The short answer: most of the lost weight comes back
When people stop a GLP-1 receptor agonist prescribed for weight, body weight rises again, and in every controlled trial that has looked, most of what was lost returns within about a year. This is not a fringe finding or a signal from a single study. It has now been shown three different ways: by following people after a trial ended, by randomising people who had already lost weight to continue or to switch to placebo, and by observing what happens in the weeks after a drug with a long half-life clears the system.
The size of the effect is large enough that it changes how the treatment should be understood. In the extension of the STEP 1 trial of semaglutide, the average participant went from about 17 percent below starting weight while on the drug to about 6 percent below it a year after stopping. In SURMOUNT-4, people who had lost roughly a fifth of their body weight on tirzepatide and were then switched to placebo put back about 14 percent over the following year, while those who stayed on the drug lost a little more.
None of this means the drug stopped working or that the loss was somehow fake. It means the loss was being held in place by ongoing pharmacology, and when that is withdrawn the physiology that produced the weight in the first place is still there, unopposed. That is an uncomfortable framing for anyone hoping for a defined course of treatment, but it is what the data show.
What the STEP 1 extension actually measured
STEP 1 randomised adults with obesity, or overweight with a weight-related condition and without diabetes, to weekly semaglutide 2.4 mg or placebo for 68 weeks alongside lifestyle counselling. Mean weight change was about 14.9 percent with semaglutide against 2.4 percent with placebo, the result that anchored the drug's approval for chronic weight management.
The extension followed a subset of those participants, several hundred people across both arms, for a further 52 weeks with no study drug and no continued intervention. This is a follow-up study rather than a randomised withdrawal, so it cannot separate the effect of stopping the drug from the effect of the trial's structure and contact ending at the same moment. Its value is that it is long, and that it measured people who had genuinely reached a large weight loss before the drug was removed.
The result: participants who had been on semaglutide were about 17.3 percent below baseline at week 68 and about 5.6 percent below baseline at week 120. That is roughly two-thirds of the loss reversed in a year. The placebo arm, which had lost little to begin with, was back to essentially its starting weight. Waist circumference followed the same path. Most cardiometabolic variables that had improved on treatment reverted toward their week-zero values, with HDL cholesterol the notable exception.
STEP 4 and SURMOUNT-4: randomised withdrawal, which is the stronger design
A follow-up study can always be criticised for confounding. A randomised withdrawal design cannot be criticised in the same way, because everyone in it has already lost the weight before randomisation, and the only thing that differs afterwards is whether the drug continues. Two such trials exist in this drug class, and they are the cleanest evidence available.
STEP 4 ran a 20-week open-label lead-in in which participants escalated to semaglutide 2.4 mg and lost about 10.6 percent of body weight. Those who completed it were randomised either to continue semaglutide or to switch to placebo for the next 48 weeks, with lifestyle support maintained in both arms. Continuing lost a further 7.9 percent. Switching to placebo produced a regain of about 6.9 percent. Expressed against original baseline, the two groups finished the trial at roughly 17.4 percent and 5.0 percent below their starting weight.
SURMOUNT-4 used the same architecture with tirzepatide and pushed the lead-in much further: 36 weeks of open-label treatment at the maximum tolerated dose of 10 or 15 mg weekly, producing a mean reduction of about 20.9 percent. Roughly three hundred participants were then randomised to continue or to switch to placebo for 52 weeks. The continuation arm lost another 5.5 percent. The placebo arm regained about 14 percent. At the end, total change from the original baseline was about 25.3 percent for continuation against 9.9 percent for withdrawal.
The three trials agree with each other despite differing in drug, dose, lead-in length and depth of weight loss, and that consistency matters more than any single number. It also rules out the most appealing hypothesis, that a deeper or longer initial loss buys durability: SURMOUNT-4 participants had lost more than a fifth of their body weight over nine months before withdrawal, and they still regained steadily. Neither withdrawal arm returned all the way to baseline within its observation window, so some benefit persisted at one year off drug; whether it persists at two or three years has not been measured.
The cardiometabolic gains regress with the weight
Weight is a convenient endpoint but rarely the reason the drug was prescribed. The more consequential question is what happens to the things weight loss improves. Across the withdrawal trials the answer is uniform: they follow the weight back up.
In the STEP 1 extension, systolic and diastolic blood pressure, triglycerides, total cholesterol, glycated haemoglobin, fasting insulin and C-reactive protein had all improved at week 68 and had all moved back toward baseline by week 120. In SURMOUNT-4, waist circumference, blood pressure, lipids and glycaemic measures deteriorated in the placebo-switch arm while continuing to improve or holding steady in the continuation arm. The improvements were not banked; they were rented.
This has a specific implication for people whose GLP-1 was started for a reason other than the number on the scale, such as prediabetes reversion, sleep apnoea, fatty liver disease or cardiovascular risk reduction. SELECT showed that semaglutide 2.4 mg reduced major adverse cardiovascular events by roughly 20 percent in people with established cardiovascular disease and overweight or obesity but without diabetes, over a median of about three years of continuous treatment. There is no randomised off-treatment phase in that trial, so nothing is known about how long any of that risk reduction outlives the prescription.
Why the weight returns: the biology was never reset
The regain is not a peculiarity of incretin drugs. It is the standard response of a defended body weight system to the removal of whatever was opposing it. Diet-induced weight loss shows the same pattern without any drug involved, which is the single most important context for interpreting these trials.
