September 1, 2026 · Science · Nutrition
What the Review Reported
Annals of Internal Medicine published an updated systematic review of GLP-1 receptor agonists and co-agonists for weight loss in adults with overweight or obesity and without diabetes on September 1, 2026. It covers 38 randomized controlled trials and 25,816 participants, adding 14 new trials and roughly 11,000 participants to the authors' earlier review. The search ran across MEDLINE, Embase and the Cochrane Central Register of Controlled Trials from October 5, 2024 through March 25, 2026, and included randomized trials with a treatment duration of at least 16 weeks.
Among commercially available therapies, placebo-subtracted weight loss reached up to -5.8% for liraglutide, with a 95% confidence interval of -8.0% to -3.6%; -14.8% for subcutaneous semaglutide, interval -16.2% to -13.4%; -14.3% for oral semaglutide, interval -17.2% to -11.4%; -12.4% for orforglipron, interval -15.1% to -9.7%; and -19.0% for tirzepatide, interval -21.6% to -16.4%.
Emerging multiagonists produced numerically greater reductions: -23.9% with amycretin, interval -29.3% to -18.5%, and -22.1% with retatrutide, interval -24.9% to -19.3%. Head-to-head data showed greater weight loss with semaglutide and JNJ-64565111 than with liraglutide, and greater weight loss with tirzepatide and cagrilintide-semaglutide than with semaglutide. The review reports no funding and is registered with PROSPERO as CRD42024505558.
The short version
Why "Up To" Is Doing Heavy Lifting
A systematic review and a meta-analysis are different instruments, and the distinction decides how much weight a number carries. A meta-analysis pools trials into one estimate with a combined confidence interval, which is only valid when the trials are similar enough to be treated as measuring the same thing. This review's authors concluded they were not, and said so: heterogeneity precluded quantitative synthesis.
What that leaves is a catalogue of the best result each agent achieved somewhere in the evidence base. The confidence intervals published alongside each figure are the intervals from those individual results, not pooled intervals across all trials of that drug. A reader comparing -19.0% for tirzepatide against -14.8% for subcutaneous semaglutide is comparing two peaks that may come from trials of different length, in different populations, at different doses. The review does report direct head-to-head comparisons separately, and those are the sounder basis for ranking, because within a single trial the two groups are comparable by design.
None of this makes the figures wrong. It makes them ceilings observed under trial conditions rather than expectations for an individual. Trial participants receive structured support, monitoring and free medication, and they are selected by eligibility criteria that exclude many people who would seek treatment in practice.
What This Means for a Longevity Platform Subscriber
Several of the platforms benchmarked here fold GLP-1 prescribing into a subscription that also sells lab panels and a dashboard. That bundling makes an efficacy figure from a trial easy to read as a forecast, and this review is a useful corrective in three specific ways.
The first is that the number quoted in marketing is usually a ceiling, not a median, and usually not labelled as placebo-subtracted. The second is that the most impressive figures in the current literature attach to agents that are not generally available, so a comparison between a marketed program and the biggest number in the news is not comparing like with like. The third is that this review measured weight and adverse events. It did not measure cardiovascular events, diabetes incidence or mortality, so it does not speak to whether a given amount of weight loss changes those outcomes.
Our guide to insulin resistance and weight loss covers the metabolic markers these programs typically track, and our metabolic reversal hub compares the structured programs that approach the same problem without medication. Whether a specific drug is appropriate for a specific person depends on their history, other conditions and current medications, and that assessment belongs with a prescribing clinician.
What Would Change This Read
Consistent safety reporting would change it most. The authors identify inconsistent reporting of safety outcomes as a limitation of the evidence base rather than of their own methods, and a reassurance about serious events drawn from trials that recorded them differently is weaker than the same reassurance from trials with a common standard.
Trials with clinical endpoints would change what the percentages mean. Weight is the measured outcome throughout this review. Evidence on whether these agents reduce cardiovascular events or mortality comes from separate trials designed around those endpoints, and that evidence is what would convert a weight figure into a health claim.
Enough homogeneous trials of a single agent would allow pooling. If future trials converge on comparable designs and durations, a quantitative synthesis becomes possible and the resulting pooled estimate would be a firmer number than any figure in this review.
