proportional hazard models

Summary Background Metabolic–bariatric surgery delivers substantial weight loss and can induce remission or improvement of obesity-related risks and complications. However, more robust estimates of its effect on long-term mortality and life expectancy—especially stratified by pre-existing diabetes status—are needed to guide policy and facilitate patient counselling. We compared long-term survival outcomes of severely obese patients who received metabolic–bariatric surgery versus usual care. Methods We did a prespecified one-stage meta-analysis using patient-level survival data reconstructed from prospective controlled trials and high-quality matched cohort studies. We searched PubMed, Scopus, and MEDLINE (via Ovid) for randomised trials, prospective controlled studies, and matched cohort studies comparing all-cause mortality after metabolic–bariatric surgery versus non-surgical management of obesity published between inception and Feb 3, 2021. We also searched grey literature by reviewing bibliographies of included studies as well as review articles. Shared-frailty (ie, random-effects) and stratified Cox models were fitted to compare all-cause mortality of adults with obesity who underwent metabolic–bariatric surgery compared with matched controls who received usual care, taking into account clustering of participants at the study level. We also computed numbers needed to treat, and extrapolated life expectancy using Gompertz proportional-hazards modelling. The study protocol is prospectively registered on PROSPERO, number CRD42020218472.