Absolute Coronary Artery Calcium Scores are superior to MESA percentile rank in predicting obstructive coronary artery disease

Original Paper

DOI: 10.1007/s10554-008-9305-5

Cite this article as:
Akram, K. & Voros, S. Int J Cardiovasc Imaging (2008) 24: 743. doi:10.1007/s10554-008-9305-5

Abstract

Background Coronary artery calcium scoring (CAC) is an excellent non-invasive method to evaluate coronary atherosclerotic burden. To better predict the risk of future events in an individual, their absolute CAC score is compared to an age- and gender-matched cohort in order to assign a percentile rank. However, it is unknown whether absolute CAC or percentile rank is better in predicting obstructive coronary artery disease (CAD). We hypothesized that absolute CAC is superior to percentile rank in predicting obstructive CAD. Methods 210 consecutive patients referred to our institution for CAC and coronary artery computed tomography angiography (CTA) were included. CAC scores were expressed as Agatston score; percentile rank as published by the Multi-Ethnic Study of Atherosclerosis. Coronary artery stenoses were graded semi-quantitatively. Receiver operating characteristics curves (ROC) were used to assess the performance of CAC to predict obstructive CAD. Results In the overall group, the area under the curve (AUC) was significantly greater for absolute CAC compared to MESA percentile rank in predicting obstructive CAD (0.80 vs. 0.72, P = 0.006). Subgroup analysis revealed similar findings: AUC for absolute CAC was greater than for MESA percentile rank in males (0.82 vs. 0.71, P = 0.008), females (0.78 vs. 0.72, P = 0.085), symptomatic patients (0.78 vs. 0.72, P = 0.067) and in asymptomatic subjects (0.89 vs. 0.74, P = 0.05). Conclusion Absolute CAC is superior to MESA percentile rank in predicting obstructive CAD. This finding was seen in both symptomatic and asymptomatic patients as well as in males and females.

Keywords

Coronary Artery Calcium Scores Percentile rank Test performance Obstructive coronary artery disease Coronary artery computed tomography 

Abbreviations

AUC

Area under the curve

CAC

Coronary artery calcium score

CAD

Coronary artery disease

CT

Computed tomography

CTA

Coronary tomography angiography

MDCT

Multi detector computed tomography

MESA

Multi-Ethnic Study of Atherosclerosis

MPI

Myocardial perfusion imaging

ROC

Receiver operating characteristics

Introduction

Despite significant advances in the diagnosis and the treatment of cardiovascular disease, it remains the leading cause of death worldwide [1]. Acute coronary syndromes are the first manifestation of atherosclerotic disease in more than 50% of subjects [2, 3]. Therefore, screening for subclinical atherosclerosis with non-invasive imaging modalities is an area of growing interest. Carotid intima-media thickness [4, 5] and coronary artery calcium score (CAC) [6, 7] have been developed to quantify the extent of atherosclerotic burden and have been validated to predict future cardiovascular events. The incremental value of CAC exceeds the traditional risk factors by up to 7 times [8].

The coronary artery calcium score was developed for asymptomatic patients and is significantly influenced by age and gender. Therefore, in order to compare calcium scores between individuals of different age and gender and to better predict future cardiovascular events, percentile ranges have been established from large, population-based cohorts [9, 10]. The percentile range is commonly reported along with the total CAC score.

More recently, it has been shown that ethnicity also significantly influences CAC scores. Therefore, the Multi-Ethnic Study of Atherosclerosis (MESA) investigators published the distribution of absolute CAC scores based on age, gender and ethnicity from a large population-based cohort [11]. Compared to previous studies, the MESA cohort included a variety of ethnic groups such as Caucasians, African-Americans, Hispanics and Asian populations. Furthermore, an online calculator is available for the conversion of absolute CAC into continuous percentile rank instead of a percentile range (http://www.mesa-nhlbi.org/Calcium/input.aspx). It is thought that the percentile rank based on age, gender and ethnicity can better predict cardiovascular events than the absolute score alone.

Even though not designed for this purpose, the correlation between CAC and obstructive CAD is modest and CAC has been used to predict obstructive CAD [12]. However, it is unknown whether absolute CAC or percentile rank is superior to predict obstructive CAD. A recent abstract showed that absolute CAC was superior to percentile rank in predicting survival [13]. In the present study we hypothesized that the absolute CAC score is superior to percentile rank calculated by the MESA criteria in the prediction of obstructive coronary artery disease.

