Across 25 randomized trials, people on GLP-1 drugs binged less and ate with less loss of control. The same drugs also pushed them to hold back from food more deliberately, and the review that pooled the trials cannot say whether that second change is healthy self-control or the start of a different problem.
The systematic review and meta-analysis ↗ was published July 15 in EClinicalMedicine and funded by the Wellcome Trust and the Medical Research Foundation. It gathered every randomized controlled trial from 2005 to April 2026 that tested a GLP-1 drug against placebo or another treatment and measured eating behavior with a validated questionnaire. The protocol was registered on PROSPERO before the analysis ran. Twenty-five trials qualified, covering 8,069 people, about two-thirds of them women and most of them White. GLP-1 drugs are the class that includes semaglutide ↗, sold as Ozempic and Wegovy, and liraglutide ↗. They work by mimicking a gut hormone that curbs appetite, and they are now the most widely used weight-loss medicines in the world.
On the behaviors you would expect them to quiet, the drugs delivered. The authors measured each shift in standardized units called Hedges' g, where roughly 0.2 counts as a small change, 0.5 as moderate, and 0.8 as large. Binge eating fell by a small-to-moderate amount (g = -0.23, from nine trials in 2,106 people). Eating disinhibition, the tendency to keep eating once you start, dropped more clearly (g = -0.54). Emotional eating eased too (g = -0.30). Loss-of-control eating, measured on a separate raw scale, showed one of the largest drops. Every one of these moved in the direction of less disordered eating. The effects were bigger in the trials that enrolled people actually diagnosed with binge eating disorder, the formal eating-disorder diagnosis in the DSM-5.
Then the number that complicates the story. Cognitive restraint, the deliberate effort to limit what you eat, went up (g = +0.31). People on the drugs reported holding back from food more on purpose. On its face that reads like self-control. In eating-disorder research it reads less cleanly, because rigid restraint is also a core feature of restrictive eating and of the binge-then-restrict cycle. The authors did not smooth this over. In their words, "whether restraint is pathological is unclear."
The mechanism makes the split plausible. A GLP-1 drug tamps down appetite and the reward pull of food, so the collapse in binging and loss of control tracks what the drug is built to do. The rise in deliberate restraint is harder to place. It could be a person eating normally with the volume turned down, or it could be the drug amplifying a restrictive pattern in someone already prone to one. A pooled questionnaire average across 8,069 people cannot tell those two apart, and the review was not designed to.
The caveats are large and the authors state them plainly. Every trial in the analysis was rated as high risk of bias or as having some concerns. The results were heterogeneous, meaning the trials disagreed with each other more than chance would explain. Publication bias could not be formally checked because there were too few studies to test it. And most of these trials were built to study weight or blood sugar, not eating disorders, so binge eating was usually a secondary measure rather than the thing the trial set out to move.
That is the honest shape of the finding. GLP-1 drugs look like they reduce binge eating, and the biggest reductions showed up exactly where you would want them, in people with diagnosed binge eating disorder. But the same drugs nudged deliberate restraint upward, and no one yet knows whether that is a benefit or a warning. The authors call for adequately powered trials that enroll people with diagnosed binge eating disorder and follow them long enough to see whether the restraint stays healthy.
The drugs act on the GLP-1 receptor ↗, the same appetite switch behind their weight-loss effect. An earlier measured-meal study found GLP-1 users ate about 270 fewer calories ↗ at a single sitting. How much you eat and how you eat are different questions, and this review is a first pooled attempt at the second one. It reads as a signal to test, not a result to prescribe on.