GLP-1 Maintenance Dose: What Lowering to 5 mg Costs You

Can You Lower Your GLP-1 Dose and Keep the Weight Off?

Most of it, yes — and there is finally a randomized trial behind the answer. In SURMOUNT-MAINTAIN, people who cut tirzepatide from their top dose down to 5 mg kept 67.9% of the weight they had lost. People who stayed on the top dose kept 96.5%. People who stopped kept 42.8%. Lowering your dose works far better than quitting. It is not free.^[1]

Here is why you probably have not read this anywhere else. The trial published online in The Lancet on 12 May 2026. As far as we can tell, most of the plain-English pages you will find on “can I lower my tirzepatide dose” were written before it existed, which is why so many of them say, accurately for their date, that nobody had studied the question. Somebody has now. This article is our attempt to translate it without shaving off the parts that matter.^[1]

The honest answer here is two-sided, and the two-sidedness is the whole point. If you came looking for permission to cut your dose, you will find real evidence that it works better than most people assume. If you came looking for reassurance that it is free, you will not find that, and we are not going to pretend otherwise.

What Does “Maintenance Dose” Actually Mean?

A maintenance dose is the amount you settle on once you are done climbing. GLP-1 medications start low and step up on a fixed schedule so your gut has time to adjust. The dose you land on and keep taking is the maintenance dose. It is a label term with a specific legal meaning, not a vibe, and not every dose in the ladder qualifies as one.^[5]

Five words do most of the work in this article, and every one of them gets used loosely on the internet. Here they are in plain English.

If any of that is new, our beginner GLP-1 guide covers the basics from the top, including a plain FAQ on what happens when you stop. For the full tirzepatide titration schedule and how the doses relate to each other, we keep the tirzepatide reference page as the deeper hub, and there is a general peptide dosage guide for how dosing conventions work across compounds.

What Did the SURMOUNT-MAINTAIN Trial Actually Do?

SURMOUNT-MAINTAIN put 378 people who had already lost weight on tirzepatide into three groups and blinded everyone to which group they were in. One group kept their top dose, one dropped to 5 mg, and one switched to placebo. Then it followed them for another full year. It is the first randomized test of this exact question.^[1][3]

The design matters, so here it is in order. The trial ran across 20 US sites and covered 112 weeks from the first injection to the last measurement. There was a two-week screening period, then a 60-week open-label lead-in in which everybody knew they were on tirzepatide and everybody climbed to their maximum tolerated dose, then a 52-week double-blind maintenance period in which the three groups split apart.^[1]

441 people started the lead-in. 378 made it to randomization at week 60 and were split three-to-three-to-two. Getting that far was not automatic: to be randomly allocated you had to have lost at least 5% of your bodyweight during the lead-in and to have tolerated a dose of at least 10 mg, so these were not simply everyone who finished the sixty weeks. 372 actually received maintenance-period study drug, and 345 of the 378 (91%) completed the trial. A 91% completion rate over more than two years is unusually good, and it is one of the reasons this trial is worth taking seriously.^[1]

Who was in it: adults with a BMI of 30 or higher, or 27 or higher with at least one weight-related health condition. Mean age was 46.6 years, 65% were women, 67% were White, mean starting BMI was 40.1, and mean starting weight was 113.8 kg, or 250.9 lb. Mean HbA1c — average blood sugar over about three months — was 5.64%, which is a non-diabetic range. People with type 2 diabetes were excluded outright, which is the single most important thing to know about who this trial does and does not describe.^[1]

The headline result

The primary endpoint was percent change in body weight from week 0 to week 112 — that is, from before anyone had taken a single injection, all the way to the end.

Negative numbers mean weight lost. Figures use the trial's main counting rule, which keeps people in their assigned group even if they later needed rescue treatment — so the numbers reflect the whole strategy, not just the people who stuck with it. A 95% confidence interval is the band the true answer is very likely to sit inside; a narrow band means the estimate is tight. Compared with stopping, the 5 mg group did 6.6 percentage points better (95% CI −8.3 to −5.0) and the top-dose group did an estimated 12.0 percentage points better. All comparisons were statistically significant at p<0.0001. The 138/142/90 counts are the efficacy analysis population — everyone randomly allocated who received at least one dose in the maintenance period — which is why they are smaller than the 378 randomly allocated. Pounds are converted from the trial's reported kilogram figures; the confidence intervals shown belong to the percentages.^[1]

Read that table once and you get the ranking: staying beats lowering, lowering beats stopping, and the gaps are large. Read it twice and you should get suspicious, because those numbers do not answer the question you actually asked. That is the next section, and it is the most important one on this page.

