Author + information
- Received February 27, 2017
- Revision received May 24, 2017
- Accepted May 26, 2017
- Published online August 7, 2017.
- Søren Zöga Diederichsen, MDa,∗ (, )
- Mette Hjortdal Grønhøj, MD, PhDb,c,
- Hans Mickley, MD, DMScb,
- Oke Gerke, MSc, PhDd,e,
- Flemming Hald Steffensen, MD, PhDf,
- Jess Lambrechtsen, MD, PhDg,
- Niels Peter Rønnow Sand, MD, PhDh,i,
- Lars Melholt Rasmussen, MD, DMScc,j,k,
- Michael Hecht Olsen, MD, PhD, DMScc,k,l and
- Axel Diederichsen, MD, PhDb,c,k
- aDepartment of Cardiology, Rigshospitalet, Copenhagen University Hospital, Blegdamsvej, Copenhagen, Denmark
- bDepartment of Cardiology, Odense University Hospital, Odense, Denmark
- cCentre for Individualized Medicine in Arterial Diseases, Odense University Hospital, Odense, Denmark
- dDepartment of Nuclear Medicine, Odense University Hospital, Odense, Denmark
- eCentre of Health Economics Research, University of Southern Denmark, Odense, Denmark
- fDepartment of Cardiology, Sygehus Lillebælt Vejle, Vejle, Denmark
- gDepartment of Cardiology, Svendborg Hospital, Svendborg, Denmark
- hDepartment of Cardiology, Hospital of South West Denmark, Esbjerg, Denmark
- iInstitute of Regional Health Services Research, University of Southern Denmark, Odense, Denmark
- jDepartment of Clinical Biochemistry and Pharmacology, Odense University Hospital, Odense, Denmark
- kCardiovascular Centre of Excellence, University of Southern Denmark, Odense, Denmark
- lCardiology Section, Department of Internal Medicine, Holbæk Hospital, Holbæk, Denmark
- ↵∗Address for correspondence:
Dr. Søren Zöga Diederichsen, The Heart Centre, Section 2013, Rigshospitalet, Copenhagen University Hospital, Blegdamsvej 9, 2100 Copenhagen, Denmark.
Objectives This study sought to determine the incidence and progression of coronary artery calcification (CAC) in asymptomatic middle-aged subjects and to evaluate the value of a broad panel of biomarkers in the prediction of CAC growth.
Background CAC continues to be a major risk factor, but the value of biochemical markers in predicting CAC incidence and progression remains unresolved.
Methods At baseline, 1,227 men and women underwent traditional risk assessment and a computed tomography (CT) scan to determine the CAC score. Biomarkers of calcium-phosphate metabolism (calcium, phosphate, vitamin D3, parathyroid hormone, osteoprotegerin), lipid metabolism (triglyceride, high- and low-density lipoprotein, total cholesterol), inflammation (C-reactive protein, soluble urokinase-type plasminogen activator receptor), kidney function (creatinine, cystatin C, urate), and myocardial necrosis (cardiac troponin I) were analyzed. A second CT scan was scheduled after 5 years. General linear models were performed to examine the association between biomarkers and ΔCAC score, and additionally, sensitivity analyses were performed in terms of binary and ordinal logistic regressions.
Results A total of 1,006 participants underwent a CT scan after 5 years. Among the 562 participants with a baseline CAC score of 0, 189 (34%) had incident CAC, whereas 214 (48%) of the 444 participants with baseline CAC score >0 had significant progression (>15% annual increase in CAC score). In the multivariate models (n = 1,006), age, sex, hypertension, diabetes, dyslipidemia, and smoking were associated with ΔCAC, whereas the strongest predictor was baseline CAC score. Low-density lipoprotein and total cholesterol levels were independently associated with CAC incidence (n = 562; incidence rate ratio [IRR]: 1.47; 95% confidence interval [CI]: 1.05 to 2.05; and IRR: 1.34; 95% CI: 1.01 to 1.77, respectively), whereas phosphate level was associated with CAC progression (n = 444; IRR: 3.60; 95% CI: 1.42 to 9.11).
