Pooled data tie bariatric surgery to a 34% lower risk of major heart events and GLP-1 drugs to 15%

Doctors already have good evidence that weight loss surgery lowers the risk of heart trouble, and good evidence that GLP-1 medicines do too. What has never been settled is which one does more. A new analysis published on September 5, 2026 in the journal Endocrinology, Diabetes and Metabolism pooled 43 studies covering 932,380 patients to try to answer that.

The main result: bariatric surgery was linked to a 34 percent lower risk of a major heart event (hazard ratio 0.66, 95 percent confidence interval 0.60 to 0.74). GLP-1 receptor agonists were linked to a 15 percent lower risk of the same thing (hazard ratio 0.85, 0.77 to 0.94). Both were compared against people who got neither treatment.

A hazard ratio below 1.0 means lower risk. The confidence interval is the range the true number is likely to sit in. When that range does not cross 1.0, researchers call the result statistically significant, which means it is unlikely to be a fluke of chance. It does not mean the effect is large or that the treatment caused it.

"Major heart event" here is a standard research bundle called MACE. In this analysis it counted death from any cause, plus the first heart attack, stroke, hospital admission for heart failure, or procedure to reopen a blocked artery.

Surgery showed lower risk on all three heart outcomes, the drugs on only one

The researchers also looked at heart failure and heart attack on their own.

For heart failure, surgery was linked to a 55 percent lower risk (hazard ratio 0.45, 0.38 to 0.53). GLP-1 medicines came in at 0.87, but the range ran from 0.73 to 1.05. Because that range crosses 1.0, the result was not statistically significant.

For heart attack, surgery was linked to a 47 percent lower risk (hazard ratio 0.53, 0.45 to 0.63). GLP-1 medicines came in at 0.91, range 0.77 to 1.08, again not statistically significant.

That last part depends on which statistical model is used, and the paper is honest about it. Under a second, less conservative model, the GLP-1 numbers for heart failure (0.89, 0.83 to 0.96) and heart attack (0.90, 0.84 to 0.97) did reach significance. So the fair reading is not that GLP-1 drugs do nothing for those two outcomes. It is that the evidence for them is weaker and less stable than the evidence for surgery.

The surgery evidence and the drug evidence do not come from the same kind of study

This is the most important thing to understand about the comparison, and the authors flag it themselves.

Every one of the 28 studies comparing surgery to no surgery was an observational cohort. That means researchers watched what happened to people who had surgery and people who did not, rather than assigning anyone. People who reach the operating room have usually been screened for fitness, cleared by several specialists, and have shown they can follow a program. People who do not have surgery may be sicker, poorer, or less able to access care in the first place.

Some of the difference in heart outcomes may therefore reflect who gets surgery rather than what surgery does. Researchers call this residual confounding, and statistical adjustment reduces it but does not remove it.

The drug side is the reverse. Most of that evidence came from randomized controlled trials, where a coin flip decides who gets the medicine. Randomization removes the selection problem, which is why randomized trials tend to produce smaller and more believable effect sizes than observational studies of the same question.

Comparing a number from an observational study against a number from a randomized trial is not a level comparison.

Only four studies compared the two treatments directly

Of the 43 studies, four compared surgery patients to GLP-1 patients head to head, covering 18,880 people. All four were observational. The rest of the comparison is indirect.

Indirect means the math works out surgery versus drugs by lining up each one against the untreated group, the way you might compare two runners who never raced each other by looking at their times against a third runner. The technique is called a network meta-analysis. It is a reasonable way to use the evidence that exists, and the authors ran a check called node splitting that found the direct and indirect estimates agreed with each other. It is still not the same as a trial that randomly assigned people to surgery or a drug. No such trial exists.

The search stopped in September 2024, before the newest medicines had heart data

The authors searched the literature through September 30, 2024. Anything published after that is not in here.

The authors also note that the randomized trials in the drug arm used mostly first-generation GLP-1 medicines. That matters because the newer drugs produce considerably more weight loss than the older ones, and weight loss is thought to be part of how these medicines protect the heart. Whether the newer agents close the gap with surgery is an open question this analysis cannot answer.

Results varied widely from study to study

The paper reports a heterogeneity statistic between 66 percent and 93 percent depending on the outcome. Heterogeneity measures how much the individual studies disagreed with each other. Numbers in that range are high, and they mean the pooled averages sit on top of a lot of scatter. The authors say so directly and call for cautious interpretation.

They also could not run the follow-up analyses that would explain the scatter. There were no breakdowns by type of operation, by which GLP-1 drug, by diabetes status, or by starting BMI, because the source studies did not report the data consistently enough. The review was also not registered with PROSPERO in advance, a step that helps show researchers did not change their plan after seeing results.

What a patient can reasonably take from this

Both treatments were associated with fewer major heart events than no treatment. That is the finding with the most support behind it.

Beyond that, this analysis is a reason to ask your own clinician a better question, not a reason to change treatment. It cannot tell you what will happen to you, and it does not account for surgical risk, your other conditions, cost, or what you are able to stick with. The authors themselves end on individualized choice rather than a winner.