What Happens When You Stop GLP-1 Therapy
The withdrawal trials and real-world cohorts now quantify what comes back when GLP-1 therapy stops, and the honest answer mixes measured regain with questions the published record has not yet asked.
The withdrawal trials and real-world cohorts now quantify what comes back when GLP-1 therapy stops, and the honest answer mixes measured regain with questions the published record has not yet asked.
I · The weight comes back, and the trials measured how much
The average participant in the STEP 1 extension regained about two-thirds of the weight they had lost within a year of stopping semaglutide, which makes the drug’s effect time-limited rather than permanent.
The question of what happens after the last injection now has a controlled answer, because John Wilding and his colleagues ran an extension of the STEP 1 trial that followed 327 participants for one year after the 68-week main phase ended, and the arithmetic of the return is straightforward. The semaglutide group had fallen to −17.3% (SD 9.3) at week 68, and by week 120 their net position was −5.6% (SD 8.9), so the 11.6 percentage points (SD 7.7) they regained equaled roughly two-thirds of the ground they had covered, whereas the placebo group sat at −2.0% (SD 6.1) over the same stretch of the main trial.1
In the STEP 1 trial extension, 327 participants followed for one year after stopping semaglutide regained an average of two-thirds of the weight they had lost, moving from −17.3% at week 68 to a net −5.6% at week 120.
Wilding et al., Diabetes, Obesity and Metabolism, 2022
The context that belongs beside that number is that the structured lifestyle intervention ended at the same week 68, so the regain reflects the withdrawal of both the drug and the coaching support that ran alongside it, which makes the two-thirds figure a description of the whole program stopping, with the pharmacology and the structured support coming off the table together. The same extension recorded cardiometabolic improvements moving back toward baseline after the injections stopped, and the contrast with continuous use shows up in STEP 5, where Garvey and colleagues kept participants on semaglutide for two years and measured −15.2% at week 104.23
The extension analyzed 327 of the 1,961 original STEP 1 participants, a non-random subset selected by site country and recruitment volume, and the authors treated the analyses as exploratory, so the two-thirds figure is a group average with an explicit sampling caveat attached.
The practical takeaway is that the semaglutide weight-loss record is time-limited, so anyone reading −17.3% needs the second half of the curve alongside it, because that is the half the extension trial actually measured.
II · Most people stop within the first year
Real-world cohort studies report that most people are no longer on the drug a year after starting it, with the exact rate depending heavily on whether they have diabetes.
The controlled trials describe what happens when therapy ends, but the real-world record shows how often that ending arrives on its own, and the two biggest cohorts now agree on the shape of the curve. Pablo Rodriguez and colleagues analyzed a Truveta electronic health record cohort of 125,474 new users of dual-labeled GLP-1 receptor agonists, and they found that 64.8% (95% CI 64.4 to 65.2) of people without type 2 diabetes discontinued within 12 months, whereas 46.5% (95% CI 46.2 to 46.9) of those with diabetes did the same, with 1-year reinitiation running at 36.3% versus 47.3%.4
A separate Evernorth claims study led by Do and colleagues, covering 195,915 new users, arrived at a similar shape with different cut points, since 50.3% of obesity-only users had stopped by month 12, compared with 35.8% of diabetes-only users and 34.2% of those with both conditions. The same population pattern shows up in persistence data from Uzoigwe and colleagues, since 67.0% of diabetes patients on semaglutide were still filling prescriptions at 360 days.56
In a Truveta EHR cohort of 125,474 new GLP-1 users, 64.8% of people without diabetes and 46.5% with diabetes discontinued within 12 months, and 1-year reinitiation ran at 36.3% versus 47.3%.
Rodriguez et al., JAMA Network Open, 2025
A national survey of people who did stop, published by DiStefano and colleagues, adds texture to the cohort numbers, since 79.7% of discontinuers had stopped within 12 months of starting and 54.6% within 6 months, with cost (36.1%) and side effects (33.3%) leading the reasons given ahead of insurance coverage (28.2%). The survey denominator matters, because it counts only people who discontinued, so its percentages describe that group, and that is the definitional care the cohort numbers also demand.7
The two big cohorts disagree on the denominator because one reads electronic health records and the other reads insurance claims, and the gap thresholds used to define a discontinuation range from 45 to 90 days, so a 50% to 70% stop-within-a-year range is the honest summary rather than any one point estimate.