After substantial weight loss, circulating leptin falls, ghrelin rises, and several other appetite-regulating signals shift in the direction of restoring lost mass. Sumithran and colleagues showed in 2011 that these hormonal changes were still present a full year after a ten-week very-low-energy diet, alongside persistently elevated subjective hunger. Resting energy expenditure also tends to fall below what body composition alone predicts, a phenomenon documented most dramatically in the long-term follow-up of Biggest Loser contestants, in whom metabolic adaptation persisted six years after the competition.
A GLP-1 receptor agonist works against this system rather than dismantling it. It slows gastric emptying, acts on hypothalamic and hindbrain circuits that govern satiety, and for many people quiets the intrusive preoccupation with food that patients describe as food noise. What it does not appear to do is lower the weight the body defends. When drug exposure falls, appetite signalling returns to its post-weight-loss state, which is a state biased toward eating more, and intake rises accordingly.
That is why the honest framing is maintenance rather than cure, and why the regain should not be read as the drug having failed. An antihypertensive that stops lowering blood pressure once you stop taking it is behaving exactly as designed.
How quickly it happens after the last dose
Semaglutide has a terminal half-life on the order of a week, so meaningful drug exposure persists for roughly a month to five weeks after a final injection. Tirzepatide's half-life is around five days, giving a similar if slightly shorter tail. Liraglutide, dosed daily, clears within a day or two. The practical consequence is that appetite does not switch back on the morning after a missed dose; it returns over weeks as concentrations decline.
In the randomised withdrawal trials, weight in the placebo-switch arms began climbing early and continued climbing throughout the observation period rather than rising sharply and then flattening. Neither STEP 4 nor SURMOUNT-4 showed a clear plateau within its window, which is one reason the one-year figures should not be read as an endpoint. They are a snapshot of a trajectory still in motion.
Side effects follow the opposite curve. Nausea, reflux, early satiety and altered bowel habit fade as exposure declines, and there is no withdrawal syndrome in the pharmacological sense. These drugs are not dependence-forming, and stopping them does not produce the physiological rebound seen with abrupt beta-blocker cessation. What returns is hunger, which people understandably describe as though it were a withdrawal effect.
Does tapering help? The evidence does not exist
The idea that stepping down the dose gradually allows the body to adjust and prevents regain is widespread in patient forums and in clinic marketing. It has not been tested. No randomised trial has compared a taper with abrupt discontinuation for any GLP-1 receptor agonist, with weight regain as the endpoint.
Both withdrawal trials stopped the drug outright by switching participants to placebo, which is exactly the comparison a taper would need to beat. A pharmacokinetic argument against tapering mattering much is also available: because these are long-acting molecules, discontinuation is already a gradual decline in exposure. Halving a weekly dose for a month is a modest change against a background of a drug that takes a month to clear anyway.
There is a separate and more plausible question that also lacks data, which is whether a reduced maintenance dose, continued indefinitely, holds weight better than stopping. That is not tapering to zero but chronic therapy at lower intensity. It is a reasonable hypothesis and it is untested in a randomised trial. Anyone describing either approach as evidence-based is describing something the literature does not contain.
What the liraglutide data add
Liraglutide is the oldest of the three and frames the same finding from a different angle. SCALE Maintenance enrolled people who had already lost at least 5 percent of body weight on a low-calorie diet and randomised them to liraglutide 3.0 mg or placebo for 56 weeks. The liraglutide arm lost about 6 percent more while the placebo arm held roughly flat, showing that pharmacotherapy changed the maintenance trajectory while it was being given.
The three-year SCALE Obesity and Prediabetes trial included a 12-week off-treatment observation after 160 weeks of dosing. In that short window, participants coming off liraglutide regained more weight than those coming off placebo, narrowing the gap between the groups. Twelve weeks is far too brief to characterise the full trajectory, but the direction was identical to what the semaglutide and tirzepatide withdrawal trials later showed at scale.
Taken together, the class behaves consistently. Three molecules, spanning a decade of trials and a wide range of efficacy, all deliver weight loss that is contingent on continued exposure.
Why people stop anyway, and why that is a different question
Trial withdrawal is deliberate and monitored. Real-world stopping usually is not. Claims-based analyses in the United States have repeatedly found that a large share of people who begin a GLP-1 receptor agonist for weight management are no longer filling prescriptions a year later, driven by cost and coverage changes, supply interruptions during the shortage years, gastrointestinal intolerance, pregnancy, and the belief that the goal has been reached.
This gap between trial populations and real-world populations matters when reading the regain figures. Trial participants had structured support and free drug; the people stopping in practice often have neither, and frequently stop abruptly and without a plan for what follows. The withdrawal trials, if anything, describe a best case.
It also reframes what a clinical conversation about stopping is actually about. The evidence does not say a person must stay on these drugs. It says that if the drug is stopped, weight will most likely rise and the associated metabolic improvements will most likely regress, and that this is predictable rather than a personal failure. What anyone should do with that information depends on their circumstances and belongs with their own clinician.
How to read the evidence without overreading it
Three limits are worth holding onto. First, these are group means with a wide spread around them. The STEP 1 extension reported that a minority of participants maintained most of their loss a year after stopping, and nobody can currently predict who those people are. Second, the withdrawal arms received no replacement intervention beyond routine lifestyle counselling; none tested whether a structured behavioural or resistance-training programme begun at the point of discontinuation changes the trajectory.
Third, the follow-up is short. One year off drug is the longest randomised observation available, and the curves had not settled. Whether weight eventually overshoots baseline, stabilises somewhere below it, or returns to the pretreatment level is not established, and the answer probably differs by individual and by how long treatment ran.
What can be said with confidence is narrow and useful: the loss is real, it is drug-dependent, that dependence has been shown in randomised designs across two molecules, and the metabolic benefits are not banked.