This study aims to characterize the association between multiple biomarkers related to CV event risk and the CAD phenotype, as defined by the CAC score and the CAD-Reporting and Data System (CAD-RADS) score. Furthermore, this study aims to define whether a comprehensive biomarker signature has the potential to add incremental predictive value to the presence of significant CAD at the stage of primary prevention, before undergoing coronary CTA.

This study is part of a prospective, ongoing, single-center registry on the use of advanced CTA-based imaging analysis for the improvement of patient’s care (ACTION registry).

The registry is currently enrolling an all-comers population with low-intermediate pretest probability referred by general cardiologists for suspected CAD. Patients with a past-medical history of CV events are not excluded from the ACTION registry, but this represented an exclusion criterion for the present analysis. Patients with low-quality CT scans were excluded from the ACTION registry.

All patients enrolled in the registry provided written informed consent and agreed to the research use of their data.

Each consenting patient completed a standardized baseline questionnaire and had blood drawn immediately before undergoing coronary CTA. Blood samples were stored for assessment of the following parameters: lipid profile [(total cholesterol, triglycerides, HDL, LDL, ApoA-1, ApoB, Lp(a)], glycated hemoglobin [HbA1c], high-sensitivity C-reactive protein [hs-CRP]), creatinine, electrolytes, and complete blood count. Estimated glomerular filtration rate (eGFR) was calculated using the chronic kidney disease-EPI 2021 equation. These biomarkers were selected for two main reasons: first, they are widely available, standardized, inexpensive, and are already integrated or have strong potential to be integrated into CV risk stratification in primary prevention. Second, they capture the three principal metabolic drivers of atherosclerosis: impaired glucose metabolism, disordered lipid metabolism, and systemic inflammation. eGFR was added as an additional adjustment factor, given the well-established impact of chronic kidney disease on atherosclerosis and vascular calcification. The ACTION registry received ethical approval from the local ethics committee (Identification code: CA2792).

Data were prospectively collected by the Cardiovascular Research Team in the Clinical Research Facility in a high-volume University Hospital. This study included consecutive patients enrolled in the ACTION registry from January 2024 to June 2025. Demographics and clinical data of each patient, including—among others—age, gender, ethnicity, past medical history, reason for referral by cardiologist, and CV risk factors were prospectively collected by the Cardiovascular Research Team members using a standard questionnaire, as well as vital parameters at the time of the CTA scan, including noninvasive blood pressure. One of the cardiologist researchers of the Cardiovascular Research Team had to be always present during the questionnaire administration to ensure that the clinical data were correctly entered. The SCORE2, SCORE2-OP, and SCORE2-DM were calculated for each patient, respectively for nondiabetic patients <70 years old, patient >70 years old, and diabetic patients. The resulting mean/median percentage of estimated 10-year risk was reported for all three subgroups under the generic term “SCORE2” in the tables. A SCORE2<2.5% ≥2.5% but <7.5%, and ≥7.5%, were defined respectively as low, intermediate, and high-risk scores.

Questionnaires were securely stored in the Clinical Research Facility according to the General Data Protection Regulation (GDPR), and data were transferred into an electronic case report form on RedCap. Blood samples were collected within 30 minutes before the CTA. Samples were then processed on site for most of the tests, except for Lp(a), ApoA-1, and ApoB, which required samples to be sent and processed by Tallaght University Hospital Laboratories. Both laboratories have ISL15189:2012 accreditation.

All coronary CTA scans were performed using a 256-slice multidetector Revolution Apex CT scanner (GE Healthcare, Waukesha, WI) with prospective ECG gating. Collimation of 256 × 0.625 mm with a z-coverage of 16 cm was used with a display field of view of 25 cm. Gantry rotation time was between 350 and 280 ms, dependent on patient size. Tube voltage was between 80 and 120 kV based on patient size (Auto Prescription, GE Healthcare, Waukesha, WI), and an adaptive tube current (SmartmA, GE Healthcare, Waukesha, WI) was used. Patient heart rates were reduced with oral and/or intravenous beta-blockers (Metoprolol) if average heart rate was over 55BPM on arrival to the imaging department, and provided no contraindications were present. Sublingual nitroglycerine (800 mcg) was given routinely immediately before the CT scan, provided no contraindications were present. An amount of 40 to 70 mL of nonionic iodinated contrast media, Iohexol 350 mg/mL (Omnipaque) was administered through an 18 g cannula at a rate of between 4.0 and 7 mL/sec, dependent on scan kV. A 40 mL saline bolus was administered following the contrast at the same infusion rate.

Bolus tracking (Smart Prep, GE Healthcare, Waukesha, WI) was used to monitor the contrast at the level of the carina, and the scan was started manually by the radiographer at peak opacification in the ascending aorta.

