P592 Higher predictive power of epigenetic signatures for response to vedolizumab and ustekinumab in anti-TNF naïve patients with active Crohn’s disease

Abstract

Abstract Background Despite the significant progress in IBD care with biological therapies, the current 'trial-and-error' method remains suboptimal. Previous results from our EPIC-CD study demonstrated that two distinct epigenetic panels of 25 and 68 DNA methylation markers, respectively, were associated with (non-)response to vedolizumab (VDZ) and ustekinumab (USTE) in active Crohn’s disease (CD)1. Data from clinical trials have demonstrated that response rates in patients naïve to biological treatment are higher. We therefore aimed to evaluate the performance of our models in bionaive and anti-TNF exposed subpopulations. Methods We prospectively measured peripheral blood DNA methylation profiles from 184 adult CD patients prior to treatment with VDZ or USTE in a discovery (n=126) and independent validation (n=58) cohort using the Illumina EPIC BeadChip array. Between 26 and 52 weeks, patients were categorized as responders or non-responders based on a combination of endoscopic response (≥50% reduction in SES-CD score), steroid-free clinical response (≥3-point drop in Harvey-Bradshaw Index (HBI) or HBI ≤4 AND no systemic steroids), and/or biochemical response (≥50% reduction in C-reactive protein (CRP) and fecal calprotectin or a CRP ≤5 g/mL and fecal calprotectin ≤250 µg/g). Epigenetic biomarker identification and classification analyses were conducted through stability selection gradient boosting on the discovery cohort. Subsequently, the performance of the identified epigenetic biomarkers was assessed independently on the validation cohort, with patients stratified by prior anti-TNF exposure. Results Out of the 58 patients in the validation cohort, 25 received treatment with VDZ, and 33 with USTE, with 52% (n=13) and 36% (n=12) respectively naïve to anti-TNF treatment. Baseline characteristics between anti-TNF naive and anti-TNF exposed patients showed no significant differences, despite a slight significant age difference in the vedolizumab group, where anti-TNF exposed patients were significantly younger (table 1). In the overall independent validation cohort, the prediction models demonstrated an AUC of 0.75 for both VDZ and USTE. However, after stratification by prior anti-TNF exposure, we observed increased performance of our models among anti-TNF naive patients for both VDZ (AUCnon-exposed=0.85 versus AUCexposed=0.66) and USTE (AUCnon-exposed=0.97 versus AUCexposed=0.63 (figure 1). Conclusion Our findings demonstrate that prior anti-TNF exposure leads to somewhat reduced predictive power of our epigenetic models for Vedolizumab and Ustekinumab response. The predictive models may therefore find its optimal application in bio-naive patients, although they remain significant in anti-TNF exposed patients.