What the Maintenance Trials Showed After the Drug Was Stopped

Answer first: two trials answered this directly, and both went the same way

There is a trial design built for the question people actually have. Give everyone the drug openly for several months. Then randomize them: half continue, half switch to placebo, and neither the patients nor the investigators know which. Follow both groups for a year. Whatever separates them from that point on is the drug.

Two of the five labels behind these products contain such a trial. On the tirzepatide weight label it ran 88 weeks in total. On the semaglutide weight label it ran 68 weeks. Both report the same shape of result: the group that continued kept losing slowly, and the group switched to placebo gained.

In the tirzepatide trial, from the point of randomization to the end, the placebo group's body weight rose by 14.0 percent while the continuing group's fell a further 5.5 percent. In the semaglutide trial, the placebo group rose 6.9 percent while the continuing group fell 7.9 percent. Two molecules, two designs, one direction.

That is the evidence base. It is smaller and more specific than the confident claims made around it, and it comes with a caveat the labels themselves print, which we get to below.

What 'maintenance' means on these labels — two different things

The word does double duty and it causes confusion. A maintenance dosage is a stage in a schedule, a matter for the prescriber and not for this article. Long-term maintenance of weight reduction is an outcome, and it is written into the indication itself: these products are approved to reduce excess body weight and maintain weight reduction long term.

Both weight labels title the main part of their Clinical Studies section accordingly — weight reduction and long-term maintenance studies. So the maintenance question is not an afterthought bolted onto a weight-loss claim. It is half of what these drugs are approved to do, and it is the half the withdrawal trials were designed to test.

That framing is also why the labeling treats this as an ongoing treatment for a chronic condition rather than a course with a finish line. The trials below are the reason.

The tirzepatide withdrawal trial, in detail

The trial ran 88 weeks and enrolled 783 adults with obesity, or with overweight and at least one weight-related condition. People with type 2 diabetes were excluded. Everyone received the drug openly for the first 36 weeks, alongside the same diet and activity program used in the other weight trials.

By week 36, 14.4 percent of the original 783 had already stopped, with adverse events the most common reason at 6.8 percent. The remaining 670 were then randomized one-to-one to continue or to switch to placebo for 52 weeks.

The open-label phase worked: mean body weight fell from 107.3 kilograms at the start to about 85 kilograms at randomization, an average reduction of 20.9 percent. Then the two groups separated. From randomization to week 88, the placebo group's weight rose 14.0 percent and the continuing group's fell 5.5 percent, a difference of 19.4 percent with a confidence interval from 17.7 to 21.2.

One more number from that trial is worth knowing. After randomization, 17.9 percent of the placebo group stopped taking their assigned treatment against 10.4 percent of the group that continued the drug.

The semaglutide withdrawal trial, in detail

This one ran 68 weeks and enrolled 902 adults with obesity, or overweight with at least one weight-related condition, again excluding type 2 diabetes. Everyone received the drug during a 20-week run-in.

By the end of the run-in, 99 of the 902 — 11 percent — had permanently stopped, most commonly because of adverse reactions at 5.3 percent. The remaining 803 were randomized two-to-one to continue or to switch to placebo for the next 48 weeks.

Mean body weight fell from 107.2 kilograms at the start to roughly 96 kilograms at randomization. From randomization to week 68, the placebo group rose 6.9 percent and the continuing group fell 7.9 percent, a difference of 14.8 percent with an interval from 13.5 to 16.

The smaller placebo rise here than in the tirzepatide trial is not a comparison between the drugs. The two trials had different lead-in lengths, different randomization ratios, different follow-up windows and different amounts of weight lost before randomization. They are two separate experiments that happen to agree on direction.

Weight was not the only thing that moved back

This is the part almost never quoted and probably the most useful. Both trials tracked cardiometabolic measures across the randomized phase, and in both, the group that switched to placebo saw several of them travel back toward where they started.

