"So this is forever."
That sentence gets said in exam rooms every week, usually flat, usually looking at the floor. For a lot of people it lands as a verdict rather than a treatment plan, and it arrives on top of years of being told that the last thing they tried should have worked.
Here is what makes it worse. When these medications do stop, whether because of a copay, a formulary change, or a job change, the person very often does not come back to the office. The stop was economic; it gets recorded as a clinical outcome, and the person carries it as a personal failure.
That binary, stay on indefinitely or stop and lose everything, is doing real damage. It keeps some people from starting at all. It leaves others feeling that one insurance letter has erased two years of work. And it shapes benefit design in a way that serves neither the person taking the medication nor the organization paying for it.
The published evidence supports a third option. Not the one most people imagine, though. The third path is not getting off the medication. It is finding the lowest dose and the longest interval that holds the result, with monitoring.
What the trials measure, and what clinics see
The frightening statistic comes from trial withdrawal arms. When semaglutide was withdrawn at the end of the STEP 1 trial, participants regained about two thirds of what they had lost over the following year, ending 5.6% below baseline rather than 17.3% below (Wilding et al., Diabetes Obesity and Metabolism, 2022). In SURMOUNT-4, 82% of those switched to placebo regained more than a quarter of their loss. A 2026 meta-regression across six withdrawal trials puts it at roughly 60% of lost weight regained at one year (Budini et al., eClinicalMedicine, 2026). Those findings are solid.
The simple version of this, which I have written myself, is that when people stop the weight comes back. That holds in the trials. It does not describe what a clinic sees, and that gap is worth exploring because it demonstrates what is possible.
Now look at what happens in practice. A Cleveland Clinic analysis followed 7,938 adults for a year after they stopped semaglutide or tirzepatide within the first year of starting it. Average weight change in the group with obesity was plus 0.5%. 45% either maintained their loss or kept losing. In the group with type 2 diabetes, 56% did (Gasoyan et al., Diabetes Obesity and Metabolism, 2026).
That average is worth looking at, because it describes almost nobody. The standard deviation was 9.9%. The group contains people who gained a great deal and people who kept losing, and the mean sits between them.
What accounts for the difference? Look at what those patients did next. 27% switched to a different medication. 20% restarted the original one. 14% had a lifestyle-modification visit with a clinician.
Trial withdrawal arms measure abandonment. Clinical practice measures transition. In a trial, the participant receives a placebo pen and nothing else. In a clinic, most move to something else, the clinician wants to follow up. Those are different events, and they produce different outcomes.
What is actually reversing, underneath the number on the scale
Weight is the least interesting thing that changes when someone stops. It is simply the easiest thing to see.
A 2025 meta-analysis of 18 randomized trials followed people for at least 12 weeks after discontinuation. In the groups with obesity, systolic blood pressure rose by about 4 mmHg, HbA1c by 0.25 percentage points, LDL cholesterol and triglycerides both drifted upward, and fasting glucose rose (Tzang et al., eClinicalMedicine, 2025). Look at those numbers as direction rather than precision: the heterogeneity across studies was extreme, and diastolic pressure and HDL did not move significantly at all.
The more useful finding came from sorting people by how much they regained. Researchers went back to the SURMOUNT-4 withdrawal group and split it into four bands. Waist circumference rose by 0.8, 5.4, 10.1 and 14.7 centimetres across those bands. Among people who regained less than a 25% of their loss, waist circumference, non-HDL cholesterol and fasting insulin all stayed where they had been on the day the drug stopped.
Metabolic benefit is along a spectrum, not all or nothing. Someone who maintains most of a reduction still gets most of the benefit.
That should change how the word "regain" is considered. Maintaining your weight as best as you can is worth doing even if you do not maintain all that weight lost. A partial result still counts as a real result.
One limit here, cuts against the argument. Blood pressure still rose in every band, including the people who maintained their weight best, by nearly 7 mmHg even there. Graded benefit has been shown for waist, lipids and insulin. It has not been shown for blood pressure.
Two things people say about stopping that the evidence does not support
Both come up constantly, including from clinicians, and both are worth getting right.