Conclusions In this prospective study, a large part of participants had incident CAC or progression of prevalent CAC at 5 years of follow-up. Low-density lipoprotein and total cholesterol were associated with CAC incidence and phosphate with CAC progression, whereas 12 other biomarkers had little value.
Coronary artery calcification (CAC) score as detected by noncontrast computed tomography (CT) is a feasible and accurate measurement of subclinical atherosclerotic burden (1). Because of the large population studies on risk factors for coronary heart disease (CHD), MESA (Multi-Ethnic Study of Atherosclerosis) and HNRS (Heinz Nixdorf Recall Study), it has been demonstrated that baseline CAC score is the most robust marker in prediction of outcomes and reclassification of subjects (2,3).
The onset and progression of CAC has been associated with traditional cardiovascular (CV) risk factors in population studies (4–6), but discrepancy exists between CAC score and traditional risk scores (7). The strongest predictor of CAC growth may be the baseline CAC score (5,6,8), and recent research suggests that the prognostic information of CAC progression is essentially contained in the latest CAC assessment (9).
CAC may itself represent a therapeutic target in CHD prevention, but trials aiming to decelerate CAC progression by lipid-lowering therapy have had little success (10,11); instead, statins may increase calcified plaque formation (12). In search for pathophysiological mechanisms and to identify possibly modifiable risk factors, a growing number of possible biomarkers have been discovered (13). Markers of calcium-phosphate (14,15) and lipid metabolism (16), inflammation (17,18), kidney function (19), and myocardial necrosis (20,21) have been associated with CAC status. However, their predictive values in CAC incidence and progression remain unresolved, especially in asymptomatic subjects (22). In the current prospective observational study, we aimed to investigate the association between 15 such biochemical markers and incidence and progression of CAC in asymptomatic subjects.
In 2009 and 2010, a random sample of 1,825 men and women from the general population, born in either 1949 or 1959 and living in southern Denmark, were invited to participate in a screening in 1 of 4 regional centers (7). Exclusion criteria were patient-reported previous CV disease (e.g., myocardial infarction, revascularization, atrial fibrillation, heart valve disease, stroke, and peripheral atherosclerosis). Inclusion rate at baseline was 67%. Accordingly, 1,221 men and women were examined in 2009 and 2010. Re-examination was performed after 5 years of follow-up.
At baseline, CV risk factors were assessed. Blood pressure was measured 3 times after 5 min of supine rest, and the last 2 values were averaged for analyses. Hypertension was defined as systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg or antihypertensive treatment. Body mass index was calculated. Diabetes was defined as nonfasting blood glucose level ≥11.1 mmol/l or fasting blood glucose levels ≥7 mmol/l on 2 separate days or treatment with antidiabetic agents. Dyslipidemia was defined as low-density lipoprotein (LDL) concentration >3 mmol/l, cholesterol concentration >5 mmol/l, or treatment with lipid-lowering drugs. Family history of CHD was defined as a first-degree relative (men age <55 years and women age <65 years) with history of CHD.
Assessment of CAC score at baseline and follow-up
At baseline and after 5 years of follow-up, a noncontrast CT scan was obtained. To measure the CAC score, the Agatston score (23) was calculated by experienced cardiologists in each of the 4 centers (Online Appendix). The dependent variable in our main analyses was ΔCAC, defined as the numerical differences between CAC scores at follow-up and those at baseline. For sensitivity analyses, among participants with CAC score = 0 at baseline, incident CAC was defined as CAC score >0 at follow-up, and among participants with CAC score >0 at baseline, CAC progression was defined as >15% annual increase in CAC score and CAC score ≥20 at follow-up (24,25).
