Accepted answer
1 mg a week is 0.143 mg a day averaged out and 52 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 1 mg is which arm it corresponds to: if a programme ran 1 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 1 mg a week a 10 mg vial is 10 weeks and you will need about 6 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.
The underlying point is that if the result deteriorates on a lower dose, returning to the previous one is straightforward and does not require re-titration from the bottom provided the gap has been short.
Dose-response for weight in the trial programmes was real but flattening at the upper end, which is consistent with a lower maintenance requirement.
A noisy weight signal makes premature conclusions easy, in both directions.
Glycaemic maintenance gives a faster signal than weight maintenance.