European Journal of Radiology 147 (2022) 110134

Contents lists available at ScienceDirect

European Journal of Radiology

journal homepage: www.elsevier.com/locate/ejrad

Research article

The ratio of the max-to-mean coronary artery calcium score in the most
calcified vessel is associated with the presence of coronary artery disease
Wenya Chen a, b, 1, Hongwei Li a, b, 1, Zhijiao Lu c, Qi Guo a, b, Xiao Liu a, b, Runlu Sun a, b,
Jie Zhang a, b, Jingjing Huang a, b, Qian Chen a, b, Junjie Wang a, b, Jun Shen c, *, Yuling Zhang a, b, *
a
Department of Cardiology, Sun Yat-sen Memorial Hospital, Sun Yat-sen University, No. 107 Yanjiang West Road, Guangzhou 510120, China
b
Guangdong Province Key Laboratory of Arrhythmia and Electrophysiology, Guangzhou 510120, China
c
Department of Radiology, Sun Yat-sen Memorial Hospital, Sun Yat-sen University, No. 107 Yanjiang West Road, Guangzhou 510120, China

A R T I C L E I N F O A B S T R A C T

Keywords: Purpose: This study aimed to clarify the relationship between the severity of the calcium burden in the most
Coronary artery calcium score calcified coronary vessel and coronary artery disease (CAD).
Agatston score Method: Of 2150 patients, 376 examined by both coronary computed tomographic angiography and invasive
Volumetric score
coronary angiography (ICA) within 30 days at Sun Yat-sen Memorial Hospital between March 2011 and July
Mass score
Coronary computed tomographic angiology
2020 were included. Three coronary artery calcium scores (CACSs), including the Agatston score, volume score,
Coronary artery disease and mass score, and other clinical variables were recorded. The ratio of max-to-mean CACS in the most calcified
vessel (CACSmax:mean) was defined as the CACS in the most calcified vessel/average CACS of the four major
epicardial coronary arteries. Logistic regression and least absolute shrinkage and selection operator (LASSO)
analyses were performed to assess the relationship between CACSmax:mean and CAD.
Results: CACSmax:mean was higher in 81.1% of subjects diagnosed with CAD than in subjects without CAD. In
multivariate logistic regression analysis, CACSmax:mean determined by the Agatston score, volumetric score, and
mass score was associated with CAD. In LASSO analysis, Agatston scoremax:mean (not the total Agatston score or
other CACSmax:mean) had the strongest correlation with CAD (β = 0.125). AUCs in the training set and the
validation set were 0.811 and 0.789, respectively. Increased age, diabetes and hypertension correlated with
higher Agatston scoremax:mean.
Conclusions: In addition to total CACS, CACSmax:mean may be a novel diagnostic parameter for CAD, showing the
calcium burden severity.

1. Introduction intermediate-risk patients, with a score ≥ 100 guiding the initiation of

statin therapy [3] Overall, addition of the CACS to a prediction model
The coronary artery calcium score (CACS), as obtained by coronary based on traditional risk factors might improve risk classification for
computed tomographic angiology (CTA) and used to quantify coronary CAD [4]. There are three commonly used CACS algorithms. Developed
artery calcification burden, has been demonstrated to be positively in 1990, the Agatston score was the first quantitative method; it is the
associated with an elevated risk for the presence of coronary artery most common and is calculated by multiplying the calcified plaque
disease (CAD) [1,2]. The CACS is recommended as a risk modifier for density and area [1]. The volumetric score was established in 1998 and

Abbreviations: Agatston scoremax:mean, ratio of the max-to-mean Agatston score in the most calcified vessel; AUC, area under the curve; CAD, coronary artery
disease; CACS, coronary artery calcium score; CACSmax:mean, ratio of the max-to-mean CACS in the most calcified vessel; CI, confidence interval; CTA, computed
tomography angiography; ICA, invasive coronary angiography; LASSO, least absolute shrinkage and selection operator; Mass scoremax:mean, ratio of the max-to-mean
mass score in the most calcified vessel; OR, odds ratio; ROC, receiver operating characteristic; Volumetric scoremax:mean, ratio of the max-to-mean volumetric score in
the most calcified vessel.
* Corresponding authors at: Department of Cardiology, Sun Yat-sen Memorial Hospital, Sun Yat-sen University, No. 107 Yanjiang West Road, Guangzhou 510120,
China (Y. Zhang). Department of Radiology, Sun Yat-sen Memorial Hospital, Sun Yat-sen University, No. 107 Yanjiang West Road, Guangzhou 510120, China (J.
Shen).
E-mail addresses: [email protected] (J. Shen), [email protected] (Y. Zhang).
1
Wenya Chen and Hongwei Li have equal contributions.

https://doi.org/10.1016/j.ejrad.2021.110134
Received 24 August 2021; Received in revised form 18 December 2021; Accepted 24 December 2021
Available online 29 December 2021
0720-048X/© 2021 Elsevier B.V. All rights reserved.
W. Chen et al. European Journal of Radiology 147 (2022) 110134