Assessment of biochemical markers at baseline
At baseline, a broad panel of biomarkers was analyzed, including calcium-phosphate metabolism (calcium, phosphate, and the calcium-phosphate product [CPP], vitamin D [including D2 and D3], parathyroid hormone [PTH], and osteoprotegerin [OPG]), lipid metabolism (triglyceride, high-density lipoprotein [HDL], LDL, and total cholesterol), inflammation (C-reactive protein [CRP], soluble urokinase-type plasminogen activator receptor [suPAR]), kidney function (creatinine, estimated glomerular filtration rate [eGFR]), cystatin C and urate, and myocardial necrosis (cardiac troponin I [cTnI]). (For details about biochemical analyses, see the Online Appendix.)
Assessment of coronary events
To assess survival bias, information on hospitalizations and deaths occurring between baseline and the scheduled follow-up visit was obtained. A coronary event was defined as acute myocardial infarction, coronary revascularization, or death from CHD. (For details, see the Online Appendix.)
Variables are presented as n (%), mean ± SD, or median (quartile 1, quartile 3) where appropriate. To test for differences between baseline variables according to baseline CAC status (CAC score = 0 or CAC score >0), 2-sided p values were calculated for proportions compared by the chi-square test, means compared by the Student t test, and medians compared by the Mann-Whitney U test. The ΔCAC score was set to 0 in subjects with a decrease in CAC score from baseline to follow-up.
The ΔCAC score was first examined as the dependent variable in univariate general linear models, using negative binomial distribution with log-link function and model-estimated value of the dispersion parameter (26). The same properties were then used in multivariate general linear models with ΔCAC as the dependent variable and CV risk factors (sex, age, body mass index, diabetes, hypertension, dyslipidemia, lipid-lowering drugs, family history of CHD, smoking status, and baseline CAC score) as independent variables. The primary model was established after backwards selection to a significance level of 5% and included sex, age, diabetes, hypertension, dyslipidemia, smoking status, and baseline CAC score. In goodness-of-fit analyses, the primary model was repeated without 1 of the independent variables, 1 by 1, and comparisons of the effect in the log likelihood function were used to identify the variable responsible for most of the model’s predictive value.
The biomarkers were then analyzed 1 by 1 as independent variables along with the variables from the primary model and ΔCAC as the dependent variable. Biomarkers of lipid metabolism were only modeled among participants not treated with lipid-lowering drugs at baseline, and these models were not adjusted for dyslipidemia. Tests for interaction between baseline CAC status (baseline CAC score 0 vs. >0) and biomarkers as well as sex and biomarkers were performed to assess whether stratification by baseline CAC status or sex was necessary. The level of significance was 5% without adjustment for multiple testing due to the exploratory nature of this study. Analyses were performed with SPSS statistical software version 22 (IBM, Armonk, New York).
The following sensitivity analyses were performed to confirm the findings of the general linear models: multivariate binary logistic regression was performed with both CAC incidence and significant CAC progression as the dependent variable, as defined earlier; the binary logistic regression method (25). Among participants with prevalent CAC at baseline, the quartiles of the annualized difference between the square root of baseline CAC score and square root of follow-up CAC score were modeled in univariate Jonckheere-Terpstra nonparametric test for trends, and multivariate ordinal logistic regressions were performed with logit link function and the quartiles as the dependent variable; the square root method (9,24,27).
The study was conducted in accordance with the Second Helsinki Declaration and was approved by the Regional Scientific Ethical Committee for Southern Denmark (project-IDs S20080140 and S20130169). Written informed consent was obtained from all participants.
