The trial evidence widens. The surgery gap doesn't close.
A 45,000-patient real-world cohort found sleeve gastrectomy still winning on combined outcomes, while two medical journals pushed seven incretin trials into new populations, comparisons, and insulin combinations.
One read on the month, with every source graded and numbered.
A real-world cohort put a number on the surgery gap
A retrospective cohort study drew on Epic Cosmos, a de-identified electronic health record dataset spanning 1,633 hospitals and nearly 38,000 clinics, to compare adults with obesity and type 2 diabetes who completed at least a year on semaglutide, on tirzepatide, or who underwent sleeve gastrectomy [1]. Among the 45,093 patients who met the study's criteria, the adjusted probability of hitting the combined target (at least 20% bodyweight loss and HbA1c below 5.7% at one year) was 24.0% after sleeve gastrectomy, 13.2% on tirzepatide, and 3.0% on semaglutide. Emergency-department visits and new prescriptions for reflux or nausea were both more frequent after surgery.
The authors are careful about what the numbers can and cannot say: the surgery group started younger, heavier, and with better baseline glucose control than the drug groups, the study has no randomisation, and it carried no outside funding. A companion commentary published in the same issue calls this "the debate" between medication and surgery still open [2]. What the cohort does establish is a baseline: at current dosing and current follow-up, the newest drugs have not closed the gap with surgery on a combined outcome, in a dataset large enough that the difference is not plausibly noise.
Basal insulin got two different answers, for two different patients
Two trials this month addressed basal insulin at two different moments in care. ACHIEVE-5 asked about intensification: it added oral orforglipron, Lilly's non-peptide GLP-1 pill, on top of insulin glargine already being titrated in 546 adults with type 2 diabetes across five countries. All three doses beat placebo on HbA1c at 40 weeks (−1.58 to −1.88 percentage points versus −0.79%, p<.001 for every dose) and on weight (up to −5.4% versus +0.2%), without an increase in hypoglycaemia [3]. A companion editorial frames both this trial and the same issue's mazdutide data as extending who the drugs work for, rather than settling anything new about whether they work [4].
COMBINE 4 asked about initiation instead: for 485 adults with type 2 diabetes who had never taken insulin and were managing on oral medications alone, is it better to start once-daily glargine, or start IcoSema, a fixed once-weekly combination of basal insulin icodec and semaglutide? Starting IcoSema won on HbA1c (−3.32 versus −2.44 percentage points, p<.001), on weight (patients lost 0.79 kg on IcoSema while gaining 3.81 kg on those starting glargine), and on hypoglycaemia, which was roughly half as frequent [5]. Adding a pill to insulin already in use, and choosing a weekly combination over insulin at the point of starting, both worked. Nothing published this month tells a clinician which moment matters more, or compares the two situations directly.
The evidence bar moved from beating placebo to beating the leading drug
CagriSema, Novo's fixed combination of the amylin agonist cagrilintide and semaglutide, has cleared placebo before. This month it cleared a harder bar. In REIMAGINE 2, a 2,713-patient phase 3 trial in adults with type 2 diabetes on metformin, CagriSema beat semaglutide 2.4 mg alone on HbA1c (−1.91 versus −1.75 percentage points; treatment difference −0.16, 95% CI −0.27 to −0.05, p=0.0035) [6]. The 2.4 mg semaglutide arm is the trial's protocol dose: semaglutide is approved up to only 2.0 mg for diabetes, and 2.4 mg is the dose approved for obesity. It is the first completed phase 3 result showing the combination beating semaglutide alone on HbA1c specifically, in people with type 2 diabetes — a narrower comparator than placebo. The margin is modest: 0.16 percentage points is below the 0.3–0.5-point range often used as a threshold for a clinically meaningful HbA1c difference, so a prescriber weighing CagriSema against semaglutide alone is weighing a small average advantage against the added cost and injection burden. A companion trial, REIMAGINE 1, confirmed the more conventional placebo-controlled result in a separate 189-patient population with milder, diet-and-exercise-managed diabetes (HbA1c differences of −1.7 and −1.3 percentage points across the two doses, both p<0.0001) [7].