Related Coverage
A federal obesity guideline reviewed the body fat measurements and endorsed none of them, which is the measurement problem underneath any percent-body-weight endpoint. A separate briefing covers what the evidence does and does not establish about GLP-1 drugs and psychiatric risk, and a third looks at a genetic study linking GLP-1 receptor activity to male pattern hair loss in men who were never recorded as taking the drugs. The nutrition hub collects the guides these questions start from.
Sources
- Moiz A, Filion KB, Samuels AE, Tsoukas MA, Yu OHY, Peters TM, Eisenberg MJ, "Efficacy and Safety of Glucagon-like Peptide-1 Receptor Agonists and Co-agonists for Weight Loss Among Adults Without Diabetes: An Updated Systematic Review," Annals of Internal Medicine, published September 1, 2026, DOI 10.7326/annals-25-05519. Abstract retrieved via Europe PMC (accessed September 1, 2026). Source of every figure on this page: the 38 trials and 25,816 participants, the 14 new trials and roughly 11,000 added participants, the database and date range of the search, the minimum 16-week treatment duration, all placebo-subtracted weight loss percentages and their confidence intervals for liraglutide, subcutaneous and oral semaglutide, orforglipron, tirzepatide, amycretin and retatrutide, the head-to-head comparisons, the 76.0% and 40.1% gastrointestinal adverse event rates, the 10.7% and 3.4% discontinuation rates, the 6.5% and 5.2% serious adverse event rates, the 0.1% and 0.0% death rates, the statements that heterogeneity precluded quantitative synthesis and that safety outcomes were inconsistently reported, the absence of funding, and the PROSPERO registration.
- MedicalXpress, "GLP-1 RAs produce significant weight loss in adults without diabetes, with newer treatments showing greater results," September 1, 2026, medicalxpress.com (accessed September 1, 2026). Detection source for this briefing. No numeric figure on this page is taken from it.
- The publisher's full text at acpjournals.org returned an access error at the time of writing. No figure on this page comes from the full text, and the classification of individual agents as commercially available or emerging follows the abstract's own wording.
Frequently Asked Questions
What does placebo-subtracted weight loss mean?
It is the difference between the treatment group and the placebo group, not the total the treatment group lost. People assigned to placebo in these trials also lose weight, because they receive diet and activity support and because enrolling in a trial changes behavior. A placebo-subtracted figure of -19.0% means the treated group lost 19.0 percentage points more of their body weight than the placebo group did. It is the part attributable to the drug.
Is this review a meta-analysis?
No, and the authors are explicit about why. The abstract states that heterogeneity precluded quantitative synthesis. That means the trials differed too much in design, dosing and population to be pooled into a single combined estimate. The figures reported are the highest values observed for each agent across individual trials, described as reaching up to those levels, rather than a pooled average across all of them.
Which of these drugs can someone actually get?
The review evaluated 17 agents. Liraglutide, semaglutide in subcutaneous and oral forms, and tirzepatide are established commercially available options. Orforglipron, amycretin, retatrutide and cagrilintide-semaglutide appear in the review as investigational agents at various stages. The largest numbers in the review belong to agents in that second group. Availability, indication and coverage differ by country and change over time, so the current status of any specific drug is a question for a prescribing clinician or the relevant regulator.
What were the side effects?
Gastrointestinal adverse events were common, occurring in 76.0% of people on a GLP-1 receptor agonist against 40.1% on placebo. Discontinuation because of adverse events was 10.7% against 3.4%, and the review notes it was numerically higher with some oral agents. Serious adverse events were 6.5% against 5.2% and deaths were 0.1% against 0.0%. The authors report no new safety signals, and also note that safety outcomes were inconsistently reported across trials.
Do these trials tell you what happens after you stop?
Not from what this review reports. It included randomized trials with a treatment duration of at least 16 weeks and measured weight change during treatment. Nothing in the reported results describes weight trajectory after discontinuation. That is a separate question with its own evidence base, and it is one of the more consequential things a person weighing a long-term medication would want to know.
Does more weight loss mean better health outcomes?
This review does not answer that. It measured weight, and the outcomes it reports are weight change and adverse events, not cardiovascular events, diabetes incidence or mortality. Percent body weight lost is an intermediate measure. Trials designed around clinical outcomes exist for some of these agents and are a different evidence base from the one summarized here.