Coronary events, CAC growth, and population characteristics
Of the 1,221 participants undergoing CT scanning at baseline, 215 opted out of re-examination and were thus excluded from the analyses of CAC growth. Among these participants, 6 (2.8%) experienced a coronary event compared with 13 (1.3%) in the study population (p = 0.13). In the study population of 1,006 participants, 377 (37%) had equal CAC scores at baseline and at follow-up, whereas 86 (8.5%) had a decrease in CAC median score of −7 (quartile 1: −17, quartile 3: −2). As shown in Table 1, among the 562 participants without CAC at baseline, 34% experienced incident CAC, and among the 444 participants with prevalent CAC at baseline, 48% had significant CAC progression. The 13 coronary events occurred among participants with prevalent CAC at baseline. As shown in Figure 1 and Online Table 1, all traditional CV risk factors from Table 1, except for body mass index, lipid-lowering drugs, and family history of CHD, were independently associated with ΔCAC in the entire population after backward selection. Goodness-of-fit analyses revealed baseline CAC score to be the variable with the largest effect on log likelihood (data not shown), and which CV risk factors independently associated with ΔCAC varied after stratification by baseline CAC status (Figure 1). Stratification by sex and exclusion of participants with decreasing CAC at follow-up or in lipid-lowering treatment, respectively, revealed comparable results (not shown elsewhere).
Association with biochemical markers
In cross-sectional analysis, the 444 participants with prevalent CAC at baseline had higher calcium (p = 0.036), triglyceride (p = 0.001), LDL (p = 0.022), cTnI (p < 0.001), creatinine (p = 0.001), cystatin C (p = 0.007), and urate (p < 0.001) concentrations than the 562 participants free of CAC at baseline, whereas vitamin D3 and HDL were lower (p = 0.01 and 0.017, respectively), and OPG, PTH, phosphate, CPP, total cholesterol, suPAR, CRP, and eGFR were statistically indifferent. Group comparisons according to CAC incidence or progression are shown in Table 1.
For univariate models of the association between biomarkers and ΔCAC, see Online Table 2. Tests for interactions between baseline CAC status and the biomarkers were significant in all models of ΔCAC (p < 0.001 for interaction for all biomarkers), and thus, the models were stratified by baseline CAC status. No interactions between sex and biomarkers were found in any model adjusting for baseline CAC score.
As shown in Table 2, phosphate and CPP were independently associated with CAC growth among participants with prevalent CAC at baseline (n = 444; incidence rate ratio [IRR]: 3.60 [95% confidence interval (CI): 1.42 to 9.11] and 1.53 [95% CI: 1.05 to 2.23]). This was confirmed by the sensitivity analyses (Online Tables 3 to 5), and for phosphate also after further adjustment for calcium, PTH, vitamin D, and eGFR (not shown elsewhere). LDL and total cholesterol were independent markers for incident CAC (n = 562; IRR: 1.47 [95% CI: 1.05 to 2.05] and 1.34 [95% CI: 1.01 to 1.77], respectively). No other biomarkers were consistently and independently associated with CAC growth in both the main and sensitivity analyses. Furthermore, models of suPAR, CRP, cTnI, urate, cystatin, triglyceride, LDL, and HDL were repeated after log-transformation of the biomarkers without changing our findings (not shown elsewhere). Stratification by sex, exclusion of participants with decreasing CAC at follow-up or in lipid-lowering treatment, respectively, revealed no changes in terms of significance.
In this prospective multicenter study of randomly sampled middle-aged low-risk subjects free of CV disease, we made 2 major findings. First, we confirmed that CAC incidence and progression are highly common and associated with traditional CV risk factors, although baseline CAC score was the strongest predictor. Second, in measurements of calcium phosphate and lipid metabolism, inflammation, kidney function, and myocardial necrosis, most biomarkers had little or no relationship with CAC growth, whereas LDL and cholesterol were independently associated with CAC incidence and phosphate with CAC progression.
Growth of CAC, coronary events, and traditional CV risk factors
In comparison with our finding of incident CAC in 34% of participants free of CAC at baseline, the MESA and HNRS trials reported cumulative incidence rates of only 15.5% and 25.0%, respectively (4,6). Thus, compared with those studies, we seem to have found a CAC with higher incidence, which could be due to higher prevalence of hypertension in our cohort (28) and longer interscan period. On the other hand, among participants with prevalent CAC at baseline, we found annual increases in median CAC score of only 8 U for women and 11 U for men, compared with 14 U for women and 21 U for men in MESA (4). Furthermore, considering all 1,221 participants scanned at baseline in our cohort, only 1.6% had a coronary event during 5 years of follow-up, whereas the corresponding rates in MESA and HNRS were 3.9% and 3.2%, respectively, applying the same event definition and follow-up duration (28). These differences could be due to our population being 5 years younger and having lower prevalences of dyslipidemia and diabetes.