Japan and China got trials built for them
Three trials this month tested incretin drugs in populations the original pivotal programmes barely included. ACHIEVE-J followed 401 Japanese adults with type 2 diabetes on oral orforglipron for 52 weeks specifically to characterise long-term safety in an East Asian population, whose "distinct pathophysiological characteristics" the trial's own background section cites as the reason it was run at all [8]. Discontinuations from adverse events ran higher at the top dose (14% at 36 mg versus 5% at 3 mg), consistent with the drug's known gastrointestinal profile rather than a new signal.
STEP 12 did the same for semaglutide in mainland China and Taiwan, enrolling adults against regionally defined BMI thresholds (24–<28 kg/m² with a weight-related comorbidity, or 28–<30 without one, lower than the thresholds used in Western trials). Weight loss reached −12.1% against −2.2% on placebo [9]. And GLORY-2 tested mazdutide, a dual GLP-1/glucagon agonist, in 461 Chinese adults with obesity, producing the largest weight-loss number of the month: −16.65% versus −1.50% on placebo at 60 weeks, though more than half the treated group reported vomiting [10]. GLORY-2 is sponsored by Innovent Biologics, a Chinese company, which makes it the one major trial this month not funded by Novo Nordisk or Lilly. That is a small but real sign that the next tier of incretin evidence won't all be generated, or paid for, in the West.
The benefit doesn't depend on the scale, and industry chased everything except a bigger number
A prespecified pooled analysis combined participant-level data from three landmark semaglutide trials (SELECT, FLOW, and SOUL) covering 30,787 people with cardiovascular disease or chronic kidney disease, to ask whether semaglutide protects the kidney across that combined population. It does: a composite of major kidney decline, kidney failure, or kidney- and cardiovascular-related death occurred less often on semaglutide than placebo (hazard ratio 0.84, 95% CI 0.77–0.91) [11]. The authors write that the benefit "might not be explained only by its glycaemic effects, weight-management effects, or both." A companion piece in Cell Metabolism makes the general version of that argument: that GLP-1 drugs act through neural, immune, and inter-organ pathways whose benefits don't reduce to pounds lost [12]. A separate commentary calls this month's flurry of dual and triple agonists targeting glucagon receptors a "renaissance," explicitly framing the mechanism hunt as bigger than any one drug's weight-loss number [13].
Industry's month pointed the same direction. Roche, through its Genentech subsidiary, paid Hanmi $190M upfront and up to $2.3B in potential milestones for HM17321, a urocortin-2 analog still in Phase 1 that works through a non-incretin pathway and is designed to spare muscle while burning fat [14][15]. Lilly's Mounjaro won FDA approval for a second indication, reducing cardiovascular risk in type 2 diabetes, on the strength of the 13,299-patient SURPASS-CVOT trial, which compared tirzepatide against dulaglutide, a drug already approved for the same cardiovascular benefit. The composite rate of cardiovascular death, heart attack, or stroke was numerically lower on tirzepatide (12.2% versus 13.1%), but that difference did not reach statistical significance for superiority (hazard ratio 0.92, 95.3% CI 0.83–1.01, p=0.09); the approval rests on tirzepatide matching dulaglutide's established benefit rather than exceeding it [16][17]. And Lilly's long-acting insulin Onswik received a NICE recommendation ahead of a likely UK approval [18]. The month's least flattering item belongs to Amgen, which discontinued its Phase 1 obesity candidate AMG 513 even after the FDA lifted a clinical hold that had paused it. The company offered only a general statement about its bar for obesity medicines, without detailing the specific reason, leaving MariTide as its only obesity asset [19]. On the tolerability side, a JAMA news article reported lower fracture risk among people with type 2 diabetes taking a GLP-1 drug — an association, not a trial outcome, and worth tracing to its underlying study before it's repeated as a benefit [20]; a review of gastroparesis in the same journal lists GLP-1 receptor agonists among the medications that can delay gastric emptying enough to require discontinuation [21]. And regulators moved on the monitoring side too: the FDA authorised the first wearable that continuously tracks both blood ketones and glucose in the same device [22].
