P305 The serum protein profile across the IBD spectrum: Results from the COLLIBRI consortium

Abstract

Abstract Background Inflammatory bowel disease (IBD) is a heterogeneous disorder. Both subtypes, i.e., Crohn’s disease (CD) and ulcerative colitis (UC), differ in disease behaviour and inflamed gastrointestinal segments. Despite this, randomized controlled trials stratify patients based on CD and UC. Molecular characterization could uncover subtype-specific differences that could guide treatment and thereby overcome current therapeutic limitations. Therefore, we aimed to examine differences in serum inflammatory protein profiles across the IBD spectrum. Methods This was a cross-sectional multicentre study of adult patients (≥18 years) with IBD from one Belgian and eight Swedish hospitals in the COLLIBRI consortium. IBD diagnosis and classification was based on international criteria, according to the Montreal classification. Relative serum protein levels were assessed using proximity extension assay technology (Olink Proteomics, Uppsala, Sweden; inflammation panel). We adopted smoothly clipped absolute deviation penalized logistic regression models to discriminate CD and UC patients. Using fitted CD vs UC logistic models, we estimated probability scores of CD vs UC for each patient based on their serum protein profiles. Scores ranged from 0 to 1, where lower scores indicated a higher molecular resemblance to UC. We evaluated the performance using leave-one-out cross-validation and the area under the curve (AUC). Results Relative levels of 86 serum inflammatory proteins were available from 1,551 patients with IBD (CD, N=883; UC, N=639 and IBD-U, N=29) (Table 1). CD vs UC probability scores based on protein estimates for patients with UC, IBDU and different CD phenotypes (ileal CD, L1; colonic CD, L2; ileocolonic CD L3) are shown in Figure 1A. We observed a spectrum of IBD patients based on their CD vs CD probability scores with most pronounced differences between ileal CD and UC. Probability scores also differed significantly between colonic CD and ileal CD, but not between ileal and ileocolonic CD. The model performance to discriminate CD and UC yielded an AUC of 0.75. Restricting the samples to only one CD phenotype vs UC respectively resulted in the highest AUC for ileal CD (0.81), followed by ileocolonic CD (0.75) and colonic CD (0.65). Key proteins in the CD vs. UC model with higher protein estimates in UC were IL-17A, MMP10, FGF19. Contrary, CSF1, and SLAMF1, were higher in CD (Figure 1B). Conclusion Our results on inflammation related serum proteins advocate for a more nuanced classification of CD into ileal-predominant and colonic-predominant subtypes. Such stratification could advance our understanding of IBD pathophysiology and may provide guidance for future therapeutic approaches.