Integrating inflammatory and coagulation biomarkers for surgical risk stratification and treatment benefit assessment in Crohn’s disease

Introduction Inflammatory and coagulation abnormalities are closely linked to disease progression in Crohn’s disease (CD). However, whether integrating these biomarkers can improve long-term surgical risk stratification and inform treatment decision-making remains unclear. Methods A total of 1,060 patients with CD were enrolled. Candidate predictors, including inflammatory and coagulation biomarkers, were selected using LASSO-Cox regression and the Boruta algorithm. Surgery-free survival was assessed using Kaplan-Meier analysis. Eight machine learning models were developed and evaluated using five-fold cross-validation. Shapley additive explanations (SHAP) were used to interpret the best-performing model. Inverse probability weighting was applied to reduce confounding and selection bias and to assess the benefit of biologic therapy across risk strata. Results Inflammatory and coagulation biomarkers, together with disease-related features, were major determinants of long-term surgical risk in CD. Among the eight models tested, the gradient boosting machine (GBM) achieved the best performance, with a C-index of 0.816 (95% confidence interval [CI], 0.789-0.843), significantly outperforming the Cox model (0.680, 95% CI, 0.643-0.717; P < 0.001). The model showed robust time-dependent discrimination, with 1-, 3-, 5-AUC values up to 0.836, 0.851, 0.832, and an integrated Brier score of 0.103. SHAP analysis indicated that inflammatory and coagulation markers together contributed approximately 70% of GBM model explainability. Consistent with these findings, dual-mediator analysis showed that fibrinogen accounted for 44.5% of the inflammation-associated increase in surgical risk, whereas D-dimer mediated 6% of the excess risk. Compared with the Cox model, the GBM improved 5-year risk reclassification, with a net reclassification improvement of 0.425 and an integrated discrimination improvement of 0.181. After inverse probability weighting, biologic therapy was associated with significant benefit in the intermediate- (pooled hazard ratio (HR) = 0.44, 95% CI: 0.26-0.75, pooled P = 0.003) and high-risk groups (pooled HR = 0.51, 95% CI: 0.30-0.88, pooled P = 0.016), but not in the low-risk group (pooled HR = 0.87, 95% CI: 0.47-1.61, pooled P = 0.657). An online platform was developed to support individualized risk stratification and treatment assessment. Conclusion Integrating inflammatory and coagulation biomarkers improves surgical risk stratification in Crohn’s disease and may help identify patients most likely to benefit from biologic therapy.