STEP 1: A Year Without Semaglutide
The STEP 1 extension offers the clearest one-year case of what can happen after semaglutide treatment stops. Its value lies in the sequence: participants first received 68 weeks of semaglutide or placebo with lifestyle counseling, then spent 52 weeks without active study treatment or structured lifestyle support. Reading that sequence carefully shows what the weight figures mean, what residual benefit looks like, and why this observational extension should not be treated as a universal stopping experiment.
The case is especially useful because it places two questions side by side. Did semaglutide produce a substantial loss while it was being used? Yes. Did participants keep all of that loss after the treatment and structured support ended? No, on average. Those answers are not in conflict. The extension is a test of durability after a large treatment response, not a test of whether the original response occurred.
The design created a clear treatment-to-withdrawal boundary
The main STEP 1 trial enrolled adults with overweight or obesity who did not have diabetes. Participants were randomized to once-weekly subcutaneous semaglutide 2.4 milligrams or placebo for 68 weeks, alongside lifestyle counseling. This randomization created the original comparison between an active GLP-1RA treatment and placebo. It did not, however, randomize participants at the end of week 68 to different ways of stopping.
At week 68, the main phase ended. Both the assigned treatment and the structured lifestyle intervention stopped. A selected group of people who had completed the treatment period then entered an off-treatment extension and were followed until week 120, another 52 weeks later. The extension included 327 participants from selected sites and countries. The selection was intended to provide a useful follow-up group, but it still meant that the extension represented people who had completed the first phase and agreed to continue rather than everyone who began the original trial.
This design answers a specific question: what happened to these treatment completers during a year in which the trial no longer supplied semaglutide or its structured lifestyle counseling? It does not answer the question of what would happen if the same participants had been randomly assigned at week 68 to continue semaglutide, reduce the dose, or stop under a planned maintenance program. There was no such continuation comparison in this extension. The absence of that comparison is not evidence that continued treatment would produce any one particular result; it simply limits what this study can establish.
The withdrawal boundary also includes more than the disappearance of a drug effect. Participants were not given an active replacement program, and structured lifestyle support ended at the same time. That makes the extension informative about an off-treatment period in which support was no longer supplied by the trial. It also means that the study cannot cleanly separate the influence of stopping semaglutide from the influence of stopping the counseling, changing routines, or moving through a longer period after the initial loss. The design describes a realistic combination of changes, not an isolated laboratory switch.
The timeline is therefore the essential fact. Week 0 is the pre-treatment reference point. Week 68 is the end of active treatment and counseling. Week 120 is the end of one year off treatment. A weight at week 120 represents the accumulated result of the original loss and the subsequent regain; it is not a measurement of the withdrawal period alone unless the comparison begins at week 68. A careful reading keeps both baselines in view.
The extension is best read as a change-of-state study. During the first state, the groups received their assigned treatment and lifestyle support. At the boundary, those supports stopped. During the second state, researchers observed what remained and what changed. That framing prevents a common mistake: treating the week-120 number as if it were simply another treatment endpoint. It is the result of treatment followed by withdrawal.
What the numbers show—and what they do not
During the first 68 weeks, the semaglutide group lost an average of 17.3% of body weight. The placebo group lost 2.0%. These are changes from the original baseline, not percentages of an already reduced weight. The difference establishes that the two groups entered the extension with very different treatment histories. The later comparison must therefore be read against the amount each group had changed before withdrawal.
During the following year, the semaglutide group regained an average of 11.6 percentage points of body weight. The placebo group regained 1.9 points. By week 120, the semaglutide group therefore retained a net mean loss of 5.6% from its original baseline, while the placebo group’s net mean loss was 0.1%.
The first interpretation is straightforward: after a large treatment-associated loss, much of the weight came back when semaglutide and structured support stopped. The paper describes 11.6 points regained against 17.3 points lost as roughly two-thirds of the prior loss. The two numbers use the same baseline frame, so their ratio communicates the share of treatment-associated change that was reversed. The remaining 5.6% is the net difference from the beginning, not the amount that escaped regain expressed as a fraction of the original loss.
A simple reconstruction makes the arithmetic visible. Suppose a person began at 100 kilograms and lost 17.3 kilograms during treatment. Regaining 11.6 kilograms would amount to about 67% of that loss and leave the person 5.7 kilograms below baseline. The actual trial reports group averages rather than this hypothetical individual, but the calculation explains why “two-thirds regained” and “5.6% below baseline” are compatible descriptions. The key is to ask whether a figure refers to the amount reversed during withdrawal or the net change across the full 120 weeks.