focuses mainly on the volume of calcified plaques [5]. The mass score scores.
was proposed in 2003, and it has been used to calculate the total calcium
mass of individual lesions corrected by the density of calcium hy­ 2.3. Clinical variables
droxyapatite, with good reproducibility [6,7].
Despite various efforts to increase the diagnostic and prognostic Details of clinical variables are provided in the supplementary ma­
value of the CACS, the total CACS cannot reflect the number of calcified terials. Briefly, demographic data, biochemical examination and echo­
vessels and the severity of calcium burden in the most calcified vessel, cardiography were recorded. Hypertension, diabetes and other
which is often associated with the risk of coronary events. For example, cardiovascular risk factors were analysed for cardiovascular risk esti­
Blaha et al demonstrated that the addition of the calcified coronary mation. Biochemical tests, including myocardial enzymes, lipids, in­
vessel number improves the predictive ability of the traditional Agatston flammatory biomarkers and echocardiography, were recommended
score for cardiovascular events [8]. Therefore, we propose a novel according to clinical guidelines [12,13]. These important clinical vari­
quantitative parameter called the ratio of the max-to-mean CACS in the ables were compared between patients with and without CAD, and po­
most calcified vessel (CACSmax:mean), as calculated by the CACS in the tential confounders for CACSmax:mean were selected.
most calcified vessel/average CACS of the four main vessels (the left
main coronary artery (LM), the left anterior descending branch (LAD), 2.4. Definition of CAD
the left circumflex branch (LCX), and the right coronary artery (RCA)).
Moreover, a recent study using machine learning methods combining CAD was defined as stenosis ≥ 50% of the diameter in any of the four
the total CACS and other clinical variables showed better performance major coronary arteries detected by ICA [14].
than traditional symptom-based pretest probability scores in estimating
the risk of obstructive CAD [9]. However, the correlation between 2.5. Statistical analysis
CACSmax:mean and CAD remains unclear, and a machine learning model
with good feature selection capacity may help to clarify this correlation. Details of statistical analysis are provided in the supplementary
Hence, this study aimed to investigate the relationship between CACS materials. Briefly, missing values of continuous variables were imputed
max:mean and the presence of CAD by using a logistic regression model as by the random forest regression method. CACS and other clinical vari­
well as a parsimonious model selected by the least absolute shrinkage ables were then compared between patients with and without CAD.
and selection operator (LASSO) method. Logistic and LASSO regression analyses were employed to estimate odds
ratios of the association between CACS and the presence of CAD.
2. Methods
3. Results
2.1. Study population
3.1. Baseline characteristics of patients with or without CAD
Patients who underwent both coronary CTA and invasive coronary
angiology (ICA) within 30 days at the Department of Radiology in Sun There were 2150 patients who underwent both coronary CTA and
Yat-sen Memorial Hospital between March 2011 and July 2020 were ICA at Sun Yat-sen Memorial Hospital between March 2011 and July
included. Previous studies suggest that the diagnostic accuracy of cor­ 2020, and 376 subjects were ultimately included in this study (Fig. 1).
onary CTA for CAD is high, indicating that both coronary CTA and ICA The median age was 62 (55, 68) years, and 60.6% of the patients were
consistently reflect the condition of coronary vessels [10,11]. Patients male. The numbers of patients with and without CAD were 305 (81.1%)
who met the exclusion criteria were those with the following conditions: and 71 (18.9%), respectively. Compared to patients without CAD, those
1) acute myocardial infarction and coronary CTA examination after with CAD were significantly older and had increased rates of diabetes
coronary revascularization; 2) structural heart disease, including rheu­ and hypertension but lower SBP at baseline (P < 0.050). In addition,
matic heart disease, cardiomyopathy, severe valvular heart disease; 3) elevated BUN, AAO, LA and IVSd were found in patients with CAD.
ongoing hormone replacement treatment; 4) alanine transaminase > There was no significant difference in DBP, HDL-C, LDL-C, total
120 IU/L and alanine transaminase:aspartate transaminase > 1; 5) cholesterol, or triglycerides between patients with and without CAD
estimated glomerular fraction rate (eGFR) < 30 ml/(min⋅1.73 m2); 6) (Table 1). Patients with CAD had a higher use of lipid-lowering drugs,
severe infection, hyperthyroidism, malignant tumour, autoimmune antiplatelet drugs, and β-blockers than those without CAD (Table S1).
disease, or rheumatic disease; and 7) demographic data or admission For CACS-related parameters, the CACS in the most calcified vessel and
note unavailable. The study protocol was approved by the Ethics Com­ the total CACS based on the Agatston score, the volumetric score and the
mittee of Sun Yat-sen Memorial Hospital (No. SYSEC-KY-KS-2020–114), mass score were elevated in patients with CAD (P < 0.001). Moreover,
and informed consent was waived for this retrospective study. Agatston scoremax:mean [1.65 (1.07, 1.98) vs. 1.00 (0.00, 1.46), P <
0.001], volumetric scoremax:mean [1.56 (1.07, 1.93) vs. 0.00 (0.00, 1.26),
2.2. The CACS of coronary CTA P < 0.001], and mass scoremax:mean [1.61 (1.00, 2.00) vs. 0.000 (0.00,
1.30), P < 0.001] were all higher in patients with CAD than in those
All computed tomography (CT) scans were performed using a dual- without CAD (Table 2).
source CT (Somatom Definition Force, Siemens, Germany; rotation
time, 0.25 s; 0.4 mm reconstruction; spatial resolution, 0.6 mm). Results 3.2. CACSmax:mean and the presence of CAD are increased with an
were calculated by 3 CACS algorithms, namely, the Agatston score, the elevated number of calcified coronary arteries
volumetric score, and the mass score of the four major coronary arteries,
including the LM, LAD, LCX, and RCA, as obtained by syngo.via: CACSmax:mean, the CACS in the most calcified vessel, the total CACS,
04.02.0000.0009 (Siemens Healthcare, Germany). The total CACS was and the presence of CAD all increased with an increase in the number of
defined as the sum of the CACS of the four major epicardial coronary calcified coronary arteries (Table 3). Moreover, the frequency distribu­
arteries. CACSmax:mean was calculated as the CACS in the most calcified tion of the most calcified vessel was the LAD (n = 172), followed by the
vessel/average CACS of the four major epicardial coronary arteries RCA (n = 74), the LM (n = 30), and the LCX (n = 29) (Figure S1).
[CACSmax:mean = the CACS in the most calcified vessel/(total CACS/
number of calcified coronary arteries)]. CACSmax:mean is defined as zero 3.3. CACSmax:mean is positively associated with the presence of CAD
in subjects without CAC. All CACS calculations were performed in the
same manner based on the results of the Agatston, volumetric and mass Agatston scoremax:mean [OR = 3.762 (2.613, 5.415), P < 0.001], the

