RESUMEN
OBJECTIVES: Crohn's 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 of a subsequent flare remains undefined. METHODS: We analyzed the long-term dynamics of microbial composition in a prospective observational cohort of patients with quiescent CD (45 cases, 217 samples) over 2 years or until clinical flare occurred, aiming to identify whether changes in the microbiome precede and predict clinical relapse. Machine learning was used to prioritize microbial and clinical factors that discriminate between relapsers and nonrelapsers in the quiescent phase. RESULTS: Patients with CD in clinical, biomarker, and mucosal remission showed significantly reduced microbial richness and increased dysbiosis index compared with healthy controls. Of the 45 patients with quiescent CD, 12 (27%) flared during follow-up. Samples in quiescent patients preceding flare showed significantly reduced abundance of Christensenellaceae and S24.7, and increased abundance of Gemellaceae compared with those in remission throughout. A composite flare index was associated with a subsequent flare. Notably, higher individualized microbial instability in the quiescent phase was associated with a higher risk of a subsequent flare (hazard ratio 11.32, 95% confidence interval 3-42, P = 0.0035) using two preflare samples. Importantly, machine learning prioritized the flare index and the intrapersonal instability over clinical factors to best discriminate between relapsers and nonrelapsers. DISCUSSION: Individualized microbial variations in quiescent CD significantly increase the risk of future exacerbation and may provide a model to guide personalized preemptive therapy intensification.