Personalizing bariatric metabolic surgery: Predictors of weight-loss success and risk of weight recurrence

Background: Bariatric metabolic surgery (Roux-en-Y gastric bypass [RYGB] and sleeve gastrectomy [SG]) effectively treats obesity and type 2 diabetes; however, weight loss varies, necessitating predictive factors. Methods: We analysed 12- and 24-month weight loss data from 811 patients (RYGB or SG). Factor Analysis of Mixed Data and neural network (NN) modelling identified distinct patient phenotypes and predicted weight-loss patterns. A comparative analysis evaluated weight loss and recurrence between the two procedures. Findings: RYGB showed significantly greater weight loss than SG at both 12 (30.3% vs. 25.4%; p < 0.001) and 24 months (26.3% vs. 21.4%; p < 0.001). SG revealed greater variability with bimodal weight loss distributions. Unsupervised clustering of SG patients highligheted three phenotypes: the highest responders were women with favourable metabolic profiles; the lowest responders were mostly men with insulin resistance and diabetes. A NN achieved an overall accuracy of 72.5% in predicting 12-month weight loss from baseline characteristics. In RYGB, clustering was less distinct, though baseline metabolic health influenced weight trajectories. A NN predicted weight recurrence versus sustained loss with 74% accuracy. Poor outcomes were associated with higher baseline glucose, insulin resistance, and dyslipidemia; younger age and absence of diabetes predicted better responses. RYGB was superior to SG, even for metabolic high-risk individuals. Interpretation: Baseline metabolic health predicts weight-loss outcomes and recurrence risk. RYGB offered greater and more consistent mid-term weight loss, especially benefiting metabolically high-risk patients. Procedure choice must be individualized accounting for specific risk profile and potential complications. These results advocate for a precision-medicine approach in bariatric procedure selection.