OP19 Pre- and Post-diagnostic Metabolomic Biomarker Profiling of Over 700,000 Individuals in Three National Biobanks Enables Prediction of Inflammatory Bowel Disease Onset and Complications

Abstract Background We currently lack sensitive and specific prognostic biomarkers of inflammatory bowel disease (IBD) onset and severity. In three large, population-scale cohorts, we investigated metabolomic biomarkers for predicting disease onset and course. Methods We measured 250 metabolomic biomarkers using nuclear magnetic resonance (NMR) in ~500,000 individuals from the UK Biobank [1]. 2036 individuals had prevalent Crohn’s disease (CD) and 4029 had ulcerative colitis (UC) at recruitment. 1242 participants developed incident CD and 2283 incident UC during 10-year follow-up, allowing us to study the pre-diagnostic metabolic profile of IBD. We investigated associations between individual metabolomic biomarkers with prevalent and incident IBD using logistic and proportional hazards regressions. Among the incident IBD cohort we used LASSO with five-fold cross-validation on 50% of the cohort to build predictive models of future IBD onset. The resulting predictive model was validated in the Estonian Biobank and the Finnish THL Biobank (N ~218,000), confirming its robustness in multiple cohorts. Finally, in the UK Biobank we identified individuals with prevalent IBD who had IBD-related emergency admission, surgery or death and used these to develop a disease course score. Results GlycA, a biomarker of systemic inflammation, was the strongest association for both incident and prevalent UC and CD. We also identified associations between non-inflammatory metabolomic markers and IBD, such as albumin, branched-chain amino acids, cardiovascular lipid factors, fatty acids and creatinine. Notably, the perturbation in metabolites was markedly greater in CD versus UC, showing that CD is a more systemic illness (Fig. 1). Metabolomics scores were predictive of CD (AUC=0.69) and UC (0.62) onset up to 10 years prior to diagnosis, with performance for CD improving closer to diagnosis (2y follow-up AUC: CD 0.76, UC 0.62). Patients in the highest decile of incident metabolomic score were substantially more likely to develop the disease, with this observation being consistent in other two biobanks (Fig. 2). Metabolite profiles were predictive of IBD complications in prevalent patients (AUC 0.68 for CD; 0.62 for UC), outperforming both genetic and blood biomarker scores. Conclusion Leveraging cohort-wide profiling of population-scale biobanks, we present the largest metabolomics study and the largest pre-diagnostic study of IBD to date. Pre-diagnostic metabolomic scores are predictive of disease onset. These results warrant further investigation of multi-omic prediction of the onset, comorbidities and disease course of IBD. References * A.S, K.S. contributed equally. # T.J., J.C.B., C.L. contributed equally. [1] Nightingale Health Biobank Collaborative Group, Barrett, J. C., Esko, T., Fischer, K., Jostins-Dean, L., Jousilahti, P., ... & Estonian Biobank Research Team. (2023). Metabolomic and genomic prediction of common diseases in 477,706 participants in three national biobanks. medRxiv, 2023-06. Figure 1: Associations between clinically validated biomarkers and prevalent IBD Figure 2: UK Biobank-derived score consistency across the three biobanks