Tirzepatide: a dual GIP and GLP-1 receptor agonist

The molecule pulls two levers where the older drugs pull one

Your gut releases hormones when you eat. Two of them matter here. GLP-1 slows the stomach, tells the pancreas to release insulin when glucose is high, and signals fullness to the brain. GIP does related work on insulin and on how fat tissue handles energy.

Semaglutide copies GLP-1. Tirzepatide copies both — one molecule that fits both docking points. That is the whole design difference, and in head-to-head weight numbers the two-lever version comes out ahead.

It is a once-weekly injection under the skin. Doses run 2.5 mg to 15 mg, escalated slowly over about 20 weeks, because the side effects are worst while the dose is climbing.

Twenty-point-nine per cent, and the four numbers that matter more than the headline

SURMOUNT-1 randomised 2,539 adults with a body mass index of 30 or more — or 27 or more with a weight-related problem — excluding diabetes, to 5 mg, 10 mg, 15 mg or placebo for 72 weeks. Average starting weight was 104.8 kg.

Weight change at week 72:

• 5 mg: −15.0%
• 10 mg: −19.5%
• 15 mg: −20.9%
• placebo: −3.1%

Those are averages, and averages hide the thing a person actually wants to know, which is their odds. The trial reported those too. Losing 5% or more of body weight: 85%, 89% and 91% across the three doses, against 35% on placebo. Losing 20% or more: 50% at 10 mg and 57% at 15 mg, against 3% on placebo.

Read the last pair again. More than half the people on the higher doses lost a fifth of their body weight. On placebo, three in a hundred did. There is no other compound documented on this site with a result of that shape.

Side effects were mostly stomach and bowel — nausea, diarrhoea, vomiting — mostly mild to moderate, and concentrated in the dose-climbing phase.

The diabetes trial got a smaller number, and that is the expected pattern

SURMOUNT-2 ran the same design in 938 adults who had obesity and type 2 diabetes, at 10 mg and 15 mg for 72 weeks. Starting weight 100.7 kg, average HbA1c 8.02%.

Weight change at week 72: −12.8% at 10 mg, −14.7% at 15 mg, against −3.2% on placebo. Between 79% and 83% lost at least 5%, against 32% on placebo. Serious adverse events occurred in 7%; fewer than 5% stopped because of side effects.

Less weight comes off in people with diabetes than in people without it, on the same drug at the same dose. That holds across this drug class and it is worth knowing before comparing your own result to a headline from the wrong trial.

Sleep apnoea is the second thing it was proven to change, and the effect is large

Two 52-week randomised trials in adults with moderate-to-severe obstructive sleep apnoea and obesity: one in people not using a CPAP machine, one in people using one. Average starting severity was about 50 breathing interruptions per hour of sleep — severe.

Breathing interruptions per hour at week 52 fell by 25.3 in the first trial and 29.3 in the second, against 5.3 and 5.5 on placebo. Every pre-specified secondary measure improved as well: weight, the oxygen debt built up overnight, C-reactive protein (an inflammation marker), systolic blood pressure, and how people rated their own sleep.

That second list matters as much as the breathing number. A drug approved on an apnoea endpoint that also moves blood pressure and an inflammation marker in the same trial is doing more than one thing, and all of it was measured under placebo control.

A 2026 post-hoc analysis of those same sleep trials went further and sorted participants by how they described themselves at the start — fatigued or not, sleepy or not, snoring or not, sleeping well or badly — then tracked both the machine measurements and what people said about their own days. That is the analysis worth reading for anyone whose complaint is exhaustion rather than a number on a sleep report, because it asks whether the people who felt worst got the most back.

Stop taking it and the weight comes back: the trial that tested exactly that

SURMOUNT-MAINTAIN is the most useful trial for anybody deciding whether to start. Its 441 participants first lost weight on the maximum tolerated dose for 60 weeks. Then 378 of them were randomly assigned to one of three arms for another 52 weeks: stay on that dose, drop to 5 mg, or switch to placebo.

Weight change from baseline at week 112:

• Stayed on the maximum tolerated dose: −21.9%
• Dropped to 5 mg: −16.6%
• Switched to placebo: −9.9%

The placebo arm had already lost the weight. They put more than half of it back over a year without the drug. A dose reduction cost about a third of the result; stopping cost more than half.

That is what "chronic treatment" means in practice, and it is the part that gets left out of the marketing. The comparison to make is not this drug against nothing, it is this drug against this drug plus the plan for what happens when you stop.

What the first few weeks predict about the rest

A 2026 post-hoc analysis pooled SURMOUNT-1 and SURMOUNT-2 and sorted participants by how much weight they had lost early, then followed what happened to each group: total weight change, cardiometabolic markers, tolerability and low-blood-sugar events.

The reason to know this exists: it means early response carries information. Somebody four weeks in, seeing very little movement, is not reading noise — that early trajectory tracks with where they land. It also cuts the other way. The people who respond fastest are not the ones who suffer most for it; tolerability did not simply worsen with a bigger response.

The inflammation markers moved, separately from the weight

A 2026 analysis in the Journal of the American College of Cardiology took SURMOUNT-1 and looked at cardiovascular risk markers over the long haul rather than weight: high-sensitivity C-reactive protein, interleukin-6, fibrinogen and white-cell counts — the standard readouts of low-grade, whole-body inflammation.

Fat tissue is not inert. It releases signalling molecules that keep low-grade inflammation running throughout the body, and those four markers are how that state is read in a blood test. A trial showing them fall alongside 20% of body weight is measuring something separate from the weight itself.

What it is not measuring is any symptom. These are intermediate markers, they moved in the right direction in thousands of people under placebo control, and that is the whole claim.

