Retatrutide Trial Results, Read Carefully

If you are Eli Lilly and you own a drug that makes people lose a startling amount of weight, your problem is not the drug. Your problem is that every result leaves the building as one number, and the number travels without the paragraph that explained it. So here are the retatrutide results as plainly as the facts allow. In phase 2, a 48-week study in adults with obesity but without type 2 diabetes (weekly doses from 1 mg to 12 mg, average starting BMI of 37.4) sits behind the figure that a review in Annals of Internal Medicine put at the top of its table: 24.2% weight loss at 12 mg weekly, ahead of tirzepatide at 17.8% and injected semaglutide at 13.9%.

In phase 3, TRIUMPH-1 (80 weeks, adults with obesity or overweight plus at least one weight-related condition, no diabetes) reported its topline on May 21, 2026, under an AJMC headline promising “Up to 30.3% Average Weight Loss.” By week 80, 37.5% of participants who started with class 3 obesity (BMI of 40 or more) had a BMI below 30 on the 12 mg dose. TRIUMPH-2, in adults with obesity and type 2 diabetes, reached up to 20.8%. There are also inflammation numbers and a liver story, which we will get to, and one important thing the facts here do not settle, which we will also get to.

The dumb version of reading a trial

The dumb version goes like this. You see “30.3%” in a headline. You conclude that people on retatrutide lose 30.3% of their weight. You tell a friend at dinner. Done.

This is a perfectly reasonable thing for a human to do. It is also what the headline was built for. (Headlines are not built to be footnoted.) And here is the weird part: the dumb version gets you to roughly the right conclusion, which is that this drug causes a very large amount of weight loss, more than the drugs that came before it. Hold onto that, because the sophisticated version is going to spend a lot of words arriving at the same spot.

The sophisticated version, or what endpoints mean

A trial is designed before anyone takes a dose, and the design says in advance what will count as the answer. That pre-declared answer is the primary endpoint. For obesity trials it is typically the average percent change in body weight from the start of the trial to a fixed week, compared against people who got a placebo injection. Everything else (the share of people who crossed some threshold, blood markers, liver fat, etc.) is a secondary endpoint or an exploratory one, and the further down that list you go, the less weight the trial’s statistics were built to carry.

Then there is the official description of the design. The Lancet paper on the phase 3 diabetes trial calls TRIUMPH-2 “a double-blind, parallel-group, randomised, placebo-controlled, phase 3 trial.” Translated: people were sorted by coin flip into separate groups at the same time, some got fake shots, and nobody in the room knew who got what. That sentence is doing most of the work of making a number believable.

Now look at “up to.” “Up to 30.3%” means the best-performing dose arm, on average. Other doses lost less. The average also contains people who lost much more and people who lost much less. And a percentage is a percentage of where you started, so a person starting at BMI 45 and a person starting at BMI 32 can post the same percent and end in very different places. That is why the 37.5% figure is interesting: it is not a percent of weight, it is a percent of people who crossed a line.

What the phase 2 retatrutide results actually measured

The number I keep coming back to is 24.2%.

A 2024 systematic review found that completed phase 2 trials of incretin-based obesity drugs showed mean weight loss ranging from 7.4% to 24.2%. So 24.2% is the ceiling of the entire field’s phase 2 work at that point, the single highest mean among every incretin molecule that had finished a mid-stage trial. And then the Annals review, pooling 26 randomized trials and 15,491 adults without diabetes, lists retatrutide at 12 mg weekly at exactly 24.2%. Presumably that is the same phase 2 number showing up in two places (the facts I am working from do not say so outright, and I am not going to pretend they do). Either way, one dose of one molecule defined the upper edge of a whole class. That is a lot of responsibility for a decimal.

On the other hand, the phase 2 obesity study was 48 weeks long and TRIUMPH-1 ran 80 weeks. Longer trials give a drug more time to keep working, which is part of how a 24.2% becomes an “up to 30.3%.” Comparing those two numbers as though they were the same measurement is how you end up very confident and slightly wrong.

