evolution of
(OPG) levels, and United Kingdom Prospective Diabetes Study (UKPDS)
and Framingham risk scores in predicting cardiovascular events. The 95% atherosclerotic lesions in patients with extensive disease
confidence intervals are provided. AUC ⫽ area under the curve. (10,12).
1856 Anand et al. JACC Vol. 47, No. 9, 2006
Osteoprotegerin and Coronary Calcium in Diabetes May 2, 2006:1850–7

hs-CRP and vascular calcification. Despite large studies Reprint requests and correspondence: Dr. Dhakshinamurthy
showing the relationship between hs-CRP and cardiovas- Vijay Anand, Cardiac Imaging and Research Centre, Wellington
cular risk (30), it did not predict near-term cardiovascular Hospital (South), Wellington Road, St. John’s Wood, London,
events in our study. Consistent with previous reports, NW8 9LE, United Kingdom. E-mail: vdanand@hotmail.com.
however, we found a lack of correlation between CAC
scores and hs-CRP (31,32). This may be because they REFERENCES
reflect different pathologic processes, i.e., whereas the CAC
score is a surrogate marker of the total atherosclerotic plaque 1. Rumberger JA, Simons DB, Fitzpatrick LA, Sheedy PF, Schwartz RS.
Coronary artery calcium areas by electron beam tomography and
burden, hs-CRP seems to provide an assessment of athero- coronary atherosclerotic plaque area: a histopathologic correlative
sclerotic plaque activity/stability. Overall levels of both study. Circulation 1995;92:2157– 62.
2. Shaw LJ, Raggi P, Schisterman E, Berman DS, Callister TQ.
hs-CRP and IL-6 were high in this population, and it is Prognostic value of cardiac risk factors and coronary artery calcium
possible that inflammation has a permissive role in deter- screening for all-cause mortality. Radiology 2003;228:826 –33.
mining coronary artery plaque volume or stability but that 3. Doherty TM, Fitzpatrick LA, Inoue D, et al. Molecular, endocrine
and genetic mechanisms of arterial calcification. Endocr Rev 2004;25:
other factors are responsible for driving these processes. 629 –72.
Supporting this notion, recent reports have suggested that 4. Young MJ, Adams JE, Anderson GF, Boulton AJ, Cavanagh PR. Medial
the predictive value of hs-CRP is relatively modest in arterial calcification in the feet of diabetic patients and matched non-
diabetic control subjects. Diabetologia 1993;36:615–21.
comparison with conventional risk factors (33). Because this 5. Lehto S, Niskanen L, Suhonen M, Ronnemaa T, Laakso M. Medial
study was only powered to detect strong risk associations, it artery calcification. A neglected harbinger of cardiovascular complica-
tions in non-insulin-dependent diabetes mellitus. Arterioscler Thromb
does not negate the prognostic value of hs-CRP or IL-6. It Vasc Biol 1996;16:978 – 83.
merely shows that CAC and plasma OPG are significantly 6. Maser RE, Wolfson SK Jr., Ellis D, et al. Cardiovascular disease and
better predictors of short-term CAD risk in this patient arterial calcification in insulin-dependent diabetes mellitus: interrela-
tions and risk factor profiles. Pittsburgh Epidemiology of Diabetes
group. Our findings are also consistent with data from Complications Study-V. Arterioscler Thromb 1991;11:958 – 65.
Schulze et al. (34), who found that hs-CRP levels did not 7. Nadra I, Mason JC, Philippidis P, et al. Proinflammatory activation of
predict cardiovascular events at 18 months in asymptomatic macrophages by basic calcium phosphate crystals via protein kinase C
and MAP kinase pathways: a vicious cycle of inflammation and arterial
diabetic men, and only became an independent predictor of calcification. Circ Res 2005;96:1248 –56.
cardiovascular survival by 5 years. 8. Dhore CR, Cleutjens JP, Lutgens E, et al. Differential expression of
bone matrix regulatory proteins in human atherosclerotic plaques.
Study limitations. Attending physicians were informed of Arterioscler Thromb Vasc Biol 2001;21:1998 –2003.
CAC results, which may have biased treatment received and 9. Simonet WS, Lacey DL, Dunstan CR, et al. Osteoprotegerin: a novel
assessment of outcome. However, it was not thought ethical secreted protein involved in the regulation of bone density. Cell
1997;89:309 –19.
to withhold this information given the emerging data on its 10. Schoppet M, Preissner KT, Hofbauer LC. RANK ligand and osteo-
prognostic power in the general population. Because of the protegerin: paracrine regulators of bone metabolism and vascular
function. Arterioscler Thromb Vasc Biol 2002;22:549 –53.
comparatively short follow-up period, both hard and soft 11. Nitta K, Akiba T, Uchida K, et al. The progression of vascular
end points were included in the analysis. Nevertheless, there calcification and serum osteoprotegerin levels in patients on long-term
were nine hard events, including eight myocardial hemodialysis. Am J Kidney Dis 2003;42:303–9.
12. Moran CS, McCann M, Mirko K, Norman P, Ketheesan N, Golledge
infarctions. J. Association of osteoprotegerin with human abdominal aortic aneu-
Conclusions. A high proportion (46.3%) of asymptom- rysm progression. Circulation 2005;111:3119 –25.
atic patients with type 2 diabetes have significant coro- 13. Jono S, Ikari Y, Shioi A, et al. Serum osteoprotegerin levels are
associated with the presence and severity of coronary artery disease.
nary atherosclerosis based on coronary calcium imaging. Circulation 2002;106:1192– 4.