2
W. Chen et al. European Journal of Radiology 147 (2022) 110134

Fig. 1. The flowchart of subject inclusion and exclusion.

volumetric scoremax:mean [OR = 3.035 (2.152, 4.280), P < 0.001], and 4. Discussion
the mass scoremax:mean [OR = 2.854 (2.049, 3.975), P < 0.001] were
positively associated with the presence of CAD. Furthermore, age, dia­ In this study, we propose a new CACS-related parameter, namely,
betes, SBP, CK-MB, AAO, LA, and ISVd all correlated with an increased CACSmax:mean, to represent the severity of the calcium burden of the
presence of CAD (Table S2). After adjusting for age, sex, diabetes, SBP, most calcified vessel, which differs from the total CACS that does not
CK-MB, AAO, LA, and IVSd, Agatston scoremax:mean [adjusted OR = provide this information. CACSmax:mean was independently associated
1.975 (1.983, 4.463), P < 0.001], volumetric scoremax:mean [OR = 2.556 with an increased presence of CAD and correlated with traditional car­
(1.743, 3.748), P < 0.001], and mass scoremax:mean [OR = 2.402 (1.047, diovascular risk factors, including age, diabetes, and hypertension.
3.470), P < 0.001] remained independently associated with CAD. The Moreover, CACSmax:mean was elevated with an increase in the number of
CACS in the most calcified vessel and the total CACS based on the CAC vessels, indicating that the calcium burden severity in the most
Agatston, volumetric, and mass scores were also independently associ­ calcified vessel was consistent with the development of multivessel
ated with an increased presence of CAD after adjustment (P < 0.001) calcification.
(Table 4). Consideration of the number of calcified coronary arteries has been
demonstrated to improve the diagnosis of and predictive performance
3.4. LASSO analysis between clinical and CACS variables and CAD for cardiovascular events. Tota-Maharaj et al. found that an increased
number of calcified coronary arteries, accompanied by the Agatston
Thirty-seven clinical variables and 9 CACS-related parameters were score and cardiovascular risk factors, was associated with a higher
included in LASSO analysis. Agatston scoremax:mean (β = 0.125), HDL-C mortality risk than single-vessel calcification [15]. In another cohort
(β = − 0.034), diabetes (β = 0.009) and age (β = 0.004) were selected in study, addition of the number of calcified coronary arteries improved
the final model validated by 10-fold cross-validation with the minimum the predictive capacity for cardiovascular events in survival analysis
λ = 0.035. The AUCs of the LASSO models were 0.811 (95% CI: based on the traditional Agatston score. A diffuse index calculated as 1−
0.745–0.877) in the training cohort and 0.789 (95% CI: 0.673–0.905) in (the Agatston score in most affected vessel/total Agatston score) was
the validation cohort. Calibration curves showed good diagnostic per­ also proposed in this study, but its addition did not improve the pre­
formance of the LASSO models for CAD, with an error of 0.035 ± 0.002 dictive performance of the survival analysis model [8]. CACSmax:mean
in the training set and 0.024 ± 0.001 in the validation set (Fig. 2). considers both the number of calcified coronary arteries and the CACS in
the most calcified vessel, which can reflect the concentration of the
3.5. Association between Agatston scoremax:mean and traditional calcium burden in the most calcified vessel. As illustrated in Figure S2, a
cardiovascular risk factors diffuse calcium distribution was characterized by a higher total Agatston
score with a lower Agatstonmax:mean (Figure S2F), though a concentrated
Age [OR = 1.045 (1.024, 1.067), P < 0.001], diabetes [OR = 1.730 calcium distribution was represented by a lower total Agatston score
(1.113, 2.691), P = 0.015], and hypertension [OR = 1.550 (1.015, with a higher Agatstonmax:mean (Figure S2A, C, D). The Agatstonmax:mean
2.367), P = 0.042] were positively associated with higher Agatston of patients without coronary artery calcification was defined as 0 (Fig­
scoremax:mean, whereas sex, BMI, smoking history, HDL-C, and total ure S2E). Importantly, a higher Agatstonmax:mean also reflected the
cholesterol had no significant relationship with Agatston scoremax:mean concentrated distribution of coronary artery calcification mainly in the
(Table S3). RCA (Figure S2B). Different patterns of CACS in the most calcified vessel
are presented in Figure S2. Overall, CACSmax:mean is a useful tool for
assessing coronary artery calcification.

