DOP11 A novel diagnostic serum protein signature for paediatric Inflammatory Bowel Disease: A discovery and validation study in two independent inception cohorts

Abstract Background A diagnostic delay is common in paediatric inflammatory bowel disease (PIBD) and is associated with impaired outcomes. Therefore, we aimed to identify and validate a diagnostic protein signature of PIBD in blood. Methods Using the Proximity Extension Assay technology (Olink Proteomics®), we assayed plasma proteins in an inception cohort of treatment-naïve paediatric patients referred to Uppsala University Children's Hospital, Sweden for suspected IBD and validated findings in an independent population-based paediatric inception cohort, i.e., the Norwegian IBSEN III cohort. Diagnosis was based on the ESPGHAN/Porto criteria. The false discovery rate approach was applied, and q-values were reported. Multivariable analyses and supervised machine learning were used to identify a diagnostic protein signature, and its performance was compared to clinically established biomarkers. Results The Swedish discovery cohort included 58 PIBD patients and 36 symptomatic controls without any discernible evidence of IBD, and the validation cohort 79 PIBD patients and 37 symptomatic controls (Table 1). In total, 154 proteins were included in the final analysis. Univariable analyses identified 26 differentially regulated proteins for PIBD versus symptomatic controls in the discovery cohort (q<0.05), and 15 of these were validated in IBSEN III. Using regularized regression, we identified a model of 31 proteins that distinguished PIBD from symptomatic controls in the discovery cohort (area under the curve (AUC) = 0.87; 95%CI: 0.79-0.93). The relative contribution of each protein is shown in Figure 1. The diagnostic capacity of the protein signature (AUC=0.83; 95% CI: 0.75-0.90) outperformed high sensitivity C-reactive protein (hsCRP) (AUC=0.72; 95%CI: 0.63-0.82) in the validation cohort (p=0.01, Figure 2). Similar results were obtained from our random forest models. Only 76/116 children in the validation cohort provided a stool sample. Among these children, the diagnostic capacity of the protein signature (AUC=0.81; 95%CI 0.70-0.90) was slightly lower than faecal Calprotectin (FCP) (AUC=0.93; 95%CI 0.86-0.98) (p=0.014). To assess the clinical relevance of the novel protein signature, we evaluated its capacity to rule out IBD compared to increased hsCRP (>5.0 mg/L) in the validation cohort. The protein signature demonstrated a significantly higher negative predictive value (55.3%) compared to increased hsCRP (39.5%) (p=0.002). Conclusion We identified and validated a diagnostic protein signature for PIBD in blood that is superior to CRP and has potential for better clinical utility than FCP since faecal samples are difficult to obtain. This signature could be used to guide decisions on referral to endoscopy and facilitate early diagnosis of IBD.