What STEP 4 and SURMOUNT-4 actually established

There is a widely held assumption that the correct dose is the highest approved one, and the trials give surprisingly little support to it. In SURMOUNT-1 the difference between the 10 mg and 15 mg arms was smaller than the difference between 5 mg and 10 mg, and both upper arms sat well above placebo. The dose-response relationship in this class flattens as it rises. Whether the last rung is worth its additional symptom burden is a judgement about an individual, and it is one the printed ladder quietly forecloses by treating the maximum as the destination.

When the curve flattens, and why

In the sixty-eight-week semaglutide obesity trials, weight reduction decelerated visibly from around week forty and was close to flat by week sixty. Extended follow-up to a hundred and four weeks found the reduction broadly maintained rather than extended, which is the clearest available statement that the plateau is a plateau and not a pause.¹

The mechanism is not mysterious. Energy requirement falls with mass. A person who has lost fifteen per cent of body weight requires materially less energy at rest and in movement, and the intake reduction produced by a fixed dose of a fixed drug eventually equals that lower requirement. Balance is restored and weight stops changing. The drug has not weakened; the target has moved.

Two inferences commonly drawn from a plateau are unsupported. The first is that receptors have desensitised — difficult to reconcile with the rapid and near-complete regain observed on withdrawal. The second is that the dose must therefore be increased, which in the trials produced a further step down the flattening dose-response curve rather than a resumption of the earlier slope. The plateau is predictable, is predicted, and is almost never mentioned to anyone before it happens.

The target dose and the maximum dose are different concepts and the trade treats them as synonyms. A target is where a trial aimed; a maximum is where a label stops. Somebody doing well below the target has not failed to reach anything.

Target dose against maximum dose

The SURMOUNT-1 results are the clearest available illustration of a flattening curve. At seventy-two weeks the mean weight reductions were approximately fifteen per cent at 5 mg, nineteen and a half per cent at 10 mg and twenty-one per cent at 15 mg, against about three per cent on placebo.² The step from 5 mg to 10 mg bought roughly four and a half percentage points; the step from 10 mg to 15 mg bought roughly one and a half.

Set against that, gastrointestinal adverse events and discontinuation for adverse events both rose across the dose range. The question of whether the top rung is worth climbing is therefore a genuine trade-off rather than a formality, and it will have different answers for different people.

The Journal has no view on where any individual should stop. We have a strong view on how the question should be framed: not as whether to reach the maximum, but as what the next increment is expected to add and what it is expected to cost. Framed that way, a decision to remain at an intermediate dose is a defensible reading of the dose-response data rather than a compromise.

What the withdrawal designs established

Two trials in this class were built specifically to answer what happens when treatment stops. In the semaglutide programme, participants who had escalated to the top dose over twenty weeks were then randomised to continue or to switch to placebo; those continuing lost a further eight per cent of body weight over the following forty-eight weeks while those withdrawn regained about seven per cent.³ In the tirzepatide programme, a thirty-six-week open-label lead-in was followed by randomised continuation or withdrawal, with the continuation group losing a further five and a half per cent and the withdrawal group regaining approximately fourteen per cent.⁴

These are among the most informative results in the field and they are frequently over-read. What they establish is that the effect is maintained by continued exposure and reverses without it. What they do not establish, because neither design examined it, is whether a reduced maintenance dose would hold the result. The comparison was full dose against nothing.

Given that cost is the leading reported reason for stopping, a randomised comparison of full-dose against half-dose maintenance would be one of the highest-value trials nobody has run.

Trial protocols permitted holding and delay. The schedules people now treat as rigid were operated with discretion in the studies that produced the evidence, and the escalation sections of those protocols say so.

Maintenance dosing: the honest state of the evidence

Practices described to us include remaining at the dose that produced the result, reducing by one rung after target is reached, extending the interval to ten or fourteen days, and stopping entirely with a plan to resume on regain. The first is what the trials tested. The others are extrapolations of varying boldness.

The interval-extension approach deserves a specific caution. Because exposure is governed by the ratio of half-life to dosing interval, moving from weekly to fortnightly dosing on a seven-day half-life does not halve average exposure — it reduces it and also converts a fairly smooth concentration profile into a pronounced peak-and-trough cycle. Whether appetite regulation tolerates that oscillation is an empirical question and the answer is not in the literature.

Our position is that maintenance is the largest under-studied decision in the treatment course, that the trials answer continuation against cessation and nothing in between, and that anyone presenting a specific maintenance protocol as evidence-based is overstating what exists. Readers who know of a randomised maintenance-dose comparison we have missed should write to standards@compoundjournal.com.

Five things about titration nobody can currently answer

First, the optimal escalation interval. No adequately powered randomised comparison of intervals at a fixed target dose exists for any molecule in this class.

Second, the optimal hold duration for a person who has not adapted at four weeks. Practice ranges from four to twelve weeks on no comparative evidence at all.

Third, the lowest maintenance dose that preserves a result. The withdrawal trials compared full dose with nothing.

Fourth, whether tolerability at one rung predicts tolerability at the next. Clinicians assume it does, plausibly, and the published dose-ranging data is not analysed in a way that answers the question.

Fifth, whether any measurable baseline characteristic predicts the ceiling. Nothing published does so usefully, which mirrors the situation for efficacy: mean behaviour in this class is well characterised and individual variation is not.⁵

The Journal lists these not as a complaint about researchers but as a map of where confident advice is currently outrunning its evidence. Anyone offering a precise answer to any of the five is offering an opinion, and should be read as doing so.

The Journal ends where the evidence does. Escalation intervals in this class rest on a kinetic argument that is sound and on a randomised comparison that does not exist. Dose holding rests on a documented adaptation and on clinical consensus. Maintenance rests on two withdrawal trials that answered a narrower question than the one readers ask. None of that makes the current practice wrong; it makes it provisional, and provisional practice deserves to be described as such rather than printed as a table.