Stop a GLP-1 and your appetite returns, and for most people a large share of the lost weight comes back with it. That’s not the drug failing or your willpower breaking. It’s the predictable result of withdrawing an ongoing treatment for a chronic, relapsing condition, and the withdrawal trials show it plainly.
What the trials found when people stopped
The cleanest number comes from SURMOUNT-4 (Aronne et al., JAMA 2024). Participants first reached a full maintenance dose over several months of open-label tirzepatide, then were randomized either to continue or to switch to placebo for the following year. The group that continued lost another 5.5% of body weight. The group that stopped regained about 14%. Both groups got there the same way; the only difference afterward was whether the drug continued.
Semaglutide behaves the same way. In STEP 4 (Rubino et al., JAMA 2021), people who had already lost weight over 20 weeks of semaglutide were split into two arms: keep going, or switch to placebo. Those who continued kept losing. Those who stopped regained a substantial share of what they had lost.
The design is what makes these trials persuasive rather than anecdotal. Both used a run-in: everyone took the drug first, and only the people who were actually on it got randomized to continue or stop. So the regain isn’t explained away by “these were people the drug never worked for.” It worked, they lost weight, and removing it reversed the effect. That’s about as close to a controlled test of stopping as the literature offers.
The withdrawal arms weren’t quitting the program. Participants stayed in a structured weight-management protocol with continued diet and activity guidance, and regained anyway. Lifestyle support is worth having. It isn’t, on its own, what held the weight off.
One caveat: these are group means. Individuals vary, and a mean regain of 14% doesn’t mean everyone returns all the way to their starting weight — some hold more of the loss, some hold less. But the direction is consistent enough that any plan should treat regain as the default to work against, not a risk to gamble on. For scale, the loss at stake is real: STEP 1 (Wilding et al., NEJM 2021) put mean weight loss at 14.9% at 68 weeks, and SURMOUNT-1 (Jastreboff et al., NEJM 2022) at 15% to 20.9% depending on dose. We go deeper on the mechanics of the bounce in rebound weight after a GLP-1.
The weight comes back because the biology does
A GLP-1 works only while it’s present. These drugs slow gastric emptying, act on appetite circuits in the hypothalamus and hindbrain, and enhance glucose-dependent insulin secretion; tirzepatide adds a second receptor, GIP. None of that permanently rewrites the appetite set-point. When the drug clears, the signaling it was suppressing switches back on, and the body resumes defending the higher weight it came from. Food noise and hunger return, intake rises, and weight follows. Weight loss also nudges total energy expenditure down, a well-described feature of obesity physiology, and that compounds the appetite rebound rather than offsetting it.
This is the same machinery behind ordinary diet rebound: the body treats lost fat as a deficit to correct. The set-point doesn’t know or care that a drug, rather than a diet, produced the loss. That’s why framing regain as a failure of discipline misreads what’s happening.
The return is gradual, not overnight, because these drugs clear slowly. Semaglutide’s label says to take a missed dose within five days and otherwise skip it, a window that reflects a half-life of about a week. Appetite tends to climb back over the weeks after your last dose, not the next morning. That lag is a trap: some people read the first few good weeks off the drug as proof they’ve “kept” the results, right before the hunger comes back.
The honest framing: a GLP-1 manages a condition, it doesn’t cure it. Semaglutide treats obesity the way an antihypertensive treats blood pressure — it works while you take it, and the numbers drift back when you stop. We say that plainly because the alternative framing, a course you finish, sets people up to read normal physiology as personal failure.
Does tapering down soften the landing?
Here the evidence thins fast. The withdrawal trials didn’t taper. They stopped the drug or switched people to placebo. There’s no trial showing that stepping the dose down gradually prevents or reduces regain compared with stopping outright. We don’t know whether a taper changes the trajectory. It’s biologically plausible that easing off blunts the rebound in appetite, but plausible isn’t shown. Anyone selling a “taper protocol” as proven weight-maintenance is outrunning the data.
It helps to separate two questions a taper might address. One is gastrointestinal comfort: a slower step-down may spare you the lurch of an abrupt change in gastric emptying, and that’s a reasonable tolerability goal. The other is weight durability — whether the taper keeps the pounds off — and that’s the part with no trial behind it. A taper can be sensible for the first reason while doing nothing measurable for the second.
