Web-based nomogram for predicting the risk of subtherapeutic infliximab trough levels in pediatric Crohn's disease: personalized risk stratification and data-driven management of infliximab therapy

Background To develop and validate a web-based nomogram for predicting the risk of subtherapeutic infliximab trough levels (IFX-TLs) in Crohn's disease (CD), enabling personalized risk stratification and data-driven management of infliximab (IFX) therapy. Methods This retrospective analysis included 205 IFX trough measurements from 74 pediatric CD patients. Stratified sampling by patient_id was performed to split the dataset into training datasets and testing datasets at an 8:2 ratio. Least absolute shrinkage and selection operator (LASSO) regression and Boruta regression identified independent predictors. A web-based nomogram was constructed. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves, Decision Curve Analysis (DCA), and risk stratification. SHapley Additive exPlanations (SHAP) analysis provided interpretability. Results Six independent predictors were identified: dose escalated, alkaline phosphatase (ALP), albumin (ALB), fecal calprotectin (FC), erythrocyte sedimentation rate (ESR), and Pediatric Crohn’s Disease Activity Index (PCDAI). The nomogram demonstrated good discrimination, with AUCs of 0.762 [training; 95% confidence interval (CI): 0.689–0.836] and 0.783 (testing; 95% CI: 0.609–0.956). Calibration curves showed moderate agreement, and DCA revealed a potential net clinical benefit. Risk stratification using the optimal cutoff (0.57) was able to distinguished high-risk from low-risk groups. SHAP indicated that dose escalated and ALB as protective; FC, ESR, ALP, and PCDAI as risk factors. Notably, dose escalated is an indicator of prior clinical management and should be interpreted as a surrogate of previous severity rather than a causal pharmacokinetic effect. Conclusions We successfully developed and validated a web-based nomogram to assist clinicians in predicting the risk of subtherapeutic IFX-TLs in pediatric CD, helping guide personalized and data-driven management of infliximab therapy. Compared with previously reported model, our nomogram transforms multiple predictors into an intuitive, quantifiable individual risk assessment tool, making it more convenient for physicians to use quickly in clinical settings.