The trial record, without the enthusiasm tax

The GLP-1 trial base is unusually strong for a drug class this new, and unusually badly summarised. What follows is the record, with populations attached, because the population is where most of the misreading happens. Every figure links to the primary publication.

Weight loss

STEP 1 (semaglutide 2.4mg, NEJM 2021). 1,961 adults with obesity and no type 2 diabetes, 68 weeks. Mean weight loss 14.9% against 2.4% on placebo. 86.4% lost at least 5%; 69.1% at least 10%; 50.5% at least 15%.

STEP 2 (semaglutide, Lancet 2021). 1,210 adults with BMI 27+ and type 2 diabetes, most already on other diabetes medication, 68 weeks. 9.6% at the 2.4mg dose against 3.4% on placebo. That is materially less than STEP 1, which is the expected pattern in a diabetic population and a common source of confused comparison.

SURMOUNT-1 (tirzepatide, NEJM 2022). 2,539 adults with obesity and no type 2 diabetes, 72 weeks. 20.9% at 15mg against roughly 3.1% on placebo. At that dose, 55.8% lost at least 20% of body weight.

SURMOUNT-5 (NEJM 2025). Head to head, tirzepatide against semaglutide: 20.2% versus 13.7%, a relative difference of roughly 47%. Semaglutide retains the longer real-world track record.

Durability: the part that matters most

STEP 4 (JAMA 2021) and SURMOUNT-4 (JAMA 2024) both ran the same experiment: take people who had lost weight on the drug, and withdraw it from half of them. In STEP 4, the placebo-switch group regained 6.9% of body weight between weeks 20 and 68, roughly two thirds of what the run-in had taken off. In SURMOUNT-4, the switch group gained a mean 14.0% of body weight over the year that followed.

This is the single most important finding in the set, and the one least represented in marketing. The trials do not establish that these drugs produce permanent weight loss. They establish the opposite: benefit persists while treatment does, and substantially reverses when it stops. Any programme built on the assumption that patients graduate off the medication and hold the result is contradicted by its own evidence base, unless something else was built underneath.

Beyond weight

SELECT (semaglutide, NEJM 2023). 17,604 adults with established cardiovascular disease, no diabetes, mean follow-up ~33 months. A primary cardiovascular event occurred in 6.5% of the semaglutide group against 8.0% on placebo: a 20% relative reduction, hazard ratio 0.80 (95% CI 0.72-0.90). Secondary signals in mood scores and hsCRP.

FLOW (semaglutide, NEJM 2024). 3,533 adults with type 2 diabetes and chronic kidney disease, median 3.4 years, stopped early for benefit. 24% reduction in the composite kidney endpoint, hazard ratio 0.76 (95% CI 0.66-0.88).

SURMOUNT-OSA (tirzepatide, NEJM 2024). Significant reduction in apnea-hypopnea index in moderate-to-severe obstructive sleep apnea, with and without PAP therapy. FDA approval followed in December 2024.

ESSENCE (semaglutide, NEJM 2025). MASH resolution without worsening of fibrosis in 62.9% against 34.3% on placebo at week 72, an absolute difference of 28.6 points. SYNERGY-NASH (tirzepatide, NEJM 2024) reported 73.3% resolution at 15mg against 13.2% on placebo, in a smaller phase 2 population.

What none of this establishes

Every figure above comes from studies of branded, FDA-approved product. Compounded formulations have not been evaluated in these trials. Where the outcome evidence is invoked (cardiovascular risk reduction especially), it belongs to the product that was studied, and carrying it across to a compounded preparation is not supported.

The populations are also narrower than the headlines suggest. SELECT enrolled people who had already had a cardiovascular event. FLOW enrolled people with both diabetes and kidney disease. Neither tells you what happens in a healthy 45-year-old seeking metabolic optimisation, and both are routinely cited as though they do.

What we do with it

Two conclusions we hold with some confidence. First, the durability data makes the case that medication is a tool inside a broader programme rather than the programme itself: the regain figures are what a lifestyle infrastructure exists to prevent. Second, the honest version of the cardiovascular and renal story is narrower and more specific than the version in circulation, and saying so is a cheap way to be trusted on the parts that are strong.