Over a follow-up that ran as long as 12.5 years, patients who took a GLP-1 drug for obesity and diabetes cost the health system about $109 more per month than patients who had weight-loss surgery instead. That gap holds even after you set aside what the drugs and the operations themselves cost. It is the price of everything else: hospital stays, other medications, the rest of a person's medical care.
The comparison, published July 1 in JAMA Network Open ↗, came out of Clalit Health Services, the largest health organization in Israel. Researchers pulled electronic records for adults with obesity (a body mass index of 30 or higher) and diabetes. Each had either started a GLP-1 receptor agonist ↗ or undergone a first bariatric metabolic surgery between 2010 and 2022. The team matched 2,721 pairs so the two groups looked alike on age, weight, blood sugar, and prior illness, then followed the bills through 2023.
Monthly costs after treatment averaged $415 in the drug group and $305 in the surgery group. Adjusted for the starting differences, the drug arm ran $109 higher each month per patient, a difference the authors report as statistically solid. Two line items carried most of it: hospitalizations added about $44 a month, and non-GLP-1 medications added about $53. Most of the divergence built up inside the first four years and then held.
The clinical trajectories point the same way. Surgery produced larger early drops in body mass index and in hemoglobin A1c, the standard three-month average of blood sugar. The drugs produced smaller changes that lasted. One intervention resets the body's set point in a single procedure; the other manages it continuously, which is exactly why the ongoing care around it keeps costing money.
What the study can and cannot say
This is an observational cohort, not a randomized trial. Nobody flipped a coin to assign surgery or a prescription. People who reach an operating table differ from people who fill a pharmacy order in ways that matching narrows but never erases. The authors are careful about this, and so is the honest reading: the result describes what happened to two comparable groups in one country's system, not a law about what any drug does to any budget.
The number that everyone will want to argue about is the one the study deliberately excludes. Strip out the cost of the surgery and the cost of the GLP-1 drugs, and the drug arm still spent more on everything else. Add those index costs back, and the arithmetic shifts again. A branded GLP-1 drug taken for years is not cheap, and a surgery is a one-time charge. The paper is answering a narrow question, downstream utilization, and answering it cleanly. It is not the last word on which choice a payer should fund.
It also lands against a familiar talking point. The pitch for covering GLP-1 drugs at scale often leans on the idea that they pay for themselves by preventing expensive complications. Here, among people eligible for both options, surgery was the arm that bent the cost curve down. That does not indict the drugs for the millions who are not surgical candidates. It does complicate the claim that the pills are the cheaper path for the patients who could go either way.
The drugs in this comparison are a class, not a single molecule. GLP-1 receptor agonists include semaglutide ↗, the peptide sold as Ozempic and Wegovy. The study pooled whatever patients were actually prescribed over a decade when the menu kept changing. peptidemodel hosts the individual molecules behind that label as computational cards, which is where the biology that the cost figures ride on actually lives. The economics are one layer. The receptor is another.