7.5 mg a week is 1.071 mg a day averaged out and 390 mg over a year — but "defensible" is not a property of the number, it is a property of where the number came from. A maintenance dose is defensible when a trial randomised people to it and reported what happened, and indefensible when it was arrived at by interpolation between two doses that were studied. So the question to ask of 7.5 mg is which arm it corresponds to: if a programme ran 7.5 mg as a maintenance level, there is an efficacy figure, a tolerability figure and a discontinuation rate attached to it. If it sits between two studied levels, everything said about it is extrapolation, and the burden of that is on whoever proposed it. The other half of the arithmetic is supply: at 7.5 mg a week a 10 mg vial is 1.33 weeks and you will need about 39 of them a year, which is worth knowing before the dose is settled rather than after. Maintenance doses are set by a prescriber against an individual; nothing here is medical advice.
Start with what is being maintained — weight, glycaemia or both — because they have different dose-response curves.
The withdrawal trials — STEP-4 and SURMOUNT-4 — established what happens when treatment stops entirely. They did not evaluate dose reduction, so the evidence for a lower maintenance dose is inference rather than data.
Label titration ladders, structure only
| Agent | Start | Step interval | Maintenance range | Max studied |
|---|
| Semaglutide (weight management) | 0.25 mg/wk | 4 weeks | 1.7–2.4 mg/wk | 2.4 mg/wk |
| Semaglutide (T2DM) | 0.25 mg/wk | 4 weeks | 0.5–2.0 mg/wk | 2.0 mg/wk |
| Tirzepatide | 2.5 mg/wk | 4 weeks | 5–15 mg/wk | 15 mg/wk |
| Liraglutide (weight management) | 0.6 mg/day | 1 week | 3.0 mg/day | 3.0 mg/day |
| Oral semaglutide | 3 mg/day | 4 weeks | 7–14 mg/day | 50 mg/day (trial) |
Structure is identical across the class: small start, four-week steps, a defined maintenance range, a defined ceiling.
The maintenance dose is not necessarily the same a year later, since the counter-regulatory response attenuates slowly if at all.
A noisy weight signal makes premature conclusions easy, in both directions.
The lowest dose that holds the result is the answer, and it is individual.
edited 17 Jun 2025 by fiadh_cronin — added the citation requested in comments