When the Treatment Stops: The Weight-Regain Question
This series examines what happens after treatment with a glucagon-like peptide 1 receptor agonist, or GLP-1RA, stops. The focus is not only whether semaglutide, tirzepatide, liraglutide, or related medicines can produce weight loss while they are being taken, but how much of that change remains when the drug—and sometimes the accompanying lifestyle support—is withdrawn. Understanding that transition helps readers interpret the evidence without confusing a treatment result with a permanent change in body weight.
The question matters because a successful first phase can create an easy but misleading expectation: if a medicine produces a large loss, perhaps the body will simply keep the new weight once the course is over. The evidence supplied for this series points to a different and more complicated pattern. Weight commonly returns after GLP-1RA cessation, often quickly at first and more slowly later. At the same time, the available studies suggest that some people retain part of the loss and some cardiometabolic improvements. The important issue is therefore durability, not a simple verdict of success or failure.
The result after treatment is part of the treatment story
GLP-1RAs are used as part of weight management for people with overweight or obesity. In the cited trials, newer agents such as semaglutide and tirzepatide produced weight losses in the broad range of 15% to 20% and improved measures including blood pressure, HbA1c, low-density lipoprotein cholesterol, and triglycerides. Those results explain why these medicines have changed the conversation about obesity treatment. They do not, by themselves, answer what happens after the active treatment ends.
The distinction is important because the sources treat obesity as a complex, chronic disease rather than a short episode that can necessarily be removed by one successful intervention. Weight loss can trigger biological responses that oppose keeping the loss. A medicine that reduces appetite and changes metabolic regulation may be helping a person maintain a new state while it is active. When the medicine stops, that support is no longer present, while the pressures that made weight loss difficult may persist. This is a description of a treatment transition, not a judgment about a person’s effort.
Lifestyle support also changes the meaning of “stopping.” Some studies continued diet and activity counseling during the follow-up period; others stopped both the medicine and structured support. If weight changes differ between those settings, the result cannot be attributed to the drug alone with confidence. In the STEP 1 extension, for example, semaglutide and the lifestyle intervention ended together at week 68, and participants were observed for another year without an active program. That design is useful because it shows what may happen when support is withdrawn in a relatively ordinary-looking setting, but it cannot identify the separate contribution of every part of the transition.
The broader review gathered evidence from 48 studies involving adults with overweight or obesity. It included randomized trials as well as non-randomized studies and covered several GLP-1RA treatments. Nearly all of the studies that reported meaningful treatment loss found at least some regain after cessation. The consistency of the direction is the first important signal. The amount and timing are the harder questions, and they require a careful definition of what “regain” means.
What counts as regain?
The supplied systematic review does not define regain as “return to the starting weight.” It measures the weight regained after treatment as a percentage of the weight lost during treatment. That denominator matters. Imagine, only as an illustration, a person whose starting weight is 100 kilograms and whose treatment loss is 20 kilograms. If they later regain 12 kilograms, they have regained 60% of the weight they lost, but they still weigh 8 kilograms less than at the beginning. “Most of the lost weight returned” and “the person remained below baseline” can therefore both be true.
This measure also makes studies with different starting weights and different treatment responses more comparable. It does not make their populations identical, however. A 60% regain fraction has a different absolute meaning after a 5-kilogram loss than after a 20-kilogram loss. Nor does it say when the regain occurred. A single value at the end of follow-up can conceal a steep early increase followed by a period of slower change.
To capture timing, the review fitted an exponential recovery model to six randomized controlled trials with 3,236 participants. In plain language, the model assumes that the trajectory rises rapidly after cessation and then approaches a limiting level more gradually. Its estimated maximum was 75.3% of the weight lost during treatment, with a 95% confidence interval of 68.9% to 81.6%. The estimated recovery rate corresponds to a half-life of about 23 weeks. At 52 weeks, the model estimated that about 60% of the original treatment loss had been regained.
Those figures are precise, but precision should not be mistaken for certainty. The six trials supplied longitudinal data only through 52 weeks after cessation. The estimated 75.3% is an asymptote from a fitted curve, not a weight measured in a large group years later. The model’s shape is useful for understanding the evidence: it rejects the idea that regain is necessarily a one-time jump and suggests that the fastest movement may occur early. It does not prove that every individual follows the same curve or that a plateau will occur at the predicted level.
The distinction between an observation and an extrapolation will recur throughout this series. A study can directly report what participants weighed at a specified visit. A meta-regression can summarize several trajectories and estimate a rate. It can then project beyond the period for which trials collected data. Each step may be reasonable, but each step also adds assumptions. The farther a conclusion extends beyond direct follow-up, the more carefully it must be described.
Two evidence lenses, one cautious conclusion
The first lens is a direct case: the off-treatment extension of STEP 1. In the main trial, adults with overweight or obesity but without diabetes received once-weekly semaglutide 2.4 milligrams or placebo for 68 weeks, alongside counseling on diet and physical activity. A selected group of treatment completers was then followed for 52 weeks after treatment and structured lifestyle support stopped. This extension included 327 participants.
During the treatment period, the semaglutide group lost an average of 17.3% of body weight, compared with 2.0% in the placebo group. By week 120, the semaglutide group had regained an average of 11.6 percentage points of body weight and retained a net loss of 5.6% from baseline. The paper describes this as about two-thirds of the prior loss being regained. Nearly half of the semaglutide participants—48.2%—still had at least 5% weight loss from baseline at week 120, although that was a substantial decline from the 86.4% who had reached that threshold at the end of the treatment phase.
This is a useful illustration of why the denominator and the timeline must stay together. The average result shows substantial reversal, not total reversal. It also shows why an average cannot answer what will happen to a particular person. The extension was exploratory, involved a much smaller selected subset than the original STEP 1 trial, and did not randomize participants to different ways of stopping. It also cannot tell us whether the observed slowing of regain would become a stable plateau after week 120.
The second lens is the broader systematic review and its nonlinear meta-regression. It places STEP 1 alongside other trials and observational studies and finds a consistent population-level direction: most studies report some regain after GLP-1RA cessation. In studies where participants had lost more than 3 kilograms during treatment, the amount regained at around 12 weeks ranged from approximately 10% to 80% of the treatment-induced loss. That broad range is not a personal prediction; it reflects differences in drugs, populations, treatment duration, follow-up, and study design. Studies with very small initial losses make a regain percentage less stable, and an apparent regain above 100% in one study was likely affected by a switch to insulin.
Taken together, these sources support a restrained but consequential conclusion. After GLP-1RA treatment stops, weight regain is common and is often substantial. The average trajectory appears to be faster at first and slower later. Some treatment benefit may remain below the pre-treatment baseline, but the long-term level suggested by the model is not directly observed, and individual outcomes vary. The evidence is stronger for the direction of the change than for its final destination.
That conclusion changes how the next question should be asked. It is not enough to ask whether a medicine “works” during treatment. We also need to ask what the treatment is doing while it is active, what happens when that support is removed, and how the answer was measured. A regain percentage is a statement about the portion of an earlier loss that returned; it is not a prediction that everyone returns to where they began. Learning to read that curve carefully is the foundation for understanding the trial evidence, the biology behind rebound, and the choices that follow from it.