If you're weighing surgery against a GLP-1 for a patient
A 45,000-patient real-world cohort found sleeve gastrectomy nearly twice as likely as tirzepatide, and eight times as likely as semaglutide, to produce both major weight loss and normalized A1c at one year. The surgery patients started younger and healthier, and the comparison is adjusted but not randomised, so residual confounding could mean the true causal gap is smaller than this adjusted estimate. It still hasn't closed.
Changes if a randomised trial, rather than a retrospective cohort, compares surgery against the newest drugs head to head.
If you're building or investing in a metabolic asset
Roche just paid up to $2.3B for a Phase 1 molecule specifically because it works through a different pathway and spares muscle. It hasn't been shown to beat GLP-1s on weight loss. When the leading drugs already clear the efficacy bar, mechanism and organ reach are where new money is going.
Changes if HM17321's Phase 1 data disappoints, or a rival non-incretin asset reports first.
If you manage type 2 diabetes at the point of insulin initiation or intensification
Two trials this month addressed different moments. For patients already on titrated glargine and needing more, adding oral orforglipron improved HbA1c and weight without extra hypoglycaemia. For insulin-naive patients on oral therapy alone, starting IcoSema beat starting glargine on HbA1c, weight, and hypoglycaemia. Neither trial compares the two strategies directly, because they answer different clinical questions.
Changes if a trial compares add-on intensification against initial fixed-combination therapy in the same population.
The trials this issue draws on, with what each has and has not shown.
| Trial | Year | Population | What it showed |
|---|---|---|---|
| —semaglutide vs tirzepatide vs sleeve gastrectomy, real-world | 2026 | Obesity + type 2 diabetes, retrospective EHR cohort (n=45,093) | Combined weight+A1c target at 1 year: 24.0% (surgery), 13.2% (tirzepatide), 3.0% (semaglutide). Observational; baseline groups differ. [1] |
| ACHIEVE-5orforglipron added to insulin glargine | 2026 | Type 2 diabetes on glargine (n=546) | HbA1c −1.58 to −1.88 pts vs −0.79% placebo, all doses p<.001. No increase in hypoglycaemia. [3] |
| COMBINE 4IcoSema vs insulin glargine U100 | 2026 | Insulin-naive type 2 diabetes (n=485) | HbA1c −3.32 vs −2.44 pts, p<.001; weight −0.79 vs +3.81 kg; hypoglycaemia about half as frequent. [5] |
| REIMAGINE 2CagriSema vs semaglutide alone | 2026 | Type 2 diabetes on metformin (n=2,713) | HbA1c −1.91 vs −1.75 pts (semaglutide alone), p=0.0035. First win over an active comparator. [6] |
| REIMAGINE 1CagriSema vs placebo | 2026 | Early type 2 diabetes (n=189) | HbA1c −1.7 / −1.3 pts vs placebo across two doses, both p<0.0001. [7] |
| ACHIEVE-Joral orforglipron, long-term safety | 2026 | Japanese adults, type 2 diabetes (n=401) | 52-week safety: 85% any adverse event; discontinuation 5–14% by dose. No new safety signal. [8] |
| STEP 12semaglutide 2.4mg, regional BMI thresholds | 2026 | China + Taiwan, overweight/obesity (n=242) | Weight −12.1% vs −2.2% placebo, p<0.0001. [9] |
| GLORY-2mazdutide 9mg | 2026 | Chinese adults with obesity (n=461) | Weight −16.65% vs −1.50% placebo, p<.001. Vomiting in 53% of treated group. [10] |
| SELECT + FLOW + SOULpooled kidney-outcomes analysis | 2026 | Cardio-kidney-metabolic disease, pooled (n=30,787) | Kidney composite HR 0.84 (95% CI 0.77–0.91) favoring semaglutide. [11] |
| SURPASS-CVOTtirzepatide vs dulaglutide | 2025 | Type 2 diabetes with atherosclerotic CV disease (n=13,299) | Non-inferior to dulaglutide on MACE (12.2% vs 13.1%, HR 0.92); not statistically superior (p=0.09). [17] |
| AMG 513early obesity candidate, Phase 1 | 2026 | Obesity | Discontinued, Aug 2026, after an FDA clinical hold was lifted. No efficacy data reported; no specific reason given. [19] |
Every source behind the read above, numbered to its [n] marker and graded by where its evidence comes from.
