Development and validation of a LASSO-logistic regression model for predicting subtherapeutic infliximab trough concentrations in patients with Crohn's disease.

Abstract

Background: Infliximab (IFX), a monoclonal antibody that neutralizes tumor necrosis factor-α, is widely used as a biologic treatment for Crohn's disease (CD). Despite its established efficacy, a substantial proportion of patients develop subtherapeutic IFX trough concentrations (< 3 μg/mL), leading to diminished clinical response and treatment failure. Early identification of high-risk individuals remains challenging due to the multifactorial nature of IFX pharmacokinetics.

Aim: This study integrated Least Absolute Shrinkage and Selection Operator (LASSO)-based variable selection with multivariable logistic regression to identify CD patients at risk for subtherapeutic IFX trough levels during induction.

Method: A total of 347 patients diagnosed with CD who commenced IFX induction therapy at the Sixth Affiliated Hospital of Sun Yat-sen University from January to December 2023 were retrospectively reviewed in this study. Comprehensive demographic, clinical, and biochemical data were retrieved from electronic records. Variable selection was performed using the Least Absolute Shrinkage and Selection Operator (LASSO), after which a multivariable logistic model was developed. Model discrimination, calibration and clinical usefulness were assessed through the area under the receiver operating characteristic curve (AUC), calibration curves, decision curve analysis (DCA) and clinical impact curves (CIC).

Results: Of the 347 participants, 148 (42.7%) exhibited subtherapeutic IFX trough concentrations. LASSO and multivariable logistic analyses identified four independent predictors: older age at diagnosis (> 40 years), elevated anti-drug antibody levels, higher erythrocyte sedimentation rate and reduced albumin (P < 0.05). The model demonstrated an AUC of 0.737 (95% CI 0.684-0.790), with a bootstrap-adjusted AUC of 0.726 (95% CI 0.697-0.739) based on 1000 resamples. Calibration demonstrated close alignment with observed outcomes, validated by a non-significant Hosmer-Lemeshow test (χ² = 8.447, P = 0.391). DCA and CIC analyses indicated meaningful clinical utility.

Conclusion: The proposed LASSO-logistic regression model demonstrates promising predictive performance for identifying subtherapeutic IFX exposure in CD patients. By leveraging readily available clinical data, it enables early risk stratification and individualized therapeutic decision-making, thereby facilitating more effective treatment optimization.