Tirzepatide's UK Cost-Effectiveness Case Gets Stronger, With Caveats

The health economics of tirzepatide just got a methodological upgrade, and I think it matters. A paper published in the Journal of Medical Economics (PMID: 42021523, DOI: 10.1080/13696998.2026.2646079) updates the UK cost-effectiveness model for tirzepatide in obesity by properly accounting for two outcomes that earlier models handled crudely: type 2 diabetes progression and obstructive sleep apnea remission. The result is ICERs between £8,327 and £10,157 per QALY gained across all three doses, well under the NICE £20,000 threshold. My position is straightforward: the methodology improvements are real and the directional finding is credible, but this is an Eli Lilly-funded analysis and should be read as a first estimate, not a settled verdict.

What Prior Models Got Wrong

Health economic models for obesity drugs have historically reduced comorbidity risk to a few blunt categories. If a patient had type 2 diabetes, they had it. If they didn’t, they didn’t. The clinical reality is more granular: HbA1c moves continuously, diabetes incidence risk scales with weight and duration of exposure, and remission is not a theoretical edge case, particularly following substantial weight loss.

Obstructive sleep apnea has been similarly flattened. Models have either ignored OSA altogether or applied static prevalence adjustments. This matters because OSA remission is a genuine quality-of-life driver, showing up in both the QALY numerator and the downstream healthcare cost denominator. Treating it as a binary on/off misestimates both.

The prior modelling framework also relied on Visual Basic for Applications. That is not a small technical footnote. VBA-based models carry real risks for reproducibility and audit; they are slow to adapt, prone to version-lock, and difficult to peer-review systematically. The shift to R addresses all three.

What This Model Actually Reports

Annemans and colleagues (including co-authors employed by Eli Lilly) rebuilt the simulation with three specific changes.

First, HbA1c is now tracked continuously rather than in categorical bins, so the model captures the gradual diabetes progression and weight-loss-mediated reversal that the SURMOUNT trial programme documented. The SURMOUNT-1 trial (NCT04184622), which enrolled non-diabetic adults with obesity, showed sustained weight reductions of up to 22.5% at 72 weeks on 15 mg tirzepatide; the model uses that trial’s clinical inputs.

Second, OSA remission is now modelled with explicit transition probabilities tied to weight change, pulling from the published obesity-OSA relationship literature rather than a single prevalence adjustment.

Third, lifestyle intervention discontinuation is modelled more realistically. Earlier versions assumed implausibly stable adherence to diet-and-exercise regimens over multi-year horizons. Loosening that assumption makes the comparator arm weaker over time, which mechanically improves tirzepatide’s relative QALY position. This is not a flaw in the revised model. It reflects what actually happens in real-world NHS practice. But it is worth being explicit about, because it is a parameter that sponsors can calibrate to their advantage without obviously falsifying anything.

The reported ICERs: £8,327/QALY for 5 mg, rising to £10,157/QALY for 15 mg (the paper reports a range; the dose-level breakdown comes from the abstract). All three doses evaluated (5, 10, 15 mg) are cost-effective at the £20,000 threshold. The refined OSA and discontinuation modelling specifically drove incremental costs down and QALYs up relative to the older model version.

My Read on the Evidence Shift

The directional conclusion, that tirzepatide is cost-effective in the UK obesity population, is consistent with multiple prior analyses and with the underlying clinical evidence. What this paper adds is not a reversal of prior findings but a tightening of the estimate via more defensible assumptions. That is genuinely useful.

I am more confident, after this paper, that the ICER sits well below £20,000. The OSA modelling improvement alone probably matters for a meaningful fraction of the target population; OSA prevalence in people with BMI ≥30 is high enough that ignoring remission pathways was always going to undercount QALYs.

What I am less confident about: the magnitude of the ICER reduction attributable to the discontinuation assumption shift. The paper does not, as far as the abstract reports, provide a sensitivity analysis isolating each methodological change’s contribution. Without that decomposition, I cannot tell whether the model is more accurate or just more favourable.

The Caveat That Earns Its Own Section

Co-authors include researchers employed directly at Eli Lilly and Company, tirzepatide’s manufacturer. The study carries industry sponsorship. This does not make the findings wrong, but it does establish a conflict of interest that any NHS decision-maker or independent health economist reviewing this model should note explicitly.

Sponsor-funded economic models in obesity pharmacotherapy have a consistent track record of producing ICERs that clear the relevant threshold, particularly in their initial publications. I am not alleging manipulation. I am noting a structural incentive and the observable pattern it produces, because acknowledging both is what makes commentary useful rather than decorative.

What Would Actually Settle This

Three things would move me from “credible estimate” to “high confidence.”

An independent replication of the model by a group without Eli Lilly co-authorship, ideally using the published R codebase, would be the clearest test. If the ICERs hold, they hold.

Real-world NHS adherence and discontinuation data for tirzepatide, which will accumulate as Mounjaro’s post-authorisation commitments play out, will either validate or undermine the comparator assumptions. The modelled discontinuation rates for diet-and-exercise are arguably the most consequential lever in the analysis.

Finally, cardiovascular outcome data in non-diabetic obesity patients. SURMOUNT-CVOT (NCT05556512) is ongoing and will report hard CV endpoints. Right now the model is projecting CV risk reductions from weight loss via epidemiological relationships, not from direct trial evidence in this population. If that trial shows attenuated or null CV benefit, the QALY projections will need revision.

The economics of treating obesity with tirzepatide almost certainly favour treatment at current UK pricing. This paper provides the most methodologically careful estimate yet. I would not stake NHS formulary policy on it without independent validation, and I would not wait long for that validation to arrive.