Calcium score for cardiovascular risk profiling and stratification in pan-cancer patients undergoing anti-vegfr-targeted tkis

Background VEGFR-targeted TKIs used to treat various solid tumors often cause adverse cardiovascular events, including hypertension and arterial stiffening. While hypertension is a known marker of efficacy, it lacks the sensitivity to predict vascular damage (1,2). In this context, the coronary artery calcium (CAC) score, detected on staging CT scans, emerges as a potential tool to quantify baseline vascular burden (3). Purpose To evaluate the clinical significance of coronary calcium scores detected on staging CT scans in a pan-cancer population treated with these therapies. We aimed to assess whether baseline vascular calcification correlates with treatment-related cardiovascular toxicities. Methods This retrospective study involved pan-cancer patients treated with Anti VEGFR-targeted-TKIs (e.g. bevacizumab, cabozatinib, lenvatinib) who underwent staging chest CT scans. Descriptive statistical analyses were performed using IBM SPSS Statistics software. Automatic segmentation and feature extraction was performed using open-source tools and locally developed scripts. A Random Forest classifier was trained following feature selection; SHAP values were used for feature importance interpretation, and UMAP+HDBSCAN clustering was applied to identify patient subgroups. We internally developed and validated a LightGBM classification model, achieving a test accuracy of 0.77. The model incorporates arterial hypertension and 3 parameters derived from staging CT imaging: aortic and cardiac calcium burden (see table 1), and visceral fat . Results The study included 129 patients, with a mean age of 60.2 years and a male predominance (58.1%). At baseline 38.0% of patients had hypertension, (11.6%) diabetes mellitus, (9.3%) heart disease and (7.8%) dyslipidemia. 32.8% were overweight, and 8.8% were underweight. During treatment, 29 patients (22.48%) developed hypertension. Among those evaluated for severity (n=27), 59.3% experienced Grade 1 and 37.0% Grade 2 hypertension. Clinical management necessitated treatment suspension in 34.5% of these cases. Two patients developed pulmonary embolism. No other major adverse cardiovascular events. Unsupervised clustering identified 5 patient clusters: 2 high-risk, 2 intermediate-risk, and 1 low-risk, see figure 1-2. Parameters extracted from CT demonstrated a strong ability to identify treatment-related events. A clear relationship emerged between aortic plaque, visceral fat, and the development of events, with risk increasing proportionally to their presence. Conclusion Staging CT scans allow for the extraction of cardiac and vascular organ damage indices. Coronary and aortic calcium score identifies patients at high risk for drug-related events. Furthermore, it has proven to be an effective parameter for profiling and stratifying patient risk, particularly in the absence of traditional risk factors.