The Two Different Zeros: The Number Nearly Everyone Will Get Wrong

The headline numbers from SURMOUNT-MAINTAIN are measured from week 0 — the day people started the drug, sixty weeks before anyone changed anything. So they bundle together the weight lost during the climb and whatever happened afterwards. They are not a measure of what lowering your dose cost you. For that you need a different zero: week 60, the day the dose actually changed.^[1]

Think of it as two clocks. One starts on your first injection. The other starts the day you change something. Almost every write-up of a maintenance trial reports the first clock, because that is the headline number, and almost every reader mentally converts it into the second clock, because that is the question they came with. Those are different quantities.

The number you actually want is how much of your lost weight you kept. The trial reported that separately, among the participants who had reached a genuine weight plateau before randomization. Here it is.

Retention is measured from week 60 among participants who had reached a weight plateau, and each person's own value is capped at 0–100%, so it is not simply the mirror image of regain. The confidence interval shown is the range around the estimated group average, which is why it can extend past 100 even though no individual score does. The regain column is the trial's own separate measurement, taken across all randomly allocated participants.^[1]

Now the two clocks tell different stories, and both are true. On the first clock, the 5 mg group finished down 16.6% from where they started — a very good outcome by any historical standard. On the second clock, that same group gave back about a third of the weight they had lost — roughly 7% of body weight, about 13 lb at this trial's average. Across the group, a little over two-thirds of the loss held. Individual results varied widely — that average sits on top of people who kept everything and people who kept almost none.

This is the same discipline we apply to every number on this site: ask what the number is counting before you decide how you feel about it. “Down 16.6% from the start” and “kept 67.9% of what I lost” describe the identical group of people. One sounds like a win and one sounds like a compromise. They are both accurate, and you need both to make a decision.

One more thing worth naming, and it deserves the same care we just asked of you: the group that stopped the drug entirely was still down 9.9% from baseline at week 112, and among the plateau reachers in that group, 42.8% of the loss was still intact. Those two figures describe slightly different sets of people — 90 and 76 respectively — so read them as two views of the same arm rather than one continuous story. Two-thirds of that group also restarted tirzepatide as rescue during the year, and the −9.9% is a modelled estimate of what would have happened without it. Quitting was clearly the worst of the three options tested, and it was not a full return to the starting line either. Both halves of that are worth holding.

What Does Lowering Cost, and What Does Quitting Cost?

Lowering to 5 mg costs about five percentage points of body weight compared with holding the top dose, plus about three times the rate of needing rescue treatment. Quitting costs far more on both counts. Set against each other, lowering looks much closer to staying than to stopping — but it is not the same as staying.^[1]

Start with the scale. The gap between the top-dose group and the 5 mg group was 21.9% versus 16.6%, a little over five percentage points, or about 5.8 kg (12.8 lb), taken from the trial's own kilogram figures. The gap between 5 mg and stopping was 6.6 percentage points. In other words, the distance from the top dose down to 5 mg was slightly smaller than the distance from 5 mg down to nothing. Lowering keeps you on the better side of the bigger gap.

Now the part that gets less attention and probably deserves more. Rescue tirzepatide was offered to anyone who regained at least half of what they had lost, and it was deliberately unavailable during the first 24 weeks of the maintenance period so that early regain could be observed honestly. Here is how often it was needed.

These are the corrected figures. A published erratum fixed the denominator behind the placebo group's rescue rate, and we have used the post-correction numbers throughout.^[2]

One in four people who lowered to 5 mg regained enough weight to qualify for rescue treatment within a year. That is a meaningfully worse outcome than staying put, and it is exactly the kind of number that disappears when a result gets compressed into “you can lower your dose.” It is also dramatically better than the two in three who needed rescue after stopping.

