(P24-013-20) Gut Microbiota Are Predictive of Coach-Directed Behavioral Weight Loss Success over 6 Months in Randomized Trial

Objectives: Several studies have demonstrated human gut microbiota are involved in energy balance and obesity, but little is known on how it may influence weight loss efforts. The objective of this study is to determine whether features of the gut microbiota predict weight loss at 6 months in participants enrolled in a trial of coach-directed behavioral weight loss.

Methods: We used baseline clinical and microbiome data to predict 6-month weight loss in participants (n=36) randomized to a coach-directed weight loss program that included diet and lifestyle modification. We sequenced the microbial 16S rRNA V4 region of stool samples collected at baseline and assigned reads into amplicon sequence variants (ASVs). Successful weight loss was defined as >5% weight loss from baseline to 6 months. We first used Variable Selection Using Random Forests (VSURF) on 571 ASVs that were present in >10% of samples, measures of alpha diversity (observed species and Shannon index), along with sex, age, BMI, and race to identify the strongest predictors of weight loss. The features selected by VSURF were then included with clinical characteristics in a final leave-one-out cross-validated random forest model to predict weight loss. We used log Poisson regression models with robust variance estimates to assess the direction and significance (FDR p < 0.05) of each predictor’s association with weight loss.

Results: After 6 months of coach-directed behavioral weight loss, 12 (33%) participants met the goal of losing >5% of their baseline weight. VSURF selected 30 bacterial ASVs for model interpretation. Sex, age, BMI, and race were not selected by VSURF, and when these participant characteristics were forced into the model, they were ranked last. Our final cross-validated random forest model had a kappa of 0.73 (out-of-bag error = 13.9%) for classifying 5% weight loss. Among the top predictors were ASVs from the Ruminococcaceae family (lower with weight loss; FDR p< 0.05), Eubacterium coprostanoligenes (higher with weight loss; FDR p< 0.05).

Conclusions: Gut microbiota sequencing improved prediction of successful weight loss. The top predictors of weight loss included bacteria that produce butyrate, secondary bile acids, and succinate, adding support to the hypothesis that gut microbiome composition may influence weight loss efforts.

Funding Sources: NHLBI