Development of Predictive Models for Long-Term Endoscopic Response to Ustekinumab in Crohn's Disease Based on Plasma Proteomics.

Purpose: Ustekinumab (UST) is effective for Crohn's disease (CD), yet reliable biomarkers for predicting long-term response remain scarce. This study aimed to identify novel plasma proteomic biomarkers and develop predictive models for long-term endoscopic response to UST therapy in patients. Methods: Baseline plasma inflammatory proteins were profiled using the Olink platform in 40 CD patients treated with UST (20 responders, 20 non-responders). Differentially expressed proteins (DEPs) were identified after adjusting for age, age at diagnosis, and SES-CD scores. Stable DEPs were selected via bootstrap resampling and further validated in an independent patient group from the same center (n=20) using ELISA. Predictive models were constructed using logistic regression, random forest, and support vector machine (SVM) algorithms. Results: Non-responders had elevated baseline interleukin-8 (IL8) and CD6 levels and reduced thymic stromal lymphopoietin (TSLP) compared to responders. ELISA confirmed differential expression of IL8, CD6, and TSLP, with CD6 showing the best diagnostic accuracy (AUC=0.800). The logistic regression model combining these markers achieved an AUC of 0.828 (95% CI: 0.701-0.954), outperforming random forest (AUC=0.745) and SVM (AUC=0.775). Conclusion: Baseline plasma IL8, CD6, and TSLP are potential predictive biomarkers of long-term endoscopic response to UST in CD, providing a basis for personalized treatment strategies.