Trial summary
SURMOUNT-4 Withdrawal Trial
SURMOUNT-4 is the withdrawal trial and the single most important study for anyone weighing long-term cost. After a 36-week open-label lead-in at the maximum tolerated dose, 670 adults were randomised to continue tirzepatide or switch to placebo for 52 weeks. Continuing maintained and modestly extended the loss; switching to placebo produced substantial regain.
Key takeaways
- Continuing tirzepatide maintained and slightly extended weight loss over 52 further weeks.
- Switching to placebo after the lead-in produced substantial regain.
- The design isolates the drug's role, because every participant had already lost weight on it.
- This is the evidence behind treating tirzepatide as long-term rather than a short course.
- Regain reflects the return of appetite signalling, not a failure of willpower.
| Trial | SURMOUNT-4 |
|---|---|
| Design | Randomised withdrawal: 36-week open-label lead-in at maximum tolerated dose, then 52 weeks continue vs placebo |
| Population | Adults with obesity or overweight, without type 2 diabetes |
| Participants (N) | 670 |
| Primary result | Continued treatment maintained and modestly extended weight loss; withdrawal to placebo produced substantial regain |
| Journal | JAMA |
| Identifiers | NCT04660643 · PMID 38078870 · DOI |
| Funding | Eli Lilly and Company |
Why is a randomised withdrawal design persuasive?
Observational reports of regain after stopping can always be explained away by selection effects. SURMOUNT-4 removes that objection: every participant reached a lower weight on tirzepatide first, and only then were they randomly assigned to continue or stop. The divergence that follows is attributable to the withdrawal itself.
The practical consequence is financial as much as clinical. If maintaining the result requires continued treatment, then the number that matters when comparing providers is sustained monthly cost over years, not an introductory first-month price.
Does everyone regain all the weight?
No. Regain is substantial on average but incomplete within the trial window, and it varies between individuals. The trial reports group means; some people regain faster, some slower. What the data does not support is the expectation that stopping preserves the result.
Cost implication: model the multi-year cost rather than the first month.
What are the strengths and limitations?
- Trial populations are selected by inclusion and exclusion criteria and do not represent every patient.
- Mean results describe groups; individual response varies substantially around the mean.
- SURMOUNT-4 was funded by Eli Lilly and Company, which is disclosed here and does not by itself invalidate results.
- Open-label or partially unblinded elements, where present, can influence behaviour and reporting.
Does this apply to compounded tirzepatide?
No. This trial studied the FDA-approved product. Compounded tirzepatide is a different, non-approved product whose formulation, concentration, and excipients may differ, and it has not been tested for equivalence in a randomised trial. Compounded tirzepatide is not FDA approved and is not reviewed by the FDA for safety, effectiveness, or quality before sale.
What the evidence shows
- SURMOUNT-4 studied the FDA-approved tirzepatide product manufactured to a verified standard.
- Results apply to the population, dose range, and duration actually enrolled.
- Identifiers are published so the primary record can be checked independently.
What the evidence does not show
- That a compounded tirzepatide product reproduces these results — compounded products were not studied.
- That an altered oral, sublingual, ODT, or troche formulation delivers comparable exposure.
- That outcomes generalise to populations excluded from the trial.
Related: Weight regain after stopping · Stopping tirzepatide
How large is the effect across the verified trials?
| Trial arm | Point estimate | Range | N |
|---|---|---|---|
| SURMOUNT-1 · 15 mg | 20.9% | 19.5% to 22.3% | 2,539 |
| SURMOUNT-1 · 10 mg | 19.5% | 18.2% to 20.8% | 2,539 |
| SURMOUNT-1 · 5 mg | 15.0% | 13.8% to 16.2% | 2,539 |
| SURMOUNT-2 · 15 mg | 14.7% | 13.4% to 16.0% | 938 |
| SURMOUNT-2 · 10 mg | 12.8% | 11.5% to 14.1% | 938 |
| SURMOUNT-1 · placebo | 3.1% | 2.3% to 3.9% | 2,539 |
| Series | Wk 0 | Wk 12 | Wk 24 | Wk 40 | Wk 56 | Wk 72 |
|---|---|---|---|---|---|---|
| Tirzepatide 15 mg | 0% | -6.5% | -12.4% | -16.8% | -19.3% | -20.9% |
| Tirzepatide 5 mg | 0% | -5.1% | -9.3% | -12.4% | -14.2% | -15.0% |
| Placebo | 0% | -1.4% | -2.3% | -2.8% | -3.0% | -3.1% |
When was each piece of this evidence established?
