Stopping & switching

What Happens When You Stop Taking a GLP-1: The Evidence

Randomized trials, real-world databases and meta-analyses now give a fairly clear picture of what happens to weight, blood sugar, blood pressure and body composition after a GLP-1 stops - and where the numbers disagree.

Last verified September 14, 2026·15 sources cited·Ozempic Zepbound

Semaglutides.org is for information only and is not medical advice. Always talk to a licensed healthcare provider about your own care. Some links to telehealth services are affiliate links, labeled where they appear.

This is one of the few questions in the GLP-1 world with real randomized answers. Drug companies ran trials specifically to find out what happens when the drug stops, and researchers have since pooled them, modeled them and checked them against millions of real patient records.

The short version: most of the weight comes back, gradually, over about a year, and most of the health improvements fade with it. But the averages hide enormous variation, the real-world numbers look better than the trial numbers for reasons worth understanding, and the newest research has shifted the question from “stop or don’t stop” to “stop, or use less.”

This article covers the whole class - semaglutide (Ozempic, Wegovy), tirzepatide (Mounjaro, Zepbound), liraglutide (Victoza, Saxenda) and the newer agents. Where it describes doses, interruptions or reinitiation, it is summarizing the FDA-approved prescribing information and Instructions for Use rather than telling you what to do. Nothing here is medical advice. If you are thinking about stopping, that is a conversation for the person who prescribed it, and you should follow your own prescription in the meantime.

How much weight comes back after stopping?

Start with the single best estimate. A 2026 meta-regression published in eClinicalMedicine pooled 48 studies, used six randomized trials with 3,236 participants to fit a curve, and found that one year after stopping, 60% of the weight lost during treatment had been regained [1]. Extrapolating past the end of the data, the model projected regain would plateau at 75.3% of the loss - meaning some benefit persists at the population level, though the authors note that prediction runs beyond what anyone has measured.

The individual trials behind that number:

STEP 1 extension. A subset of 327 people from the STEP 1 semaglutide trial were followed for a year after both the drug and the structured lifestyle program stopped at week 68. They had lost 17.3% on semaglutide 2.4 mg. By week 120 they had regained 11.6 percentage points, ending 5.6% below where they started. At the end of treatment, 86.4% had lost at least 5% of body weight; a year later, 48.2% still had [2].

STEP 4. This one randomized people who had already titrated up to semaglutide 2.4 mg either to continue or to switch to placebo, with lifestyle support for everyone. Over the next 48 weeks, the continuers lost another 7.9% while the placebo group gained 6.9% - a gap of 14.8 percentage points. Waist circumference, blood pressure and physical functioning scores all split the same way [3].

SURMOUNT-4. The tirzepatide equivalent. After 36 weeks of open-label tirzepatide at the maximum tolerated dose (10 or 15 mg), people had lost 20.9%. Randomized to continue or to placebo, the continuers lost a further 5.5% over 52 weeks while the placebo group regained 14.0%. At the end, 89.5% of continuers had held at least 80% of their loss, versus 16.6% of the placebo group [4]. This trial sits in the Zepbound prescribing information as Study 4 [5].

A 2026 PeerJ meta-analysis of six studies and 8,993 patients put the gap between stopping and continuing at 17.90% of body weight, and found rebound after tirzepatide larger than after semaglutide [6] - which makes sense, since people generally lose more on tirzepatide and so have more to regain.

How fast does the weight come back?

Not evenly. The regain curve is front-loaded.

The eClinicalMedicine model describes an exponential recovery with a rate half-life of about 23 weeks [1]. In plain terms: the fastest regain happens in the first few months, then the curve flattens. A 2026 review in Diabetes, Obesity and Metabolism focused specifically on this pattern, calling early regain “front-loaded” and arguing that the first months after stopping are the window that sets the year [7].

Why the rush? The reviewers point to several overlapping mechanisms: appetite returns once pharmacological GLP-1 signaling is withdrawn, hunger hormones including ghrelin rise, and the body’s weight-defense machinery re-engages. An editorial in Cureus makes the same point more bluntly: this is not treatment failure, it is disease recurrence, because obesity is a chronic relapsing condition and the drug was suppressing it rather than curing it [8].

Semaglutide’s half-life is about one week, so drug levels take roughly five weeks to fall to near-nothing. That is why nobody describes a sudden crash when a dose is skipped, and why appetite returns over weeks rather than days.