Methods

Patient enrollment

This was a retrospective, observational, single-center study. The study protocol was approved by the institutional review board of Piedmont Hospital. Consecutive patients referred to our institution for simultaneous CAC scoring and coronary artery CTA were included in the analysis. Patients with a history of revascularization (percutaneous coronary intervention or coronary artery bypass graft surgery) and inadequate study quality for accurate interpretation were excluded. Enrollment criteria were met by 210 patients. Demographic and clinical information was collected from medical records. Patients were divided into two groups on the basis of symptomatic status using the Diamond-Forrester classification system [14]. Chest pain was evaluated for three different characteristics: (1) Substernal chest pain of typical quality, (2) Exacerbation by physical or emotional stress and (3) Relieved by nitrates and/or rest in less than 10 min. Typical angina, atypical chest pain and non-cardiac chest pain were characterized by the presence of all three, two or one of the three features, respectively. Patients with typical angina and atypical chest pain were classified as symptomatic and those with non-cardiac chest pain and those without any symptoms were classified as asymptomatic.

Coronary artery CT angiography

All coronary artery CT angiography examinations were performed on a 32 × 2 multi-detector computed tomography (MDCT) system (Siemens Somatom 64; Erlangen, Germany). CAC was obtained with non-contrast enhanced scans and images were acquired using 3 mm collimation with 2 mm slice increment. Acquisition parameters included a gantry rotation of 330–375 ms, pitch 0.24, tube voltage 120 kV and tube current 250 mA. CAC scores were expressed as Agatston scores.

Coronary artery CTA was performed during end-expiratory breath-hold using retrospective ECG-gating. Oral and intravenous metoprolol was administered as needed to keep the heart rate below 60 beats per minute. After non-contrast localization image acquisition, a test bolus of 20 ml iodinated contrast (Visipaque, GE Amersham Health, USA) was administered intravenously at a rate of 3–5 ml/s to determine the delay until arrival of the contrast in the ascending aorta for optimal signal intensity. Coronary arterial image acquisition was performed using 60–80 ml of intravenous contrast (Visipaque; GE Amersham Health, USA) followed by 30 ml normal saline flush. Acquisition parameters included 32 × 2 detector rows, 0.6 mm collimation, gantry rotation 330–375 ms, pitch 0.24, tube voltage 120 kV and tube current 800–950 mA. Images were reconstructed in 0.6 mm axial slices and post-processing was performed on a dedicated workstation.

Image interpretation

Image post-processing and data analysis was performed on an off-line workstation (Siemens WIZARD; Siemens Medical Solutions, Erlangen, Germany). CAC scores were expressed as Agatston score; percentile rank as published by the MESA study using the publicly available online calculator (http://www.mesa-nhlbi.org/Calcium/input.aspx). Coronary artery stenoses were graded semi-quantitatively: no luminal stenosis, mild (<30%), intermediate (30–70%) and obstructive (>70%) luminal stenosis. One experienced reader performed all clinical interpretation (SV).

Statistical analysis

All statistical analysis was performed using the MedCalc (Version 9.2.0.1) software. Receiver operating characteristics (ROC) curve analysis was used to determine the performance of CAC and percentile rank to predict obstructive disease. Statistical significance was set at a P-value ≤ 0.05.

Results

Patient characteristics

The study group consisted of 210, 134 (64%) symptomatic and 76 (36%) asymptomatic, patients. Clinical characteristics are shown in Table 1. Mean CAC in the entire cohort of 210 patients was 220 ± 438 (median 18.3, range 0–3689). Mean CAC in symptomatic patients was significantly higher compared to asymptomatic subjects: 244 ± 500 (median 12.85, range 0-3689) versus 175 ± 294 (median 26.7, range 0–1363) (P < 0.001). Mean CAC in males was significantly higher compared to females: 318 ± 565 (median 42.7, range 0–3689) versus 132 ± 251 (median 1.8, range 0–1234) (P < 0.001). An absolute CAC of 192 (sensitivity 71%, specificity 83%) and a MESA percentile rank of 81st percentile (sensitivity 63%, specificity 76%) were the best cut-off values to predict obstructive CAD. CAC values of 1.2 and 493 and percentile ranks of 13th and 95th percentile had 90% sensitivity and 90% specificity to predict obstructive CAD, respectively.
Table 1

Baseline demographic data and prevalence of CV risk factors

 

n

%

Total

210

 

Mean age ± SD (yrs)

57.4 ± 11.8

 