The open question these cohorts leave is why the stopping happens, since the survey points to cost and side effects while the Truveta cohort found people without diabetes drifting away faster, and answering that question is what would turn a statistic into something usable.
III · The appetite returns when the drug leaves
Appetite suppression is present while the drug is being taken, and the return of hunger after stopping has now been measured in a prospective cohort, alongside a rebound in weight.
The mechanism that quiets hunger is the same one that comes back online, and a small prospective study has now documented the rebound in numbers, without leaning on anecdote. A 2026 study led by Wang and colleagues followed 28 women with obesity through 36 weeks of semaglutide and a 12-week withdrawal period, and it found that 78.5% reported an appetite rebound, defined as a ≥30% increase in hunger scores with a sustained rise of at least 300 kcal per day, while 71.4% regained weight, an average of +5.1 kg (SD 1.6), after losing −16.9 kg (SD 4.8) on drug.8
A 2026 prospective study of 28 women found that 78.5% reported a rebound in hunger within 12 weeks of stopping semaglutide, alongside weight regain in 71.4% of the cohort.
Wang et al., Diabetes, Obesity and Metabolism, 2026
The study is small and single-arm, with a female-only cohort, so it counts as hypothesis-generating in scope, but its direction matches the animal work of Shah and Ayala, in which obese mice regained weight quickly after semaglutide cessation as food intake rose and meals grew larger and more frequent, which points at the same appetite axis.9
The word people use for the return of persistent food-related thoughts is food noise, a patient-coined term that entered public discourse through appetite changes reported while people were taking GLP-1 drugs, and a 2026 commentary in Appetite by Brewis and colleagues notes that limited evidence exists regarding its underlying mechanisms and that no validated measure exists for it. That distinction matters, because the term’s popularity has run ahead of its science, so this article treats food noise as reported patient experience, and the commentary itself draws that line.10
Food noise is real to the people who experience it, but it is a patient-coined term, and the Appetite commentary flags that it lacks a validated instrument and has no mechanism study, so its return after stopping is documented anecdotally rather than measured.
The takeaway here is that the appetite rebound has a verified number behind it for the first time, and that number sits inside a small single-arm study, so the honest summary is that hunger came back for most of the women in that cohort and that the size of the effect beyond it is unquantified.
IV · The metabolism question has an on-drug answer only
The on-drug record shows no disproportionate metabolic adaptation from these agents, and no human study has measured resting metabolic rate after discontinuation, so the off-drug metabolism story is currently unwritten.
The fear that GLP-1 therapy slows the metabolism in some lasting way collides with a narrower record than the fear assumes, because the human data that exist were all collected while the drug was still on board. Any weight loss lowers absolute resting metabolic rate simply because the body is smaller, so the real question is whether GLP-1-based loss adds a disproportionate adaptation on top of that. A phase 1 trial led by Eric Ravussin and colleagues, published in Cell Metabolism in 2025, found that tirzepatide appeared to have no impact on metabolic adaptation relative to placebo, while increasing fat oxidation and reducing appetite and calorie intake.11
Tirzepatide appeared to have no impact on metabolic adaptation in people with obesity, while increasing fat oxidation and reducing calorie intake.
Ravussin et al., Cell Metabolism, 2025
Support comes from the SEMALEAN cohort, in which Alissou and colleagues measured 106 people on semaglutide with DXA and indirect calorimetry, and resting energy expenditure normalized to lean mass rose between months 7 and 12 even as weight fell 13% over the year, which is the opposite direction from a slowing metabolism. The minipig data of Bredum and colleagues point the same way, since diet restriction produced significant metabolic adaptation and lower energy expenditure than semaglutide at matched body weight, whereas the drug preserved energy expenditure.1213
The gap sits exactly where the public conversation is loudest, because no published human study has measured resting metabolic rate after the drug is discontinued, so claims that adaptation lingers or fades once injections stop are not yet evidence-based. The familiar metabolic adaptation literature, including the Biggest Loser follow-up cohort, derives from diet-based weight loss, so transferring that pattern to GLP-1 pharmacology without attribution would overstate the record, and keeping the two literatures separate is the rigorous move.14
The persistent-metabolic-adaptation claims that circulate online come from calorie-restriction cohorts, and no study of that design has been run on GLP-1 withdrawal, so the honest position is that on-drug metabolism looks unremarkable and off-drug metabolism is unmeasured.
