P192 Personalizing treatment for predicting response to infliximab in children with Crohn’s disease using proteomic biomarkers

Abstract

Abstract Background Identifying minimally invasive diagnostic biomarkers that predict and monitor response to TNF inhibitors would enable treatment to be personalized and increase knowledge of the biological pathways associated with response or lack thereof. Methods Children and adolescents with Crohn’s disease who started treatment with infliximab (IFX) were enrolled in a prospective, multi-center (n=6) trial to identify biomarkers that predict response to IFX. Electronic case report forms were prepared by Studytrax (Macon, GA). Subjects <18 years of age were diagnosed with Crohn’s disease using standard clinical/pathologic criteria. Clinical evaluation and serum samples were obtained prior to initiation of IFX, after induction (dose 4), at six months, and after one year or at the time of treatment withdrawal. There were 119 subjects enrolled and 102 subjects completed induction. Non-responders (n=15) were defined prior to the 4th infusion as having a PCDAI > 10, an elevated Physician Global Assessment (PGA) or requiring dose escalation or medication change. Responders at the 4th infusions were either in remission (PUCAI ≤ 10) or were described as quiescent by PGA. Using the 7k SomaScan (SomaLogic; Boulder, CO) proteomics platform, quantitative protein profiles were generated from banked pre-treatment (V1) and after the three-dose induction (V2) to identify candidate protein biomarkers predictive of therapeutic response. Associations between the 7,310 proteins detected with SomaScan and response to IFX were assessed using t test. Response predictor models were developed using linear regression with backward selection. We report our preliminary findings from a discovery cohort of 56 subjects who completed induction. Results There were 15/56 (27%) non-responders and 41/56 (73%) responders at V2 after IFX induction. SomaScan analysis of the 56 individuals identified sets of differentially expressed proteins discriminating between non-responders and responders at both V1 and V2 and also defined protein signatures that changed from V1 to V2 for responders and non-responders, respectively. Strikingly, only responders demonstrated significant decreases in pro-inflammatory pathways at V2 compared to V1, thereby correlating well with anticipated reduced inflammation in responders. A 6-protein linear regression model performed with an AUC of 0.99 for predicting IFX response at pre-treatment V1. Conclusion In a discovery cohort from a multicenter, prospective trial of children and adolescents treated with infliximab, a comprehensive proteomics platform was able to identify serum proteins that are highly associated with response to infliximab and accurately predict response prior to treatment.