Who Was Actually in the GLP-1 Weight-Loss Trials — and Who Was Kept Out

Answer first: the headline numbers come from a narrower group than the label's approved population

The approved use for these medications covers adults with obesity, or adults with overweight who have at least one weight-related condition. The trials that produced the famous percentages were run in a subset of that population, chosen by entry criteria that are printed in the label and almost never printed anywhere else.

The biggest single filter is diabetes. Reading every exclusion sentence in the Clinical Studies sections of the tirzepatide, semaglutide and orforglipron labels, the trials conducted in adults with obesity or overweight excluded patients with type 2 diabetes — with one exception, noted below. The trials that were run in people with diabetes are reported separately, and they produced visibly smaller weight results.

The second filter is who volunteered and was enrolled. These trial populations skew heavily female, and the cardiovascular and sleep apnea trials for the same drugs skew heavily male. The racial and ethnic composition swings enormously from one trial to the next, even within one drug's program.

None of this is hidden. All of it is in Section 14 of the prescribing information. It just does not travel with the percentage.

Type 2 diabetes was the standard exclusion, and the exception proves it was deliberate

In the tirzepatide program, the two main weight trials in adults with obesity or overweight both state that patients with type 2 diabetes were excluded, as does the trial built around an intensive lifestyle lead-in, and as do both obstructive sleep apnea trials. In the semaglutide program, the same exclusion appears in the two original injection trials, the withdrawal trial, the tablet trial, and the newest high-dose trial. In the orforglipron program it appears in the trial conducted in adults without diabetes.

There is one clear exception in that set. A semaglutide trial conducted in 401 patients in Japan and South Korea carries no diabetes exclusion, and the label reports that 24.7 percent of its patients had type 2 diabetes at baseline. That trial also used different entry thresholds from the others. It is worth naming precisely, because a claim that every one of these trials excluded diabetes would be wrong.

The cardiovascular outcomes trial went further than any of the weight trials. It excluded patients with a history of type 1 or type 2 diabetes, not just type 2.

Where a trial was deliberately run in people with diabetes, a second exclusion appears. Patients already taking insulin or an injectable GLP-1 receptor agonist were kept out of the tirzepatide diabetes-population trial, and patients on injectable glucose-lowering agents including insulin, GLP-1 receptor agonists or pramlintide were kept out of the corresponding orforglipron trial. So the people in these trials were, by design, not already on a drug from this class.

Diabetes changes the number by roughly a third, on the same drug

This is the practical consequence of that exclusion, and it shows up the same way on both labels that report it.

In the tirzepatide program, the trial in adults without type 2 diabetes reported a mean 20.9 percent weight reduction at the highest dosage studied. The companion trial in adults with type 2 diabetes, using the same two upper dosages, reported 14.7 percent.

In the semaglutide program, the main injection trial in adults without diabetes reported 14.9 percent. The companion trial in adults with type 2 diabetes reported 9.6 percent. In the newest pair of semaglutide trials, the obesity trial reported 18.8 percent at the higher amount studied and the diabetes trial reported 13.2 percent.

Three independent comparisons, two different molecules, the same direction and roughly the same magnitude of gap. If you have type 2 diabetes, the figures generated in the trials that excluded diabetes are not the figures generated in the trials that studied people like you — and the labels report both, side by side, for exactly this reason.

The weight trials are mostly women; the cardiovascular and sleep trials are mostly men

In the tirzepatide trial in adults without diabetes, 68 percent of patients were female. In the semaglutide injection trials, the figures were 74 percent and 81 percent female; in the tablet trial, 79 percent were women; in the newest obesity trial, 74 percent. The orforglipron trial in adults without diabetes was 64 percent female. The adolescent semaglutide trial was 38 percent male.

Now the same drugs, different indications. The cardiovascular outcomes trial was 72 percent male. The two obstructive sleep apnea trials were 67 percent and 72 percent male.

