Why a Triple Agonist Is Not Simply Better Than a Dual

The intuition, and why it fails

One receptor, then two, then three. It reads like a version number. Each generation should beat the last.

The published trials do not line up that way, and the reason is not that triple agonism fails. It is that the number attached to each drug was produced by a different experiment.

Four things differ between any two of these trials

Population. SURMOUNT-1 enrolled adults with obesity or overweight without diabetes¹. TRANSCEND-T2D-1 enrolled adults with type 2 diabetes inadequately controlled by diet and exercise². Weight loss is systematically smaller in people with type 2 diabetes across this class.

Duration. SURMOUNT-1 ran 72 weeks¹. Survodutide's phase 3 ran 76³. Retatrutide's phase 2 ran 48⁴, and TRANSCEND-T2D-1 ran 40².

Phase. Retatrutide's headline −24.2% is a phase 2 figure from 338 people⁴. Tirzepatide's −20.9% is a phase 3 figure from 2,539¹.

Estimand. This is the one that gets skipped, and it changes numbers by percentage points on its own. More on it below.

The estimand problem, made concrete

Survodutide's phase 2 reported a mean weight change of −14.9% at the highest dose at week 46. Its primary analysis was based on the dose assigned at randomisation, with data censored for COVID-19-related discontinuations⁵. Only 233 of 386 treated participants — 60.4% — completed the 46-week treatment period⁵.

Survodutide's phase 3, SYNCHRONIZE-1, reported −12.2% and −13.0% at week 76 for its 3.6 mg and 6.0 mg arms. Its primary analysis used a treatment-regimen estimand, which the paper describes as incorporating the effects of early discontinuation, use of protocol-prohibited obesity medications, and a prolonged dose-escalation period³.

Read those two paragraphs again. The phase 3 number is lower, and part of the reason is that the phase 3 counted things the phase 2 analysis set aside. How much of the gap is estimand and how much is efficacy is not something either paper separates out — which is exactly why "the number dropped" is the wrong summary.

What the class actually looks like, annotated

Semaglutide 2.4 mg — one receptor. −14.9% at 68 weeks, 1,961 adults without diabetes, phase 3⁶.

Tirzepatide 15 mg — two receptors, GIP on. −20.9% at 72 weeks, 2,539 adults without diabetes, phase 3¹.

Survodutide 6.0 mg — two receptors, glucagon on. −13.0% at 76 weeks, 725 adults without diabetes, phase 3, treatment-regimen estimand³.

Maridebart cafraglutide — two receptors, GIP *off*. −12.3% to −16.2% at 52 weeks, 465-participant obesity cohort, phase 2⁷.

Retatrutide 12 mg — three receptors. −24.2% at 48 weeks, 338 adults, phase 2⁴; and −15.3% at 40 weeks, 537 adults with type 2 diabetes, phase 3².

The largest figure on that list belongs to the smallest trial. The second-largest belongs to a phase 3 with 2,539 people. Both facts are true and neither ranks the drugs.

Adding a receptor adds a side-effect profile too

In retatrutide's phase 2, gastrointestinal events were the most common adverse events, were dose-related, and were partially mitigated by starting at a lower dose — 2 mg rather than 4 mg. The trial also reported dose-dependent increases in heart rate that peaked at week 24 and declined afterwards⁴.

In survodutide's phase 3, gastrointestinal symptoms were reported by 80.9% of the 3.6 mg group and 89.7% of the 6.0 mg group, against 47.9% on placebo³.

Neither of those is a reason to dismiss the mechanism. They are the other half of the trade a receptor buys, and a comparison that reports the efficacy column without the tolerability column is not a comparison.

The one comparison that would settle it

A head-to-head. One trial, one population, one duration, one estimand, participants randomised between the two molecules.

Those trials exist in this class. Semaglutide has been randomised against tirzepatide in obesity, and semaglutide at its maximum tolerated dose of 1.7 or 2.4 mg reached −13.7% at 72 weeks against −20.2%⁸. That is a real ranking, because it came from one experiment.

For triple against dual, a registered trial is running: NCT06662383 lists retatrutide compared with tirzepatide, active and not recruiting as of August 2026, with a listed enrollment of 800 and a listed completion date in December 2026. Until a trial like that reports, the triple-versus-dual question is being answered by arithmetic across trials that were never designed to be compared.

How to read a new number when it lands

Ask which population. Ask how long. Ask which phase, and how many people. Ask which estimand, and whether the paper says what it did with people who stopped early.

Then ask what the comparator was. A drug that beat placebo has not beaten anything else.

Our incretins explainer covers what each receptor does. The retatrutide page, tirzepatide page and survodutide page carry each programme's full record, and weight loss lines them up against the goal rather than against each other.