Trial results describe what happens under close supervision with motivated volunteers. A study published in August in The Lancet Diabetes & Endocrinology asked a harder question: among 45,093 real patients with obesity and type 2 diabetes, how many actually hit a meaningful combined target a year after starting treatment? The answer for semaglutide was 3.0%. For tirzepatide, 13.2%. For sleeve gastrectomy, 24.0%.
What was measured
The researchers used Epic Cosmos, a de-identified electronic health record dataset drawn from 1,633 hospitals and 37,900 clinics across 280 US health systems. They identified 468,712 patients with at least 365 consecutive days of semaglutide or tirzepatide prescriptions, and 238,028 who had undergone sleeve gastrectomy. After applying inclusion criteria — BMI of at least 35, HbA1c above 6.4%, and both weight and HbA1c recorded in a 9-to-15-month follow-up window — 45,093 patients remained: 33,482 on semaglutide, 4,178 on tirzepatide, 7,433 who had surgery.
The bar was deliberately demanding, and this is the key to reading the numbers. The primary outcome required both at least 20% body weight loss and an HbA1c below 5.7% — that is, substantial weight loss and blood sugar back into the non-diabetic range, simultaneously, at one year. Missing either meant not counting. Results were adjusted for age, sex, race, baseline HbA1c and BMI, comorbidity burden, social vulnerability and baseline medication load.
Adjusted probabilities of hitting that combined target: 3.0% (95% CI 2.8–3.2) for semaglutide, 13.2% (12.2–14.3) for tirzepatide, and 24.0% (22.9–25.1) for sleeve gastrectomy. On safety, emergency department visits within a year were more common after surgery than after either drug, and new prescriptions for reflux and nausea were higher in absolute terms after surgery too.
Why this is not a fair fight
The authors are direct about it. This is a retrospective cohort, not a randomised trial — nobody was assigned a treatment, patients and their doctors chose. The surgery group started with higher BMI, was younger, and had lower baseline HbA1c, all of which affect the odds of reaching the target independently of what treatment was given. Statistical adjustment reduces that problem but cannot remove it, and the authors name residual confounding explicitly. The safety comparison was also narrow in scope.
The 3.0% figure in particular needs care. It is not the proportion who lost meaningful weight on semaglutide — it is the proportion who cleared a high double bar in routine care, in a population with diabetes, which is a group that reliably loses less. Read it as evidence that trial averages overstate typical outcomes, not as evidence the drug rarely works. The study received no funding.
What it does support is a comparison rarely made honestly: for people with both obesity and type 2 diabetes, surgery remained substantially more likely to produce combined remission-level results at one year than either drug, at the cost of more emergency visits. That trade is the actual decision, and it deserves better than the framing that drugs have made surgery obsolete. What the drugs cost while you find out is set out in our pricing breakdown.