That is not a flaw in either set of trials — it reflects who has the conditions being studied and who enrolls. But it means the evidence base for weight reduction in this class was built disproportionately in women, and the evidence base for the cardiac and sleep apnea indications was built disproportionately in men. Anyone reading a single percentage should know which of those two populations produced it.

Racial and ethnic composition swings hard between trials of the same drug

Within the semaglutide weight program: one trial was 75 percent White with 13 percent Asian and 6 percent Black or African American patients; another was 76 percent White with 19 percent Black or African American and 2 percent Asian; the tablet trial was 92 percent White with 1 percent Asian and 7 percent Black or African American; the East Asian trial was entirely Asian.

Hispanic or Latino ethnicity swings just as wide. The tirzepatide trial in adults without diabetes was 48 percent Hispanic or Latino, and its companion diabetes trial 60 percent. The orforglipron trial in adults without diabetes was 38 percent. Meanwhile the semaglutide tablet trial was 7.8 percent, the newest semaglutide obesity trial 5 percent, and the cardiovascular trial 10 percent.

Those are not small differences in composition. They are the difference between a trial where nearly half the participants were Hispanic or Latino and one where one in twenty were. Labels do state that weight reduction was observed irrespective of race and ethnicity, which is a finding about consistency of direction. It is not the same as saying the trials were demographically interchangeable.

Everyone in these trials was substantially heavier than the entry threshold

The approved population starts at a body mass index of 27 with a weight-related condition, or 30 without one. The people actually enrolled sat well above that. Mean baseline BMI was 38 in the main tirzepatide trial, 37.9 and 38 in two semaglutide injection trials, 38 in the semaglutide tablet trial, 39.9 in the newest semaglutide obesity trial, and 37 in the orforglipron trial in adults without diabetes.

Mean baseline body weight ran from roughly 100 to 113 kg across those trials. In the obstructive sleep apnea trials, mean baseline weight was between 112 and 117 kg.

So a person who qualifies at the lower end of the approved range — a BMI in the high twenties with one qualifying condition — is at the very edge of the population these results were produced in, not in the middle of it. The labels report that reduction was observed irrespective of baseline BMI, which is meaningful. It is still worth knowing that the average trial participant was considerably heavier than the average person who meets the criteria.

The cardiovascular trial had the strictest entry criteria of all

Its population is worth describing precisely because it is the narrowest and the most often generalized. All patients were 45 or older with a body mass index of 27 or greater, and all had established cardiovascular disease, defined as a prior heart attack, a prior stroke, or peripheral arterial disease. Diabetes of either type was excluded. Mean age was 62, with a range from 45 to 93.

At baseline, 76 percent had already had a heart attack, 23 percent had already had a stroke, 9 percent had peripheral arterial disease, and 24 percent had heart failure. Nearly all of them were already on cardiac medication: 90 percent on lipid-lowering therapy, 86 percent on platelet aggregation inhibitors, 74 percent on ACE inhibitors or angiotensin II receptor blockers, and 70 percent on beta blockers.

That last set of figures is the one people miss. Whatever benefit the trial found was found on top of a standard of care that almost everyone in it was already receiving. It is not a result about a drug used instead of those medications, and it is not a result in people who have never had a cardiac event.

What to take from this into an appointment

Two questions get most of the value here. Which trial population am I closest to — and is that the one the number I have in my head came from? If you have type 2 diabetes, the honest expectation is the figure from the diabetes trials, not the headline. If you are near the lower end of the eligibility criteria, the trials were run in people substantially heavier than you.

The second is what the trial required of everyone in it. These were year-and-a-half trials in which every participant was also enrolled in a diet and exercise program, and where anyone already on a drug from this class was, in most of them, kept out. That is the setting the numbers describe.

Entry criteria are not fine print. They are the definition of who the result is about, and they are printed in the label for anyone who wants to check which trial they most resemble.