The first is tolerance. People describe the medication "stopping working," and reach for the idea that the body gets used to it. Weight loss on these drugs does plateau, reliably, somewhere after a year. However, a plateau is not necessarily a tolerance, and there are at least two other explanations.
One is that energy balance has reached a new equilibrium. A smaller body needs less fuel, the food coming in now matches the lower fuel needs, and the drug is still doing its job.
The other is muscle. If someone has lost lean mass through inadequate protein intake and no strength training, resting metabolism falls further, and that lowers the energy requirement again. The plateau happens sooner and sits higher than it needed to.
Receptor desensitization is a real laboratory phenomenon and it does blunt some effects, such as gastric emptying. Loss of weight-lowering effect over time has not been demonstrated. This distinction matters practically, because a plateau calls for a different response than a drug that has stopped working, and one of those responses is protein and resistance training rather than a dose increase.
The second is what the regained weight is made of. The common claim is that you lose muscle on the way down and regain fat on the way up, ending worse off than you started. It is a reasonable hypothesis. It has not been measured. Neither of the two purpose-built withdrawal trials recorded body composition at all, and the one trial that did scan people stopped scanning the week treatment stopped. The 2026 systematic review notes: nobody has assessed what tissue comes back.
What would you expect? Muscle is expensive to build and expensive to keep. So in the absence of resistance training and enough protein to support it, the reasonable prediction is that most of what returns is fat. That is physiology rather than data, and it should be understood that way until somebody scans people after they stop.
The muscle half of the story is also usually told backwards. In the SURMOUNT-1 body-composition substudy, 74% of the weight lost was fat and 26% was lean mass. The placebo arm was 75% and 25%. The paper notes that dietary weight loss produces the same proportions. These medications do not preferentially strip muscle.
What is true is simple math, and it is enough. The same ratio applied to a much larger total loss is a much larger absolute loss: 5.6 kilograms of lean mass on tirzepatide against 1.2 on placebo. So the muscle question is real, the mechanism is not sinister, and the unmeasured part is what happens afterward. This is the reason for building the tissue while you still have the appetite suppression helping you, rather than discovering the question later.
The part nobody puts in the chart
There is a cost to regain that none of these outcome tables record.
Someone who has been told for twenty years that their weight is a matter of discipline, who then finds something that works, and who then watches it come back after a coverage change, does not experience that as pharmacokinetics. They experience it as confirmation of the thing they were always afraid was true.
That is not a soft observation. It is a common reason that people stop coming to appointments, and it is why the second attempt is so often harder than the first.
Regain after a treatment stops is a relapse of a chronic condition. It is not a verdict on the person, and the way it is framed at the moment it happens shapes what they do next. Naming that in advance, is part of the treatment.
Why people actually stop
Ask a room of prescribers why patients discontinue and many will say side effects. The charts say something else.
In the only study that pulled charts on people who stopped rather than surveying them, cost or insurance accounted for 47.6% of discontinuations. Side effects accounted for 14.6%. Surveys invert that ranking. Chart review does not.
Here is what likely explains the gap. Nausea is easy to say out loud. "I cannot afford this" is not. Patients tend to hand a clinician a side effect before they hand over their finances, so the affordability conversation rarely starts on its own.
There is another version of the same problem. When cost rises, people ration. They stretch the pen and often do not mention it. A self-directed stretch driven by a copay is not the same thing as a planned interval extension with monitoring, even though the dosing looks identical on paper. That is the difference between a taper and an interruption, and only one of them is being monitored.
Timing matters too. Side-effect discontinuations cluster early, during titration. Cost-driven ones cluster later in the first year, which puts the coverage conversation at around month six.
The third path now has randomized evidence
Until the spring of 2026, the literature compared only full dose against nothing. That changed with SURMOUNT-MAINTAIN, published in The Lancet in June.
After sixty weeks of treatment, participants were randomized to continue at their maximum tolerated dose, step down to 5 mg, or move to placebo. At week 112 the three groups sat at minus 21.9%, minus 16.6% and minus 9.9%. The reduced-dose group kept roughly three quarters of the full-dose effect.
