Computational modeling enables individual assessment of postprandial glucose and insulin responses after bariatric surgery.

BACKGROUND: Bariatric surgery enhances glucose metabolism, yet the detailed postprandial joint glucose and insulin responses, variability in individual outcomes, and differences in surgical approaches remain poorly understood. METHODS: We used hierarchical multi-output Gaussian process (HMOGP) regression to reveal clinically relevant patterns between persons undergoing two types of bariatric surgery by modeling the individual postprandial glucose and insulin responses and estimating the average response curves from individual data. 44 participants with obesity underwent either Roux-en-Y gastric bypass (RYGB; n = 24) or One-Anastomosis gastric bypass (OAGB; n = 20) surgery. The participants were followed up at the 6th and 12th months after the operation, during which they underwent an oral glucose tolerance test (OGTT) and a mixed meal test (MMT). RESULTS: A marked reduction in glycemia, an earlier glucose peak, and an increase and sharpening in the postprandial glucose and insulin responses are evident in both metabolic tests post-operation. MMT results in higher postprandial glucose and insulin peaks compared with OGTT. Higher glucose and insulin responses are observed after RYGB compared with OAGB, suggesting differences between the procedures that may influence the clinical practice. CONCLUSIONS: Computational modeling with HMOGP regression can thus be used to, in detail, predict the combined responses of patient cohorts to ingested glucose or a mixed meal and help in assessing individual metabolic improvement after weight loss. This can lead to new knowledge in personalized metabolic interventions.