3
W. Chen et al. European Journal of Radiology 147 (2022) 110134

Table 1 Table 1 (continued )

Baseline characteristics of patients with/without CAD. Total Patients with Patients P
Total Patients with Patients P population CAD without CAD
population CAD without CAD (n = 376) (n = 305) (n = 71)
(n = 376) (n = 305) (n = 71)
IVSd (mm) 10.00 (9.00, 10.00 (9.00, 9.00 (9.00, 0.013*
Demographic data 10.76) 11.00) 10.00)
Age 62 (55, 68) 63 (56, 70) 57 (49, 62) <0.001* LVDd (mm) 47.313 ± 47.33 ± 4.28 47.25 ± 4.03 0.891
Male gender (%) 228 (60.6%) 186 (61.0%) 42 (59.2%) 0.776 4.23
BMI 24.64 ± 3.62 24.75 ± 3.67 24.15 ± 3.36 0.212 LVPWd (mm) 9.00 9.02 9.00 0.134
Smoking history (%) 131 (34.8%) 111 (36.4%) 20 (28.2%) 0.190 (9.00,10.99) (9.00,10.00) (8.44,10.00)
Hypertension (%) 239 (63.6%) 204 (66.9%) 35 (49.3%) 0.006* EF (%) 67.65 ± 5.91 67.57 ± 6.06 67.99 ± 5.26 0.592
Diabetes (%) 118 (31.4%) 105 (34.4%) 13 (18.3%) 0.008*
Median (the 1st interquartile, the 3rd interquartile) is used for nonnormally
Family history of 24 (6.4%) 21 (6.9%) 3 (4.2%) 0.409
CAD (%) distributed variables, mean ± standard deviation for normally distributed var­
SBP (mmHg) 133.36 ± 134.83 ± 127.04 ± 0.003* iables, and cases (%) for categoric variables. Normally distributed variables are
19.905 20.71 17.95 compared by t test, while non-normally distributed variables are compared by
DBP (mmHg) 79.00 79.00(71.50, 79.00(70.00, 0.658 Wilcoxon-Mann-Whitney tests. Categorical variables were compared by the χ 2
(71.00, 85.50) 84.00) test.
85.00) *P < 0.05
Biochemical Examination AOR: aortic diameter; AAO: aortic ascendens; CAD, coronary artery disease;
HDL-C (mmol/L) 1.12 (0.95, 1.10 (0.95, 1.14 (0.96, 0.210
BMI: body mass index; CK: creatine kinase; CK-MB: creatine kinase-isoenzyme;
1.32) 1.31) 1.34)
LDL-C (mmol/L) 3.24 (2.54, 3.24 (2.51, 3.25 (2.73, 0.960
DBP: diastolic blood pressure; EF: ejection fraction; eGFR: estimated glomer­
3.72) 3.76) 3.66) ular filtration rate; HDL-C: high-density lipoprotein-cholesterol; hsCRP: high-
Total cholesterol 5.07 (4.12, 5.08 (4.10, 5.00 (4.35, 1.000 sensitive C reactive protein; LA: left atrium; IVSd: interventricular septal
(mmol/L) 5.72) 5.72) 5.63) thickness at diastole; LDH: lactate dehydrogenase; LDL-C: low-density
Triglycerides 1.50 (1.14, 1.47 (1.14, 1.57 (1.17, 0.429 lipoprotein-cholesterol; LVDd: left ventricular dialostic dimension; LVPWd: left
(mmol/L) 2.05) 2.07) 1.94) ventricular posterior wall dimension; RVDd: right ventricular dialostic diameter;
Apolipoprotein A 1.15 (1.04, 1.15 (1.04, 1.15 (1.05, 0.743 SBP: systolic blood pressure; SOD: superoxide dismutase; UA: uric acid.
(mmol/L) 1.28) 1.28) 1.32)
Apolipoprotein B 0.94 (0.76, 0.94 (0.75, 0.93 (0.77, 1.000
(mmol/L) 1.07) 1.08) 1.03)
BUN (mmol/L) 5.40 (4.60, 5.44 (4.60, 5.20 (4.50, 0.043*
Table 2
6.40) 6.45) 5.90) CACS-related parameters between patients with CAD and patients without CAD.
Creatinine (μmol/L) 85.00 84.00 86.00 0.438 Total Patients with Patients P
(74.16, (73.50, (77.00, population (n CAD (n = without CAD
97.00) 96.00) 98.00) = 376) 305) (n = 71)
eGFR (ml/min⋅1.73 77.52 77.42 78.09 0.792
m2) (65.22, (64.65, (67.51, Agatston scores
86.74) 86.45) 89.25) Agatston scoremax: 1.55 (1.000, 1.65 (1.07, 1.00 (0.00, <0.001*
UA (μmol/L) 390.02 ± 391.16 ± 385.11 ± 0.623 mean 1.95) 1.98) 1.46)
93.28 93.41 93.22 Agatston score in 81.00 (10.88, 100.50 0.40 (0.00, <0.001*
hs-CRP (mg/L) 1.67 (0.81, 1.71 (0.81, 1.49 (0.84, 0.429 the most 227.63) (32.15, 35.80)
4.39) 4.50) 4.10) affected vessel 251.85)
SOD (U/ml) 140.00 140.00 141.00 0.409 Total Agatston 111.00 (12.60, 150.00 0.40 (0.00, <0.001*
(130.00, (130.00, (129.00, score 311.68) (43.20, 42.60)
150.00) 149.00) 155.00) 401.95)
Fibrinogen (g/L) 2.93 (2.60, 2.93 (2.62, 2.91 (2.52, 0.792 Volumetric scores
3.45) 3.45) 3.45) Volumetric 1.43 (1.00, 1.56 (1.00, 0.00 (0.00, <0.001*
D Dimer (mg/L FEU) 0.26 (0.15, 0.28 (0.16, 0.21 (0.13, 0.081 scoremax: mean 1.86) 1.93) 1.26)
0.49) 0.52) 0.38) Volumetric score 44.30 (0.38, 72.00 0.00 (0.00, <0.001*
White blood cell 7.20 (6.01, 7.08 (5.93, 7.68 (6.31, 0.065 in the most 168.00) (14.40, 26.80)
(×109/L) 8.49) 8.51) 8.45) affected vessel 194.75)
Platelet (×1012/L) 238.94 ± 238.29 ± 241.75 ± 0.720 Total volumetric 66.95 (0.38, 98.00 0.00 (0.00, <0.001*
73.18 76.08 59.52 score 236.93) (16.70, 36.30)
Hemoglobin (g/L) 135.75 ± 134.84 ± 139.65 ± 0.131 298.75)
24.18 25.99 13.47 Mass score
Neutrophils-to- 2.25 (1.69, 2.24 (1.65, 2.26 (1.73, 0.792 Mass scoremax: 1.47 (0.12, 1.61 (1.00, 0.00 (0.00, <0.001*
lymphocytes ratio 3.05) 3.06) 2.86) mean 1.92) 2.00) 1.30)
CK (U/L) 86.00 86.00 81.16 0.949 Mass score in the 11.47 (0.01, 16.53 (2.48, 0.00 (0.00, <0.001*
(66.00, (65.00, (66.00, most affected 40.34) 46.09) 5.60)
115.75) 115.50) 117.00) vessel
CK-MB (U/L) 12.00 12.00 12.00 0.285 Total mass score 14.99 (0.01, 23.11 (3.43, 0.00 (0.00, <0.001*
(10.00, (10.00, (10.00, 53.87) 67.78) 6.26)
15.00) 15.00) 14.00)
*P < 0.05.
LDH (U/L) 179.00 181.00 174.00 0.073
Agatston scoremax: mean: the ratio of max-to-mean Agatston score in the most
(159.00, (160.00, (159.00,
209.750) 215.00) 194.00) calcified vessel; CACSmax: mean: the ratio of max-to-mean CACS in the most
Echocardiography calcified vessel; Mass scoremax: mean: the ratio of max-to-mean mass score in the
AOR (mm) 20.58 20.68 20.00 0.188 most calcified vessel; Volumetric scoremax: mean: the ratio of max-to-mean
(20.00, (20.00, (19.00, volumetric score in the most calcified vessel.
21.26) 21.90) 21.00)
AAO (mm) 32.00 32.26 31.00 <0.001*
(30.00, (30.35, (29.00, The LASSO regression model permitted a large number of covariates
34.00) 32.26) 32.03) to be included, with a unique feature penalization for the shrinkage of
LA (mm) 33.66 ± 3.53 33.86 ± 3.55 32.84 ± 3.36 0.028* coefficients and automatically removed unnecessary covariates
RVDd (mm) 20.21 ± 1.91 20.28 ± 1.87 20.00 ± 2.10 0.258 [16,17,18]. Bundy et al. established a parsimonious LASSO regression
model, including age, weight, current smoking, N-terminal pro B type