It is also important not to compare the semaglutide group’s 11.6 points with the placebo group’s 1.9 points as though the groups began the extension at the same weight. The semaglutide group had lost far more during the first phase, so it had more treatment-associated weight available to regain. The placebo group’s smaller regain occurred after a much smaller average initial change. The placebo arm provides context for the trial sequence, but it does not turn the extension into a simple comparison of two identical starting states.
The second interpretation is more nuanced: residual benefit was common enough to matter, but it was not equivalent to maintaining the full treatment response. At week 120, 48.2% of participants in the semaglutide arm still had at least 5% weight loss from baseline. That is a clinically meaningful threshold for many weight-management studies, yet it represented a substantial fall from the 86.4% who had reached it at the end of active treatment in the main trial. A treatment can leave a measurable net benefit while losing much of its original effect.
The threshold result should also be kept in its proper frame. It tells us how many participants still met a predefined level of net loss at week 120; it does not say that the other participants returned exactly to baseline, nor does it show that everyone above the threshold had the same health profile. A group average and a threshold proportion answer different questions. The average describes the center of the observed change, while the proportion shows how widely a clinically relevant level of loss was distributed.
The extension also supports a cautious interpretation of individual variation. Participants who lost more during the treatment period tended to regain more afterward, but they also retained greater net loss at week 120 than participants who lost less. The steepest regain appeared among those who had lost at least 20% of baseline weight. This does not mean that a large initial loss mechanically causes a fixed amount of regain. It means that greater initial success can create both a larger amount available to regain and a larger residual loss after the regain observed in this study.
The cardiometabolic results reinforce the same pattern without making it identical. After withdrawal, blood pressure and HbA1c moved upward in the semaglutide group, and several improvements moved toward baseline. Some relative benefit remained at week 120. Those findings matter because the treatment state included more than body weight, but their interpretation belongs alongside the separate limits of biomarker measurement and glycaemic classification. A scale result should not be used as a perfect proxy for every health outcome.
The boundary between a case and a forecast
STEP 1 is persuasive because its chronology is easy to follow and its treatment-associated loss was substantial. It is also limited in ways that affect how widely the result can be applied. The extension was exploratory and had no formal power calculation. It followed a selected subset of people who completed the main treatment period, not everyone originally randomized. It used selected sites and countries rather than a new sample designed to represent every setting in which semaglutide might be prescribed.
The participants had overweight or obesity without diabetes at entry to the main trial. The results therefore do not directly establish the trajectory for people with type 2 diabetes, people who stop after a shorter course, people taking another dose, or people who move to another treatment. They also do not tell us how a gradual dose reduction would compare with abrupt cessation. Those are not minor variations on the same experiment; they change the conditions under which the post-cessation trajectory develops.
There is a temptation to turn the STEP 1 figures into a rule: lose a certain amount, stop, and expect a certain percentage to return. The design does not support that use. The numbers are averages from a defined group under a defined sequence. Individual changes varied, and subgroup analyses suggested that participants who lost more during treatment tended to regain more afterward while still retaining greater net loss. That observation is informative about the relationship between initial loss and later outcome, but it does not prove that initial loss causes regain or predict an individual’s path.
The missing continuation arm is especially important. Because the extension did not keep a randomized group on semaglutide after week 68, it cannot measure the difference between stopping and continuing under otherwise comparable conditions. It shows what followed withdrawal in the selected extension group. It does not show what would have happened if the treatment had remained active, if only the lifestyle support had stopped, or if the dose had been reduced. Those counterfactuals require different study designs.
The more transferable lesson is about treatment durability. A large on-treatment response can be real and clinically meaningful while still depending on continued pharmacological support. Stopping exposes what happens when that support is removed, but the result is shaped by the amount lost, the duration of treatment, the availability of lifestyle support, other medicines, and the person’s circumstances. The extension makes that dependency visible; it does not reduce it to a single number.
STEP 1 therefore supplies a strong observed example within one year and one treatment context. It shows substantial average regain, a remaining average loss below baseline, and movement in several health measures toward their starting state. It does not establish a final lifetime weight, a universal plateau, or the best way to stop. The case is most useful when read as a carefully bounded observation: after 68 weeks of semaglutide plus counseling, one year without either was associated with a major but incomplete reversal of the weight loss.
That concrete finding now has to be placed beside the broader evidence. The next task is not to repeat the STEP 1 numbers, but to decide which conclusion survives when a single extension, a pooled trajectory, and the limits of both are considered together.