Coronary artery calcium quantification on routine contrast-enhanced CT-scans in patients with cancer receiving immune checkpoint inhibitors: a prospective cohort study
A Strijdhorst, K Froeling, T C Dormans, S Vyawahare, D Karkalousos, N S Nurmohamed, R A P Takx, I Isgum, N EsAbstract
Background
Patients with cancer are at increased risk of atherosclerotic cardiovascular (ASCVD) events, yet guidelines provide limited guidance on risk stratification. Coronary artery calcium (CAC) on chest computed tomography (CT) scans routinely performed in oncology care may help identify high-risk patients.
Objective
To compare CAC volume quantified on contrast-enhanced chest CT scans versus dedicated non-contrast, ECG-gated cardiac CT scans.
Methods
Data were used from ITHACA, a prospective cohort study including adult patients with cancer who underwent both dedicated non-contrast, ECG-gated cardiac CT (reference) and routine contrast-enhanced, non-ECG gated chest CT. CAC volume was quantified on the reference scan using standard software, and on contrast-enhanced CT scans using manual annotation, with application of a scan-specific threshold to account for the presence of contrast by a standardized method (Figure 1). The correlation between CAC volume measured on contrast-enhanced scans and the reference scan was assessed using Spearman’s correlation. Secondary analyses included agreement by Bland-Altman analysis, risk categorization using regression-based volume thresholds (reflecting Agatston risk groups), and interclass correlation.
Results
A total of 103 contrast-enhanced CT scans from 56 patients were included. Median age was 67 (IQR 59-72), and 80% were male. Median time between reference and contrast-enhanced CT was 85 days (IQR, 44-147). Median CAC volume was 191.58 mm³ (IQR, 10.15–834.64) on non-contrast CT and 84 mm³ (IQR, 13–559) on contrast-enhanced CT, with 84% and 79% having a volume >0 mm³, respectively. A strong correlation was observed between volume on reference CT and manually annotated volume on contrast-enhanced CT (ρ=0.78, 95% CI 0.68-0.85) (Figure 2). Bland Altman analysis showed minimal systemic bias and no evidence of proportional bias. Interobserver reliability was excellent (intraclass correlation coefficient (ICC) = 0.98, 95% CI 0.96-0.98). Agreement in Agatston risk categories, assigned by applying regression-based volume thresholds to CAC volume, was substantial (Cohen’s kappa 0.76), with most discrepancies in patients with low CAC volume.
Conclusion
CAC volume measured on contrast-enhanced chest CT scans performed in routine cancer care strongly correlates with CAC volume on dedicated non-contrast cardiac CT. Further studies should focus on developing and validating methods for automated CAC quantification on contrast-enhanced scans.