A cost-effectiveness model that finds a drug both cheaper and better is making an unusual claim. Health economists call it dominance, and it is the strongest result such a model can produce. It is worth understanding where the money in it comes from, because it does not come from the price — the same distinction that decides what a GLP-1 actually costs once you stop reading the sticker.
What the model did
Using data from SURMOUNT-5, the head-to-head trial of the two drugs in people without type 2 diabetes, a patient-level simulation projected costs and outcomes over a lifetime from a US societal perspective. [1] It tracked weight, systolic blood pressure, HDL, HbA1c and total cholesterol, and used them to predict complications — cardiovascular disease, obstructive sleep apnea, type 2 diabetes and others — then attached costs and quality-of-life values to each.
The result was per-patient cost savings of $41,688 with tirzepatide, 0.506 additional quality-adjusted life years, and an incremental net health benefit of 0.784. Per 1,000 patients the model predicted 70 fewer developing type 2 diabetes and 10 fewer developing cardiovascular disease, with semaglutide patients spending 3.07 more years living with moderate or severe sleep apnea.
Where the saving actually sits
Not in the purchase. The model does not claim tirzepatide has a lower price; it claims that over a lifetime the complications it prevents cost less than the difference. Those are avoided hospital admissions, avoided diabetes care, avoided productivity losses — the societal perspective counts things no individual ever sees as money.
The counter tells a different story. Across the 380 sellers here publishing both molecules by injection, read September 2026, tirzepatide runs a median of $60 more a month. A buyer pays that difference monthly and in cash, and receives the modeled saving never — because it accrues to insurers, employers and the health system over decades. You can see the live spread in the price check.
The horizon problem
This is the objection that matters most, and this site has already published the evidence for it. The model assumes treatment continues. Real persistence does not. In a clinic program where the drug cost patients nothing, median persistence was 10.7 months.
A lifetime of avoided complications requires a lifetime of treatment, or at least enough of one for the metabolic changes to persist. Somebody who stops at month eleven keeps neither the weight nor the modeled savings. So the honest reading of a dominance result is conditional: tirzepatide dominates for the patient who stays on it for decades, and that patient is rarer than the model assumes.
Who funded it
This desk cannot say. The indexed record carries no conflict-of-interest statement and lists no grant agency, so the funding is simply unknown to us rather than absent. That is worth stating plainly on a page about a model that favors the more expensive molecule — not as an accusation, but because provenance is part of how a modeled result should be weighed, the way it is on the MASH model whose authors are named.
How to hold it
As one input among several, pointing the same direction as some evidence and against other evidence. A randomized cardiovascular trial recently found tirzepatide noninferior rather than superior to an older drug on hard endpoints. A lifetime model finds it dominant on cost-effectiveness. Both can be true; they measure different things over different horizons.
For somebody buying a month at a time, neither settles anything. The question at that scale is what the extra costs now and what the trial actually measured, which is where the premium gets weighed on its own evidence.