One thing genuinely went the other way, and it belongs here. The trial reported fewer gastrointestinal side effects in the 5 mg group, which the authors read as better tolerability at the lower dose. They noted that in practice a prescriber might reduce the dose specifically to make the medication easier to live with. If side effects are what is pushing you toward stopping altogether, that is the trade-off this trial actually speaks to.^[1]

Beyond averages, the trial reported how the results were distributed. Among the plateau reachers, 77.5% of the top-dose group (90 of 116), 42.4% of the 5 mg group (50 of 118) and 10.4% of the placebo group (8 of 76) held onto at least 80% of their lead-in weight loss. This was a pre-specified key secondary endpoint, controlled for multiple testing, with p<0.0001 for both comparisons against placebo.^[1]

How does this square with what we already knew about stopping? It fits. The extension of the STEP 1 semaglutide trial found participants regained roughly two-thirds of their lost weight within a year of stopping the drug.^[7] Here, the group that stopped tirzepatide regained 15.2% of their body weight over the following 52 weeks, which left them holding 42.8% of what they had lost. Different drug, different trial, same direction. Our tirzepatide versus semaglutide comparison sets the two drugs' regain data side by side and reads it as a class effect rather than something that separates them.

Who Does This Trial Not Apply To?

SURMOUNT-MAINTAIN describes a specific group of people: adults without type 2 diabetes, taking tirzepatide for weight, who had climbed to 10 or 15 mg and had already plateaued. If you are outside that description, the trial does not transfer to you automatically, and in one case it very clearly does not transfer at all.^[1]

• You have type 2 diabetes. Diabetes was an exclusion criterion. Average blood sugar in the trial sat in the non-diabetic range. Nothing here tells you what lowering a dose does to blood sugar control, because that question was designed out of the study.^[1]
• You take Zepbound for obstructive sleep apnea. This is the distinction almost nobody is drawing, and it is the most consequential one on the page. 5 mg is an approved maintenance dosage for weight reduction only. For obstructive sleep apnea, Zepbound's recommended maintenance dosages are 10 mg or 15 mg. If you are on it for sleep apnea, a weight-maintenance trial is not a license to go lower, and that conversation belongs entirely with your prescriber.^[5]
• You are on semaglutide, not tirzepatide. This trial tested one molecule. There is no equivalent randomized trial for lowering a semaglutide maintenance dose, so the result does not carry across. Our semaglutide reference page covers what is known about its own dosing ladder.
• Your top tolerated dose is 7.5 mg or 12.5 mg. In this trial, “maximum tolerated dose” meant 15 mg or 10 mg and nothing in between. If your ceiling sits at one of the intermediate rungs, you were not represented in either comparison group.^[1]
• You have not plateaued yet. The retention figures — the 96.5 / 67.9 / 42.8 numbers — were measured among people whose weight had changed less than 5% between weeks 48 and 60. Someone still actively losing is in a different situation.^[1]
• You were already on a GLP-1 when you started. The trial excluded anyone who had used a GLP-1 or GIP/GLP-1 receptor agonist within 12 months of screening, so everyone entered incretin-naive and climbed from scratch. Type 1 diabetes was also an exclusion.^[3]
• You are hoping 2.5 mg counts. It does not. The Zepbound label states plainly that “the 2.5 mg dosage is for treatment initiation and is not approved as a maintenance dosage.”^[5]

One more scope note worth stating out loud: this was 52 weeks of maintenance. It is the longest randomized look we have at this question, and it is still only a year. It does not tell you what year three looks like on 5 mg.

Does Spacing Doses Out or “Microdosing” Work for Maintenance?

Nobody knows, and this trial does not help. SURMOUNT-MAINTAIN tested a lower weekly dose. It did not test a longer gap between doses. Everyone in the 5 mg group still injected once a week, on schedule, at a full approved maintenance dosage. Injecting every other week is a different intervention, and it has not been put through a comparable randomized trial.^[3]

This distinction is getting lost fast, so it is worth being blunt about. When people say “microdosing” a GLP-1, they usually mean one of two things: taking a fraction of a standard dose, or stretching a standard dose over a longer interval. The trial supports neither. The lowest arm tested was 5 mg weekly, which is a real maintenance dosage in its own right, one-third to one-half of the top dose rather than a sliver of it. And even that gave back about a third of the lost weight.

It is also worth understanding why spacing is not a small change. Tirzepatide's dosing interval is built around how long it stays active in the body, which is what makes once-weekly work in the first place.^[5] Doubling the gap between injections does not deliver half the drug evenly; it delivers a peak followed by a trough. Whether that trough matters for appetite and weight is an empirical question, and the answer is that we do not have the study. Be precise about what that means, though: untested is not the same as disproven. Nobody has shown that spacing doses out fails for maintenance. Nobody has shown it works either, and that is the honest state of it.