The extent of CAC was a powerful predictor of cardio- 14. Kiechl S, Schett G, Wenning G, et al. Osteoprotegerin is a risk factor
for progressive atherosclerosis and cardiovascular disease. Circulation
vascular event-free survival. Of the biochemical markers 2004;109:2175– 80.
studied, only elevated OPG levels predicted subclinical 15. Stevens RJ, Kothari V, Adler AI, Stratton IM, Holman RR, on behalf
atherosclerosis and near-term cardiovascular events. of the United Kingdom Prospective Diabetes Study Group. The
UKPDS risk engine: a model for the risk of coronary heart disease in
Since conventional risk factors and inflammatory biomar- type 2 diabetes (UKPDS 56). Clin Sci 2001;101:671–9.
kers do not provide adequate risk stratification in diabetic 16. Szulc P, Hofbauer LC, Heufelder AE, Roth S, Delmas PD. Osteo-
protegerin serum levels in men: correlation with age, estrogen, and
subjects, measurement of plasma OPG merits further testosterone status. J Clin Endocrinol Metab 2001;86:3162–5.
investigation as a simple test for identifying high-risk 17. Yano K, Tsuda E, Washida N, et al. Immunological characterization
individuals. of circulating osteoprotegerin/osteoclastogenesis inhibitory factor: in-
creased serum concentrations in postmenopausal women with osteo-
porosis. J Bone Miner Res 1999;14:518 –27.
18. Khosla S, Arrighi HM, Melton LJ III, et al. Correlates of osteopro-
tegerin levels in men and women. Osteoporos Int 2002;13:394 –9.
Acknowledgments 19. Oh ES, Rhee EJ, Oh KW, et al. Circulating osteoprotegerin levels are
The authors are grateful to Paul Bassett, Stats Consultancy, associated with age, waist-to-hip ratio, serum total cholesterol, and
London, for performing the statistical analysis, Usha Raval low-density lipoprotein cholesterol levels in healthy Korean women.
Metabolism 2005;54:49 –54.
and Kiran Nagar for technical assistance, and Linda Pon- 20. Browner WS, Lui LY, Cummings SR. Association of serum
tello and Latif Firdoussi for study coordination. osteoprotegerin levels with diabetes, stroke, bone density, fractures
JACC Vol. 47, No. 9, 2006 Anand et al. 1857
May 2, 2006:1850–7 Osteoprotegerin and Coronary Calcium in Diabetes

and mortality in elderly women. J Clin Endocrinol Metab 28. Min H, Morony S, Sarosi I, et al. Osteoprotegerin reverses osteopo-
2001;86:631–7. rosis by inhibiting endosteal osteoclasts and prevents vascular calcifi-
21. Rhee EJ, Lee WY, Kim SY, et al. The relationship of serum cation by blocking a process resembling osteoclastogenesis. J Exp Med
osteoprotegerin levels with coronary artery disease severity, left 2000;192:463–74.
ventricular hypertrophy and C-reactive protein. Clin Sci 2005;108: 29. Price PA, June HH, Buckley JR, Williamson MK. Osteoprotegerin
237– 43. inhibits artery calcification induced by vitamin D. Arterioscler Thromb
22. Knudsen ST, Foss CH, Poulson PL, Andersen NH, Mogensen CE, Vasc Biol 2001;21:1610 – 6.
Rasmussen LM. Increased plasma concentrations of osteoprotegerin in 30. Ridker PM. Clinical application of C-reactive protein for cardio-
type 2 diabetic patients with microvascular complications. Eur J vascular disease detection and prevention. Circulation 2003;107:
Endocrinol 2003;149:39 – 42. 363–9.
23. Schoppet M, Sattler AM, Schaefer JR, Herzum M, Maisch B, 31. Redberg RF, Rifai N, Gee L, Ridker PM. Lack of association of
Hofbauer L. Increased osteoprotegerin serum levels in men with
C-reactive protein and coronary calcium by electron beam computed
coronary artery disease. J Clin Endocrinol Metab 2003;88:1024 – 8.
tomography in post menopausal women: implications for coronary
24. Avignon A, Sultan A, Piot C, Elaerts S, Cristol JP, Dupuy AP.
artery disease screening. J Am Coll Cardiol 2000;36:39 – 43.
Osteoprotegerin is associated with silent coronary artery disease in
high-risk but asymptomatic type 2 diabetic patients. Diabetes Care 32. Folsom AR, Pankow JS, Tracy RP, et al., on behalf of the Investigators
2005;28:2176 – 80. of the NHLBI Family Heart Study. Association of C-reactive protein
25. Ueland T, Jemtland R, Godang K, et al. Prognostic value of osteo- with markers of prevalent atherosclerotic disease. Am J Cardiol
protegerin in heart failure after myocardial infarction. J Am Coll 2001;88:112–7.
Cardiol 2004;44:1970 – 6. 33. Danesh J, Wheeler JG, Hirschfield GM, et al. C-reactive protein and
26. Shin VY, Johnson C, Jenny N, Wong N, Romero E, Detrano R. Serum other circulating markers of inflammation in the prediction of coronary
osteoprotegerin and coronary calcium. Circulation 2005;112:II799A. heart disease. N Engl J Med 2004;350:1387–97.
27. Bucay N, Sarosi I, Dunstan CR, et al. Osteoprotegerin-deficient mice 34. Schulze MB, Rimm EB, Li T, Rifai N, Stampfer MJ, Hu FB.
develop early onset osteoporosis and arterial calcification. Genes Dev C-reactive protein and incident cardiovascular events among men with
1998;12:1260 – 8. diabetes. Diabetes Care 2004;27:889 –94.