4
W. Chen et al. European Journal of Radiology 147 (2022) 110134

Table 3
The comparison of CACS-related parameters and the presence of CAD in different numbers of calcified coronary arteries.
The number of calcified vessels (numbers) 0 (n = 65) 1 (n = 57) 2 (n = 90) 3 (n = 94) 4 (n = 70) P

Agatston score
Agatston scoremax: mean 0.00 (0.00, 1.00 (1.00, 1.00) 1.66 (1.35, 1.87) 1.84 (1.54, 2.22) 2.10 (1.77, 2.66) <0.001*
0.000)
Agatston score in the most affected vessel 0.00 (0.00, 0.00) 30.00 (5.60, 74.00 (34.68, 136.60 (79.58, 293.48) 264.25 (129.88, <0.001*
95.05) 150.75) 445.85)
Total Agatston score 0.00 (0.00, 0.00) 30.00 (5.60, 91.45 (48.28, 224.40 (129.15, 490.85 (237.05, <0.001*
95.05) 202.45) 529.88) 897.13)
Volumetric score
Volumetric score max: mean 0.00 (0.00, 0.00) 1.00 (1.000, 1.00) 1.60 (1.25, 1.81) 1.77 (1.46, 2.10) 1.93 (1.59, 2.53) <0.001*
Volumetric score in the most affected 0.00 (0.00, 0.00) 16.70 (4.40, 45.40 (21.40, 96.43) 105.85 (51.50, 245.95) 194.20 (59.88, 342.55) <0.001*
vessel 48.30)
Total volumetric score 0.00 (0.00, 0.00) 16.70 (4.40, 64.15 (28.58, 171.30 (82.50, 444.40) 354.15 (127.03, <0.001*
48.30) 143.05) 692.88)
Mass score
Mass score max: mean 0.00 (0.00, 0.00) 1.00 (1.00, 1.00) 1.67 (1.26, 1.87) 1.82 (1.50, 2.22) 2.04 (1.71, 2.55) <0.001*
Mass score in the most affected vessel 0.00 (0.00, 0.00) 3.24 (0.63, 12.79) 11.72 (3.62, 24.77) 24.12 (11.40, 50.64) 47.49 (15.39, 77.14) <0.001*
Total mass score 0.00 (0.00, 0.00) 3.24 (0.63, 12.79) 13.51 (5.76, 30.11) 38.25 (17.99, 94.08) 77.43 (29.84, 153.26) <0.001*
The presence of CAD (%) 30 (46.2) 42 (73.7) 77 (85.6) 88 (93.6) 68 (97.1) <0.001*

*P < 0.05.
Agatston scoremax: mean: the ratio of max-to-mean Agatston score in the most calcified vessel; CACSmax: mean: the ratio of max-to-mean CACS in the most calcified vessel;
CAD: coronary artery disease; Mass scoremax: mean: the ratio of max-to-mean mass score in the most calcified vessel; Volumetric scoremax: mean: the ratio of max-to-mean
volumetric score in the most calcified vessel.

