Development and internal validation of a web-based risk assessment tool for chemotherapy-induced peripheral neuropathy: A retrospective cohort study.

Abstract

e24154 Background: Chemotherapy-induced peripheral neuropathy (CIPN) is a common, dose-limiting toxicity that can impair quality of life and lead to dose reductions, treatment delays, or discontinuation, thereby undermining the effectiveness of anticancer therapy. This study aimed to identify key risk factors for CIPN and to develop a clinically translatable prediction model for early risk stratification to inform prevention and management. Methods: We developed a prediction model for chemotherapy-induced peripheral neuropathy (CIPN) in 371 cancer patients receiving chemotherapy at Zhongshan Hospital, Xiamen University (January 2023–September 2025). Missing data were handled using multiple imputation by chained equations. A multivariable logistic regression model was derived with predictors selected by stepwise AIC and/or LASSO, and coefficients were pooled using Rubin’s rules. A random-forest model was trained for benchmarking and interpreted using SHAP values. Internal validation used bootstrap resampling to assess discrimination (AUC) and calibration; clinical utility was evaluated by decision-curve analysis. The final model was implemented as an online risk calculator. Results: Among 371 patients receiving chemotherapy, CIPN occurred in 215 patients (57.95%). The final multivariable logistic model, derived using multiple imputation with pooled estimates, retained 5 predictors (Cancer Type, Triglycerides, Calcium, Globulin, Prothrombin Time). On bootstrap internal validation, the model showed good discrimination (AUC=0.805, 95%CI [0.771–0.860]) and satisfactory calibration (calibration slope =0.8856; intercept =0.0060; bootstrap-corrected calibration curve closely aligned with the ideal line). Decision-curve analysis demonstrated net clinical benefit across a clinically relevant threshold range (0-0.6). A web-based calculator was implemented to provide individualized risk estimates and facilitate bedside use. Conclusions: The proposed model demonstrated good discrimination and calibration on internal validation and may serve as a practical, user-friendly tool to support early identification and risk stratification of patients at high risk for CIPN.