OP32 Stool microbiome communities predict remission in treatment-naïve Pediatric Crohn's Disease patients. (27th May 2021)

• Record Type: Journal Article
• Authors: Verburgt, C; Dunn, K; Bielawski, J; Otley, A; Heyman, M; Sunseri, W; Shouval, D; Levine, A; de Meij, T; Hyams, J; Denson, L; Kugathasan, S; Benninga, M; de Jonge, W; Van Limbergen, J
• Is Part Of: Journal of Crohn's and colitis. Volume 15(2021)Supplement 1
• Page Start: S030
• Page End: S030
• Publication Date: 2021-05-27
• Subjects: Inflammatory bowel diseases -- Periodicals; 616.344005
• DOI: 10.1093/ecco-jcc/jjab075.031
• Languages: English
• ISSNs: 1873-9946

Abstract

Abstract: Background: Early relapse in paediatric Crohn's Disease (CD) is associated with severe disease course that heavily impairs quality of life. Changes in gut microbiome composition have been linked to active CD and disease course. This has led to development of microbiome-based prediction models for diagnosis and response to treatment. Our aim was to identify community-level microbiome signatures of treatment-naïve children with mild-to-moderate CD who did not require anti-TNF or surgery at diagnosis, with the goal of predicting need for re-induction or treatment escalation within the first year after diagnosis. Methods: We selected de novo, treatment-naïve paediatric CD patients from the RISK cohort(Gevers 2014). Taxonomic labels were assigned to the 16 s rRNA amplicon data using QIIME and closed OTU-picking. A hierarchical Bayesian model for microbial community structure was used to learn how baseline gut microbiomes differed according to treatment outcome. Model predictions were assessed using a leave-one-out analysis. We compared 16 S rRNA sequences of CD patients with non-IBD controls(Gevers 2014) and healthy siblings of CD patients(Turpin 2016). Results: Metadata and 16 S rRNA amplicon data were available from 197 stool samples of de novo paediatric CD patients from the RISK cohort. We selected 44 out of 197 samples of patients that were treatment-naïve. Prior to treatment, PCDAI scores were similar between patients reaching remission and those that did not at 6 months. Bayesian analysis characterized 4 assemblages that accounted for 93% of the posterior probability distribution. The Bayesian model on pre-treatment stool microbiomes was able to predict 6-month outcome of patients that maintained remission and those that did not from the pre-treatment microbiome in 81% and 75% of samples (AUC=0.79). When comparing CD samples to 28 non-IBD controls (many with GI symptoms but negative for IBD during endoscopy, e.g. Irritable Bowel Syndrome), 6 assemblages were characterized with 44% of distributions shared between groups (AUC=0.61). In contrast, in CD samples compared to 728 healthy sibling samples (with increased genetic susceptibility), shared distribution within 4 characterized assemblages was less than 1% (AUC=1). Conclusion: A Bayesian approach predicted clinical course in treatment-naïve children with CD in the first year after diagnosis with high accuracy, when ensuring only treatment-naïve faecal samples in the analysis. This classification level is comparable to previous findings using mucosal samples. Further study is needed to validate these pre-treatment microbiome signatures of newly diagnosed paediatric CD patients to allow identification of patients with mild-to-moderate disease who are most likely to require treatment escalation.