The open question is a simple one for trial designers: measure resting metabolic rate after the drug is withdrawn, because that single measurement would settle a debate that currently runs on inference.
V · The documented muscle loss and the unmeasured recovery
Most of the weight lost on these drugs is fat, and no published trial has measured whether the lean portion returns after stopping, so the recovery half of the story is currently unwritten.
The lean tissue lost while taking these drugs is the question every withdrawal trial walked past, since the studies that measured weight regain never scanned body composition, and the on-drug half of the record is the only half that exists. Real-world body-composition studies document the picture, since Angelopoulos and colleagues measured 51 adults on tirzepatide for 12 weeks and found −8.75 kg of body weight made up of −6.87 kg of fat and −1.73 kg of soft lean tissue, so roughly 78% of the loss came from fat. The CRAVE study led by Babazadeh and colleagues estimated, using bioimpedance in 28 people, that about one-quarter of the weight lost on these drugs is skeletal muscle.1516
The SEMALEAN cohort adds the timing detail, with lean mass falling about 3 kg by month 7 on semaglutide before stabilizing, which suggests the lean loss front-loads while the fat loss continues. The off-drug half is a vacuum, because neither the STEP 1 extension nor SURMOUNT-4 scanned body composition after discontinuation, and no cohort study found in the literature measured lean mass after GLP-1 therapy stopped, so any statement that lost muscle does or does not come back would go beyond published evidence.17
These estimates come from bioimpedance in small real-world cohorts, so they are approximations, one tier below DXA-grade precision, and the one-fifth to one-quarter range for the lean share of loss is the defensible summary across the sources.
The honest summary is that the composition question remains open, and the gap matters because lean tissue is metabolically and functionally meaningful, so the trial that measures it after withdrawal would close a major hole in this literature.
VI · SURMOUNT-4 and the controlled test of stopping tirzepatide
Participants switched to placebo regained 14.0% of body weight over 52 weeks, while those who continued on tirzepatide lost 5.5% more, which makes the difference between stopping and continuing a −19.4 percentage point divergence.
The STEP program answered the stopping question for semaglutide, and SURMOUNT-4 ran the same experiment for tirzepatide with a randomized control that stayed on the drug, so the divergence between the two arms is direct and large. Louis Aronne and his colleagues randomized 670 participants, after a 36-week open-label lead-in during which mean weight reduction reached 20.9%, to continue tirzepatide or switch to placebo for 52 weeks. From week 36 to week 88 the continued group lost a further 5.5% whereas the placebo group regained 14.0% of body weight, a between-group difference of −19.4 percentage points (95% CI −21.2 to −17.7, P<.001).18
In SURMOUNT-4, participants who had lost 20.9% over the lead-in and were switched to placebo regained 14.0% of body weight over the next 52 weeks, ending at −9.9% net, whereas those who continued lost 5.5% more to finish at −25.3%.
Aronne et al., JAMA, 2024
The maintenance statistic tells the same story from the other side, since 89.5% of those who stayed on tirzepatide kept at least 80% of their lead-in weight loss, compared with 16.6% of those switched to placebo, and the week 0 to week 88 contrast stood at −25.3% versus −9.9%. The trial that most closely mirrors this design, STEP 4 led by Domenica Rubino, used a 20-week semaglutide run-in with a −10.6% starting loss before randomizing participants, and the placebo switch then regained +6.9% from week 20 to week 68 while continued semaglutide lost another 7.9%, a −14.8 percentage point difference (95% CI −16.0 to −13.5).19
The SURMOUNT-4 lead-in was open-label with two dose levels, 10 mg and 15 mg, so the 20.9% starting point describes a selected population rather than an average beginner, and the placebo-arm regain of 14.0% counts from week 36, the point where randomization began.
The takeaway is that these trials measure group means with tight confidence intervals, so they describe the group, and individual paths vary widely around the mean, and the consistent direction across two different drugs is the durable finding.
VII · Retatrutide and the edge of the published record
The 12 mg arm of the phase 2 retatrutide trial produced a mean 24.2% weight loss at 48 weeks, and the published record stops there, because the extension of that trial has no peer-reviewed publication.
