Temporal Alignment and Validation in Machine-Learning Prediction of Secondary Loss of Response to Infliximab [Letter]

Abstract

We read with interest the recent study by Guo et al, who developed an interpretable gradient boosting machine (GBM)- based survival model for predicting secondary loss of response (SLOR) to infliximab (IFX) in patients with Crohn’s disease. 1 By integrating therapeutic drug monitoring (TDM), clinical variables, and SurvSHAP analysis, the study provides a useful framework for individualized risk assessment during long-term IFX therapy. Nevertheless, three methodological issues concerning predictor timing, outcome ascertainment, and model evaluation may further refine interpretation of the findings.