P1022 Machine learning for predicting secondary non-response of Infliximab treatment in Crohn’s disease:a single-center retrospective study

Abstract Background Infliximab (IFX) is effective at inducing and maintaining clinical remission and mucosal healing in patients with Crohn’s disease (CD).However, about 30–50% of patients experiences secondary non-response to IFX treatment. We aimed to develop and test an individualized machine learning model to assess risk factors and predict the rate of secondary non-response in these patients. Methods From January 2018 to December 2022, this study collected data on CD patients who secondary non-response of Infliximab treatment from Chongqing General Hospital. Mean decline accuracy (MDA) and Mean decrease Gini (MDG) was used for recognizing key clinical variables. The optimal predictive model was chosen among seven machine learning algorithms on the basis of area under curve. SHapley Additive exPlanations was employed to interpret the outcome of the predictive model. Results This study ultimately comprised 195 individuals with secondary non-response of Infliximab treatment in Crohn’s disease. The Random Forest model was chosen with the highest area under curve for 0.824. According to SHapley Additive exPlanations results, the most related factors for prediction model were hypersensitive C-reactive protein concentration of Infliximab, disease course, white blood cell count and hemoglobin. Conclusion This study developed seven machine learning models to predict secondary non-response of Infliximab treatment in CD. Compared to other algorithms, the Random Forest model performed significantly better. Conflict of interest: Xiang, Lingya: Song, Xiaomei: No conflict of interest Liu, Yan: No conflict of interest Wei, Tan: No conflict of interest Guo, Hong: No conflict of interest