Predictive capacity of the HFA-ICOS risk stratification scale for cardiovascular adverse events in patients treated with proteasome inhibitors

Introduction The development of proteasome inhibitors has transformed hematologic malignancy management, significantly improving cancer patient outcomes. Despite their benefits, these drugs are associated with cardiovascular adverse events (CVAEs) in a significant proportion of patients. In 2020, HFA-ICOS developed a cardiovascular risk stratification scale to assess baseline CVAE risk in proteasome inhibitor-treated patients. However, to date, no published data validates its predictive ability, despite the fact that accurate risk stratification is essential for optimizing follow-up for high-risk patients. Purpose To validate the accuracy of the HFA-ICOS risk stratification scale for predicting CVAEs in proteasome inhibitor-treated cancer patients in routine clinical practice. Methods A retrospective, single-center study included patients treated with proteasome inhibitors between 2019 and 2024. Baseline demographics, lifestyle, cardiovascular risk factors, cardiac biomarkers, and preexisting cardiovascular diseases or prior cardiotoxic cancer treatments were recorded in accordance with the HFA-ICOS scale. Patients were classified into four risk groups: low, medium, high, and very high risk. CVAEs included cardiovascular death, arrhythmias (atrial fibrillation/flutter), heart failure, worsening/new arterial hypertension, and thromboembolic or ischemic events during treatment. All CVAEs were cardiologist-assessed. Baseline variables were compared between patients with and without CVAEs. Survival and ROC curves were generated to assess the scale’s predictive accuracy. Results A total of 93 patients were included; 80 (86%) received bortezomib and 13 (14%) received carfilzomib. Seventy-six percent of patients were classified as high or very high risk, leaving only 24% in the low or moderate-risk groups. Fifteen patients (16.1%) developed CVAEs, with a higher proportion of patients treated with carfilzomib compared to those without cardiovascular events (40.0% vs. 9.0%, p = 0.002) (Table 1). Survival analysis showed no statistically significant differences in cardiac events between the risk groups (p = 0.750), but low-risk and very high-risk patients were more accurately stratified than medium- and high-risk patients (Fig. 1A). The ROC curve for predicting CVAEs yielded an AUC of 0.609 (95% CI: 0.461–0.757), indicating limited discriminative ability (Fig. 1B). While the model performs better than random classification, its predictive accuracy remains suboptimal. Conclusions In our population, the HFA-ICOS risk stratification scale demonstrated limited discriminatory ability. Low-risk and very high-risk patients were more accurately classified but it struggled to differentiate between medium- and high-risk groups. The variables included in the scale should be refined to improve the stratification of high-risk patients.