We taper rather than stop. At goal, we hold your effective dose for a few months first, then step down one rung every 8–12 weeks, watching what the weight does at each stop. The lowest dose that holds you is the destination, and you find it by walking down until the weight tells you to stop. If it drifts up over a month at the new rung, go back one — that is not a failure of the taper, it is the taper working. You found the floor.
Your history on the way up is data. If appetite was already loud at 1.0 mg during titration, 1.0 mg is unlikely to hold you on the way down.
Be clear about which part of this is evidence and which is judgement. The tolerability rationale is sound and mechanistic. The idea that tapering protects the weight better than stopping outright has no trial behind it.
The other option is not stopping
Continuing is the intervention the trials actually validate. In both STEP 4 and SURMOUNT-4, the arm that stayed on treatment held or extended its loss. If keeping the weight off is the goal, staying on the medication is the option with direct trial support. That’s why how long you stay on a GLP-1 is a clinical decision, not a fixed course with a finish line.
Staying on doesn’t have to mean staying at the top dose. Many members step down to a lower maintenance dose once they reach their goal. The caveat: the trials continued people at their treatment dose, not a reduced one, so how well a lower dose holds the line isn’t well characterized by the data. It’s a reasonable clinical practice, not a trial-proven one. And true microdosing, sub-therapeutic amounts marketed as a maintenance hack, has no trial support at all. We don’t know that it works. The marketing is well ahead of the evidence.
On the safety of staying on for years, which is the natural next worry: SELECT (Lincoff et al., NEJM 2023) followed 17,604 people for a mean of more than three years. It’s the largest long-term safety dataset we have, and semaglutide there cut major adverse cardiovascular events by 20% (hazard ratio 0.80). Long-term treatment is, so far, well tolerated. That doesn’t settle every question, but it argues against treating “I’ve been on it a while” as its own reason to quit.
One milligram is the lowest dose we have seen hold weight reliably. Below that, people tend to drift — not immediately, and not everyone, but often enough that we would not plan around 0.5 mg as a maintenance dose.
That is an observation from our own panel, not a trial finding, and it is worth saying plainly: no study has established a minimum effective maintenance dose. It also means “I have been on this a while” is not by itself a reason to go lower. The reason to step down is that a lower rung still holds you, and you find that out by trying it under supervision, not by assuming it.
If you’re stopping anyway, do it on a plan
People stop for good reasons: side effects that never settle, a planned pregnancy (GLP-1s are not for use in pregnancy), an interruption in access, or simply wanting to test how their own appetite regulation holds without the drug. Stopping is a legitimate choice. Stopping by accident is not a plan. The difference is mostly preparation: knowing what’s coming and deciding in advance how you’ll respond to it. If you’re coming off, build in these:
- Protect your muscle. DXA substudies show a meaningful share of GLP-1 weight loss is lean mass, not just fat. Resistance-train at least twice a week and hold protein around 1.2–1.6 g/kg per day — before, during, and after you come off — so the weight you keep is muscle. The long-term functional consequences of that lean-mass loss aren’t yet characterized, which is why you defend against it rather than wait and see.
- Expect appetite back within weeks. Set up meals and environment for the return of hunger, not for the artificial quiet of your last few dosed weeks.
- Set a re-engagement threshold in advance. Pick the weight or waist measurement at which you’ll talk to your clinician about restarting, and write it down while you’re still off the daily-scale rollercoaster.
- Time it properly around pregnancy. If you’re stopping to conceive, your clinician sets how far ahead of trying you should come off.
- Keep the door open. Restarting is common and isn’t a relapse in any moral sense. If you’re re-evaluating your options, the two-minute quiz routes you back to a clinician who can work from your history.
What we still don’t know
The open question isn’t whether stopping causes regain — the trials settle that. It’s whether anything changes the landing. No head-to-head trial has compared an abrupt stop against a structured taper against a defined low maintenance dose for the durability of weight loss. Until someone runs it, “how do I come off and keep the most weight?” has an answer built from mechanism and clinical experience, not a trial-grade one. That gap, and the still-uncharacterized long-term functional cost of the lean-mass share of the loss, are what to watch.
