78 Background: Colorectal cancer (CRC) remains a major health burden in Korea. Oxaliplatin is widely used in CRC treatment, but its dose-limiting toxicity is chemotherapy-induced peripheral neuropathy (CIPN), a debilitating condition that may persist long-term and impair quality of life. Current management strategies, such as dose reduction or discontinuation, are largely empirical. Predictive models integrating patient-specific clinical and comorbidity data are lacking. This study aimed to develop and validate a machine learning (ML) model to predict CIPN risk using real-world data. Methods: A retrospective cohort study was conducted using the Korea-Clinical data Utilization network for Research Excellence (K-CURE), a nationwide registry. Eligible patients were ≥18 years, diagnosed with CRC, and treated with oxaliplatin-based chemotherapy. Exclusion criteria included prior use of neuropathy-related drugs, non-oncology/neurology prescriptions, or missing chemotherapy records. CIPN was defined as a new prescription of pregabalin or gabapentin. Predictors included demographics, BMI, cancer stage, chemotherapy dose, comorbidities, and polypharmacy. Four ML models (XGBoost, Random Forest, logistic regression, LightGBM) were trained, and feature importance was assessed using SHapley Additive exPlanations (SHAP). Subgroup analyses were performed on established risk factors. Results: From 53,485 CRC patients diagnosed between 2012–2020, 10,682 met inclusion criteria. CIPN occurred in 5,523 (51.7%). The Random Forest model showed the best performance (AUROC 0.90, accuracy 0.91, F1-score 0.91). SHAP analysis highlighted cumulative chemotherapy dose, diabetes, cardiovascular disease, and polypharmacy as major predictors. Subgroup analyses confirmed higher CIPN risk in older patients, those with diabetes, and those receiving higher cumulative oxaliplatin doses. Conclusions: A clinically interpretable ML model was developed to predict CIPN risk in CRC patients. The model demonstrated strong accuracy and identified key clinical predictors. Implementation as a clinical decision support system could help oncologists identify high-risk patients before treatment, enabling proactive strategies to maintain therapy and preserve quality of life.
Cho et al. (Sat,) studied this question.