A Machine-Learning Assisted Genetic Risk Score Identifies Improved Weight Loss After Endoscopic Sleeve Gastroplasty.

Abstract

Background and aims: Obesity remains an epidemic associated with significant health consequences. Endoscopic sleeve gastroplasty (ESG) can lead to significant and sustained weight loss, with heterogeneous response. We previously reported a machine-learning (ML) assisted genetic risk score (GRS) for calories to satiation that predicts response to anti-obesity medications. Here, we evaluated the performance of novel GRSs for emotional hunger (EH), and calories to satiation (CTS, also high or low CTS_GRS) to predict weight loss after ESG.

Methods: Individuals treated with ESG at two separate endobariatric centers completed genetic testing using MyPhenome^® test (Phenomix Sciences, Menlo Park, CA). This test uses a machine-learning (ML) assisted GRS for high or low CTS_GRS and a GRS score combined with survey responses for EH. The primary outcome was total body weight loss (TBWL) after ESG at 12 and 24 months. Last observation carried forward (LOCF) analysis was used for missing values. Statistical analysis was performed using ANOVA analysis for multiple groups, and Tukey's HSD for pairwise analysis.

Results: Forty individuals completed testing. The low CTS_GRS group had a greater TBWL than both other groups at all observed time points (3 to 24 months). TBWL in the low CTS_GRS group was most significant at 12 months using LOCF analysis (21.4% vs. 13.7% in EH; p = 0.0153, and 14.9% in high CTS_GRS; p = 0.0305) and persisted to 24 months.

Conclusion: We report that a machine-learning assisted GRS is associated with significantly greater weight loss after ESG. Identifying individuals more likely to have superior weight loss response may improve selection of patient candidates for ESG.