Male (56.5 ± 11.2 yrs)

99

47

Female (58.3 ± 12.2 yrs)

111

53

Symptomatic

134

64

Asymptomatic

76

36

Hypertension

140

67

Diabetes Mellitus

37

18

Smoker

50

24

Hypercholesterolemia

144

69

CV = cardiovascular; SD = standard deviation; yrs = years

Performance of absolute CAC versus MESA percentile rank in the overall group

Receiver operating characteristics (ROC) curve for absolute CAC in the entire group showed an area under the curve (AUC) of 0.80, indicating that CAC is an overall good test to identify obstructive disease. ROC curve for percentile rank as obtained by the MESA calculator showed an AUC of 0.72. Absolute CAC was significantly superior to percentile rank to predict obstructive disease (P = 0.006) (Table 2, Fig. 1).
Table 2

AUC comparison and corresponding P-values

 

Total

Male

Female

Symptomatic

Asymptomatic

CAC

MESA

CAC

MESA

CAC

MESA

CAC

MESA

CAC

MESA

AUC

0.80

0.72

0.82

0.71

0.78

0.72

0.78

0.72

0.89

0.74

P-value

0.006

0.008

0.085

0.067

0.05

AUC = area under the curve; CAC = coronary artery calcium score

Fig. 1

ROC curves for all patients. CAC = coronary artery calcium, MESA = multi-ethnic study of atherosclerosis

Performance of absolute CAC versus MESA percentile rank in symptomatic patients

In symptomatic patients, the AUC under the ROC curve was higher for absolute CAC compared to the MESA percentile rank (0.78 vs. 0.72; P = 0.06), indicating that absolute CAC is a better predictor of obstructive CAD in symptomatic patients. Although this did not reach statistical significance, it trended in the same direction as in the overall cohort (Table 2, Fig. 2).
Fig. 2

ROC curves for symptomatic (left) and asymptomatic (right) patients. CAC = coronary artery calcium, MESA = multi-ethnic study of atherosclerosis

Performance of absolute CAC versus MESA percentile rank in asymptomatic subjects

In asymptomatic subjects, the AUC under the ROC curve was significantly higher for absolute CAC compared to the MESA percentile rank (0.89 vs. 0.74; P = 0.05), indicating that absolute CAC is a better predictor of obstructive CAD in asymptomatic subjects (Table 2, Fig. 2).

Performance of absolute CAC versus MESA percentile rank in male patients

In males, the AUC under the ROC curve was significantly higher for absolute CAC compared to the MESA percentile rank (0.82 vs. 0.71; P = 0.008), indicating that absolute CAC is a better predictor of obstructive CAD in males (Table 2, Fig. 3).
Fig. 3

ROC curves for male (left) and female (right) patients. CAC = coronary artery calcium, MESA = multi-ethnic study of atherosclerosis

Performance of absolute CAC versus MESA percentile rank in female patients

In females, the AUC under the ROC curve was higher for absolute CAC compared to the MESA percentile rank (0.78 vs. 0.72; P = 0.09), indicating that the absolute CAC is a better predictor of obstructive CAD in females Although it did not reach statistical significance, it trended in the same direction as in the overall cohort and in the other subgroups (Table 2, Fig. 3).

Correlation with invasive angiography

Invasive angiography confirmed obstructive disease on coronary CTA in 80% of cases.

Discussion

The present study has an important novel aspect. To our knowledge, we are the first to compare the performance of absolute CAC and percentile rank to predict obstructive CAD. We hypothesized that absolute CAC is superior to percentile rank in the prediction of obstructive CAD as defined by the reference standard, coronary CT angiography. In the overall group, the AUC for absolute CAC was significantly higher than for the MESA percentile rank (0.80 vs. 0.72; P = 0.006). Therefore, we were able to reject the null-hypothesis and were able to confirm our primary hypothesis. Our findings show the superiority of the absolute CAC in predicting obstructive CAD compared to MESA percentile rank. This difference was significant in the overall group, in asymptomatic individuals and in males and trended in the same direction in the symptomatic group and in females.

Although not designed for this purpose, CAC has been correlated with presence of obstructive CAD on invasive angiography, mostly in symptomatic patients [15, 16, 17]. Also, CAC has been used in symptomatic patients who present to the emergency department with chest pain. A negative CAC was used to exclude obstructive (CAD) and it has been proposed that such patients can be safely discharged from the emergency department [18].

