# The Cost-Effectiveness Case, and What Is Inside It

By Ruth Calder, Enforcement Editor (https://glpwatchdog.org/authors/ruth-calder) | Published: 2026-08-20 | Updated: 2026-10-30

Canonical page: https://glpwatchdog.org/research/the-cost-effectiveness-case

> A model projects tirzepatide saving $41,688 per patient and adding half a quality-adjusted life year. Nobody measured that. It was computed, and the adherence assumption is doing most of the work.

## Key points

- A modeled $41,688 saving and +0.5 QALYs versus semaglutide. It is a simulation, not a measurement.
- Models are fed TRIAL efficacy; ordinary care produces a third to a half of it.
- The total includes absenteeism and presenteeism, real societal costs that a payer's arithmetic leaves out.
- Always establish the perspective a savings figure was computed from.

A cost-effectiveness analysis reports that tirzepatide saves **$41,688 per patient** and adds half a quality-adjusted life year compared with semaglutide. It takes the head-to-head SURMOUNT-5 trial and projects it over a lifetime.[1] Nobody measured that figure. It was computed from assumptions, and one of those assumptions does most of the work.

## What a cost-effectiveness model does

It takes efficacy from clinical trials, assumes a level of adherence, projects the consequences across years or decades, attaches costs to each event avoided, and discounts it all back to a present value. The output is a single number that looks like a measurement but is a simulation.

That is a legitimate and necessary exercise, because payers have to decide something and waiting decades for outcome data is not an option; [two public payers in North Carolina](https://glpwatchdog.org/research/north-carolina-medicaid-and-state-health-plan-glp1) have reached opposite decisions on the same drugs. It also means the result is only as good as its inputs, and the inputs are choices. Even the unit of benefit, the quality-adjusted life year, is a modeled quantity that nobody has ever observed. The same is true of the [projected 22% cut in heart risk for people without cardiovascular disease](https://glpwatchdog.org/research/a-benefit-nobody-observed): nobody in that study took the drug.

> Nobody measured the $41,688 saving. It was projected over a lifetime from one head-to-head trial.

## The assumption that undoes it

Models like this are fed trial efficacy. Outside a trial, these drugs achieve **5.4% to 9.1%** weight loss at one year, against the 14.9% and 20.9% their pivotal trials reported (see [what these drugs do outside a trial](https://glpwatchdog.org/research/what-these-drugs-do-outside-a-trial)).

The reasons are in [many people never reach the dose](https://glpwatchdog.org/research/most-people-never-reach-the-dose). Only 58% of people reach maintenance dosing after four weeks, 21% are still on the starting dose, and discontinuation runs from 37% to 56% depending on the drug.

A model that assumes trial adherence computes savings that ordinary care does not generate. The events avoided are avoided in the simulation because the simulated patients kept taking the drug at the studied dose. Four in ten or more of real ones do not.

> **This is not an accusation of manipulation.** Using trial efficacy is standard practice and is what the source data supports. The point is narrower and it applies to every model of this kind: the number describes a world in which the drug is taken as tested. Whether that world resembles yours is a separate question, and it is the one that decides whether the savings appear.

## What is inside the total

The analysis includes reductions in **absenteeism** (days lost from work) and **presenteeism** (being at work while unwell and less productive).[1] Both are legitimate societal costs and neither appears on a payer’s balance sheet.

That matters for how the figure is used. From a societal perspective, productivity losses are real money. From an insurer’s perspective, they are somebody else’s money, and an analysis that includes them will always look more favorable than the one a payer runs. Whenever you see a savings figure, first establish whose perspective it was computed from. We apply that test to ten published models in [our cost-effectiveness reading list](https://glpwatchdog.org/research/lists/top-glp1-cost-effectiveness-studies).

## How to read any of these

- **Whose costs?** Societal analyses include productivity; payer analyses do not.
- **What adherence was assumed?** Trial adherence produces trial savings.
- **Over what horizon?** Lifetime models bank decades of avoided events from a year or two of data.
- **Who funded it?** Economic models are commissioned, and the commissioning party chooses the comparator. Three of this model's nine authors, the first among them, list Eli Lilly, the maker of tirzepatide, as their affiliation.[1]
- **Does it beat doing nothing, or beat the alternative?** Those are different claims and get conflated.

None of this means tirzepatide is not more cost-effective than semaglutide. It may well be, since it produces more weight loss in trials and in ordinary care, and that is the substance underneath the model. The specific dollar figure, though, is a projection with a large adherence assumption inside it, and should be quoted as one.

## Frequently asked questions

### Is tirzepatide more cost-effective than semaglutide?

A modeling study concluded so, projecting lifetime savings of $41,688 per patient and an additional half a quality-adjusted life year. It is a simulation built from SURMOUNT-5 trial efficacy and assumptions rather than an observed result, and three of its authors, including the first, list Eli Lilly, tirzepatide's maker, as their affiliation.

### Why does the adherence assumption matter so much?

Because models are fed trial efficacy, and ordinary care produces roughly half to a third of trial weight loss, and only 58% of people reach maintenance dosing. Savings computed on trial adherence do not appear if the drug is not taken as tested.

### What are absenteeism and presenteeism doing in a cost calculation?

They are productivity losses: days missed, and days worked while unwell. Both are real societal costs, and neither appears on an insurer's balance sheet, so including them makes any analysis look more favorable than a payer's own arithmetic.

### Is the model wrong then?

Not wrong, but bounded. Using trial efficacy is standard and defensible. The number describes a world where the drug is taken as tested, and whether that matches reality determines whether the savings materialize.

### What should I check when I see a savings figure?

Whose costs it counts, what adherence it assumed, how long a horizon it projects over, who commissioned it, and whether it is claiming to beat no treatment or to beat the alternative drug.

## References

1. Johansson E, Wilding JPH, Upadhyay N, et al. Cost-effectiveness of tirzepatide versus semaglutide for patients with obesity or overweight in the US. Journal of Medical Economics. 2026. PMID: 42012820. https://pubmed.ncbi.nlm.nih.gov/42012820/

---

Cite as: GLP Watchdog, Ruth Calder, "The Cost-Effectiveness Case, and What Is Inside It", published 2026-08-20, updated 2026-10-30. https://glpwatchdog.org/research/the-cost-effectiveness-case

Canonical HTML page: https://glpwatchdog.org/research/the-cost-effectiveness-case

Medical disclaimer: Nothing here is medical advice. We report on this subject; we do not treat anyone. Whether to begin a medication, supplement, or treatment, whether to adjust it, and whether to stop are questions for a clinician licensed to answer them, who knows your history.
