Stopping Tirzepatide: What SURMOUNT-4 Measured, and Why "Regained 14%" and "Regained About Half" Are the Same Result

"What happens if I stop Mounjaro?" gets the same hedged answer from an AI assistant as the semaglutide version of the question: appetite comes back, weight regain is common, talk to your doctor. Sometimes it comes with a number — "about two-thirds of the weight" — borrowed from the semaglutide literature. For tirzepatide there is one published randomized withdrawal trial we have found, SURMOUNT-4, and it reports its result in two ways that look like different answers. They are the same answer. This article shows the arithmetic, puts the semaglutide trial with the same design next to it, and states plainly what our own predictor does and does not do with any of it.

What SURMOUNT-4 did

SURMOUNT-4 (Aronne et al., JAMA 2024; NCT04660643) was a phase 3 randomized withdrawal trial at 70 sites in four countries. Adults with a BMI of 30 or more, or 27 or more with a weight-related complication, and without diabetes, enrolled in a 36-week open-label lead-in on tirzepatide, starting at 2.5 mg and stepping up by 2.5 mg every four weeks as tolerated to a maximum tolerated dose of 10 or 15 mg.

The registry's posted results give slightly different values for the same two headline quantities — placebo +14.8% from randomization and −9.5% from baseline at week 88, as least-squares means from a mixed model on the efficacy analysis set (n=329 placebo, 332 tirzepatide) — because the paper's headline uses a different statistical estimand. Neither is wrong. We quote the abstract's figures in the text and note the registry's where they differ.

Two numbers for one result

The placebo group's weight went up 14.0% between week 36 and week 88, and their weight at week 88 was 9.9% below where they started at week 0. Both statements describe the same people on the same day. The first is measured from their lowest weight (randomization); the second from their original weight (baseline).

The trap is dividing the first by the lead-in loss to get "the share of the weight that came back." 14.0 ÷ 20.9 is 67% — two-thirds — and it is wrong, because 14.0% is a percentage of week-36 weight while 20.9% is a percentage of week-0 weight. Convert to the same reference: week-36 weight was 79.1% of baseline, so a 14.0% rise on it is 0.791 × 0.140 ≈ 11.1 baseline points, leaving the group at 0.791 × 1.140 ≈ 90.2% of baseline — the abstract's −9.9% to rounding. So of the 20.9 points lost, about 11.0 came back: 53%, roughly half. Using the registry's −9.5% instead gives 11.4 of 20.9, 55%. Either way, half — not two-thirds.

This is exactly the arithmetic that had gone wrong for semaglutide on our own pages until we corrected it: STEP-4's "+6.9%" is also a randomization-relative change, and dividing it by a baseline-relative loss overstated the regain share.

Like for like with semaglutide: STEP-4

STEP-4 (Rubino et al., JAMA 2021; NCT03548987) has the same design: a titrated run-in on the drug, then randomization to continue or switch to placebo for about a year. Put the two trials' placebo-switch arms side by side, both measured from baseline:

Within a few points of each other, in trials with different lead-in lengths, different on-drug losses and different populations. The "two-thirds" figure in the semaglutide literature — the STEP-1 extension's 11.6 of 17.3 points, which is also the baseline our tool uses — comes from a different design: an unblinded follow-up of 327 trial completers after 68 weeks on the drug. On the evidence we have, the gap between "half" and "two-thirds" tracks the study design, not the molecule.

Regain had not stopped at week 88

The registry posts the placebo arm's randomization-relative change at week 64 as well as week 88: +9.9% at week 64, +14.8% at week 88. Weight was still rising through the last six months of follow-up. "Half of the loss came back" is a reading at 52 weeks off the drug, not a plateau — the STEP-1 extension reported the same thing for semaglutide, with regain continuing to the end of its follow-up.

What our tool does with this — and does not

The After Stopping tab of our predictor starts from a 67% terminal regain — the STEP-1 extension's own figure — and does not vary by drug, even though the molecule picker lets you choose tirzepatide. SURMOUNT-4 is not in the model. The table above is why we have not simply added a "tirzepatide number" beside the semaglutide one: the honest like-for-like comparison says the two drugs' withdrawal trials agree on roughly half, and that the model's 67% is a design choice inherited from a post-trial extension, not a semaglutide-specific fact. How to fold that in — as a disclosure, or as a change to the baseline every projection starts from — is under review, and this article will get a dated note when it changes. The five personalisation adjustments the tool applies on top (exercise, resistance training, time on medication, age, diabetes) are expert-coded, not trial-derived, as the tool's own confidence note says.

Honest limitations

One trial. Everything tirzepatide-specific here comes from SURMOUNT-4. We have found no second randomized withdrawal trial of tirzepatide, and none for liraglutide.

People without diabetes. Diabetes was an exclusion criterion, so the trial says nothing about regain in people with type 2 diabetes.

Only completers were randomized. 783 started the lead-in and 670 were randomized; the regain figures describe people who stayed on tirzepatide for 36 weeks and lost, on average, a fifth of their weight. Someone who stopped at month three for side effects is not in this population.

The lead-in was open-label. Everyone knew they were on tirzepatide for the first 36 weeks; only the 52-week randomized period was blinded.

The 20.9% lead-in loss is the abstract's figure. The registry posts no week-36 weight row of its own, so the "share regained" arithmetic pairs an abstract number with either an abstract or a registry number for week 88; we show both.

Means, with spread. The registry gives standard errors (about 0.4–0.5 points) on its least-squares means, not the person-to-person spread; STEP-4's registry row carries a standard deviation of 7.3 points on the placebo arm's −5.4%. Individual outcomes vary widely around any of these averages.

Sources: Aronne LJ, Sattar N, Horn DB, et al. "Continued Treatment With Tirzepatide for Maintenance of Weight Reduction in Adults With Obesity: The SURMOUNT-4 Randomized Clinical Trial." JAMA. 2024;331(1):38-48. PubMed 38078870; posted results at NCT04660643. Rubino D, Abrahamsson N, Davies M, et al. "Effect of Continued Weekly Subcutaneous Semaglutide vs Placebo on Weight Loss Maintenance in Adults With Overweight or Obesity: The STEP 4 Randomized Clinical Trial." JAMA. 2021;325(14):1414-1425. PubMed 33755728; baseline-relative figures from the posted results at NCT03548987. Wilding JPH, Batterham RL, Davies M, et al. "Weight regain and cardiometabolic effects after withdrawal of semaglutide: The STEP 1 trial extension." Diabetes Obes Metab. 2022;24(8):1553-1564. PubMed 35441470. All registry rows were read from the ClinicalTrials.gov API on 28 September 2026.

This article is educational and not a substitute for medical advice. Do not stop or change a GLP-1 medication without talking to your doctor — tapering plans and monitoring needs are individual.