Real-World Comparative Weight Loss of GLP-1 Receptor Agonists in the All of Us Research Program: A Retrospective...

To assess the real-world associations of five glucagon-like peptide-1 receptor agonists (GLP-1RAs) with weight loss in a diverse US population of adults with obesity, with and without type 2 diabetes, and whether the relative ordering of weight loss observed in trials is mirrored in routine practice.

We conducted a retrospective, new-user, active-comparator cohort study using the National Institutes of Health All of Us Research Program. We included 14,046 adults with obesity (body mass index [BMI] ≥30 kg/m²) initiating exenatide, liraglutide, dulaglutide, semaglutide, or tirzepatide. Groups were compared after multivariable adjustment for baseline confounders. Participants were followed from the index date until the outcome, death, or loss to follow-up; percent weight and BMI change was assessed at 12 months (±45 days). Cox proportional hazards models estimated the relative likelihood over time of achieving weight-loss thresholds (≥5%, ≥10%, ≥15%) and ≥1 BMI-category reduction; linear regression assessed the magnitude of weight change, with liraglutide as reference.

Tirzepatide showed the highest likelihood of ≥15% weight loss (adjusted hazard ratio [aHR] 5.57; 95% CI, 3.66-8.49) and the greatest mean weight loss at 12 months (-13.0%), followed by semaglutide (aHR 1.77; 95% CI, 1.56-2.01; -5.9%); both were significantly greater than the older agents in adjusted analyses. Exenatide, dulaglutide, and liraglutide showed comparable, modest weight loss (approximately -4.0%). This ordering was consistent regardless of type 2 diabetes status, although weight loss was greater without diabetes for semaglutide but similar for tirzepatide.

In this large, diverse real-world cohort, tirzepatide and semaglutide were associated with significantly greater weight loss than older agents. These findings are consistent with the relative ordering reported in trials, though the observational design cannot confirm it. Residual confounding cannot be excluded, and the small tirzepatide sample (n=386) and limited follow-up warrant cautious interpretation.