GLP-1 stands for glucagon-like peptide-1. It is a short signaling molecule that the gut releases after a meal, and it plays a central role in how the body manages blood sugar. Over the past decade, researchers have developed synthetic versions of this peptide and tested them extensively in people with type 2 diabetes. What has emerged from those trials goes well beyond glucose control: multiple studies have suggested these compounds may also affect the heart and blood vessels.
A comprehensive systematic review and network meta-analysis published in the journal Diabetes, Obesity and Metabolism brought together data from fifteen randomized controlled trials and nearly 97,200 participants. The goal was to produce updated, agent-level comparisons of cardiovascular outcomes across the GLP-1 therapy class, using a statistical approach called a hazard ratio-based network meta-analysis. This method lets researchers rank and compare individual compounds against one another, not just against placebo, even when those compounds have never been tested head-to-head in a single trial.
The findings offer the clearest picture yet of how this peptide class performs on outcomes that matter most to cardiologists and endocrinologists: death from any cause, death from cardiovascular causes, major adverse cardiovascular events (commonly called MACE), non-fatal heart attack, and non-fatal stroke.
How the analysis was built
The research team searched three major databases, PubMed, Cochrane Library, and Scopus, through December 2025. They included only randomized controlled trials that enrolled adults with type 2 diabetes and that reported time-to-event cardiovascular data. Time-to-event reporting is important because it accounts for differences in how long participants were followed, giving a more accurate picture of risk over time.
Fifteen trials met the inclusion criteria, collectively enrolling 97,173 participants. The analysis then ran in two complementary ways. First, standard pairwise meta-analysis compared each GLP-1-based therapy against placebo across all trials. Second, the frequentist random-effects network meta-analysis connected all agents into a single web of comparisons, allowing researchers to estimate how individual peptides ranked relative to each other on every cardiovascular endpoint.
Class-level cardiovascular signal
When the pairwise analyses grouped all GLP-1-based therapies together and compared them against placebo, the results were statistically significant across three key endpoints: all-cause mortality, cardiovascular mortality, and MACE. In plain terms, the pooled evidence suggested that people assigned to a GLP-1-based therapy were less likely to die from any cause, less likely to die from a cardiovascular event, and less likely to experience a major adverse cardiovascular event than those assigned to placebo.
The literature describes this as a class-level benefit, meaning the signal appears to be a shared property of GLP-1-based peptides rather than an effect tied to one particular agent. That distinction matters for researchers trying to understand the underlying mechanism. If every tested member of a compound family shows a similar pattern, the mechanism is more likely rooted in the shared biology of that class rather than in a unique chemical feature of one molecule.
Agent-level differences in the network analysis
The network meta-analysis painted a more nuanced picture when it came to comparing individual agents against each other. For mortality endpoints, several agents showed estimates that favored benefit, but most of those between-agent comparisons did not reach statistical significance. The researchers interpreted this as meaning the mortality data, while directionally consistent, did not have enough precision to confidently separate one agent from another on that specific endpoint.
For MACE, the picture was sharper. Three agents stood out with the most favorable comparative profiles in the network: efpeglenatide, albiglutide, and injectable semaglutide. These three showed the strongest signals for reducing major adverse cardiovascular events relative to other agents in the analysis. MACE is a composite endpoint that typically combines cardiovascular death, non-fatal heart attack, and non-fatal stroke, so a favorable MACE signal carries considerable clinical weight in trial design.
One agent, albiglutide, also showed a statistically meaningful reduction in non-fatal myocardial infarction compared to placebo. For non-fatal stroke, the estimates across all agents were described as imprecise, meaning the data did not converge clearly enough to draw firm conclusions about stroke specifically.
What network meta-analysis adds
A traditional meta-analysis can only compare agents that have been tested head-to-head in the same trial. That limitation is significant in cardiovascular research, where each large outcomes trial is expensive, takes years, and often tests only one compound against placebo. A network meta-analysis solves this by using placebo as a common comparator, creating an indirect chain of evidence that links agents even when they have never been randomized against each other directly.
The frequentist random-effects approach used here is considered a rigorous method for this type of indirect comparison. The random-effects component acknowledges that the underlying true effect may vary somewhat across trials due to differences in populations, follow-up duration, or background treatment. This tends to produce more conservative and more honest confidence intervals than a fixed-effects model, which assumes all trials are measuring exactly the same effect.
The authors were transparent about a key limitation of the agent-level comparisons: most between-agent differences in mortality were not statistically significant, which means the ranking of individual agents on mortality should be interpreted cautiously. The MACE findings carried more statistical weight, suggesting that particular endpoint may be where individual compound differences are most detectable.
Possible mechanisms behind the cardiovascular findings
GLP-1 receptors are found not only in pancreatic tissue but also in heart muscle, vascular endothelium, and the brain. Early data points at several pathways that could explain cardiovascular effects beyond glucose lowering. Researchers have studied reductions in inflammation, improvements in endothelial function, modest decreases in blood pressure, and favorable shifts in lipid profiles as potential contributors.
The fact that cardiovascular benefits appear in trials even when glucose control differences between groups are relatively small has led many researchers to hypothesize that direct effects on the heart and vessels, rather than indirect effects through blood sugar alone, may be driving part of the signal. This remains an active area of mechanistic research, and the current meta-analysis does not resolve it. What the analysis does confirm is that the cardiovascular signal persists across a broad population and multiple agents, which strengthens the case that the effect is real rather than a statistical artifact.
Scope and limitations of the findings
All fifteen included trials enrolled adults with type 2 diabetes, many of whom had established cardiovascular disease or were at high cardiovascular risk. This means the findings apply most directly to that population. Whether the same cardiovascular patterns would emerge in people without diabetes or in lower-risk populations is not something this analysis addresses.
The imprecision of the non-fatal stroke estimates is also worth noting. Stroke is a heterogeneous outcome, with different underlying mechanisms for different stroke subtypes, and the available trial data may not have been sufficiently powered to separate agent-level stroke effects from background noise. The researchers flagged this as an area where further evidence is needed.
Overall, the systematic review and network meta-analysis represent some of the most comprehensive evidence yet assembled on GLP-1 peptides and cardiovascular outcomes. With 97,173 participants across fifteen randomized trials, the dataset is large enough to support meaningful class-level conclusions while also beginning to distinguish between individual agents, particularly on the MACE endpoint.