natriuretic peptide, CACS, and cardiac troponin-T, for 5-year prediction

Table 4
of atrial fibrillation, with a high c-statistic of 0.802 [19]. In our study,
Univariable and Multivariable Logistic analyses between CACS and the presence
only 4 variables, including Agatston scoremax:mean and 3 traditional risk
of CAD.
factors, namely, HDL-C, diabetes and age, were selected in the final
Univariable Multivariable
LASSO model. Agatston scoremax:mean, but not volumetric scoremax:mean
OR (95 % P Adjusted OR P or mass scoremax:mean, remained as the necessary covariate to avoid high
CI) (95 %CI) collinearity, confirming that Agatston scoremax:mean was necessary and
Agatston score stable for assessing the calcium burden severity of the most calcified
Agatston scoremax: mean 3.762 <0.001* 1.975 (1.983, <0.001* vessel and the overall risk of CAD. Hence, Agatston scoremax:mean should
(2.613, 4.463)
be carefully regarded when assessing the risk of CAD by the CACS based
5.415)
Agatston score in the 1.013 <0.001* 1.009 (1.005, <0.001* on coronary CTA. Age, diabetes, and hypertension were also found to be
most affected vessel (1.008 1.014) positively associated with increased Agatston scoremax:mean, similar to
1.017) the risk factors for the total Agatston score [20]. Nevertheless, there was
Total Agatston score 1.009 <0.001* 1.006 (1.003, <0.001* no significant association between blood lipids and CACSmax:mean or the
(1.005, 1.010)
presence of CAD, which might be attributed to higher use of lipid-
1.012)
Volumetric score lowering drugs among patients with CAD.
Volumetric scoremax: 3.035 <0.001* 2.556 (1.743, <0.001* There were some limitations in this retrospective study. First, se­
mean (2.152, 3.748) lection bias might exist due to the inclusion of patients receiving coro­
4.280)
nary CTA and ICA within 30 days. The duration of participant inclusion
Volumetric score in 1.015 <0.001* 1.012 <0.001*
the most affected (1.009, (1.005,1.018) was 10 years, which might partly mitigate selection bias. Further pro­
vessel 1.021) spective study designs mitigating statistical bias are needed to validate
Total volumetric score 1.010 <0.001* 1.007 (1.003, 0.001* the relationship between CACSmax:mean and cardiovascular events. Sec­
(1.005, 1.012) ond, a cross-sectional design with a small sample size and lack of an
1.014)
external validation set might weaken the credibility of the LASSO
Mass score
Mass scoremax: mean 2.854 <0.001* 2.402 (1.663, <0.001* regression model; further prospective studies to verify the diagnostic
(2.049, 3.470) and predictive performance of the LASSO regression model are required.
3.975)
Mass score in the most 1.063 <0.001* 1.047 (1.021, <0.001*
affected vessel (1.036, 1.073) 5. Conclusions
1.090)
Total mass score 1.041 <0.001* 1.030 (1.013, 0.001* Agatston scoremax:mean is an advantageous variable correlating
(1.023, 1.048) independently with CAD among all CACS-related parameters, but a risk
1.060)
prediction algorithm was not developed due to the cross-sectional
*P < 0.05. design of this study. Agatstonmax:mean may be sensitive in reflecting
The multivariate model was adjusted by age, gender, diabetes, SBP, CK-MB, the severity of coronary artery calcification burden, whereas the routine
AAO, LA, IVSd. total Agatston score may underestimate this risk. Prospective cohort
Agatston scoremax: mean: the ratio of max-to-mean Agatston score in the most
validation by using diagnostic models along with clinical variables for
calcified vessel; CACSmax: mean: the ratio of max-to-mean CACS in the most
cardiovascular risk calculation is needed.
calcified vessel; CI: confidence interval; Mass scoremax: mean: the ratio of max-to-
mean mass score in the most calcified vessel; OR: odds ratio; Volumetric
scoremax: mean: the ratio of max-to-mean volumetric score in the most calcified 6. Ethics statement
vessel.
The authors are responsible for all aspects of the work and for
ensuring that questions related to the accuracy or integrity of any part of

5
W. Chen et al. European Journal of Radiology 147 (2022) 110134

Fig. 2. LASSO regression analysis in the training

set and validation set. (A) Variation in the coeffi­
cient increased with log(λ) in the training set. The
horizontal axis at the bottom represents log(λ);
that at the top represents the number of variables
with coefficient > 0 with the longitudinal axis of
coefficient. Agatstonmax:mean, HDL-C, diabetes,
and age were four significant variables included in
the LASSO regression model. (B) Tenfold cross-
validation of the LASSO regression model in the
training set. The horizontal axis at the bottom
represents log(λ); that at the top represents the
number of variables with coefficient > 0 with the
longitudinal axis of mean-squared error. The
minimum λ = 0.035. (C) ROC curve of the LASSO
regression model in the training set (AUC =
0.811). (D) ROC curve of the LASSO regression
model in the validation set (AUC = 0.789). The
horizontal axis represents (1 − Specificity); the
longitudinal axis represents sensitivity. (E) The
calibration curve of the LASSO regression model
in the training set. (F) The calibration curve of the
LASSO regression model in the validation set. The
horizontal axis represents the predicted probabil­
ity; the longitudinal axis represents the actual
probability. The solid curve close to the dotted
line demonstrates good predictive performance of
the LASSO regression model for the actual prob­
ability. Agatstonmax:mean: the ratio of the max-to-
mean Agatston score in the most calcified vessel;
AUC: area under the curve; HDL-C: high-density
lipoprotein-cholesterol; LASSO: least absolute
shrinkage and selection operator; ROC: receiver
operating characteristic.

