When GLP-1 medication stops, one specific thing stops with it: the appetite effect the drug was producing. Everything else about a maintenance phase — the eating pattern, the training, the muscle that has been kept or lost — carries on exactly as it was, which is why what has been built during treatment matters more at this point than at any other. The withdrawal trials show weight returning in the months after stopping, though they measured body weight rather than appetite. This article covers what those trials actually measured, what they did not, how fast weight comes back, and what has to do the work the medication was doing.
Maintenance phase: the period after active weight loss ends, when the goal changes from losing more to holding what has been lost. After medication stops, it is also the period when behaviour has to hold the result on its own.
What actually stops when the medication stops?
The drug clears from the body over a matter of weeks, and the appetite effect it was producing stops being produced.
The pharmacology is straightforward. Semaglutide has an elimination half-life of about one week and is present in the circulation for roughly five to seven weeks after a last 2.4 mg dose; tirzepatide's elimination half-life is about five to six days (DailyMed). A half-life describes drug concentration rather than the duration of any downstream physiological effect, so those numbers say when the medicine is gone, not when its consequences finish unwinding.
Here is the part that gets stated more confidently than the evidence allows. No GLP-1 withdrawal trial measured appetite. The STEP 1 extension, STEP 4 and SURMOUNT-4 all measured body weight and cardiometabolic markers; none of them measured hunger, craving, food intake or fullness after treatment ended. So "appetite comes back when you stop" is a reasonable mechanistic inference from a drug leaving the system — it is not a trial finding. We go through that gap in detail in appetite after stopping GLP-1.
What this means practically is that the size of the change a person experiences is not something the literature can predict for them.
What the withdrawal trials actually measured
Body weight, waist and cardiometabolic markers — and the direction is consistent across all of them.
In the STEP 1 extension, participants who had reached a 17.3% mean weight loss by week 68 came off treatment and regained 11.6 percentage points of that loss over the following year — roughly two-thirds of it — leaving them 5.6% below their original starting weight at week 120. Cardiometabolic improvements seen during treatment reverted towards baseline for most variables over the same period (Wilding 2022). The trial reports a mean with a standard deviation of 7.7 percentage points, so the two-thirds figure describes a group average around which individuals varied considerably.
STEP 4 and SURMOUNT-4 used randomised-withdrawal designs, which isolate the medication's contribution more cleanly. In STEP 4, a 20-week run-in produced a mean 10.6% loss; participants were then randomised either to continue semaglutide or to switch to placebo. Over the next 48 weeks the placebo group gained 6.9% of body weight while the continuing group lost a further 7.9% (Rubino 2021). That 6.9% is measured from the already-reduced week-20 weight, not from where they started. SURMOUNT-4 ran the same design with tirzepatide after a 36-week lead-in: 14.0% regained on placebo against a further 5.5% lost on continued treatment, with the placebo group still 9.9% below their original baseline at week 88 (Aronne 2024).
A meta-analysis pooling eight of these trials across 2,372 participants confirms the pattern across the drug class (Berg 2025). It is worth knowing what that review does not contain: it pooled body weight, waist circumference and BMI only. There is no fat-mass or lean-mass data in it, and no withdrawal trial has reported what the regained weight is made of.
Does the regain keep going?
It decelerates. That is the most useful thing the pooled data adds.
A nonlinear meta-regression across six cessation trials and 3,236 participants found about 60% of the weight lost during treatment was back at one year post-cessation, with a modelled half-life of 23.0 weeks — meaning the fastest regain happens early and the curve flattens after that. The model projects a plateau at 75.3% of the weight lost, still below pre-treatment baseline (Budini 2026). That plateau figure is an extrapolation of the regression beyond the 52 weeks of observed data, and the paper says so; treat it as a modelled projection rather than something that was measured.
Two readings follow from this, and both are worth holding at once. The first six months after stopping are where most of the movement happens, which makes them the part of a maintenance phase that deserves the most attention. And the trials do not describe a return to square one — in every dataset above, the groups that came off treatment finished below where they had started.
What is known about people who hold a loss?
More than most people assume, and less securely than it is usually reported.
The main source is the National Weight Control Registry, and its design has to be stated before its findings. It is a self-selected group of volunteers who had already succeeded — entry required having lost at least 13.6 kg and kept it off for five years — so survivorship bias is built into it. It describes what successful maintainers report doing. It cannot show that those behaviours caused the success.