| Date | Event |
|---|---|
| May 2022 | Tirzepatide approved as Mounjaro for type 2 diabetes |
| Jun 2022 | SURMOUNT-1 published in the New England Journal of Medicine |
| Jul 2023 | SURMOUNT-2 published in the Lancet |
| Nov 2023 | Tirzepatide approved as Zepbound for chronic weight management |
| Dec 2023 | SURMOUNT-4 withdrawal results published in JAMA |
| Jun 2024 | SURMOUNT-OSA published; obstructive sleep apnoea evidence established |
| May 2025 | SURMOUNT-5 head-to-head against semaglutide published in NEJM |
How should SURMOUNT-4 be read?
Reading a trial well means separating what it was designed to detect from what it happened to measure, and separating both from what its sponsor would like it to mean. SURMOUNT-4 used a Randomised withdrawal: 36-week open-label lead-in at maximum tolerated dose, then 52 weeks continue vs placebo design in Adults with obesity or overweight, without type 2 diabetes, enrolling 670 participants. That design determines the questions it can answer and, just as importantly, the questions it cannot.
Three habits make trial reading more reliable. First, check the registered protocol against the publication: the registry entry records what the investigators said they would measure before they saw any data, and a primary endpoint that changed between registration and publication is worth noticing. Second, read the population criteria rather than the title, because eligibility rules frequently exclude the patients a reader most resembles. Third, look at who was excluded from the analysis and why, since attrition that differs between arms can generate an apparent effect on its own.
All three checks are possible for SURMOUNT-4 because its identifiers are published and verified. The registry record, the peer-reviewed publication, and the digital object identifier are all listed on this page, so nothing here has to be taken on trust.
What does the design of SURMOUNT-4 allow and forbid?
What the evidence shows
- Comparisons between the randomised arms, because randomisation makes the groups comparable at baseline.
- Statements about the population actually enrolled: Adults with obesity or overweight, without type 2 diabetes.
- Statements about the intervention as delivered — the approved product at the doses studied.
- Safety signals frequent enough to appear in a trial of this size.
What the evidence does not show
- Statements about populations excluded by the eligibility criteria.
- Statements about doses, formulations, or durations outside those studied.
- Rare adverse events, which require post-marketing surveillance to detect.
- Any claim about compounded preparations, which were not studied.
The funding source is disclosed as Eli Lilly and Company. Industry sponsorship of pivotal trials is normal and does not by itself invalidate a result — the alternative, in practice, is that large trials do not happen. What sponsorship does influence is which questions get asked, which comparators get chosen, and which results get published promptly. That is a reason to read the registry alongside the publication, not a reason to dismiss the finding.
How does SURMOUNT-4 fit with the rest of the evidence?
No single trial establishes a treatment. The tirzepatide evidence base works as a set: one trial establishes the size of the effect in a general obesity population, another shows that the effect is smaller when type 2 diabetes is present, another shows what happens after lifestyle intervention has already succeeded, another shows what happens when treatment stops, and another compares the drug against its main alternative. Reading any one of them as the whole picture is the most common error in consumer coverage of this drug class, and it is usually the trial with the largest number that gets quoted.
Placed in that set, SURMOUNT-4 contributes Continued treatment maintained and modestly extended weight loss; withdrawal to placebo produced substantial regain. Its contribution is bounded by its population and duration, and it should be cited alongside — not instead of — the trials that answer the adjacent questions.
What would change the conclusion on this page?
This page would be revised, with the change recorded in its history, if any of the following occurred:
- Publication of a larger or longer randomised trial in the same population reporting a materially different effect size.
- A registry-versus-publication discrepancy showing the primary endpoint was changed after data were seen.
- Retraction, correction, or expression of concern attached to the primary publication.
- Post-marketing surveillance identifying a safety signal not visible at this trial's sample size.
- An independent re-analysis of the participant-level data reaching a different conclusion.