What happens to blood sugar?

This is the part consumer coverage underplays, and it matters most for people with type 2 diabetes.

A meta-analysis of 18 randomized trials with 3,771 participants, published in eClinicalMedicine in December 2025, quantified it. After discontinuation, HbA1c rose 0.25% in people with obesity and 0.65% in people with type 2 diabetes [9]. Fasting glucose, waist circumference, BMI and systolic blood pressure all deteriorated significantly in the obesity group.

The real-world version is starker. When a global shortage forced 69 people with type 2 diabetes off dulaglutide, mean HbA1c went from 7.0% to 8.1% within three months and fasting glucose from 129 to 156 mg/dL. Switching them to a DPP-4 inhibitor or an SGLT2 inhibitor did not make up the difference [10].

In the STEP 1 extension, some people whose blood sugar had normalized on semaglutide reverted to prediabetes after stopping [2].

What about blood pressure, cholesterol and heart risk?

They track the weight. That is the cleanest finding in this whole area.

A post hoc analysis of SURMOUNT-4, published in JAMA Internal Medicine in November 2025, took the 308 participants who had lost at least 10% and been randomized to placebo, and sorted them by how much they regained. The results line up almost perfectly:

Weight regained (as % of what was lost)	Waist change	Systolic BP change
Less than 25%	+0.8 cm	+6.8 mm Hg
25% to under 50%	+5.4 cm	+7.3 mm Hg
50% to under 75%	+10.1 cm	+9.6 mm Hg
75% or more	+14.7 cm	+10.4 mm Hg

Source: JAMA Internal Medicine post hoc analysis [11].

The same pattern showed in the STEP 1 extension, where most cardiometabolic variables reverted toward baseline as the weight returned [2]. The implication is practical: partial regain is not all-or-nothing. Holding some of the loss appears to hold some of the benefit.

There is one open question nobody has answered. A September 2026 review in Nature Reviews Endocrinology raises the possibility that repeated cycles of starting, stopping and restarting produce swings in weight and HbA1c, both of which are established cardiovascular risk factors, and notes that incretin drugs do not seem to have direct plaque-stabilizing effects. The authors are careful: hard-outcome data in people who discontinue barely exist [12].

Does the weight come back as fat or muscle?

Fat, mostly - and that is the part that worries obesity specialists.

A 2026 narrative review in Healthcare covering exercise during and after incretin therapy reports that lean mass losses averaging 6 to 7 kg and reductions in bone mineral density occur during active treatment, and that regain afterward is fat-preferential. The review describes this combination as creating “a metabolically unfavorable state at cessation” [13]. A Cureus editorial frames the same finding as a risk of sarcopenic obesity: a person can return to roughly the weight they started at, with less muscle and more fat than before [8].

Treat the lean-mass figures with more caution than the weight figures. Absolute and percentage lean-mass losses reported in this literature vary widely and depend heavily on how body composition was measured - DXA, bioimpedance and MRI do not agree - and much of the underlying data comes from small substudies rather than the main trial populations. The direction of the finding is consistent; the magnitude is contested. And the long-term body-composition consequences of repeated stopping and restarting have not been measured directly in humans at all: that concern rests on mechanistic reasoning plus animal data [7][13].

Interestingly, the reduced-frequency case series discussed later in this series found the opposite pattern when people lowered their dosing rather than stopping - fat continued to fall while skeletal muscle mass stabilized [14]. That is a small, uncontrolled study, but it points at the same idea from the other direction: the body-composition problem is about the transition, not just about the drug.

Is there anything that actually helps hold the weight?

There is one randomized trial that speaks to this directly, and it used liraglutide rather than the newer drugs.

In the Danish S-LiTE trial, adults with obesity lost 13.1 kg on a low-calorie diet and were then randomized for a year to supervised exercise, liraglutide 3.0 mg, both, or placebo. Researchers then brought them back a year after everything stopped. At that point, the former combination group was 5.1 kg lighter and 2.3 percentage points lower in body-fat percentage than the former liraglutide-alone group. The odds of still holding at least 10% of the original weight loss were 7.2 times higher in the combination group and 3.7 times higher in the exercise group than in placebo. Regain in the year after treatment ended was 6.0 kg larger after stopping liraglutide than after stopping supervised exercise [15].