the work are appropriately investigated and resolved. Validation. Runlu Sun: Funding acquisition, Validation. Jie Zhang:
Visualization. Jingjing Huang: Investigation. Qian Chen: Data cura­
tion, Funding acquisition. Junjie Wang: Supervision. Jun Shen:
CRediT authorship contribution statement
Conceptualization, Supervision. Yuling Zhang: Conceptualization,
Funding acquisition, Supervision, Writing – review & editing.
Wenya Chen: Investigation, Data curation, Formal analysis, Writing
– original draft. Hongwei Li: Investigation, Methodology, Visualization,
Writing – original draft. Zhijiao Lu: Data curation, Investigation. Qi
Guo: Funding acquisition, Visualization. Xiao Liu: Methodology,

6
W. Chen et al. European Journal of Radiology 147 (2022) 110134

Declaration of Competing Interest of different thresholds and quantification methods, Radiology 227 (3) (2003)
795–801.
[8] M.J. Blaha, M.J. Budoff, R. Tota-Maharaj, Z.A. Dardari, N.D. Wong, R.A. Kronmal,
The authors declare that they have no known competing financial J. Eng, W.S. Post, R.S. Blumenthal, K. Nasir, Improving the CAC score by addition
interests or personal relationships that could have appeared to influence of regional measures of calcium distribution: multi-ethnic study of atherosclerosis,
the work reported in this paper. JACC Cardiovasc. Imaging. 9 (12) (2016) 1407–1416.
[9] S.J. Al’Aref, G. Maliakal, G. Singh, A.R. van Rosendael, X. Ma, Z. Xu, O.A.
H. Alawamlh, B. Lee, M. Pandey, S. Achenbach, M.H. Al-Mallah, D. Andreini, J.
Acknowledgements J. Bax, D.S. Berman, M.J. Budoff, F. Cademartiri, T.Q. Callister, H.J. Chang,
K. Chinnaiyan, B.J.W. Chow, R.C. Cury, A. DeLago, G. Feuchtner, M. Hadamitzky,
J. Hausleiter, P.A. Kaufmann, Y.J. Kim, J.A. Leipsic, E. Maffei, H. Marques, P.
This work was supported by the National Natural Science Foundation A. Gonçalves, G. Pontone, G.L. Raff, R. Rubinshtein, T.C. Villines, H. Gransar, Y. Lu,
of China (NSFC) [grant numbers 81970388, 81700397, and 81903299] E.C. Jones, J.M. Peña, F.Y. Lin, J.K. Min, L.J. Shaw, Machine learning of clinical
the Natural Science Foundation of Guangdong [grant number variables and coronary artery calcium scoring for the prediction of obstructive
coronary artery disease on coronary computed tomography angiography: analysis
2019A1515011682], the Development Funding of the International from the CONFIRM registry, Eur. Heart J. 41 (2020) 359–367.
Science and Technology Innovation Center in the Greater Bay Area of [10] G.L. Raff, M.J. Gallagher, W.W. O’Neill, J.A. Goldstein, Diagnostic accuracy of
the 2020–2022 Science and Technology Project of Guangdong Province noninvasive coronary angiography using 64-slice spiral computed tomography,
J. Am. Coll. Cardiol. 46 (3) (2005) 552–557.
[grant number 2021A0505030021], and the Yat-sen Launching Project [11] A. Arbab-Zadeh, J.M. Miller, C.E. Rochitte, M. Dewey, H. Niinuma, I. Gottlieb,
[grant number YXQH202014]. N. Paul, M.E. Clouse, E.P. Shapiro, J. Hoe, A.C. Lardo, D.E. Bush, A. de Roos,
C. Cox, J. Brinker, J.A.C. Lima, Diagnostic accuracy of computed tomography
coronary angiography according to pre-test probability of coronary artery disease
Appendix A. Supplementary material
and severity of coronary arterial calcification. The CORE-64 (Coronary Artery
Evaluation Using 64-Row Multidetector Computed Tomography Angiography)
Supplementary data to this article can be found online at https://doi. International Multicenter Study, J. Am. Coll. Cardiol. 59 (4) (2012) 379–387.
org/10.1016/j.ejrad.2021.110134. [12] J. Knuuti, W. Wijns, A. Saraste, D. Capodanno, E. Barbato, C. Funck-Brentano,
E. Prescott, R.F. Storey, C. Deaton, T. Cuisset, S. Agewall, K. Dickstein,
T. Edvardsen, J. Escaned, B.J. Gersh, P. Svitil, M. Gilard, D. Hasdai, R. Hatala,
References F. Mahfoud, J. Masip, C. Muneretto, M. Valgimigli, S. Achenbach, J.J. Bax, Group
ESCSD, 2019 ESC Guidelines for the diagnosis and management of chronic
[1] A.S. Agatston, W.R. Janowitz, F.J. Hildner, N.R. Zusmer, M. Viamonte, R. Detrano, coronary syndromes, Eur. Heart J. 41 (2020) 407–477.
Quantification of coronary artery calcium using ultrafast computed tomography, [13] F.L.J. Visseren, F. Mach, Y.M. Smulders, D. Carballo, K.C. Koskinas, M. Back,
J. Am. Coll. Cardiol. 15 (4) (1990) 827–832. A. Benetos, A. Biffi, J.M. Boavida, D. Capodanno, B. Cosyns, C. Crawford, C.
[2] T.S. Genders, E.W. Steyerberg, M.G. Hunink, K. Nieman, T.W. Galema, N.R. Mollet, H. Davos, I. Desormais, E. Di Angelantonio, O.H. Franco, S. Halvorsen, F.D.