The relationship between CAC growth and different traditional CV risk factors is inconsistent across studies (4,5,8,22). In our case, traditional CV risk factors appeared to be related to the risk of CAC progression, more than CAC incidence in adjusted analyses (Figure 1). Beyond traditional CV risk factors, we found baseline CAC score to have a high effect on follow-up CAC score, implying that prevalent CAC is an important risk factor itself and likewise that absence of CAC or low CAC score predicts less CAC growth (5,8). The prevalence of incident CHD was higher among subjects with CAC progression, consistent with the MESA study (29).
Growth of CAC in relation to biochemical markers
This is the first study to prospectively investigate the association between calcium-phosphate metabolism and CAC growth in healthy subjects. We found that phosphate was a marker of CAC progression independent of traditional CV risk markers, baseline CAC score, and calcium, and remained so after stratification by sex. On the other hand, as previously described, phosphate was not associated with baseline CAC score, even after adjustment (15). This suggests that phosphate could indeed be a marker for future CAC progression, rather than just another association to current CAC status. A previous study on young subjects has indicated an effect of baseline phosphate; however, without knowledge about CAC status at baseline (30). Our finding is compatible with data from patients with chronic kidney disease, where phosphate is often dysregulated and vascular calcification may be extreme (31). In the laboratory setting, it is well-known that increased phosphate induces vascular smooth muscle calcification (32). Furthermore, we found an association between CAC progression and CPP, and our data suggest that calcium may have an effect on CAC incidence (Table 2). In our data, PTH, vitamin D, and OPG did not affect CAC growth, which is in line with previous cross-sectional studies (14,15,33). All in all, it appears that plasma phosphate may be of particular importance for the progression of plaque calcifications, but its effect on plaque vulnerability (34) and whether phosphate-dependent CAC progression may be a target for treatment remain to be determined.
We found LDL and total cholesterol to be markers of CAC growth, but only among participants without CAC at baseline, which is in line with findings from the MESA study (4). We thus contribute to the understanding that lipid markers play a role in the onset of CAC, but the effect may be overruled by other factors hereafter. As previous trials aiming to decelerate CAC progression by statin treatment have failed (10,11), it has been debated whether lipid-lowering drugs instead increase CAC score, possibly by stabilizing soft plaques (12). In our population, statin treatment was relatively rare (10.9%), and we did not see an effect on CAC progression. We found no prospective effect of triglyceride, although triglyceride levels have been associated with CAC score in a large population study (16).
Inflammatory markers are associated with current CAC score (17,18), but we found no independent effect of CRP or suPAR on CAC growth. Although Alman et al. (35) reported inflammatory markers to be associated with CAC progression, other studies have reported negative results (4,18,21). Likewise, both urate and cTnI have been independently associated with current CAC score (19–21), although we found these markers to have no effect on CAC growth. Finally, we found no value in creatinine, eGFR, or cystatin C markers in our healthy population, and the MESA investigators suggested that their finding of higher CAC incidence in participants with low creatinine was a false positive (4). In other, selected populations of patients with diabetes or chronic kidney disease, markers of kidney function or OPG may be useful in predicting CAC progression (36,37).
Although a growing number of biochemical markers has been associated with CAC in cross-sectional studies (13–21), our prospective study revealed only few associations. This may have several explanations. First, such biomarkers may not have any causative connection with the formation of calcified plaques and may not fulfill clinical criteria for biomarker application (38). Second, the biomarkers may be more closely related to traditional CV risk markers than to CAC, which is well in line with our previous findings of the discrepancy between CAC score and traditional risk scores, as well as electrocardiographic markers (7,39). Third, the lack of associations could possibly be due to challenges in statistical modeling of CAC (5,9,22,24,25,27,29), as discussed in the following section.
