CyFRASS-TR: A practical clinical predictive model for acute irinotecan-induced toxicity to enhance patient safety in solid tumor management

Abstract

Irinotecan-based chemotherapy regimens are mainstay treatments for various solid tumors but are frequently complicated by acute toxicities that can impair treatment adherence and clinical outcomes. Personalized risk prediction is essential to optimize patient tolerance and safety. This study sought to develop and validate the CyFRASS-TR model, a clinical predictive tool designed to identify patients at high risk of acute irinotecan-induced toxicity, thereby enabling personalized preventative interventions. A retrospective cohort study was conducted on 429 patients with solid tumors treated with irinotecan at a single cancer center. Multivariable logistic regression was utilized for model construction, followed by rigorous internal validation via cross-validation procedures. The CyFRASS-TR scoring model demonstrated robust discriminatory capacity with an area under the receiver operating characteristic curve (AUC-ROC) of 0.8923. At a 27-point cutoff, the model yielded a sensitivity of 94.23%, a specificity of 60.00%, and a negative predictive value (NPV) of 79.0%. Significant predictors included female gender, impaired renal function (eGFR <80 mL/min), elevated serum alkaline phosphatase (ALP ≥200 IU/L), primary tumor site, surgical history, chemotherapy regimen, and treatment cycle. The CyFRASS-TR model serves as an effective screening tool for identifying high-risk patients, facilitating targeted supportive care such as atropine premedication, particularly in clinical environments where genetic testing is unavailable.