Does maintenance at a fraction of the therapeutic dose actually work, and is there any evidence for it?

Accepted answer

There is no randomised trial of reduced-dose maintenance for either molecule. Every maintenance trial that exists compared the full dose against placebo, and both showed the full dose works. Reduced-dose maintenance is an inference from the dose-response curve plus community practice, and it should be labelled as such.

What the trials did and did not test

• STEP 4 randomised participants after a run-in to either continuing semaglutide 2.4 mg or placebo. Continuing produced further loss; placebo produced regain [1]. No reduced-dose arm.
• SURMOUNT-4 randomised after a 36-week lead-in to continuing max tolerated tirzepatide or placebo. Continuing gave a further 5.5% loss; placebo gave 14.0% regain [2]. No reduced-dose arm.
• STEP 1 extension observed one year off treatment entirely [3]. No reduced-dose arm.

So the evidence base is binary: full dose or nothing. Note also that in both withdrawal trials the continued-treatment arms kept losing weight, which is direct evidence that the full dose delivers more appetite effect than maintenance requires. That is the strongest available argument for reduced-dose maintenance, and it is indirect.

The dose-response reasoning

The relationship between dose and weight effect is monotonic and roughly saturating rather than binary. Approximate placebo-adjusted weight change by dose in the pivotal programmes:

Read that table with maintenance in mind. A dose that produces 15% loss in a treatment-naive person is producing far more appetite suppression than is needed to hold a stable weight in someone already at that weight, because the counter-regulatory drive you are opposing at maintenance is smaller than the drive plus the deficit you were opposing during loss. The inference that a fraction of the dose suffices is reasonable. It is still an inference.

What people report

Consistent patterns across reports, offered as reported practice and not as recommendation:

• Stepping down one or two levels and holding, for example 12.5 to 7.5 or 5 mg, is the most commonly described approach and the one most consistent with the pharmacology.
• Reduced dose at the same weekly interval is reported to work better than the same dose at an extended interval, because exposure stays flat rather than cycling. People on fortnightly intervals frequently describe appetite returning in the last three or four days.
• The minimum effective maintenance dose varies a great deal between individuals, and people who lost the most weight tend to report needing more of it, which is what you would expect if counter-regulation scales with the size of the loss.
• Failures are usually described as gradual rather than sudden: nothing changes for six weeks, then weight begins a slow, steady climb. That slow onset is exactly why pre-set stopping rules matter.

Finding your own minimum without uncontrolled regain

Set the rules before you start, then follow them mechanically:

1. Establish a baseline. Hold your current dose and a stable weight for at least eight weeks, recording rolling seven-day averages. You cannot detect a change against an unstable baseline.
2. Define your ceiling in advance. Pick a number, for example baseline plus 2 kg on a rolling average sustained for two weeks. Write it down.
3. Step down one level. Wait six weeks for the new steady state, then collect three more weeks of trend data. Nine weeks per step. This is slower than anyone wants and it is the reason people get unreliable answers.
4. If the ceiling is breached, step back up immediately. Not next month. The 2 kg ceiling exists so that the cost of a failed step is 2 kg, which is trivially recoverable, rather than 8 kg, which is not.
5. If stable, consider one more step down. Stop when you find the level where stability fails, then hold one level above it.

Expect this to take six to nine months to complete properly. That is the honest timescale, and rushing it produces a wrong answer plus regain.

When discontinuation is clinically driven rather than chosen

Distinct from all of the above, some reasons to stop are not optional and are not about weight. Worth knowing so that they are recognised rather than negotiated:

• Pregnancy or planning pregnancy. Labels for this class specify discontinuation well in advance of a planned conception, and the interval reflects the long half-life.
• Suspected pancreatitis. Severe persistent abdominal pain radiating to the back is a stop-and-seek-care situation, not a dose-adjustment situation.
• Symptomatic gallbladder disease, which is more frequent with rapid weight loss.
• Planned surgery or any procedure under sedation, because delayed gastric emptying raises aspiration risk and anaesthetic guidance on timing is specific.
• Persistent vomiting with signs of dehydration or acute kidney injury.
• Progressive loss of lean mass or functional decline, particularly in older people, where continuing may be causing more harm than the remaining weight.
• Reaching a weight where further loss is not clinically desirable, which is a real and under-recognised endpoint.

None of that is a taper question. Those are clinician conversations, and several of them are urgent ones.

edited 27 May 2026 by linnea_wahlberg — corrected a unit error in the worked example