In the tirzepatide trial, from randomization to week 88, the placebo group's waist circumference rose 7.8 centimeters while the continuing group's fell 4.3. Systolic blood pressure rose 8.2 millimeters of mercury on placebo against a 2.0 rise on continued treatment. Diastolic pressure rose 3.2 against a 0.7 fall. HbA1c rose 0.3 percentage points against no change. Triglycerides rose 13.5 percent against a 4.8 percent fall.

In the semaglutide trial, from randomization to week 68, the placebo group's waist circumference rose 3.3 centimeters against a 6.4 centimeter fall on continued treatment. Systolic pressure rose 4.4 against a 0.5 rise. Total cholesterol rose 11.4 percent against a 4.9 percent rise, and LDL cholesterol rose 7.6 percent against a 1.1 percent rise.

For anyone whose reason for treatment was never only the number on a scale, that is the relevant table. It is also, notably, the set of measures a prescriber is most likely to be watching.

One measure moved the other way, and it belongs in the record

Heart rate. Both of these drugs raise it modestly, and the semaglutide weight label carries a Warnings and Precautions subsection on heart rate increase.

In both withdrawal trials, the group that switched to placebo saw pulse fall further than the group that continued. In the tirzepatide trial, pulse rate fell 5.2 beats per minute on placebo against 2.1 on continued treatment — and the same table shows pulse had risen 5.0 beats per minute during the open-label phase when everyone was on the drug. In the semaglutide trial, heart rate fell 5.3 on placebo against 2.0 on continued treatment.

So the heart rate effect these drugs produce appears to reverse on stopping, in the same trials where the weight effect reversed. Reporting only the numbers that point one way would be a smaller kind of dishonesty than inventing them, but it would still be one.

The caveat both labels attach, and why it changes the number

Everyone in the randomized part of these trials had already done two things: tolerated the drug for months, and reached the highest amount studied. Anyone who stopped during the escalation, or who never got that far, was not eligible.

One label says so outright. Because patients who discontinued during titration and those who did not reach the highest weekly amount were not eligible for the randomized-treatment period, the results may not reflect the experience of patients in the general population who are first starting treatment.

The numbers behind that sentence are printed on the same pages: 11 percent of the 902 in the semaglutide trial never reached randomization, and 14.4 percent of the 783 in the tirzepatide trial stopped before it. So roughly one in eight people who started these trials was filtered out before the maintenance question was even asked. The results describe people who got through that filter.

That cuts both ways for a reader. It means the reported loss during the lead-in is flattering relative to a general population. It also means the reported regain after stopping describes people who had responded well, which is arguably the group most likely to be told they can come off.

Two of these labels contain no such trial at all

Worth stating precisely, because absence is easy to assume and hard to prove. Searching the full text of all five labels in one pass, the words maintenance and withdrawal appear zero times on the orforglipron label and zero times on the tirzepatide diabetes label — while the word discontinued appears 13 and 17 times on those same two files in the same pass. The ruler reads them. There is simply no maintenance or withdrawal trial on either.

The semaglutide diabetes label has 15 mentions of maintenance and none of withdrawal, and its Clinical Studies section contains glycemic, cardiovascular and kidney trials rather than a withdrawal design.

So the direct evidence on what happens when treatment stops covers two products. For the newest tablet in this class, that trial has not been done, or at least has not reached the label.

What this is and is not useful for

This is useful for setting an expectation before you start, and for having a specific conversation rather than a vague one. Two randomized trials found that stopping did not hold the loss, and that several measures beyond weight moved back as well. That is a real finding from a design built to produce it.

It is not a prediction about you. A trial reports a group average over a fixed window in a filtered population, and the spread behind an average like 14.0 percent is not printed anywhere. Nor is it a reason to do anything in particular — stopping, continuing, and every decision about a schedule belong to the person who wrote the prescription, made with you.

The useful version of this in an appointment sounds like a question rather than a plan. If treatment ends, what does follow-up look like and which of these measures would you want checked? That question has an answer specific to your history, and the trials above are only the reason for asking it.