It is important to point out that the majority of obstructive stenoses identified by CT angiography were confirmed by invasive X-ray angiography (80%; there were 20% false positive stenoses identified by coronary CT angiography).

The coronary artery calcium score was developed to quantify the total burden of atherosclerotic plaque by measuring the total amount of calcification in the coronary tree. This approach was pioneered by Agatston et al., who developed a reliable method to quantify coronary artery calcification by taking into account both the number of pixels representing calcium and the density of the calcification as well [19]. This approach was found to have prognostic value; higher Agatston scores are associated with worse outcomes [8]. This is presumably due to the fact that more atherosclerosis, in general, translates into worse outcomes. Higher CAC scores are associated with worse outcomes both in men and women. The CAC score has independent predictive value over the Framingham risk score. Arad et al. showed that CAC was superior to the Framingham risk score in predicting coronary events, by providing information that was independent of, and additive to, the Framingham risk score [20]. Elevated absolute CAC is associated with increased risk of cardiovascular events both in males and females [8, 21]. Therefore, the absolute CAC reflects the total amount of calcium in a given individual’s coronary tree and is independent of the population.

In contrast to the absolute CAC score, the CAC percentile rank quantifies the total amount of calcification in an individual’s coronary arterial tree compared to an age-, gender-, and ethnicity-based cohort. Therefore, rather than expressing the absolute amount of calcium, it reflects whether the given individual has more or less calcium than an average subject of the same age, gender and ethnicity. Percentile ranges and ranks were introduced on the premise that since CAC is influenced by race, gender and ethnicity, the predictive value of CAC with regards to events can be improved by placing it in the context of these factors. For example, the same CAC score (e.g., a CAC score of 100) in a 50-year-old Caucasian male (percentile rank 80%) indicates more advanced, more significant atherosclerosis than in a 80-year-old African-American female (percentile rank 62%).

Therefore, the absolute CAC is an indicator of total plaque burden in a given individual, while percentile rank compares that value to the population. Against this background, it is not surprising that we found that absolute CAC predicts obstructive CAD better than percentile rank.

Interestingly, while it is expected that percentile rank predicts events better than the absolute CAC score, Gopal et al. showed that the absolute CAC was better than percentile rank in predicting events in a large population [13]. They showed the superiority of the absolute CAC score over percentile rank in predicting mortality in a group of 25,252 individuals who were followed for 8 years. In multivariate analysis, absolute CAC was a stronger indicator compared to percentiles in predicting mortality [13]. Our findings extend these observations and might explain these perhaps unexpected findings.

These data taken together suggest that the absolute CAC score, therefore the actual total amount of atherosclerosis, might be more important for the prediction of obstructive CAD. Compared to absolute CAC, percentile rank might overestimate the risk for a given CAC when the patient is young and might underestimate the risk for future events in older patients. This might explain the reported findings by Gopal et al. [13] and by us, showing that the absolute CAC was superior to the MESA percentile rank.

Reports evaluating the predictive value of CAC in predicting obstructive CAD showed a moderate correlation between the absolute CAC and angiographic stenosis [12]. This is due to the fact that the absolute score represents the total amount of plaque in the entire coronary tree and has no localizing ability. Interestingly, preliminary data in our laboratory indicates that vessel-based or lesion-based scores may be more accurate in predicting obstructive CAD.

Limitations

Limitations of our study include its retrospective design. Also, our reference standard was coronary CTA and not invasive X-ray angiography. In that regard, it is well described that coronary CT angiography might overestimate the severity of stenosis, especially in heavily calcified lesions. However, the interpretation of the CAC and the coronary CTA were performed independently. Also, 80% of obstructive stenoses identified by coronary CTA were confirmed by invasive X-ray angiography, in the remaining 20% coronary artery CTA overestimated the degree of stenosis. Overall, our findings complement the available data in the literature regarding the predictive value of CAC in the prediction of obstructive CAD.

Conclusions

Absolute CAC is superior to percentile rank calculated based on the MESA database in the prediction of obstructive CAD. Therefore, while for risk-stratification purposes percentile rank remains an important predictor, for the purposes of predicting obstructive CAD and for further diagnostic referral, the absolute CAC might be preferred.

Copyright information

© Springer Science+Business Media, B.V. 2008

Authors and Affiliations

  1. 1.Department of Internal MedicineAtlanta Medical CenterAtlantaUSA
  2. 2.Fuqua Heart Center of Atlanta, Piedmont HospitalAtlantaUSA

Personalised recommendations