GLP-1 receptor agonists are a class of peptides that mimic a hormone the gut releases after eating. That hormone, glucagon-like peptide-1, signals the brain to reduce appetite, slows how quickly the stomach empties, and nudges the pancreas to manage blood sugar. Over the past decade these peptides have become central to obesity research, but a key question kept going unanswered: do they work equally well in people who have type 2 diabetes versus those who do not?
A subgroup analysis published in Surgical Endoscopy set out to answer that question by pooling data from 45 clinical trials and observational studies published between 2010 and 2025. All together the dataset covered 56,580 patients followed for at least one year. The researchers used mixed-effects meta-regression, a statistical method that lets scientists untangle which factors independently predict an outcome, to identify what actually drives weight loss when someone uses a GLP-1 peptide.
The short answer is that diabetes status matters enormously, but adherence and the specific agent used also play measurable roles. The findings push toward what the authors call a phenotype-based approach, meaning that who the patient is biologically shapes what kind of result the therapy can realistically achieve.
The size of the gap between diabetic and non-diabetic outcomes
The headline number from the meta-analysis is stark. In what the researchers called the mixed cohort, meaning all patients regardless of whether they finished their full treatment course, non-diabetic participants lost an average of 15.7 percent of their baseline body weight over 12 months. Diabetic participants in the same cohort lost 5.1 percent over the same period.
That is roughly a three-fold difference from the same class of therapy. When the researchers ran their meta-regression to control for other variables, diabetes status showed up as a strong, independent negative predictor of weight loss. The statistical coefficient was negative 1.77 with a p-value below 0.001, which means the association is very unlikely to be a chance finding.
The literature suggests several reasons the gap might exist. People with type 2 diabetes often have higher baseline levels of circulating insulin and altered gut hormone signaling. The metabolic environment in a diabetic body may blunt the appetite-suppressing signal that GLP-1 receptor agonists produce, or it may reduce how much the body draws on fat stores in response to reduced calorie intake.
How adherence shapes the numbers
The study evaluated outcomes across three different population definitions. The mixed cohort included everyone who started treatment. The intention-to-treat group tracked all enrolled participants regardless of whether they completed the protocol. The treatment-adherent group included only people who stayed on the medication as prescribed.
Results were similar between the mixed and intention-to-treat groups, though the ITT numbers were modestly lower, which is typical when dropouts are folded into the average. The adherent group, however, showed the largest reductions across the board. This pattern is not surprising to researchers who study long-term medication use, but it reinforces a practical point: the peptide can only do what the biology allows when it is actually reaching the receptor consistently over time.
Adherence is a known challenge with injectable therapies in general. Side effects, cost, access, and motivation all influence whether someone stays on a regimen for the multi-year periods that produce the largest effects. The meta-analysis did not drill into the reasons for dropout, but the data confirm that adherent patients represent a meaningfully different outcome group.
Agent-specific differences in the data
GLP-1 receptor agonists are not a single compound. They are a family of peptides that share a mechanism but differ in molecular structure, half-life, and dosing schedule. The meta-analysis was able to compare two agents that appeared frequently enough across studies to support statistical comparison: semaglutide and liraglutide, both referred to by their generic names here.
After adjusting for other variables, the analysis found that semaglutide produced significantly greater weight reductions than liraglutide. The regression coefficient was positive 1.67, again with a p-value below 0.001. Researchers have proposed that the longer half-life of semaglutide, which allows once-weekly dosing compared to liraglutide's daily injections, may improve adherence and produce more sustained receptor activation. The structural modifications that give semaglutide its longer action may also enhance binding affinity.
The study does not address newer dual or triple agonist peptides, which target additional receptors alongside GLP-1. Those compounds are the subject of separate ongoing research. What this meta-analysis captures is the within-class variation that existed across a 15-year window of published data.
The role of baseline BMI
One finding that may seem counterintuitive is that higher baseline body mass index was inversely associated with the percentage of weight lost. The regression coefficient for baseline BMI was positive 0.46, which in this model's framing means that as BMI goes up, the percent weight change becomes less favorable.
In practical terms this suggests that people with more weight to lose do not automatically lose a larger share of it. Absolute weight loss in kilograms may still be greater for higher-BMI individuals, but the percentage reduction appears smaller. Early data from other metabolic research points in a similar direction, suggesting that more severe obesity involves hormonal and inflammatory changes that can reduce responsiveness to appetite-based interventions.
This finding adds another dimension to the personalized prescribing argument the authors make. Diabetes status, adherence potential, BMI, and agent selection all interact. A one-size prediction for any individual based on population averages will miss meaningful variation.
What the research design can and cannot show
A meta-analysis pools results from many different studies, which increases statistical power but also introduces heterogeneity. The 45 studies included in this analysis used different protocols, measurement intervals, and patient populations. The researchers used mixed-effects meta-regression specifically to account for that variation, but some residual noise is inevitable.
The study also cannot establish causation in the way a single head-to-head randomized controlled trial would. The finding that non-diabetic individuals lose more weight is consistent across the included studies, but the reasons remain partly inferential. It is possible that unmeasured differences between diabetic and non-diabetic participants, such as baseline diet, concurrent medications, or activity levels, contribute to the gap alongside the biological factors researchers have proposed.
The publication in Surgical Endoscopy is notable because that journal focuses on minimally invasive procedures and metabolic surgery. GLP-1 research increasingly intersects with bariatric surgery planning, where understanding who responds to peptide therapy matters for decisions about surgical candidacy and sequencing of treatments.
Relevance to peptide research more broadly
GLP-1 receptor agonists sit within a larger landscape of peptide research that examines how signaling molecules can influence metabolic processes. Researchers studying other peptide classes, including growth hormone secretagogues, mitochondrial peptides, and anti-inflammatory short-chain peptides, face similar questions about responder heterogeneity. Who responds, how much, and why are central to the science regardless of the specific compound.
This meta-analysis contributes to that broader understanding by demonstrating that even within a single well-studied peptide class, outcomes vary by a factor of two to four depending on patient phenotype. The literature suggests this kind of variability is not an exception but a rule in peptide pharmacology, and it underlines why researchers continue to study subgroup differences rather than relying only on population averages.
For anyone following peptide science, the takeaway from this study is methodological as much as it is numerical. Large-scale subgroup analyses that stratify by phenotype, track adherence separately, and compare agents within a class produce a richer and more actionable picture than single trials alone. That approach is increasingly standard in the field, and this paper adds a substantial dataset to the foundation.