The triple-agonist question pushes the same stopping math into newer territory, and the published record draws a sharp line at week 48, which makes this the place where the evidence and its boundaries are most visible. Ania Jastreboff and her colleagues ran the phase 2 obesity trial of retatrutide, a triple-hormone-receptor agonist, in 338 participants, and the 12 mg dose delivered a least-squares mean weight change of −24.2% at 48 weeks, with the 8 mg arm at −22.8% and the 4 mg arm at −17.1%, while the placebo arm moved −2.1%.20
In the phase 2 obesity trial of the triple agonist retatrutide, the 12 mg dose produced a mean 24.2% weight loss at 48 weeks, and the post-48-week extension of the trial has no peer-reviewed publication on the record.
Jastreboff et al., New England Journal of Medicine, 2023
Within the 12 mg group, 100% of participants lost at least 5% of body weight at 48 weeks, with 93% crossing the 10% threshold and 83% crossing 15%, which gives a sense of the consistency of the response in addition to its mean. The stopping question for retatrutide remains a matter of the record’s boundaries, because no peer-reviewed publication of the 72-week extension exists in the indexed literature as of August 2026, so the ~26% to 27% post-48-week figures circulating online are conference-level material without a citable record. The last verified number on the shelf is the 22.1% placebo-subtracted 48-week effect of the 12 mg dose reported in the systematic review by Moiz and colleagues.21
Retatrutide is not FDA-approved, so this article reports its trial data as trial data, and the editorial posture of this site treats anything past 48 weeks as unpublished, without repeating conference numbers as fact.
The open question is when the extension reaches a citable form, because a published 72-week extension record would extend the class-level picture that STEP 1 and SURMOUNT-4 have already drawn.
VIII · Where the evidence stops and the open questions begin
The verified record covers weight regain and discontinuation alongside appetite rebound, and it stops before muscle recovery and off-drug metabolism can be answered, with follow-up capped at two years, so the honest summary is a list of measured averages beside a list of unmeasured questions.
Pulled together, the withdrawal literature answers the weight question with group averages and leaves the composition and metabolism questions open, and the shape of that answer is worth stating in one place. The verified record assembles into a short set of group averages, since the STEP 1 extension measured two-thirds of lost weight returning within a year, SURMOUNT-4 measured a 14.0% placebo-arm regain against a 20.9% lead-in loss, and the Truveta and Evernorth cohorts put 12-month discontinuation at 64.8% versus 46.5% and at 50.3%. The 28-woman prospective cohort documented a hunger rebound in 78.5% of participants, and the phase 2 retatrutide trial anchors the newest class at 24.2% at 48 weeks with its extension still unpublished.14581820
The record’s absences are equally precise, since no withdrawal trial scanned body composition after discontinuation and no human study has measured resting metabolic rate off drug, while follow-up stops at two years, which leaves muscle recovery and metabolic persistence open, and the definitional sensitivity of the real-world rates means a range is the honest framing where a single number would be false precision.22
Regain figures are population statistics with standard deviations and confidence intervals in the original papers, so individual outcomes vary widely around them, and this article avoids translating any average into a personal prediction, which is the line between reporting and advice.
The final takeaway is that stopping a GLP-1 receptor agonist brings back part of the measured effect while the rest of the story waits for trials that measure the right things, and any decision about starting or stopping therapy belongs with a clinician, because this article is editorial reference material rather than medical instruction.
- Wilding, J.P.H. et al. “Weight regain and cardiometabolic effects after withdrawal of semaglutide: The STEP 1 trial extension.” Diabetes, Obesity and Metabolism, 2022;24(8):1553-1564. PMID 35441470. 327 participants regained an average of two-thirds of lost weight within one year of stopping.
- Wilding, J.P.H. et al. “Once-Weekly Semaglutide in Adults with Overweight or Obesity.” New England Journal of Medicine, 2021;384(11):989-1002. PMID 33567185. STEP 1 main trial: −14.9% versus −2.4% at week 68.
- Garvey, W.T. et al. Nature Medicine, 2022;28(10):2083-2091. PMID 36216945. STEP 5: −15.2% versus −2.6% at 104 weeks of continuous semaglutide.
- Rodriguez, P.J. et al. “Discontinuation and Reinitiation of Dual-Labeled GLP-1 Receptor Agonists Among US Adults With Overweight or Obesity.” JAMA Network Open, 2025;8(1):e2457349. PMID 39888616. Truveta EHR cohort of 125,474 new users: 64.8% versus 46.5% 12-month discontinuation by diabetes status.
