The scale stops moving, the dose has not changed, and nothing obvious has gone wrong. The usual advice at this point is to eat less and try harder, which misreads what is happening. A weight-loss plateau is the predicted end state of a control system, not evidence of failure — and the modeling work explains precisely why GLP-1 medications reach it later than dieting does, and why they reach it at all.
Why a plateau happens at all
The clearest account comes from mathematical modeling of the real trajectories. Using a validated model of energy metabolism and body-composition dynamics, one analysis simulated the mean weight-loss curves for diet restriction, semaglutide 2.4 mg, tirzepatide 10 mg and Roux-en-Y gastric bypass, fitting each to two parameters: a persistent shift of the system away from its baseline equilibrium, and the strength of a feedback circuit linking weight loss to increased appetite.[1]
Both parameters matter, and they explain different things. Gastric bypass produced a persistent effect more than threefold greater than diet restriction and about double that of tirzepatide and semaglutide. But the finding that explains the plateau is the second one: every intervention except diet restriction substantially weakened the appetite-feedback control circuit, and it is that weakening which produces an extended period of weight loss before the plateau arrives.[1]
Read that as a tug of war. As you lose weight, appetite regulation pushes back with increasing force. Diet restriction does nothing to that force, so the pushback catches up quickly and the curve flattens early. GLP-1 medications blunt the force itself, so the curve runs much longer before the two sides balance. But they blunt it rather than abolish it, so a balance point still exists. Reaching it is the system working as described, not the drug failing.
Where the trial curves actually flatten
This is the part that reframes most people’s plateau. The pivotal trials did not run for twelve weeks; they ran for well over a year, and substantial weight loss kept accruing late.
+7.9%
Additional weight lost between week 20 and week 68 on continued semaglutide
STEP 4
+5.5%
Additional weight lost between week 36 and week 88 on continued tirzepatide
SURMOUNT-4
25.3%
Total mean weight reduction by week 88 with continued tirzepatide
SURMOUNT-4
In STEP 4, participants completed a 20-week run-in with a mean weight loss of 10.6%, and those randomized to continue semaglutide lost a further 7.9% over the following 48 weeks.[2] In SURMOUNT-4, a 36-week open-label lead-in averaged 20.9%, and participants who continued tirzepatide lost a further 5.5% by week 88, reaching 25.3% overall.[3] Both of those late-phase losses happened after the point at which most people conclude they have stalled.
| Trial | Loss by the interim point | Further loss on continued treatment | Period |
|---|---|---|---|
| STEP 4 (semaglutide 2.4 mg) | 10.6% at week 20 | 7.9% | weeks 20–68 |
| SURMOUNT-4 (tirzepatide 10 or 15 mg) | 20.9% at week 36 | 5.5% | weeks 36–88 |
What to check before calling it a plateau
Three things are worth ruling out, in this order, and none of them is willpower.
Time. Weekly weight is noisy; fluid shifts alone can mask several weeks of real change. A plateau is a trend over months, not a flat fortnight. Against trial curves that were still descending at week 68 and week 88, a stall at week 14 is not yet a plateau at all.
Dose. Both labels approve more than one maintenance dose, and titration is deliberately slow. If you settled at a lower rung early because of side effects, the ladder still has room — though the returns diminish near the top, which is set out in GLP-1 maintenance dosing.
Composition. Scale weight conceals what is being lost. If resistance training and protein intake have gone up, the number can stall while body composition improves — and preserving lean mass matters independently, as covered in GLP-1s and lean-mass loss and sarcopenia treatment.
The honest bottom line
Plateaus are built into how body weight is regulated. The modeling shows GLP-1 medications work partly by weakening the appetite-feedback circuit that causes them, which buys a much longer descent than dieting does but not an unlimited one. The practical error is calling it early: the trial curves were still falling at week 68 and week 88, and both continuation trials recorded meaningful loss in exactly the window where most people decide it has stopped working. Check the calendar and the dose before you conclude anything about yourself.