Background/Purpose: Within the past decade, there have been several major discoveries in cross-sectional gut microbiome studies suggesting that dysbiosis of the gut microbiota is a key hallmark in Rheumatoid arthritis (RA). However, the association of gut microbiome with improvement in disease activity in RA patients remains unknown. In this study, we aimed to investigate the association of minimum clinically important improvement (MCII) in RA disease activity with gut microbiome of RA patients. In addition, we generated a machine-learning model, incorporating gut microbiome and clinical data, that could predict the course of RA irrespective of the treatment strategy.

Methods: Illumina based DNA shotgun sequencing was performed on 72 stool samples, which were collected at two time-points, from 36 RA patients. Disease activity and gut microbiomes were assessed in order to investigate the association of gut microbiome with MCII (i.e., CDAI change of at least 1 for low (CDAI less than 10); of 6 for moderate (CDAI 10–22); and of at least 12 for high (CDAI greater than 22) disease activity) in RA disease activity. Machine-learning models based on XGBoost were also generated based on gut microbiome composition and clinical data at baseline to predict the future disease activity in RA patients.

Results: We found that distributions of gut microbiome profiles were significantly different in patients of MCII⁺ (i.e., patients who show improvement in disease activity) and MCII^– groups (i.e., patients who did not show improvement). At baseline, Fisher’s alpha diversity and species richness were significantly higher (Fig.1b), and five microbial taxa, including Negativicutes, Selenomonadales, Prevotellaceae, Coprococcus, and Ruminococcus sp., were significantly more abundant among patients of MCII⁺ group than those of MCII^– group (Fig.1b). Biochemical pathway-level analysis of baseline stool metagenomes suggested that eight MetaCyc pathways, including ornithine biosynthesis, arginine biosynthesis, and rhamnose degradation were significantly decreased in RA patients of MCII^– group (Fig.1c). For the machine-learning model, the Pearson correlation coefficient between predicted and actual disease activity score (CDAI) at follow-up visits was 0.77 (P = 2.9 × 10^-12), which shows that our machine-learning model, incorporating gut microbiome and clinical data, is very effective in forecasting the disease activity in RA patients.

Conclusion: Our findings confirm the association of gut microbiome with MCII in disease activity in RA patients, and highlights the importance of gut microbiome in predicting the course of RA irrespective of the treatment strategy.

« Back to ACR Convergence 2020

ACR Meeting Abstracts - https://acrabstracts.org/abstract/minimum-clinically-important-improvement-in-patients-with-rheumatoid-arthritis-associates-with-gut-microbiome/