For patients with regular heart rates below 60 BPM, the scan was triggered at 75% of one R-R interval during a breath hold. For faster or irregular heart rates, a systolic acquisition at 45% of one R-R interval was also acquired. Both phases utilized padding to allow for digital motion artefact correction (Snap Shot Freeze, GE Healthcare).

All scans were reconstructed at 0.625 mm using a standard soft tissue kernel with DLIR (TrueFidelity, GE Healthcare) set to high.

Scans were analyzed by two highly experienced cardiologists. The scans were reported according to the CAD-RADS recommendations.

The study population was divided into two main subgroups—patients with CAC score = 0 and patients with CAC score >0. Differences within these two subgroups were tested between patients with or without plaque and patients with CAC score ≥ or <101, respectively. Continuous variables were tested for identifying normal distributions according to the Shapiro–Wilk test and Q-Q plot visual examination. They were compared with an unpaired Student t test or with the Mann–Whitney U nonparametric test, when appropriate. Data were expressed as means ± standard deviation or median and interquartile range. Categorical variables were compared using χ ² or Fisher’s exact test as appropriate and expressed as frequencies and percentages. Continuous variables were compared across more than two groups using the Kruskal–Wallis test. When the overall test was statistically significant, pairwise comparisons were performed with Bonferroni-adjusted post hoc tests to account for multiple testing. An unsupervised k-means clustering algorithm was applied to standardized values of laboratory variables (HbA1c, Lp(a), apoA-1, apoB, ratio, LDL, HDL, triglycerides, hs-CRP, eGFR) to identify natural groupings within the study population. k-means is a way to group data into a number of clusters by finding “centers” in the average values of the data provided. The optimal number of clusters was set to three to ensure interpretability of the subsequent statistical comparisons and to avoid creating clusters that were too small. Quantitative clustering diagnostics, including elbow and silhouette analyses, supported K = 3 as a reasonable compromise between cluster compactness and interpretability. Inspection of the elbow plot revealed a clear inflection around K = 3, beyond which further increases in cluster number resulted in less relevant reductions in within-cluster variance (Supplementary Figure S1). The silhouette graph (Supplementary Figure S2) showed that an increased number of clusters would not have reduced clusters overlap, and, instead, would have just increased it for K = 5 or 6. Clustering stability was assessed using a resampling approach. The k-means clustering procedure ( K = 3) was repeated on multiple randomly selected subsets of the study population using the same biomarker variables. Agreement between original and resampled cluster assignments was quantified using the adjusted Rand index (ARI). After a repeated random subsampling generating 15 independent resamples (each resample included 250 randomly selected patients), the following results showed a moderate-to-good stability (mean ARI: 0.51; median ARI: 0.47; ARI min-max range: 0.30-0.83). k-means with K = 3 splits patients into 3 groups (clusters) based on overall similarity across the values of biomarkers. Each group has a centroid (the average profile). In other words, patients in one cluster are more alike than those in other groups, in terms of biomarker values. Each patient was assigned to one of the three resulting clusters for subsequent analyses. To simplify data interpretation, the binary values of CAD-RADS score ≥3 and CAC score ≥101 were chosen to explore the presence and severity of CAD in the three different clusters. In order to test the consistency of the results, a sensitivity analysis was performed, repeating the clustering and the subsequent analyses after excluding patients on lipid-lowering therapy at the time of the coronary CTA.

A multivariable logistic regression model was built to test the independent association of the cluster classification with the presence of CAD-RADS score ≥3.

Three other separate multivariable logistic regression models were built for the overall population and the subgroups of patients with a CAC score of 0 and those with a CAC score >0, to explore significant predictors of the presence of pure lipid plaque and the presence of a CAC score ≥101, respectively, using the single biomarkers.

Variable selection for multivariable model building was performed using the least absolute shrinkage and selection operator (LASSO) regression. All candidate laboratory variables were standardized prior to analysis. The penalty parameter ( λ ) was chosen via 10-fold cross-validation to minimize prediction error, and only variables with nonzero coefficients in the optimal model were retained for inclusion in the final multivariable logistic regression. Odds ratio with 95% confidence interval (CI) was used as the summary effect size. Missing values did not exceed 6% for any variable and occurred exclusively in laboratory results, due to rare errors in blood sample collection. The percentage of missing values for each laboratory variable is reported in Supplementary Figure S3.