The rescue rule inside the maintenance trial, which is the detail that makes it honest

The maintenance trial did something most withdrawal studies do not: it wrote in an escape hatch. From week 84 — 24 weeks after the switch — any participant whose regain passed 50% of what they had lost could be given tirzepatide again as rescue. Of the 378 randomised, 345 (91%) completed.

That design choice matters twice over. Ethically, it means nobody was held in a placebo arm watching all of their result reverse. Statistically, it means the placebo arm's final figure of −9.9% is if anything generous to stopping: the people regaining fastest were pulled out and re-treated, and the analysis assumed they gained no further benefit from their assigned arm. The real cost of stopping is at least as large as the number printed.

What the dose climb actually involves

The trials did not start people at the dose that produced the headline result. SURMOUNT-1 included a 20-week escalation period inside its 72 weeks, meaning more than a quarter of the trial was spent getting to the top dose. Doses run 2.5 mg, 5 mg, 7.5 mg, 10 mg, 12.5 mg, 15 mg.

Two consequences follow directly. The first is that the stomach and bowel side effects concentrate in that quarter of the timeline rather than spreading evenly, which is why the trials report them as mostly mild to moderate and mostly early. The second is that anyone comparing their own week-8 result to a 20.9% figure is comparing themselves to a 72-week endpoint reached at a dose they have not started yet.

Population matters, and so does the setting

A 2026 systematic review and meta-analysis compared results in Asian against non-Asian adults with obesity and without diabetes across the phase 3 trials, because body composition and cardiometabolic risk differ between those groups at the same body mass index. The trials themselves were run across seven countries in the diabetes arm and included Chinese, Japanese, Russian, Brazilian and Argentine sites in the first — this is not a single-country evidence base.

There is also a five-year pragmatic trial running in UK primary care, SURMOUNT-REAL UK, enrolling roughly 3,000 adults with class I obesity, delivered as an addition to ordinary care and tracked through the health record rather than through clinic visits. That is the study that answers the question a trial never can: what happens when the drug is used by ordinary people in an ordinary practice, rather than by trial volunteers under supervision. It has not reported.

The muscle question, which is real and unresolved

When people lose 20% of body weight, some of what leaves is muscle. That is true of dieting, of surgery, and of this drug class, and it is the one cost of a large weight loss that does not show up in the headline number.

The trials measured body weight, not body composition, as their main endpoint. Sub-studies using scans have reported that the proportion of loss coming from lean mass looks broadly similar to other weight-loss methods, which is reassurance rather than resolution.

What follows practically is not a hedge: resistance training and adequate protein during the loss phase are the standard countermeasures, and there is no version of this drug that removes the need for them.

What is approved, what is compounded, and why the label on the vial matters

Tirzepatide is an FDA-approved medicine. It is sold as Mounjaro for type 2 diabetes and Zepbound for weight management and for obstructive sleep apnoea with obesity, both from Eli Lilly, both prescription-only.

Everything else in circulation is a different legal object. Compounded tirzepatide, "research use only" vials, and grey-market powder are not the approved product, have not been through the approval process, and carry no guarantee that the contents match the label. The trials on this page were run with the manufacturer's molecule at a controlled dose. A vial from another source has no established link to those numbers, and the correct assumption is that dose, purity and sterility are unknown until somebody tests them.

That distinction is sharper for this compound than for the others on this site, precisely because an approved version exists. With BPC-157 there is no approved product to compare against. Here there is, and buying around it means giving up the one thing the evidence is attached to.

How the side effects actually behave

Across the trials the pattern is consistent. Nausea, vomiting, diarrhoea and constipation dominate, they cluster during dose escalation, most are mild to moderate, and discontinuation for side effects ran under 5% in the diabetes trial.

Two further points that come from the trial designs rather than the results. The 20-week escalation exists because going up faster produces worse symptoms — the slow ramp is part of the drug, not a suggestion. And gallbladder problems and pancreatitis appear as recognised risks across this whole class, uncommon but serious, which is a reason the approved product is prescription-only and supervised.

Where this sits next to the other compounds on this site

Set the evidence side by side and the contrast is the point.

Tirzepatide: two 72-week randomised trials with 3,477 people between them, two 52-week trials in sleep apnoea, a 112-week maintenance trial, regulatory approval, and a defined dose.

BPC-157: no completed randomised trial in a person, a large animal literature, and a counted self-reported record.

TB-500: the molecule sold is not the molecule studied, and no human study has measured it.

That gap is not a reason to prefer one over another — they are aimed at different things and measured to different standards. It is a reason to be exact about which kind of claim you are relying on. Here, the claim is weight, and it is proven.

One more thing is worth stating plainly, because it is the difference between this page and most writing about this drug. Every figure above is a group average from a randomised comparison, which is the only kind of number that survives contact with a placebo. The 3.1% the placebo group lost in SURMOUNT-1 is what participation in a trial does on its own — the attention, the food diaries, the fortnightly weigh-ins. Any account of a weight-loss compound that does not carry its placebo number is not reporting a result, it is reporting an anecdote with a sample size. That is the reason a compound with 3,477 people in two trials sits on the same site as compounds whose whole record is forty posts, and the reason the two are never described in the same register here.

What is still open on this compound

Three things are not settled by anything above. How long the effect holds beyond the two years these trials cover — the longest ran 112 weeks, and obesity is measured in decades. What the loss is made of in the long run, since body composition was a sub-study rather than a primary endpoint. And what happens in ordinary care rather than trial conditions, which is what the five-year pragmatic trial exists to answer and has not yet.

There is also a gap that no trial fixes. The evidence on this page belongs to a manufactured product given at a controlled dose under supervision. The further a vial sits from that description, the less of this page applies to it, and nothing in the literature tells you how much less.

Nine published trials and analyses are attached to this page. Every number above comes from one of them.