- Do, D. et al. “GLP-1 Receptor Agonist Discontinuation Among Patients With Obesity and/or Type 2 Diabetes.” JAMA Network Open, 2024;7(5):e2413172. PMID 38787563. Evernorth claims cohort of 195,915 new users: 50.3% obesity-only discontinuation at 12 months.
- Uzoigwe, C. et al. Diabetes Therapy, 2021;12(5):1475-1489. PMID 33837922. Optum claims: 67.0% 360-day semaglutide persistence in type 2 diabetes.
- DiStefano, M.J. et al. “Discontinuation of glucagon-like peptide-1 receptor agonists among US adults: A national survey.” Journal of Managed Care & Specialty Pharmacy, 2026;32(8):903-910. PMID 42504813. 79.7% of discontinuers stopped within 12 months of starting.
- Wang, N. et al. “Post-semaglutide weight regain in females with obesity: Associations with gut microbiota, bile acid metabolism, and central nervous system.” Diabetes, Obesity and Metabolism, 2026;28(5):3903-3916. PMID 41705640. 78.5% appetite rebound and 71.4% weight regain in 28 women after withdrawal.
- Shah, H. and Ayala, J.E. Diabetes, 2026;75(2):288-300. PMID 41329075. Obese mice regained weight rapidly after semaglutide cessation via increased food intake, meal size, and frequency.
- Brewis, A. et al. “Food noise: Conceptual, methodological, and ethical considerations.” Appetite, 2026;226:108700. PMID 42392316. Commentary: limited evidence on food noise mechanisms; no validated measure.
- Ravussin, E. et al. “Tirzepatide did not impact metabolic adaptation in people with obesity, but increased fat oxidation.” Cell Metabolism, 2025;37(5):1060-1074.e4. PMID 40203836. Phase 1 trial: no impact on metabolic adaptation versus placebo.
- Alissou, M. et al. “Impact of Semaglutide on fat mass, lean mass and muscle function in patients with obesity: The SEMALEAN study.” Diabetes, Obesity and Metabolism, 2026;28(1):112-121. PMID 41068996. REE normalized to lean mass rose from month 7 to 12; weight −13% at 12 months; lean −3 kg at month 7.
- Bredum, S.K. et al. American Journal of Physiology, Endocrinology and Metabolism, 2026;330(4):E401-E410. PMID 41671030. Obese minipigs: diet restriction caused significant metabolic adaptation; semaglutide preserved energy expenditure.
- Metabolic adaptation after weight loss is documented in diet-based cohorts, including the Biggest Loser follow-up study; that literature does not transfer to GLP-1 pharmacology without attribution, and no human study has measured resting metabolic rate after GLP-1 discontinuation.
- Angelopoulos, N. et al. Nutrition, 2026;150:113286. PMID 42275945. 51 adults on tirzepatide for 12 weeks: −8.75 kg total, −6.87 kg fat, −1.73 kg soft lean, about 78% of loss from fat.
- Babazadeh, D. et al. “The CRAVE study.” Obesity Pillars, 2026;19:100292. PMID 42440974. About one-quarter of weight loss estimated as skeletal muscle by bioimpedance in 28 people.
- No published trial (STEP 1 extension, STEP 4, SURMOUNT-4, or any located cohort) measured body composition after GLP-1 discontinuation; SEMALEAN lean-mass details as in fn-12.
- Aronne, L.J. 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. PMID 38078870. Placebo arm +14.0% regain versus continued −5.5%; 89.5% versus 16.6% maintained at least 80% of lead-in loss.
- Rubino, D. 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. PMID 33755728. Placebo switch +6.9% regain from week 20 to 68.
- Jastreboff, A.M. et al. “Triple-Hormone-Receptor Agonist Retatrutide for Obesity: A Phase 2 Trial.” New England Journal of Medicine, 2023;389(6):514-526. PMID 37366315. 12 mg: −24.2% at 48 weeks; no peer-reviewed publication of the 72-week extension located.
- Moiz, A. et al. Annals of Internal Medicine, 2025;178(2):199-217. PMID 39761578. Systematic review: 22.1% placebo-subtracted 48-week effect for retatrutide 12 mg.
- Research gaps as of verification 2026-08-19: no body-composition measurement after discontinuation, no off-drug resting metabolic rate data in humans, no >2-year off-drug follow-up, and definition-sensitive real-world discontinuation rates (45-90 day gap thresholds).