For the k-means clustering analysis, listwise deletion was applied, and patients with any missing laboratory value were excluded. Patients excluded due to missing of complete laboratory values ( n = 91, 11.5%) showed no clinically meaningful differences in age, sex, BMI, SCORE2 risk, or prevalence of CAD-RADS ≥3 compared with included patients (Supplementary Table S1). Interobserver variability between the two expert analysts for the adjudication of CAD-RADS ≥ 3 was assessed in a sample of 35 patients, yielding a kappa coefficient of 0.83 (95% CI 0.69-0.98) and a percentage of agreement of 95%.

All data analyses were performed using STATANoW version 18.0 (Stata Corporation, LLC, College Station, TX).

Patients’ clinical characteristics across the four subgroups are presented in Table 1 , while biomarker values for these subgroups are shown in Table 2 . Patients with plaque despite a zero CAC score were comparable with patients without plaque regarding almost all the baseline clinical characteristics, including the median 10-year risk percentage according to the SCORE2. In terms of biomarkers concentration, patients with plaque showed higher HbA1c levels (39 vs 37 mmol/mol, P =.009), despite a similar percentage of diabetic patients, a higher Lp(a) concentrations (16.5 vs 11.5 mg/dL, P =.030), despite a comparable LDL and HDL level, and lower eGFR (88 vs 90 mL/min/1.73 m², P =.022) compared with those without plaque. In the subgroup with CAC score >0, patients with a CAC score≥101 ( n = 171) were older (64 vs 60 years, P <.001), more frequently male (70% vs 56%, P =.003), and had a higher SCORE2 compared with those with CAC score between 1 and 100. In terms of biomarkers, patients with CAC score ≥101 had lower HDL (1.3 vs 1.5 mmol/L, P =.009) and lower ApoA-1 (1.6 vs 1.7 g/L, P =.043) compared with those with CAC score between 1 and 100, with similar values of all the other biomarkers.

Table 2

Laboratory biomarker values in each of the four subgroups.

	Patients with zero calcium score		P -value	Patients with coronary calcification		P -value
	Zero CAC score without plaque ( n = 340)	Zero CAC score with plaque ( n = 42)		CAC score ≥1 and ≤100 ( n = 234)	CAC score ≥101 ( n = 171)
Hb1Ac (mmol/mol)	37 (35-39)	39 (36-41)	.009	39 (37-41)	40.0 (37.0-43.5)	.119
Hs-CRP (mg/L)	1.1 (0.6-2.6)	1.6 (1.1-2.5)	.101	1.4 (0.7-2.7)	1.6 (0.9-3.7)	.140
Total cholesterol (mmol/L)	4.9 (4.2-5.6)	5.0 (3.8-5.9)	.255	4.8 (4.0-5.5)	4.4 (3.7-5.3)	.069
Triglycerides (mmol/L)	1.1 (0.8-1.8)	1.2 (0.9-1.7)	.736	1.4 (0.9-1.8)	1.5 (1.0-2.0)	.092
HDL (mmol/L)	1.5 (1.2-1.9)	1.5 (1.2-1.9)	.754	1.5 (1.2-1.7)	1.3 (1.1-1.6)	.009
LDL (mmol/L)	2.7 (2.0-3.2)	2.7 (1.9-3.6)	.426	2.6 (2.0-3.3)	2.3 (1.9-3.2)	.076
Apo A-I (g/L)	1.6 (1.4-1.9)	1.5 (1.4-1.9)	.713	1.7 (1.4-1.9)	1.6 (1.4-1.8)	.043
Apo B (g/L)	0.9 (0.7-1.0)	0.9 (0.7-1.1)	.868	0.9 (0.7-1.1)	0.9 (0.7-1.1)	.746
Apo B/AboA-1 ratio	0.5 (0.4-0.6)	0.5 (0.4-0.7)	.886	0.6 (0.4-0.7)	0.5 (0.4-0.7)	.569
Lp (a) (mg/dL)	11.5 (8.8-38.0)	16.5 (8.8-86.6)	.030	10.9 (8.8-51.9)	16.4 (8.8-65.0)	.089
eGFR (mL/min/1.73 m²)	90 (81-90)	88 (74-90)	.022	90.0 (77.5-90.0)	87 (76-90)	.346

Only gold members can continue reading. Log In or Register to continue

Share this:

• X
• Facebook

Related

Related posts:

1. Design and rationale of the clinical trial to obtain the highest efficacy of dual antiplatelet therapy after carotid artery stenting in high bleeding risk patients (CHET): A multicenter, randomized, open-label, superiority trial
2. Remote dietitian counseling with short-term meal delivery improves DASH diet adherence and lowers blood pressure in veterans with hypertension and obesity
3. A comprehensive set of per-protocol analyses of the myocardial ischemia and transfusion (MINT) trial
4. Effect of age on restrictive and liberal transfusion outcomes in patients with anemia and myocardial infarction

Stay updated, free articles. Join our Telegram channel

Join