DOP36 Gut microbial variations in patients with quiescent Crohn’s disease predict subsequent disease flare Article Swipe
YOU?
·
· 2019
· Open Access
·
· DOI: https://doi.org/10.1093/ecco-jcc/jjy222.070
· OA: W2912758266
Crohn disease (CD) is a chronic relapsing-remitting gut inflammatory disorder with a heterogeneous unpredictable course. Dysbiosis occurs in CD, however, whether microbial dynamics in quiescent CD are instrumental in increasing the risk for a subsequent flares remains undefined. We aimed to identify whether changes in the microbiome precede and predict clinical relapse. We analysed the long-term dynamics of microbial composition in a prospective observational cohort of patients with quiescent CD (45 cases, 217 samples) undergoing rigorous clinical, biochemical, and mucosal follow-up over 2 years or until a clinical flare occurred. Clinical assessment, faecal calprotectin, faecal microbial characterisation, and CRP were measured routinely every 3 months. Patient underwent video capsule endoscopy (VCE) every 6 month. 16S rRNA gene V4 variable region using Illumina adapted universal primers 515F/806R was conducted to characterise microbial variation. Machine learning was employed to prioritise microbial and clinical factors that discriminate between relapsers and non-relapsers in the quiescent phase. Of the 45 patients with quiescent CD, 12 (27%) flared during follow-up. Samples in quiescent patients preceding flare showed significant reduced abundance of Christensenellaceae and S24.7, and increased abundance of Gemellceae in comparison to those patients in remission throughout, and a composite ‘flare index’ summarising those microbial taxa, was significantly higher in patient who subsequently flared vs. those who remained in remission (p = 9.2e−11). Notably, higher individualised microbial instability in the quiescent phase was associated with higher risk of subsequent flares (hazard ratio 11.32, 95% CI 3–42, p = 0.0035) using two pre-flare samples. When prioritising clinical, demographic, and microbial factors in a supervised learning Random Forest algorithm to predict a subsequent flare, the top contributing factors were the ‘flare index’ and the intra-personal microbial instability. Those were followed by BMI, capsule endoscopy Lewis score, and microbial richness. Importantly, CRP, treatment exposure, and calprotectin were not within the top 5 contributing factors in the prediction model Individualised microbial variations in quiescent CD can precede and predict future exacerbation. These results may imply that microbiome changes during the quiescent phase may be the cause or an associated reporter of other factors upstream of the inflammatory process pre-flare that subsequently lead to a disease flare.