With that in place, what it reports is coherent. Registry members had lost a mean of 30 kg, and reported a low-fat eating pattern with about 24% of energy from fat alongside a high level of physical activity, roughly an hour a day (Klem 1997). Followed for a year, 59% of members maintained their loss and 35% gained; the factors associated with regaining were more recent losses (under two years rather than two years or more), larger losses relative to maximum weight, and higher scores for depression, dietary disinhibition and binge eating (McGuire 1999). A separate registry analysis found that members who reduced how often they weighed themselves regained 4.0 kg over a year, against 1.1 kg among those who weighed themselves more often (Butryn 2007).
The time finding is the one worth internalising. Maintenance appears to get easier with duration: a review estimates that more than 20% of people with overweight or obesity achieve an intentional 10% loss held for at least a year, and notes that once a loss has been held for two to five years the chance of longer-term success rises substantially (Wing and Hill 2001). The hardest period is the one immediately after the loss, which is also the period in which a medication is most often stopped.
On activity specifically, the ACSM position stand associates more than 250 minutes a week with better maintenance on cross-sectional and prospective evidence — while stating in the same document that no well-designed randomised trial exists to judge whether physical activity prevents regain (Donnelly 2009). Both halves belong with the number.
What has to carry the result instead?
The things that do not stop when a prescription does.
Muscle that is still there. Across 114 trials covering 4,184 people, lean mass was statistically unchanged when resistance training accompanied caloric restriction (Lopez 2022). Muscle preserved during the losing phase is muscle that is still working during the maintenance phase, and it is the one asset that cannot be assembled quickly at the end.
An eating pattern that has become ordinary. During treatment, a smaller intake is partly the medication's doing. In maintenance it has to be structural — what is bought, what is cooked, what gets ordered at a hawker centre — because the pharmacological help has stopped.
A measurement habit. Self-weighing is the cheapest early-warning system there is, and the registry association above is the reason to keep it running rather than stopping when active treatment does.
Activity that survives a bad week. The 250-minute figure is an association rather than a proven threshold, but the direction of the evidence is consistent, and a routine that only works in ideal conditions will not survive the first busy month.
Is a gentler exit possible?
Possibly, and it has not been tested.
Every published GLP-1 withdrawal trial used an abrupt switch to placebo. No randomised trial has compared reducing a dose gradually against stopping outright, so there is no evidence base for a taper as a method — not evidence that it fails, but no evidence that it works either. The nearest data is a retrospective cohort from one commercial digital clinic, in which tapering was offered as a clinician-guided personal choice rather than randomised, reporting a mean 16.7% weight loss at week 64 on individualised dosing (Seier 2025). It has no control group, and the clinic that produced the data also sells the programme being evaluated. We set out the whole picture in tapering off GLP-1.
At GetLean, our philosophy is that GLP-1 medication should act as a catalyst — not something to depend on indefinitely. The maintenance phase is where that philosophy either pays off or does not, because it is the point at which the plan around the medication has to work without it. Individual results vary, and clinical-trial figures describe the populations studied. Never change or stop a prescribed medication on your own; that is a conversation with the doctor who prescribed it.
Common questions
What is the maintenance phase after GLP-1 medication?
The period after active weight loss ends, when the goal changes from losing more to holding what has been lost. Once the medication stops, it is also the period in which whatever behaviour is already in place has to hold the result on its own.
Does appetite come back when you stop GLP-1 medication?
No withdrawal trial measured it. STEP 1's extension, STEP 4 and SURMOUNT-4 measured body weight and cardiometabolic markers, not appetite, hunger or intake. What is known is that the drug clears — semaglutide has an elimination half-life of about a week and is present in the circulation for roughly 5 to 7 weeks after a last 2.4 mg dose, tirzepatide about 5 to 6 days (DailyMed) — so the appetite effect stops being produced. That is an inference from pharmacology rather than a measured trial outcome.
How much weight comes back after stopping?
In the STEP 1 extension, participants regained about two-thirds of what they had lost within a year of stopping, ending 5.6% below their original starting weight (Wilding 2022). A meta-regression across six cessation trials found about 60% of lost weight regained at one year, decelerating rather than continuing at the same rate (Budini 2026). Individual results vary widely — the STEP 1 figure carries a standard deviation of 7.7 percentage points.
Is it better to taper the dose than to stop suddenly?
No randomised trial has tested it. Every published withdrawal trial used an abrupt switch to placebo, so there is no comparison between a gradual reduction and a sudden stop. Any dose change is a decision for the doctor who prescribed it.
What do people who keep weight off actually do?
Registry data points at a short list: a consistent eating pattern, a high level of physical activity — around an hour a day in the National Weight Control Registry (Klem 1997) — and continued self-weighing (Butryn 2007). That registry is a self-selected group of people who had already succeeded, so it describes what successful maintainers report rather than proving what causes success.