A second trial, ATTAIN-MAINTAIN, found that patients switched from an injectable to a daily oral agent held 75 to 79% of their prior loss, against 38 to 49% for those moved to placebo. One caveat belongs with that number: the comparison was against placebo, not against staying on the injection, so it shows the oral route holds a result rather than that it matches the injection. The trial was run by the maker of the oral drug.
And in a real-world programme of 2,246 patients presented at the European Congress on Obesity, 353 patients began a clinician-approved dose reduction with a median taper phase of nine weeks, with weight essentially stable throughout. That work is a conference abstract rather than a peer-reviewed paper, and the analysis was authored by staff of the clinic whose programme it evaluates, so read it as a signal rather than a finding.
So the third path exists, it is measurable, and it is mostly about dose and interval rather than cessation.
What exercise leaves behind, and the medication does not
One randomized trial asked a different question: not what happens when you swap the drug, but what is still there a year after everything stops.
In the Danish S-LiTE study, 195 adults completed an eight-week diet and were randomized to a year of supervised exercise, liraglutide, both, or placebo, then left alone for a year with no treatment and no contact (Jensen et al., eClinicalMedicine, 2024). Regain over that untreated year was 9.6 kg after liraglutide alone and 3.6 kg after exercise alone.
The body-composition result is the one that should change practice. Two years from the start, body fat percentage in the liraglutide groups was indistinguishable from placebo. The exercise groups held a 1.5 percentage point advantage.
The drug had no extra benefit. The training had.
That is not an argument against the medication, which did its job while it was being taken. It is an argument about what gets built during the window it buys.
Building the plan in the right order
Plans work better when they are built in a particular sequence, and the sequence is not obvious.
Protein and resistance training go first, at the very first visit, before the first injection. It's not because they prevent regain on their own, but because a habit built while appetite is quiet because of the drug, is a habit that survives when the medication is gone. In the clinics, it's confirmed, the people who maintain their result are not the ones who started training after they stopped taking the medication. The maintainers started training early.
A practical note from the room on why this is harder than it sounds: appetite suppression does not reduce eating evenly. Protein tends to go first, because it takes the most time and effort to buy, prepare and chew, and it often is more expensive.
Baseline measurements come next, and they include something beyond weight. Waist, blood pressure, A1c, lipids and ApoB, insulin, liver enzymes, and one measure of function such as grip strength or a chair stand. Weight alone cannot tell you whether the weight lost was fat or muscle.
Then the coverage conversation, with a renewal date written down. A renewal date is a clinical risk factor because it predicts a treatment interruption as reliably as anything else in the chart.
And through all of it, monitor the signals that move before weight does. Hunger, cravings and the return of food preoccupation shift weeks ahead of the scale, which is why a decision to hold a dose or lower it further is made on symptoms rather than on a calendar.
If you are the person on treatment
Not everyone should come off. For many people the safest and most effective plan is to stay on, and that is a legitimate outcome rather than a failure of resolve.
And if you have already stopped, whether by choice or because coverage ended, that is not a door closing. Nearly half of people in that position maintain their result at a year, and the ones who do best are usually the ones who went back and asked for a plan rather than waiting until something had gone wrong. Regular check-ins do make a difference.
Finally, what matters is that a plan exists. Six questions worth bringing to the next appointment: What is our goal, and how will we know when we reach it? What are we measuring besides weight? What is my protein target, and what does training look like for me starting now? What is our plan when coverage renews? If we lower the dose, what will we be watching? And if I need to stop, what does that transition look like?
Obesity is a chronic condition, and chronic conditions are managed, reassessed and adjusted over time. The goal was never to get everyone off these medications. It was never to put everyone on them forever either. The goal is a plan that changes when the person does.
Which brings this back to the sentence in the exam room.
"So this is forever" deserves a better answer than yes or no. The honest answer is that treatment will change over time, that the changes can be planned rather than forced, and that a good deal of what gets built during the treatment years belongs to the person rather than to the prescription. That is a different thing to hear, and it is also, as it happens, what the evidence supports.