DOP14 Identification and validation of a lipidomic signature as a novel diagnostic biomarker of paediatric Inflammatory Bowel Disease

Abstract Background Improved diagnostic biomarkers are an unmet clinical need for paediatric inflammatory bowel disease (IBD). We therefore aimed to identify a diagnostic lipidomic signature of IBD in blood. Methods We performed a non-targeted liquid chromatography-time of flight mass spectrometry-based lipidomics (UPLC-QTOFMS) analysis of plasma samples from 211 treatment-naïve children with suspected IBD (Table 1). Discovery cohort: the Uppsala Children’s Hospital cohort (n=94); Validation cohort: children from the population-based IBSEN III inception cohort (n=117). All children underwent diagnostic workup for IBD in accordance with the ESPGHAN/Porto criteria. Multivariable and supervised machine learning were used to identify a diagnostic lipidomic signature and its performance was compared to clinically established biomarkers i.e., high sensitivity C-reactive protein (hsCRP) and faecal Calprotectin. Results The discovery cohort comprised of 58 children with IBD and 36 symptomatic controls without any discernible evidence of IBD. In the validation cohort, the corresponding numbers were 80 and 37. In a multivariable analysis, using regularized regression and discovery cohort data, a lipidomic signature comprising 30 lipid species was identified as influential in distinguishing IBD from symptomatic controls (Figure 1). Further, while examining both individual lipidomic species and lipidomic signatures in the discovery cohort, the highest diagnostic accuracy was observed for a short lipidomic signature comprising of only two lipid species, LacCer(d18:1/16:0) and PC(18:0p/22:6). When applied to the validation cohort, the short signature improved the diagnostic prediction of paediatric IBD (AUC=0.87 [95%CI 0.79-0.95]) compared to high-sensitivity C-reactive protein (hsCRP) alone (AUC=0.73 [95%CI 0.63-0.82]), P-value=0.0004. Adding hsCRP to the diagnostic model did not improve its performance (AUC=0.86) (Figure 2). Only 77 (66%) children from the validation cohort provided a stool sample, and among these children, the diagnostic capacity of the short lipidomic signature (AUC=0.88 95%CI 0.80-0.95) and faecal Calprotectin (AUC=0.93 95%CI 0.87-0.99) did not differ (P-value=0.22). Conclusion This study has identified and validated a diagnostic lipidomic signature of pediatric IBD that was superior to hsCRP and did not differ from faecal Calprotectin. This signature could be developed into an accessible and scalable blood test for paediatric IBD, particularly among patients unwilling to provide stool samples.