Two more things worth saying plainly. First, we are not aware of any head-to-head trial comparing dose reduction against dose spacing for maintenance, so pages that told you nobody had compared the two were correct, and are still correct on that specific point. Second, this is one of the places where GLP-1s behave differently from most of the compounds we cover: our guide to cycling peptides notes that GLP-1 medications are typically taken continuously rather than cycled on and off, and the regain data in this trial is a good part of why.

What Do the Labels Actually Allow?

The approved range of “maintenance” got wider at both ends in 2026. At the low end, 5 mg is already an approved Zepbound maintenance dosage for weight reduction, so this trial tested something a prescriber could legally already do. At the high end, Wegovy gained a 7.2 mg dosage in March 2026. Neither change is a free pass.^[5][6]

Zepbound, at the low end

Three things the label actually says, and they are easy to blur together:

1. 2.5 mg is not a maintenance dosage. In the label's words, it “is for treatment initiation and is not approved as a maintenance dosage.” It is the on-ramp.^[5]
2. 5 mg is an approved maintenance dosage — for weight reduction. That is what makes the trial's 5 mg arm a realistic option rather than an experiment in off-label territory.^[5]
3. For obstructive sleep apnea, the recommended maintenance dosages are 10 mg and 15 mg — 5 mg is not one of them. This is the sentence that governs the sleep apnea case above, and it is the reason a weight-loss result cannot simply be generalized to every Zepbound user.^[5]

Wegovy, at the high end

On 19 March 2026, the FDA approved a higher semaglutide dosage. The label now reads: “For patients who tolerate the 2.4 mg dosage for at least 4 weeks and additional weight reduction is clinically indicated, the dosage may be increased to a maximum dosage of 7.2 mg subcutaneously once weekly.”^[6]

Read the scope carefully, because it is narrow. This applies to weight reduction in adults only. Cardiovascular risk reduction stays at 2.4 mg or 1.7 mg. Pediatric use stays at 2.4 mg or 1.7 mg. The fatty-liver indication — MASH, a form of liver disease driven by metabolic problems — is 2.4 mg only. If someone does not tolerate 2.4 mg, the label allows dropping to 1.7 mg. It is also a physically distinct product, marketed as WEGOVY HD at 7.2 mg per 0.75 mL, and there is no intermediate rung — the label goes straight from 2.4 mg to 7.2 mg, a threefold jump.^[6]

We are keeping this brief because it is context rather than the subject. The point is directional: the ceiling and the floor of what a maintenance dose can be have both moved in the past year, which means “what dose should I be on to hold this?” is a live clinical question with more real options in it than it had in 2025.

How Do You Bring This Up With Your Prescriber?

You bring it as a question with evidence attached, not as a decision you have already made. Nothing in this article is a reason to change your own dose, and changing a prescription dose without your prescriber is a genuinely bad idea. What has changed is that you can now ask about lowering with a randomized trial behind the question instead of a hunch.

Things worth having ready before that conversation:

1. Your actual weight trend over the last three months. The trial's definition of a plateau was less than 5% change between weeks 48 and 60. Whether you have genuinely plateaued is the first thing that decides whether any of this applies to you.^[1]
2. Which dose you are on, and whether it is your ceiling or just where you stopped climbing. These are different situations and the trial only studied one of them.
3. Why you are on the medication. Weight reduction, sleep apnea, diabetes and cardiovascular risk are different indications with different recommended maintenance dosages. This is the question that most changes the answer.^[5]
4. What you would want to happen if weight came back. A quarter of the 5 mg group needed rescue treatment within a year. Agreeing in advance on what triggers a change — a specific number on the scale, a specific date to reassess — is the part of this that is genuinely within your control.^[1]
5. Your honest reason for wanting to lower. Cost, side effects, supply and simply disliking being on a drug forever are all legitimate, and they point to different conversations. A prescriber can only help with the one you actually name.
6. The two numbers from this trial. Kept 67.9% of lost weight on 5 mg versus 96.5% at the top dose, and 25% versus 8% needing rescue. Those two pairs are the whole trade-off in a form a busy clinician can absorb in ten seconds.^[1]

If the broader goal is what you are weighing rather than the dose itself, our weight loss overview covers the peptide options and how they compare.

