Put the two best-known numbers side by side and the triple agonist looks like the winner: 24.2% of body weight gone at 48 weeks against 20.9% at 72 weeks. Both figures are real, correctly reported and easy to check. They also come from trials that share almost nothing except a weighing scale. One is a dose-finding study in 338 people; the other is a registration trial in 2,539. One supports a marketed prescription product, and the other supports a substance that no United States pharmacy may lawfully dispense or compound — a status covered in the retatrutide article.
The two headline results, with their designs attached
The retatrutide obesity trial was phase 2. It enrolled 338 adults with a body-mass index of 30 or higher, or 27 to under 30 with a weight-related condition, and randomized them across six regimens and placebo for 48 weeks. Its primary endpoint was the change at 24 weeks, where the 12-mg group reached −17.5% against −1.6% on placebo. The 48-week figure everyone quotes, −24.2% against −2.1%, is a secondary endpoint. A loss of 15% or more had occurred in 83% of that group against 2% of placebo.[1]
SURMOUNT-1 was phase 3. It randomized 2,539 adults with a mean baseline weight of 104.8 kg to tirzepatide 5, 10 or 15 mg or placebo for 72 weeks, with a 20-week escalation. Mean weight change was −15.0% (95% CI, −15.9 to −14.2) at 5 mg, −19.5% (95% CI, −20.4 to −18.5) at 10 mg and −20.9% (95% CI, −21.8 to −19.9) at 15 mg, against −3.1% (95% CI, −4.3 to −1.9) on placebo. A reduction of 20% or more occurred in 57% at the top dose against 3% on placebo.[2] Its numbers are unpacked in the SURMOUNT-1 article.
Four reasons the two figures are not commensurable
The clocks differ by 24 weeks. A 48-week reading and a 72-week reading are not two measurements of one quantity. The approved drug is being quoted at a later, more mature point; whether that flatters or penalizes it depends on where each weight curve was still heading at the moment it was cut, and neither abstract reports that.
The placebo arms moved by different amounts. Placebo was −2.1% in the retatrutide trial and −3.1% in SURMOUNT-1. Subtracting each trial’s own control leaves 22.1 points against 17.8 points rather than 24.2 against 20.9. That narrows the apparent gap by a quarter without touching either drug.
One dose was chosen before the trial, the other after. A phase 2 study exists to find a dose. It ran seven arms, and the number in circulation is the best-performing of them, identified once the results were in. A registration trial commits to its doses in advance and reports all of them. Reading the top arm of a dose-finding study as though it were a fixed regimen borrows a certainty the design was never built to supply.
The precision is not the same. SURMOUNT-1 reports a confidence interval on every estimate, roughly plus or minus one percentage point at the top dose. The phase 2 abstract reports least-squares means without intervals, in arms of a few dozen people each after some of the 4-mg and 8-mg groups were combined.[1] The wider point about how a second receptor changes an incretin’s behavior is in the GIP article.
The ranking inverts when the evidence is standardized
A network meta-analysis published in 2026 pooled 262 randomized trials and 99,791 participants, searched through November 2025, and reported every agent at a common one-year horizon with a GRADE certainty rating attached. Tirzepatide came out at −14.9% against lifestyle modification alone (95% CI, −16.0 to −13.9) with moderate-to-high certainty. Retatrutide appears in a separate sentence, grouped with ecnoglutide and mazdutide as emerging agents that may produce similar or greater reductions of 13.1% to 14.6%, at very low to low certainty; the abstract does not separate the three.[3]
That is a reversal, not a rounding difference. On the trial headlines, the unapproved compound leads by 3.3 percentage points. On a common one-year clock with certainty graded, it sits inside a band whose whole range falls below the approved drug’s point estimate, two certainty grades lower. Nothing about retatrutide changed between those two readings. What changed is that the second one adjusts for the duration and rates how much the evidence can bear.
The same analysis records a cost attached to tirzepatide’s lead that no weight percentage shows: it reduced fat mass the most, by 25.7%, and also reduced lean mass the most, by 8.3%.[3] Body composition is measured in only a minority of these trials, and the class-wide picture is in the muscle article.
