To Örebro University

Background: The clinical course of inflammatory bowel disease (IBD) is highly heterogeneous. Previously we identified a protein-based prognostic signature in serum using the proximity extension assay (PEA). To enable clinical translation, further assay development is required, including the establishment of standard curves and clinically actionable cut-offs. Therefore, we aimed to develop, optimise and validate a serum-based assay that quantifies absolute protein concentrations at diagnosis to identify individuals at higher risk of an aggressive course of IBD.

Methods: Proteins were selected from the commercially available Olink inflammation and oncology II panels as well as from five custom-made multiplex panels comprising 460 proteins based on genes in 163 IBD-risk loci (the IBD Character project1). Serum from 71 newly diagnosed IBD patients (Crohn’s disease, n = 29; ulcerative colitis, n = 42) (SIC-IBD) was analysed. Proteins associated with an aggressive course: I) need for biologics, ciclosporin, or surgery for flare after initial induction or II) hospitalisation, refractoriness to targeted therapy, excessive corticosteroid use or surgery within one year, were identified using penalised logistic regression. From these proteins a focused prognostic panel was developed and analytically optimized for absolute quantification. Model performance was evaluated by nested cross-validation (area under the curve [AUC]) and externally validated in 329 treatment-naïve IBD patients from the population-based IBSEN III cohort. Time to the composite outcome was analysed by estimating hazard rations (HR) using Cox regression.

Results: Thirteen proteins (CCL19, XPNPEP2, TGFα, VGFA, DLL1, PSGL1, OSM, CSF1, FURIN, IL-18, EpCAM, and TNFRSF9) met the pre-defined criteria and formed the final prognostic signature. In the discovery cohort, the model achieved an AUC of 0.76, with no interaction with age, sex or IBD subtype. External validation in the IBSEN III cohort yielded an AUC of 0.66 (95%CI 0.57-0.75), sensitivity 49% and specificity 70% at the discovery cohort derived optimal cut-off. Among patients classified as high risk, compared to low risk, the HR for the composite endpoint of aggressive disease course was 1.81 (P = 0.0002, Figure 1).

Conclusion: We developed and validated the first PEA-based serum panel enabling absolute quantitation of prognostic proteins for clinical use. At diagnosis, IBD-patients classified as high risk based on the 13-protein signature had an 81% higher relative risk of an aggressive disease course, supporting further evaluation.