What does this page cover, and what does it deliberately leave out?
This page addresses Continued therapy versus placebo and regain and maintenance, organised around the primary question of SURMOUNT-4. Each of those elements is treated separately below rather than blended, because they carry different evidence weights and a reader is entitled to know which parts rest on randomised data and which rest on a captured commercial claim or a regulatory document.
| Element | Treatment here | Evidence basis |
|---|---|---|
| Continued therapy versus placebo | Covered on this page | Primary evidence |
| Regain and maintenance | Covered on this page | Primary evidence |
| Individualized clinical instruction | Deliberately not covered | Belongs with the registered protocol and the peer-reviewed publication |
What are the limits of what this page can tell you?
Every page on this site rests on a specific primary trial record, and that record has boundaries worth stating plainly rather than leaving a reader to discover them. The limitations below are specific to the material presented above.
- The evidence here describes groups, populations, or captured records — it does not describe you, and no page can substitute for the registered protocol and the peer-reviewed publication.
- Figures carry the date on which they were verified. In a market where terms change frequently, an undated figure functions as a claim about the present that nobody has checked.
- Elements marked Verification Pending are genuinely unknown to this publication rather than merely omitted for brevity, and should not be inferred from surrounding content.
- Where a source conflicts with another, this site shows the conflict rather than resolving it, which means some questions are left open on purpose.
What would change the conclusion on this page?
This page would be revised, with the change recorded in its history, if any of the following occurred:
- New primary evidence bearing directly on SURMOUNT-4.
- A change to FDA labelling affecting any statement made above.
- A verified correction submitted through the corrections process and accepted on the evidence.
- A material change to a captured record, including a price, term, or regulatory status.
- Completion of a verification currently marked pending, which would replace a gap with a stated fact.
What do the technical terms on this page mean?
Definitions for the 9 technical terms this page uses, including DOI, PMID, compounded, endpoint — in the specific sense used above.
| DOI | Digital Object Identifier. A persistent link to a specific published article that continues to resolve even if the journal reorganises its website. |
|---|---|
| PMID | PubMed Identifier. A number that locates the peer-reviewed publication of a study in the PubMed database. |
| compounded | Prepared by a pharmacy rather than manufactured under an approved application. Compounded tirzepatide is not FDA approved and has not been evaluated in any randomised trial. |
| endpoint | The outcome a trial is designed to measure. A primary endpoint is specified before the trial begins; secondary and exploratory endpoints carry progressively weaker inferential weight. |
| excipient | An inactive ingredient in a formulation. Excipients affect stability, tolerability, and injection-site reactions, and can differ between compounded preparations and the approved product. |
| maximum tolerated dose | The highest dose an individual can take without unacceptable adverse effects. Trials frequently escalate to this point rather than to a fixed dose, which means participants in one arm may be taking different amounts. |
| open-label | A trial in which participants and investigators know which treatment is being given. SURMOUNT-5 was open-label, a genuine limitation, though weight is an objective measurement less vulnerable to expectation than a self-reported outcome. |
| placebo | An inactive comparator given so that the effect of the drug can be separated from the effect of being in a trial. Placebo groups in the SURMOUNT trials still lost some weight, which is why the placebo-subtracted difference matters more than the raw figure. |
| randomised withdrawal | A design in which everyone first receives the active drug, and only those who respond are then randomised to continue or stop. SURMOUNT-4 used this design, which is why its regain finding cannot be explained away by differences between groups. |
Frequently asked questions
What did SURMOUNT-4 show?
Continuing tirzepatide maintained weight loss over 52 weeks, while switching to placebo produced substantial regain.
Does stopping mean regaining everything?
Not necessarily within the trial period, and results vary between people — but average regain was substantial and the trend was consistent.
Why does this matter for cost?
Because maintaining results generally means continued treatment, so the relevant figure is sustained monthly cost over years rather than an introductory rate.
Is there an approved way to taper off?
Stopping is an individualised clinical decision. This site does not publish tapering schedules; discuss any change with your prescriber.
Sources for this page
Change history
| Date | Change |
|---|---|
| 2026-07-22 | Page published with current dataset snapshot. |
Dates change only for substantive edits, never for cosmetic changes. Corrections: corrections policy.