A total of 215 participants (17.6% of all scanned at baseline) opted out of the re-examination CT scan and were excluded. As described, only 6 (2.8%) of these had a coronary event during follow-up, and generally, the event rate was low, minimizing possible survival bias. The excluded participants were statistically indifferent from our study population in terms of CV risk factors, except for more often having diabetes and being smokers, and biomarkers, except for having higher suPAR, CRP, and calcium and lower vitamin D levels (not shown elsewhere). Thus, some bias exists, potentially clouding a possible effect of some of these markers on CAC growth. However, the current study is strengthened by the long interscan period in comparison with previous reports of CAC growth (4,9,24).
Lipid markers were modeled only among patients not treated with lipid-lowering drugs at baseline. Thus, the value of LDL and total cholesterol in our data is not influenced by possible bias from statin use (10,11).
Negative ΔCAC was set to zero in our analyses, but the decrease in CAC score was ≤10 U in 55 of 86 cases, comparable to that in the MESA study (4). As we repeated all analyses after exclusion of the 86 participants and found the same results in terms of significance, regarding both the association with traditional CV risk factors and biomarkers, we argue that a decrease in CAC score represents interscan variability rather than an actual decrease in calcified plaque (27).
Previous reports have struggled with difficulties in the statistical modeling of CAC growth (5,9,22,24,25,27,29). Our primary analysis estimates the IRR for increases in the dependent variable per unit increase in a covariate (e.g., presence of hypertension is associated with a 1.93-fold higher CAC growth over time) (Figure 1, Online Table 1). Three important properties led to the choice of this general linear model as the primary analysis. First, it facilitates analysis of the whole population in the same model, including interactions. Second, it can take into account the overdispersion and excess zeros in a variable like ΔCAC by use of the negative binomial distribution (26). Third, it minimizes the significant loss of information that characterizes logistic regressions with categorically instead of numerically dependent variables. However, as our study is exploratory by design, to avoid false positive findings we also conducted sensitivity analyses, including binary logistic regression (25) in all strata and ordinal logistic regression by the square root method (24,27) among participants with prevalent CAC at baseline.
We added the criterion of CAC score ≥20 at follow-up to the definition of progression (24,25) to prevent false classification of participants with low CAC score at both baseline and follow-up as having progression. However, as we repeated our analyses without this criterion, it did not change our findings. Previously, it has been argued that our methods accommodate interscan variability (24,27).
As we repeated all analyses in subdivisions according to baseline CAC status (CAC score = 0 and >0), lack of significance in these subgroups could be due to type 2 error (insufficient sample size). However, according to the current evidence of increased risk even with low CAC score (2) and the significant interactions we find between baseline CAC status and biomarkers in all models of CAC growth, we argue that these are indeed different populations in terms of risk, and should be addressed accordingly.
In this prospective multicenter study of middle-aged, asymptomatic men and women from the general population, more than one-third of participants had an incidence of CAC or significant progression of prevalent CAC after 5 years of follow-up. Traditional risk factors were associated with CAC growth, even after adjustment for baseline CAC, which was the strongest predictor. LDL and cholesterol were independently associated with CAC incidence and phosphate with CAC progression, whereas 12 other biomarkers had minimal or no effect on CAC growth.
COMPETENCY IN MEDICAL KNOWLEDGE: Beyond traditional risk factors, the strongest predictor for future calcified plaque formation is the baseline CAC score. LDL and total cholesterol are independent markers of incident CAC in subjects not treated with statins, whereas phosphate is a marker of CAC progression in subjects with prevalent CAC. Several other biomarkers seem to have no predictive value in asymptomatic subjects, despite being associated with CAC in cross-sectional studies.