The one place the designs actually match
There is exactly one pair of trials in which these two molecules were studied the same way. TRANSCEND-T2D-1 and SURPASS-1 are both 40-week, double-blind, placebo-controlled phase 3 monotherapy trials in adults with type 2 diabetes inadequately controlled by diet and exercise, both run across sites including the United States, Mexico and India, both with glycated hemoglobin as the primary endpoint. TRANSCEND-T2D-1 randomized 537 participants; SURPASS-1 randomized 478.[4][5]
On the raw result they nearly tie. Glycated hemoglobin fell 1.94% on retatrutide 12 mg and 2.07% on tirzepatide 15 mg. On the placebo-subtracted result they do not: the estimated treatment difference was −1.12 percentage points (95% CI, −1.39 to −0.85) for retatrutide and −2.11 percentage points for tirzepatide. The reason is the control groups. SURPASS-1’s placebo arm rose 0.04%; TRANSCEND-T2D-1’s placebo arm fell 0.81%.[4][5] A control group that improves by four fifths of a percentage point absorbs most of what the drug would otherwise be credited with, and the two trials’ populations were not the same either: mean diabetes duration was 2.5 years against 4.7, and mean body-mass index 35.8 against 31.9.
On weight, the closest-matched pair of trials in the entire comparison cannot be compared at all. TRANSCEND-T2D-1 reports −15.3% of body weight at 12 mg against −2.6% on placebo. SURPASS-1 reports a dose-dependent loss of 7.0 to 9.5 kg and no percentage, and its abstract does not give the baseline weight that would convert one into the other. Two trials that match on design, duration, phase and endpoint still do not report the same unit.
What a phase 2 number does and does not predict
The assumption underneath most of these comparisons is that a phase 2 result shrinks on the way to phase 3. Tirzepatide is the obvious test of that, and it does not support the rule. Its own phase 2 trial randomized 318 adults with type 2 diabetes for 26 weeks and reported glycated hemoglobin falling 1.94% at 15 mg against 0.06% on placebo and 1.21% on dulaglutide, with weight change ranging to −11.3 kg.[6] Its phase 3 monotherapy trial then reported 2.07% at the same dose over a longer 40 weeks.[5] The estimate held.
So the problem with reading retatrutide’s phase 2 result as a forecast is not that early numbers are inflated. It is that a dose-finding study answers a different question, in a smaller and more selected group, and leaves the questions a registration program exists to answer entirely open.
What is still unknown
The obesity phase 3 program has not reported. TRIUMPH comprises four randomized trials in more than 5,800 participants, nesting obstructive sleep apnea and knee osteoarthritis protocols inside two weight-management trials, with further trials in cardiovascular disease and in osteoarthritis alone.[7] Until those publish, the obesity evidence for this molecule is one 48-week trial of 338 people.
Nothing is known about how the two compare head to head, because no such trial exists. Nothing is known about retatrutide’s cardiovascular outcomes, while tirzepatide’s program has reported across diabetes, sleep apnea and heart failure. Discontinuation is not comparable either: adverse events ended treatment in 4.3%, 7.1% and 6.2% of the three tirzepatide arms against 2.6% on placebo in SURMOUNT-1,[2] and in 2% to 5% of retatrutide arms against 0% on placebo in the 40-week diabetes trial,[4] which is a different population over a shorter period. Retatrutide also produced dose-dependent increases in heart rate that peaked at 24 weeks and declined afterward,[1] an effect the whole class shares to some degree — see the heart rate article.
One of these is a prescription and one is not a product
The asymmetry that matters most is not statistical. Tirzepatide is an approved drug with a label, a dose ladder and a manufacturer answerable for what is in the vial. Retatrutide has none of those. The FDA states that it has not been found safe and effective for any condition and that it cannot be used in compounding under federal law, being no part of any approved drug.[8] A telehealth service offering to prescribe it has described something it is not permitted to do.
That distinction survives every argument about effect sizes. Compounded semaglutide and tirzepatide are already outside the approval system: they are not FDA-approved and the agency does not assess them for safety, efficacy or quality before dispensing, which is what that phrase actually means. A substance that may not be compounded at all sits a step further out again. Other triple and dual agonists in development are tracked in the survodutide and mazdutide article, and the same caution applies to each of them.