GLP-1 head-to-head: a new retrospective finds tirzepatide outperforms semaglutide on weight loss and adherence

The slide circulated on the morning of 2 July 2026 looked, at first glance, like a piece of marketing dressed up as science: a clean bar chart, two molecule names, and an unmistakable gap between them. By the end of the day it had been screenshotted into timelines across the GLP-1 conversation — the now-loud corner of medicine populated by clinicians, telehealth operators, and millions of patients who track their own weekly weigh-ins with the discipline of a day trader.

The chart claims something specific and consequential: that users of tirzepatide, the dual GIP/GLP-1 agonist sold as Mounjaro for diabetes and Zepbound for weight management, lose more weight on average than users of semaglutide, the GLP-1-only agonist sold as Ozempic and Wegovy. It also claims that tirzepatide users are less likely to fall into two of the most clinically frustrating categories — non-responders who never move the needle, and late-responders whose progress stalls before they reach a meaningful target. Both claims, if they hold up, sharpen the conversation about which drug should be prescribed first, and at what price the difference is worth having.

What the slide shows

The figure compares two cohorts of GLP-1 users stratified by molecule. The headline finding is a mean weight-loss differential favouring tirzepatide — the direction most clinicians working with both drugs already expected, given tirzepatide's dual-agonist mechanism and the head-to-head data Eli Lilly has published from its SURMOUNT programme. The second finding is the one that travels further: a smaller share of tirzepatide users appear in the non-responder tail, and a smaller share are late-responders whose curves flatten before reaching clinically meaningful weight loss. In a field where adherence is the limiting reagent, fewer dropouts is itself a result.

That distinction matters because the GLP-1 market is no longer a one-drug conversation. Two competing pharmacologies — semaglutide (Novo Nordisk) and tirzepatide (Eli Lilly) — have been moving in parallel for nearly three years, with a third wave of candidates, including oral GLP-1s and triple-agonists, now in late-stage development. Retrospective comparisons like the one circulated this week are how clinicians, payers, and self-pay patients triangulate between molecules in the absence of a clean head-to-head superiority trial run by a neutral party.

The evidence the slide is leaning on

The retrospective visual echoes the pattern set by Lilly's SURMOUNT-5 trial, the first randomised head-to-head reading of tirzepatide against semaglutide in adults with obesity but without diabetes. In that study, tirzepatide delivered a larger mean reduction in body weight than semaglutide over 72 weeks, alongside a higher share of participants losing at least 10%, 15%, and 20% of baseline weight. The structural caveat is the usual one: the trial was funded by Eli Lilly, the molecule's manufacturer, and used a dose-titration schedule that mirrors the company's labelling rather than an independent protocol.

For semaglutide, the comparator signal comes from Novo Nordisk's STEP programme, which established the 68-week weight-loss envelope for Wegovy and from which the SURMOUNT-5 results are benchmarked. Novo has since pointed to post-hoc and real-world analyses that show a meaningful share of patients losing 20% or more on semaglutide — including in head-to-head settings — and has argued that the durability and tolerability profile of its drug remains competitive. The retrospective circulating on X this week does not displace either sponsor's framing; it sits on top of them.

The structural frame

What the slide quietly captures is the moment GLP-1 prescribing stops being a single-drug decision. Through 2023 and most of 2024, "the GLP-1" was effectively semaglutide — supply permitting. Telehealth clinics were built around it, compounding pharmacies routed around it, and employer benefit plans negotiated against it. By the second half of 2025, tirzepatide had overtaken semaglutide in new prescription starts in major US pharmacy datasets, and the conversation shifted from "who can get a GLP-1" to "which GLP-1 should they get, and at what co-pay." The molecule is now the variable; the patient's phenotype, the payer's formulary, and the clinician's familiarity are the constraints.

That shift has secondary consequences the slide does not address but which follow from its findings. A drug that produces fewer non-responders and fewer late-responders is, all else equal, a drug that costs the health system less per kilogram lost. It is also a drug whose manufacturer has more leverage in formulary negotiations, more room to defend list price, and a stronger position in the parallel discourse about compounding, shortages, and the future of peptide-manufacturing capacity. Conversely, the comparator's sponsor now has an incentive to publish more granular responder analyses — to show that the average understates its drug's upside for the right patient — rather than compete on mean outcomes alone.

What the slide does not tell you

There are reasons to read the figure as a useful provocation rather than a verdict. Retrospective comparisons inherit the structure of the data they were built from: who got prescribed which drug, who stayed on it, who paid out of pocket, and who was counted in the denominator. Dosing, titration speed, and persistence — the variables that actually drive the divergence between these two molecules in real practice — are easy to flatten in a slide and hard to control in a chart. So is the underlying indication: tirzepatide's Mounjaro label covers type 2 diabetes, while Zepbound covers weight management, and prescribing patterns across the two labels are not interchangeable.

The honest reading is that the slide is consistent with what Lilly's randomised data already showed, that it adds an adherence angle the SURMOUNT programme touched only obliquely, and that it does not — on its own — settle the prescribing question. The clinical community will want to see the underlying cohort definitions, the missingness rates, and whether the responder categorisation was prespecified or post hoc. Until then, the figure is best treated as a strong hint, not a guideline.

Monexus framed this as a real-world-evidence signal inside a maturing therapeutic category rather than as a head-to-head clinical verdict — the wire frames tend to treat any GLP-1 comparison as definitive, while the underlying data still needs to be opened up.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

• https://x.com/cremieuxrecueil/status/1809297007831630000
• https://www.fda.gov/drugs/new-drugs-fda-cders-new-molecular-entities-and-new-therapeutic-biological-products/novel-drug-approvals-2024