How Good Is This Evidence, and Who Paid for It?

The evidence is strong for the narrow question it asked, and narrow in every other direction. It is randomized, double-blind, 91% complete, and it tests exactly the choice patients face. It is also a single trial of a single drug over a single year, run by the company that makes the drug. Both halves are true and neither cancels the other.^[1]

The funding, stated plainly: this trial was funded by Eli Lilly, and 7 of the 13 authors are Lilly employees and shareholders, including the corresponding author, who is also the named guarantor of the data. Of the six academic authors, five report consulting or speaking fees from Lilly and from Novo Nordisk; one reports none. This is a manufacturer-run trial of the manufacturer's own product.^[1]

That is not a reason to throw the result out. Most large obesity trials are industry-funded, the study was registered in advance, the design was published separately in 2025 before results existed, and a 91% completion rate is hard to fake.^[4] It is a reason to notice which findings serve the sponsor. Worth observing: the result that most favors Lilly commercially is “stay on the top dose,” and that is what the trial found. It is also, as far as we can tell, what the data actually shows. Hold both.

The rest of the caveats, in descending order of how much they should change your reading:

• There was a published erratum. A correction was applied online on 4 June 2026. It fixed the denominator behind the placebo group's rescue-tirzepatide rate and the y-axes on two panels of one figure. Every number on this page is post-correction. We are mentioning it because a corrected paper is a working paper, and you deserve to know which version you are reading.^[2]
• Registration, protocol and publication disagree on which outcome was primary. ClinicalTrials.gov and the 2025 design paper both pre-specified percent maintenance of weight loss among plateau reachers as the primary endpoint, and the design paper's sample size was powered on it. The published paper presents the week-0 change as primary and demotes maintenance to a key secondary. Both endpoints were pre-specified and both are reported, so nothing was hidden — but the switch happened after the protocol was published, it was not explained in the paper, and it moved the headline onto the more flattering of the two numbers. Notice it.^[3][4]
• We have left one confidence interval out on purpose. The published abstract contains an obvious typographical error in the interval around the top-dose comparison, so we have quoted the single best estimate and not the range around it. We would rather leave a visible gap than print a number we cannot verify.
• The population was heavier than average and US-only. Mean starting BMI was 40.1 across 20 US sites. That is a more severe starting point than many readers will recognize as their own.^[1]
• Fifty-two weeks is not “forever.” The question people are really asking is about the next decade. This answers the next year.

Evidence grade, in one line: the strongest randomized answer that exists to this question, from a trial with an obvious commercial interest in the answer, covering one drug, one year and one kind of patient. That is a good deal better than where this question sat twelve months ago, and a long way short of settled.

The Bottom Line

“Do I have to stay on this forever?” is one of the most common questions in weight medicine, and for years the honest answer was that nobody had tested lowering the dose. Trials had compared staying on the drug with stopping it altogether; the middle option was the untested one. Somebody has now. The answer is a middle answer, which is usually a sign it is a real one.

Lowering to 5 mg is a legitimate option that beats stopping by a wide margin. About two-thirds of the weight loss held, against just over two-fifths for the group that quit. If cost, side effects or the sheer weight of being on a top dose indefinitely are pushing you toward stopping altogether, this trial says there is a rung between here and there, and the rung holds better than most people would guess.

And it costs you. About five percentage points of body weight compared with staying put, about seven percent of body weight handed back, and about three times the rate of needing rescue treatment inside a year — 25% on 5 mg against 8% at the top dose. Anyone who tells you the trial endorses microdosing your way to maintenance has read the headline and skipped the second table.

The most useful habit this article can leave you with is not a dose. It is the two-clocks question. When you see a maintenance number anywhere — a study, a forum post, a clinic's marketing — ask which zero it is measured from. “Down 16.6% since I started” and “kept 67.9% of what I lost” are the same people. Knowing which one you are being shown is most of the skill.

If you want the underlying detail, start with the tirzepatide hub and its dosing section, then the tirzepatide versus semaglutide comparison, which lays the two drugs' regain data side by side and concludes it looks like a class effect rather than a difference between them. If you are earlier in the process, the GLP-1 guide is the place to begin. And then take the question to the person who writes your prescription, because that is where it belongs.