Source gradeRegulatoryGuidelineTrialPeer-reviewedPreprintTrade press
Real-world evidence & commentary
- Peer-reviewed 1-year outcomes of semaglutide, tirzepatide, and sleeve gastrectomy in obesity in type 2 diabetes: a retrospective cohort study
- Peer-reviewed Obesity medications versus metabolic bariatric surgery: the debate goes on
Trials — insulin combinations
- Peer-reviewed Orforglipron Added to Titrated Insulin Glargine in Type 2 Diabetes: The ACHIEVE-5 Randomized Clinical Trial
- Peer-reviewed Mazdutide and Orforglipron—New Evidence in Obesity and Diabetes
- Peer-reviewed Once-weekly IcoSema versus once-daily insulin glargine U100 in type 2 diabetes management (COMBINE 4): an open-label, multicentre, treat-to-target, randomised, phase 3b trial
Trials — CagriSema head-to-head
- Peer-reviewed Cagrilintide-semaglutide (CagriSema) versus semaglutide or cagrilintide in people with type 2 diabetes (REIMAGINE 2): a double-blind, randomised, controlled, phase 3 study
- Peer-reviewed Efficacy and safety of once-weekly cagrilintide-semaglutide (CagriSema) in adults with type 2 diabetes inadequately controlled on diet and exercise (REIMAGINE 1): a randomised, double-blind, placebo-controlled, phase 3a study
Trials — built for Japan and China
- Peer-reviewed Long-term safety of oral orforglipron in Japanese participants with type 2 diabetes (ACHIEVE-J): a multicentre, randomised, open-label, parallel-group phase 3 trial
- Peer-reviewed Efficacy and safety of once-weekly semaglutide 2·4 mg in Chinese adults with overweight or obesity (STEP 12): a randomised, double-blind, placebo-controlled, multicentre, phase 3b trial
- Peer-reviewed Treatment With 9-mg Mazdutide for Weight Reduction in Chinese Adults With Obesity: The GLORY-2 Randomized Clinical Trial
Mechanism & kidney benefit
- Peer-reviewed Effect of semaglutide on kidney outcomes in the SELECT, FLOW, and SOUL trials: a prespecified pooled analysis
- Peer-reviewed Weight-loss-independent actions of GLP-1 medicines
- Peer-reviewed The glucagon renaissance in obesity therapeutics
Industry & deals
- Trade press Hanmi Pharm Signs Exclusive Licensing Deal with Genentech for Novel Obesity Therapy
- Trade press Roche commits to lean-mass preservation with Hanmi obesity deal
- Trade press FDA approves Lilly's Mounjaro (tirzepatide) to reduce cardiovascular risk in adults with type 2 diabetes
- Peer-reviewed Cardiovascular Outcomes with Tirzepatide versus Dulaglutide in Type 2 Diabetes
- Trade press Lilly's Onswik set for broad UK coverage after pricing authority gives OK
- Trade press Amgen drops early-stage obesity drug, upping pressure on MariTide
Tolerability & devices
Radar scans the literature, registered trials, industry and deals, and tolerability and access each month, then surfaces the items that matter and grades each by where its evidence comes from. Registered trials are reported as direction of travel, never as results, and company-reported topline is labelled as such. This is issue No. 3; the source list grows as the scanner's coverage widens.
Working a diabetes question that deserves more than a headline?
For confidential scientific, pipeline, or strategic diligence inquiries in metabolic disease, start with a brief scoping conversation.
Request a confidential scoping call