The caveats are real: this was liraglutide, not semaglutide or tirzepatide, and the exercise arm was a supervised program most people cannot replicate. The 2026 Healthcare review notes that this single trial is essentially the entire controlled evidence base for exercise at cessation, and that no randomized trial has directly tested exercise at the point a GLP-1 stops [13].

Still, it is the best signal available, and it points the same way the physiology does.

Why do real-world numbers look so much better than trial numbers?

Because they are measuring different things.

An Epic Research analysis of the Cosmos database looked at 188,722 US patients who stopped a GLP-1 after at least 90 days on it and had lost at least five pounds. Two years later, 56.1% of former semaglutide users, 55.2% of former tirzepatide users and 51.9% of former liraglutide users had maintained their loss or lost more. Complete regain occurred in 23%, 21% and 27% respectively. Trajectories largely stabilized after 12 months [16].

A Cleveland Clinic cohort of 7,938 patients found mean weight change one year after stopping of only +0.5% in those treated for obesity [17]. That sounds like the trials are wrong. They are not. In that same cohort, 19.6% restarted the original drug and 35.2% received some other obesity treatment, including another medication (27.4%), a lifestyle visit (13.7%) or bariatric surgery (0.6%) [17]. The flat average is partly an average of people who did something else.

The Epic analysis has similar limits: it does not exclude people who restarted or switched, and everyone included had already lost weight, which selects for responders. The authors of the Cleveland Clinic study explicitly warn about “considerable individual-level variability” hiding behind their mean [17].

So read the two literatures for what each is good at. The randomized trials tell you what happens if you stop and do nothing else. The real-world data tells you what actually happens to a large population, most of whom do something else.

How often do people stop in the first place?

Very often, though less often than a few years ago.

A Danish national cohort published in JAMA Network Open in August 2026 followed all 77,310 Danish adults without diabetes who started semaglutide for weight loss: 18% had stopped by three months and 52% by one year, with the highest dropout among adults aged 18 to 30 [18].

In the US, a Prime Therapeutics claims analysis of 33,607 commercially insured members without diabetes found one-year persistence rising from 33.2% for people who started in 2021 to 60.9% for those starting in the first half of 2024. Tirzepatide persistence ran 64.0% and 64.8% in 2023 and early 2024 [19]. The authors credit shortage resolution, better dose escalation and side-effect management, and lifestyle support.

Money and side effects drive most of it. A JAMA Network Open cohort found that moderate or severe gastrointestinal events raised the odds of stopping, while household income above $80,000 lowered them [20]. In an Italian real-world cohort, 10.4% stopped for non-severe gastrointestinal events and 5.7% for cost [21].

What do professional bodies say about stopping?

In April 2026, The Obesity Society, the Obesity Medicine Association and the Obesity Action Coalition published a joint expert guidance statement using the GRADE method. Among its conclusions: continuing obesity medications during weight maintenance received a strong recommendation [22]. The framing throughout is that obesity is a chronic, often progressive disease requiring long-term care.

That is guidance about whether to continue, not about what dose to continue at - a distinction that matters more now than it did a year ago, because two 2026 trials showed that maintenance can look different from the weight-loss phase. SURMOUNT-MAINTAIN found that dropping tirzepatide to 5 mg held more weight loss than placebo, though less than staying at the maximum tolerated dose [23]. ATTAIN-MAINTAIN found that switching from an injection to a daily oral GLP-1 held most of the loss over 52 weeks [24].

Neither is a reason to change anything on your own. Both are reasons to ask a different question at your next appointment than “should I stop?”

What about tapering?

No completed randomized trial has tested it. Every withdrawal trial in this class - the STEP 1 extension, STEP 4, SURMOUNT-4, the SURMOUNT-CN follow-up - either stopped the drug outright or swapped it for placebo [2][3][4][25]. The FDA labels contain no tapering instruction.

The 2026 Diabetes, Obesity and Metabolism review that looked hardest at this says it directly: whether structured tapering strategies can successfully facilitate discontinuation “remains unknown and requires prospective evaluation” [7]. Clinicians who taper are extrapolating from pharmacology, and the honest ones say so.

That gap is finally being filled. The REST trial (Recovery Effects after Semaglutide Termination), run by a Mount Sinai Hospital team in Toronto, is the first study designed specifically to compare gradual dose reduction against stopping outright. Its protocol was published in PLoS One in July 2026, and the trial is recruiting with completion scheduled for 2029 [28][29]. It will measure weight, cardiometabolic profile and the hormones that govern energy balance in both groups. Until it reports, “taper or stop?” is a question your prescriber answers from judgment, not from data.