P.J. de Feyter, G.P. Krestin, H. Alkadhi, S. Leschka, L. Desbiolles, M.F. Meijs, M. R. Hobbs, M. Hollander, E.A. Jankowska, M. Michal, S. Sacco, N. Sattar,
J. Cramer, J. Knuuti, S. Kajander, J. Bogaert, K. Goetschalckx, F. Cademartiri, L. Tokgozoglu, S. Tonstad, K.P. Tsioufis, I. van Dis, I.C. van Gelder, C. Wanner,
E. Maffei, C. Martini, S. Seitun, A. Aldrovandi, S. Wildermuth, B. Stinn, J. Fornaro, B. Williams, Group ESCSD, 2021 ESC Guidelines on cardiovascular disease
G. Feuchtner, T. De Zordo, T. Auer, F. Plank, G. Friedrich, F. Pugliese, S. prevention in clinical practice, Eur. Heart J. 42 (2021) 3227–3337.
E. Petersen, L.C. Davies, U.J. Schoepf, G.W. Rowe, C.A. van Mieghem, L. van [14] J.J. Linde, H. Kelbaek, T.F. Hansen, P.E. Sigvardsen, C. Torp-Pedersen, J. Bech,
Driessche, V. Sinitsyn, D. Gopalan, K. Nikolaou, F. Bamberg, R.C. Cury, J. Battle, M. Heitmann, O.W. Nielsen, D. Hofsten, J.T. Kuhl, I.E. Raymond, O.P. Kristiansen,
P. Maurovich-Horvat, A. Bartykowszki, B. Merkely, D. Becker, M. Hadamitzky, I.H. Svendsen, M.H.D. Vall-Lamora, C. Kragelund, M. de Knegt, J.D. Hove,
J. Hausleiter, M. Dewey, E. Zimmermann, M. Laule, Prediction model to estimate T. Jorgensen, G.G. Fornitz, R. Steffensen, B. Jurlander, J. Abdulla, S. Lyngbaek,
presence of coronary artery disease: retrospective pooled analysis of existing H. Elming, S.K. Therkelsen, E. Jorgensen, L. Klovgaard, L.E. Bang, P.R. Hansen,
cohorts, BMJ 344 (2012) 10. S. Helqvist, S. Galatius, F. Pedersen, U. Abildgaard, P. Clemmensen, K. Saunamaki,
[3] S.M. Grundy, N.J. Stone, A.L. Bailey, C. Beam, K.K. Birtcher, R.S. Blumenthal, L. L. Holmvang, T. Engstrom, G. Gislason, L.V. Kober, K.F. Kofoed, Coronary CT
T. Braun, S. de Ferranti, J. Faiella-Tommasino, D.E. Forman, R. Goldberg, P. angiography in patients with non-ST-segment elevation acute coronary syndrome,
A. Heidenreich, M.A. Hlatky, D.W. Jones, D. Lloyd-Jones, N. Lopez-Pajares, C. J. Am. Coll. Cardiol. 75 (2020) 453–463.
E. Ndumele, C.E. Orringer, C.A. Peralta, J.J. Saseen, S.C. Smith Jr., L. Sperling, S. [15] R. Tota-Maharaj, P.H. Joshi, M.J. Budoff, S. Whelton, I. Zeb, J. Rumberger, M. Al-
S. Virani, J. Yeboah, 2018 AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/ Mallah, R.S. Blumenthal, K. Nasir, M.J. Blaha, Usefulness of regional distribution of
APhA/ASPC/NLA/PCNA guideline on the management of blood cholesterol: a coronary artery calcium to improve the prediction of all-cause mortality, Am. J.
report of the American college of cardiology/American heart association task force Cardiol. 115 (9) (2015) 1229–1234.
on clinical practice guidelines, Circulation 139 (2019) e1082–e1143. [16] P.J. Bickel, Y.A. Ritov, A.B. Tsybakov, Simultaneous analysis of Lasso and Dantzig
[4] T.S. Polonsky, R.L. McClelland, N.W. Jorgensen, D.E. Bild, G.L. Burke, A.D. Guerci, selector, Ann. Stat. 37 (4) (2009) 1705–1732.
P. Greenland, Coronary artery calcium score and risk classification for coronary [17] B. Efron, T. Hastie, I. Johnstone, R. Tibshirani, Least angle regression, Ann. Stat. 32
heart disease prediction, JAMA 303 (2010) 1610–1616. (2) (2004), https://doi.org/10.1214/009053604000000067.
[5] T.Q. Callister, B. Cooil, S.P. Raya, N.J. Lippolis, D.J. Russo, P. Raggi, Coronary [18] S. Lee, S. Kwon, Y. Kim, A modified local quadratic approximation algorithm for
artery disease: improved reproducibility of calcium scoring with an electron-beam penalized optimization problems, Comput. Stat. Data Anal. 94 (2016) 275–286.
CT volumetric method, Radiology 208 (3) (1998) 807–814. [19] J.D. Bundy, S.R. Heckbert, L.Y. Chen, D.M. Lloyd-Jones, P. Greenland, Evaluation
[6] M. Ferencik, A. Ferullo, S. Achenbach, S. Abbara, R.C. Chan, S.L. Booth, T.J. Brady, of risk prediction models of atrial fibrillation (from the Multi-Ethnic Study of
U. Hoffmann, Coronary calcium quantification using various calibration phantoms Atherosclerosis [MESA]), Am. J. Cardiol. 125 (1) (2020) 55–62.
and scoring thresholds, Invest. Radiol. 38 (2003) 559–566. [20] S.Y. Jang, S.M. Kim, J. Sung, S.J. Cho, Y.H. Choe, Coronary artery calcium scores
[7] C. Hong, K.T. Bae, T.K. Pilgram, Coronary artery calcium: accuracy and and cardiovascular risk factors in 31,545 asymptomatic Korean adults, Int. J.
reproducibility of measurements with multi-detector row CT–assessment of effects Cardiovasc. Imaging 32 (S1) (2016) 139–145.