Three Compounds, Compared on What Happens After

The short version

This page sets semaglutide, tirzepatide and retatrutide against each other on the dimensions that actually separate them: how many receptors each one engages, how big the reported weight change is, how mature the evidence behind it is, and — the column this desk exists for — what each programme has recorded about stopping.

The pattern is easy to state. Weight effect grows with the number of receptors engaged: one, then two, then three, in ascending order of reported reduction. The withdrawal file runs in the opposite direction. Semaglutide, with the smallest headline figure, has both a randomised withdrawal trial and an off-treatment extension behind it. Tirzepatide has a randomised withdrawal trial. Retatrutide, with the largest figure of all, has neither.

Bigger effect, thinner record on what the effect does when it ends. That inverse relationship is the single most useful thing on this page, and nothing here is medical advice or a recommendation about any compound for any person.

The comparison table

What each trial measured, and when

Semaglutide — measured longest and widest. STEP 1 recorded -14.9% from baseline to week 68 against -2.4% with placebo ^[4]. Beyond weight, SELECT counted major adverse cardiovascular events in 17,604 participants and reported HR 0.80 ^[3]; FLOW counted kidney events in 3,533 participants and reported HR 0.76 ^[2]; SUSTAIN-6 reported HR 0.74 for its cardiovascular composite alongside a higher rate of retinopathy complications, HR 1.76, in participants with pre-existing disease ^[7]. This is the only compound here whose evidence extends past the scale into event counts.

Tirzepatide — measured biggest among approved options. SURMOUNT-1 reported -15.0%, -19.5% and -20.9% at week 72 across three doses against -3.1% with placebo, in 2,539 participants ^[10]. SURMOUNT-5 put it directly against semaglutide in 751 participants and reported -20.2% against -13.7% at week 72 ^[1]. SURPASS-2 covered glycaemia in 1,879 participants with type 2 diabetes, with HbA1c reductions of 2.01 to 2.30 percentage points against 1.86 for semaglutide 1 mg ^[11]. No outcome trial has reported.

Retatrutide — measured largest, earliest, briefest. A 48-week Phase 2 obesity trial reported -24.2% at 12 mg against -2.1% with placebo ^[15]; a 36-week Phase 2 trial in type 2 diabetes reported an HbA1c reduction of 2.02 percentage points at 24 weeks and -16.94% body weight at 36 weeks ^[16]; a Phase 2 liver substudy reported an 82.4% reduction in liver fat at 24 weeks with 86% of participants reaching under 5% ^[14]. Every one of those is Phase 2, and the longest of them ran 48 weeks.

Read the three columns together and the trend is not only in the size of the number but in how long anybody watched. The compound with the largest reported effect has been observed for the shortest period, in the smallest programme, in the earliest phase.

What the withdrawal designs found

Two of the three compounds have been through a study built specifically to answer this, and the designs matter more than the headline.

Randomised withdrawal takes participants who have already responded and randomises them to continue or to switch to placebo. The semaglutide and tirzepatide programmes both used this design. Both reported the same divergence: the arm that continued maintained a different trajectory, and the arm switched to placebo regained. Because the arms enter randomisation after the same active-treatment lead-in, the contrast isolates the contribution of continuing exposure in a way no placebo-from-baseline comparison can.

Off-treatment extension withdraws the interventions and keeps watching. The semaglutide extension reported substantial regain, with cardiometabolic improvements reverting toward baseline. Pooled analyses summarized in the corpus describe regain as related to the amount initially lost.

No design at all is retatrutide's position, and it should be reported as an absence rather than filled in from the neighbours. Its evidence summary flags rebound as a class expectation and lists durability after discontinuation among the unknowns.

These conclusions remain qualitative because no direct withdrawal publication appears in the numbered source set. The comparison therefore does not attach an efficacy-paper citation to a separate stopping result.

Evidence maturity on the cessation question, ranked

Ranked strictly by how much is known about what happens after treatment ends, the order inverts the ranking by effect size.

Semaglutide, first. Two independent designs, a randomised withdrawal and an off-treatment extension, both reporting regain, plus a quantified figure for how much and over what horizon, plus the observation that cardiometabolic improvements moved with the weight rather than persisting separately.

Tirzepatide, second. One randomised withdrawal design with a clear divergence, supported by pooled class analyses that add proportionality to the picture. No dedicated off-treatment extension has reported, so the horizon beyond the withdrawal period is less well described than semaglutide's.

Retatrutide, third by a distance. Nothing measured. The expectation of rebound rests entirely on the shared mechanism, and while the inference is reasonable it remains an inference. Its glucagon arm, which adds energy expenditure rather than only suppressing intake, is the one part of the mechanism that has no measured counterpart in the withdrawal literature.

The consequence for reading any of these figures is the same in each case. A percentage attached to one of these compounds is a description of a state held open by continuing exposure, not a property the participant acquired and kept.

How to read a percentage on this desk

Three habits make the numbers on this page behave.

Attach the timestamp. Not -14.9% but -14.9% at week 68 with treatment continuing ^[4]; not -24.2% but -24.2% at week 48 in a Phase 2 trial ^[15]. A percentage without its timepoint has quietly become a claim about permanence that no trial made.

Attach the population. SELECT enrolled adults with established cardiovascular disease and without diabetes ^[3]; FLOW enrolled people with type 2 diabetes and chronic kidney disease ^[2]; SURMOUNT-1 enrolled adults with obesity and without diabetes ^[10]. Effects were measured in those groups, and a result does not automatically travel to a different one.

Attach the phase. A Phase 2 estimate and a completed outcome trial are not the same kind of object, and the compound with the largest figure on this desk is the one whose figure is least confirmed.

None of this constitutes medical advice, and nothing here recommends, discourages, schedules or doses any compound for any person.