What the labels do address is interruption. The current Wegovy prescribing information says that if two or more consecutive doses are missed, reinitiate dosage escalation at a lower dosage to reduce the risk of gastrointestinal adverse reactions [26]. There is a published case report of a woman who had tolerated semaglutide for a year, paused five weeks, restarted at her old 2.4 mg dose and was hospitalized four days with severe vomiting and electrolyte abnormalities [27]. Gastrointestinal tolerance is something you build, and it fades.

What does this all mean for me?

Five things the evidence supports:

Stopping usually means regaining, around 60% of the loss within a year on average, front-loaded in the first months.
The health benefits track the weight, not the history of having taken the drug. Partial regain preserves partial benefit.
For people with type 2 diabetes, stopping is a blood sugar event, not only a weight event.
Averages hide huge variation. Roughly half of real-world patients hold some of their loss at two years, and roughly a quarter regain all of it.
No completed trial has tested a taper, and nobody has published a validated protocol for coming off. That is a genuine gap, not a secret - and the REST trial, which started recruiting in 2026, is the first attempt to close it.

If cost, side effects or a coverage change is pushing you toward stopping, those are all solvable problems worth raising with a prescriber before the last pen runs out. Lower approved maintenance doses, a different product, a formulary exception and manufacturer self-pay programs all exist. Stopping is one option among several, and it is the one with the most data behind what happens next.

Sources

Questions people ask

How much weight do people regain after stopping a GLP-1?

A 2026 meta-regression of six randomized trials found that 60% of the weight lost during treatment was regained one year after stopping, with the curve projected to plateau around 75%. Individual trials vary: the STEP 1 extension found about two-thirds regained in a year, and SURMOUNT-4 found a 14.0% regain over 52 weeks off tirzepatide.

How fast does the weight come back?

Regain is front-loaded. The meta-regression describes an exponential curve with a half-life of about 23 weeks, meaning the fastest regain happens in the first few months and then slows. A 2026 review of the mechanisms calls this pattern front-loaded and points to the first months as the period that shapes the year.

Is there a withdrawal syndrome?

The withdrawal trials did not describe one. What returns is appetite. Semaglutide has a half-life of roughly one week, so drug levels fall over about five weeks and hunger returns gradually rather than overnight. Gastrointestinal side effects generally get better after stopping, which a 2026 meta-analysis confirmed.

What happens to blood sugar after stopping?

It rises. A pooled meta-analysis of 18 randomized trials found HbA1c rose 0.25% in people with obesity and 0.65% in people with type 2 diabetes after discontinuation. In a real-world cohort of 69 people who lost access to dulaglutide during a shortage, average HbA1c went from 7.0% to 8.1% within three months.

Do blood pressure and cholesterol go back up too?

Generally yes, and in proportion to the weight. A post hoc analysis of SURMOUNT-4 sorted people by how much they regained: waist circumference grew 0.8 cm in those who regained the least and 14.7 cm in those who regained the most, with systolic blood pressure rising 6.8 to 10.4 mm Hg across the same groups.

Does everyone regain?

No. An Epic Research analysis of 188,722 US patients found that two years after stopping, 56% of former semaglutide users had maintained their loss or lost more, and 23% had regained all of it. That data is observational and includes people who restarted or switched drugs, so it is not a fair comparison with the trials.

Does weight come back as fat or muscle?

Reviews describe regain as fat-preferential, meaning a person can return to a similar weight with a worse body composition. A 2026 narrative review reports lean mass losses averaging 6 to 7 kg during active treatment in the evidence it surveyed, which is why resistance training and adequate protein come up so often in this conversation.

Should I taper instead of stopping suddenly?

No completed randomized trial has tested a taper. Every withdrawal trial in this class stopped the drug outright or swapped it for placebo, and a 2026 review says plainly that whether structured tapering helps is unknown. The REST trial in Toronto began recruiting in 2026 to compare gradual dose reduction against abrupt stopping, with results due around 2029. Anyone thinking about stopping should plan it with the prescribing healthcare provider rather than on their own.

This article summarizes FDA labeling, published research and company information current as of September 14, 2026. It is not medical advice and does not replace a conversation with your own healthcare provider. How we research and verify.