Diabetes, inflammation and the liver

There were two phase 2 trials, one in obesity without diabetes (Study 2) and one in type 2 diabetes (Study 1). Researchers later ran what the paper calls “a post-hoc analysis of two phase 2 retatrutide trials” that “assessed lipoprotein and inflammatory biomarkers.” Post-hoc means after the fact, which in medicine is a respectable way of going back to the data for a second look and, in most other areas of life, is how you end up explaining yourself to a group chat.

What they found: in Study 2, retatrutide cut high-sensitivity C-reactive protein by 54.8% and interleukin-6 by 29.6% versus placebo. In Study 1, the diabetes population, those reductions were not statistically significant. That’s fine, and it is also the sort of detail a headline drops. Not significant does not mean the markers went up. It means the trial could not rule out chance.

On diabetes more broadly, Lilly’s own release on its first phase 3 type 2 diabetes trial, TRANSCEND-T2D-1, says retatrutide “demonstrated significant reductions in A1C and weight.” The specific A1C figure is not in the material I have, so I will leave it there. TRIUMPH-2 reached up to 20.8% weight loss in people with both obesity and diabetes, which is lower than the non-diabetes trials. People with diabetes tend to lose less weight on these drugs; tirzepatide’s own trial program has the same split between SURMOUNT and SURPASS.

The liver is where I have to be honest about the limits. A review of glucagon signalling says that in early trials, retatrutide, survodutide and mazdutide each showed “substantial efficacy for weight loss in people with obesity while improving liver health in those with MASLD.” The actual liver substudy numbers (fat fraction, how many people normalized, etc.) are not in the facts I was handed. I feel like the responsible move for a guy who writes about other people’s spreadsheets is to say that and stop.

Why everyone reads it the way they do

Everyone in this story is behaving sensibly. Lilly announces the best dose because the best dose is the news, and its job is to tell investors and doctors what the drug can do. Investigators report pre-declared endpoints and flag post-hoc ones because that is what keeps them honest. Journalists lead with the biggest number because a headline that said “Mean Percent Change From Baseline in the Highest Dose Arm” would be read by about four people (I would be one of them). And readers take the headline number because they have jobs.

Here, as an illustration I am inventing, is how it goes wrong. A reader sees 30.3% and assumes tirzepatide is now obsolete. She sees 17.8% for tirzepatide in the Annals review and 19.5% and 20.9% in SURMOUNT-1, and compares them all to 30.3%, mixing trial lengths, populations and doses. Her brother-in-law then compares 30.3% to orforglipron’s 11.2% in ATTAIN-1 and announces that pills are over. Someone else notices UBT251 hit 19.7% in only 24 weeks and extrapolates to 80 weeks, and by Thanksgiving the family has a drug that makes people disappear entirely. A statistician somewhere says, “guys, these are different trials,” and is not invited back.

This is a bit like comparing a marathon time to a half-marathon time and declaring the half-marathoner faster. Correct arithmetic, wrong race.

Where the long road ends

The proper fix for cross-trial comparison is a head-to-head trial, and there is one: TRIUMPH-5, retatrutide versus tirzepatide, with an estimated primary completion in November 2026. Results from that trial are not among the facts here, so whether the gap holds when the two drugs share a protocol is not settled.

Anyway, after all the endpoints, estimands-in-spirit, dose arms, post-hoc caveats and population splits, the sophisticated version does reach a conclusion. A 2026 review says triple agonists like retatrutide “have exhibited unprecedented weight-loss efficacy in clinical trials.” Which is what the guy at dinner said, with more syllables. The point is that the long road and the short road land in the same parking lot; the long road just knows which parking space, and that diabetes patients park a few spaces further away.

At this rate the most careful possible reading of retatrutide results is simply the headline, delivered eighty weeks later with a confidence interval.