Weight-loss drugs eased binge eating in trials. But the evidence is weak.

Meta-analysis of randomised trials. 8,069 people. 25 RCTs (24 in the meta-analysis), published 2005 to April 2026; trial lengths 6 to 104 weeks, mostly short-term.

Many trials, added up. In each one, chance decided who did what.

Who: Adults with overweight or obesity (median of trial mean BMI 35.8; median of mean age 47.3), ~two-thirds female (67.5%), mostly White (median 84.7%), 12 high/upper-middle-income countries; some trials in type 1/2 diabetes, PCOS, bipolar or psychotic disorders; only 3 trials recruited people with binge eating disorder or binge eaters

• Binge eating scorelowersmall-to-moderate
• Loss of control-18 ptson a 0–100 scale
• Holding back on foodhigherhealthy? unclear

More to know, from our read of the paper:

• Low-quality evidence: all 25 trials rated high risk of bias (12) or some concerns (13); substantial heterogeneity. Authors: 'current evidence is preliminary'
• Only 3 trials recruited people with diagnosed binge eating disorder / binge eaters; all participants had overweight or obesity; mostly White, middle-aged. No evidence for binge eating without obesity
• Mostly short-term; no evidence on long-term effects or rebound after stopping
• Standardised effect sizes (Hedges' g) are not percentages; don't translate g into '% fewer binges'. The binge eating severity effect (g -0.23) is called 'moderate' in the abstract but 'small-to-moderate' in Results; use 'small-to-moderate'
• Loss-of-control MD -18.38 is on a harmonised 0-100 scale (CoEQ item 19, post-intervention values), from only 5 trials and 362 people

This systematic review pooled 25 randomised trials with 8,069 people with overweight or obesity, mostly not diagnosed with binge eating disorder. On GLP-1 drugs, people reported a small-to-moderate drop in binge eating scores and about 18 points less loss of control eating on a 0-100 scale (5 trials) than comparison groups. They also reported holding back on food more; whether that is healthy is unclear. Every trial had some risk of bias, most were short, and most were run or funded by drug makers, so the authors call the evidence preliminary.

We read the abstract and the full text of this paper and checked every number on this page against it (17 facts, checked on 2026-10-29). The study design is named above. Cohort and other observational studies are described as linked, not proven. Found an error? Email theis@onestudyaday.com.