TRANSLATIONAL OUTLOOK: In the search for pathophysiological mechanisms and possibly modifiable risk factors for CAC, a growing number of possible biomarkers have been discovered. In the current prospective study of asymptomatic subjects, baseline CAC was the strongest predictor, whereas few biomarkers had independent association with CAC growth. Future research of biomarkers in CAC should focus on prospective findings to identify factors fulfilling the clinical criteria for biomarker application.
The authors thank the dedicated staff at the Department of Nuclear Medicine, Cardiology and Clinical Biochemistry and Pharmacology, Odense University Hospital, and the Department of Cardiology and Radiology of the hospitals in Vejle, Esbjerg, and Svendborg.
For supplemental tables and data, please see the online version of this article.
This research was supported by The Danish Heart Foundation, Region of Southern Denmark, Odense University Hospital, Odense Patient Data Explorative Network, University of Southern Denmark, the Bønnelykke Foundation, the AP Moller and Chastine Mc-Kinney Moller Foundation for general purposes, the Aase and Ejnar Danielsens Foundation, and the Herta Christensens Foundation. All authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- coronary artery calcification
- coronary heart disease
- calcium-phosphate product
- C-reactive protein
- computed tomography
- cardiac troponin I
- high-density lipoprotein
- low-density lipoprotein
- parathyroid hormone
- soluble urokinase-type plasminogen activator receptor
- Received February 27, 2017.
- Revision received May 24, 2017.
- Accepted May 26, 2017.
- 2017 American College of Cardiology Foundation
- Rumberger J.A.,
- Simons D.B.,
- Fitzpatrick L.A.,
- Sheedy P.F.,
- Schwartz R.S.
- Erbel R.,
- Möhlenkamp S.,
- Moebus S.,
- et al.
- Kronmal R.A.,
- McClelland R.L.,
- Detrano R.,
- et al.
- Min J.K.,
- Lin F.Y.,
- Gidseg D.S.,
- et al.
- Radford N.B.,
- DeFina L.F.,
- Barlow C.E.,
- et al.
- Schmermund A.,
- Achenbach S.,
- Budde T.,
- et al.
- Arad Y.,
- Spadaro L.A.,
- Roth M.,
- Newstein D.,
- Guerci A.D.
- Puri R.,
- Nicholls S.J.,
- Shao M.,
- et al.
- Grønhøj M.H.,
- Gerke O.,
- Mickley H.,
- et al.
- Moon B.S.,
- Park H.-J.,
- Lee M.-K.,
- et al.
- Sørensen M.H.,
- Gerke O.,
- Eugen-Olsen J.,
- et al.
- Gauss S.,
- Klinghammer L.,
- Steinhoff A.,
- et al.
- Olson F.,
- Engborg J.,
- Grønhøj M.H.,
- et al.
- Laufer E.M.,
- Mingels A.M.A.,
- Winkens M.H.M.,
- et al.
- McEvoy J.W.,
- Blaha M.J.,
- DeFilippis A.P.,
- et al.
- Agatston A.S.,
- Janowitz W.R.,
- Hildner F.J.,
- Zusmer N.R.,
- Viamonte M.,
- Detrano R.
- Budoff M.J.,
- Hokanson J.E.,
- Nasir K.,
- et al.
- Raggi P.,
- Callister T.Q.,
- Shaw L.J.
- Hilbe J.M.
- Budoff M.J.,
- Young R.,
- Lopez V.A.,
- et al.
- Foley R.N.,
- Collins A.J.,
- Herzog C.A.,
- Ishani A.,
- Kalra P.A.
- Hosbond S.E.,
- Diederichsen A.C.P.,
- Saaby L.,
- et al.
- Mintz G.S.
- Alman A.C.,
- Kinney G.L.,
- Tracy R.P.,
- et al.
- Pérez de Ciriza C.,
- Lawrie A.,
- Varo N.
- Maahs D.M.,
- Ogden L.G.,
- Kretowski A.,
- et al.
- Goff D.C.,
- Lloyd-Jones D.M.,
- Bennett G.,
- et al.