What the GLP-1 Head-to-Head Evidence Actually Compared

Answer first: there is no head-to-head weight-loss trial on any of these labels

Take the three products approved in the United States to reduce excess body weight — tirzepatide injection, semaglutide as injection and tablet, and orforglipron tablet — and search each label's full text for the names of the other two molecules. The tirzepatide weight label mentions semaglutide zero times and orforglipron zero times, while naming its own molecule 347 times. The semaglutide weight label mentions tirzepatide zero times and orforglipron zero times, against 327 mentions of its own. The orforglipron label mentions the other two zero times, against 286 of its own.

That is a real absence rather than a failure of the search, because the same search in the same pass over the same files found each label's own molecule hundreds of times, and found 38 mentions of semaglutide and 18 of dulaglutide on the tirzepatide diabetes label. The ruler works. There is simply nothing there.

One string does appear on two of the weight labels: liraglutide, twice each. Each of those four instances was opened and read, and all four are the same class-wide boxed warning sentence about postmarketing reports of medullary thyroid carcinoma. None is a trial. A count without a read would have made that look like evidence.

So when a page ranks these three against each other on weight loss, whatever it is doing, it is not reporting a head-to-head trial. It is placing numbers from separate trials side by side.

The head-to-head trials that do exist were run for blood sugar

The comparisons are all on the two diabetes labels, and every one of them had an HbA1c primary endpoint. On the tirzepatide diabetes label there are four: against semaglutide added to metformin, against insulin degludec added to metformin with or without an SGLT2 inhibitor, against insulin glargine added to one to three oral agents, and against dulaglutide in the long cardiovascular outcomes trial. The semaglutide diabetes label states in one sentence that its efficacy was compared with placebo, sitagliptin, exenatide extended-release and insulin glargine.

That is the whole set. Every head-to-head trial supporting any of these labels was conducted in adults with type 2 diabetes, for glycemic control or for cardiac events, not for weight reduction in people without diabetes. Two of those comparisons put one GLP-1 receptor agonist against another — tirzepatide against dulaglutide, semaglutide against exenatide extended-release — and both were run for a diabetes endpoint.

This matters more than it sounds. A trial in adults with a mean diabetes duration of eight to fifteen years, already on metformin or insulin, measuring HbA1c, is not the trial that produced anybody's headline weight percentage. Weight in those trials is a reported measure alongside the primary one, in a population that the weight trials specifically excluded.

The one direct comparison of these two molecules, read carefully

The trial people mean when they say tirzepatide beat semaglutide ran 40 weeks in 1,879 adults with type 2 diabetes whose blood sugar was inadequately controlled on metformin alone. Three amounts of tirzepatide were compared against semaglutide. The primary endpoint was change in HbA1c, and tirzepatide's two higher amounts produced a statistically significant reduction against the comparator.

Weight was also reported, and the gap is real: the comparator arm lost 5.7 kilograms, the three tirzepatide arms 7.6, 9.3 and 11.2 kilograms. Those are the numbers that circulate.

Three features of that trial change what the numbers mean. First, the comparator was given at the amount studied for blood sugar control, not the larger amount later studied for weight — so it is a comparison against semaglutide as a diabetes drug. Second, the trial was open-label, blinded only as to which amount of tirzepatide a patient received; everyone knew which molecule they were on. Third, the population was people with a mean diabetes duration of 8.6 years and a mean body mass index of 34, which is the population the weight trials excluded.

None of that makes the result wrong. It makes it a result about glycemic control in type 2 diabetes, with a weight measure attached, and not the head-to-head weight trial it gets quoted as.

The longest head-to-head found no difference on its primary endpoint

The other direct comparison is much larger and much longer, and it points the other way. Over a median of 210.1 weeks, 13,299 adults with type 2 diabetes and established cardiovascular disease received either tirzepatide or dulaglutide, an older once-weekly GLP-1 receptor agonist. The primary endpoint was a three-part composite of cardiovascular death, non-fatal heart attack and non-fatal stroke.

Tirzepatide met non-inferiority, at a hazard ratio of 0.92 with a confidence interval from 0.83 to 1.01. The label then states it directly: superiority to dulaglutide was not established.

So the two direct comparisons available say different things about the same drug. In a 40-week unblinded trial it lowered HbA1c and weight more than a single-pathway GLP-1 at that drug's glycemic amount. In a four-year blinded trial with cardiac events as the endpoint it did not outperform a different single-pathway GLP-1. Anyone quoting one of those without the other is picking.

Why the weight trials cannot be lined up against each other

The temptation is to put the headline percentages in a row and call the largest one the winner. Here is what differs between those trials, reading the design paragraphs of the three weight labels.

Duration differs: 72 weeks for the main tirzepatide trials, 68 weeks for most of the semaglutide injection trials, 64 weeks for the semaglutide tablet trial, 72 weeks again for the newest semaglutide pair. A percentage at 64 weeks and a percentage at 72 weeks are not the same measurement.

The lifestyle program differs, and in one case dramatically. Most of these trials gave everyone a roughly 500 calorie per day deficit and a target of at least 150 minutes of activity per week. One semaglutide trial instead began with an eight-week low-calorie diet of 1,000 to 1,200 calories a day, followed by 60 weeks of a 1,200 to 1,800 calorie diet with activity rising from 100 to 200 minutes per week. Everyone in that trial, on drug or placebo, was in a different program from everyone in the others.

The statistical handling of missing data differs — retrieved-dropout multiple imputation in some, a hybrid approach in others, placebo-based multiple imputation in another. So does the randomization ratio, which runs from one-to-one to five-to-one across these trials. Each of those choices moves a reported average by an amount nobody publishes alongside the average.

And the populations differ, which is the largest factor of all and is covered in its own article. The practical consequence is simple: a difference between two numbers from two different trials is not a measured difference between two drugs. It is a difference between two experiments.

One label compares two forms of the same drug, and admits a gap

There is one comparison inside a single brand worth knowing about. The semaglutide weight label covers both an injection and a tablet, and a dedicated subsection in Use in Specific Populations addresses type 2 diabetes.

It states that the tablet has not been studied for weight reduction in adults with type 2 diabetes and obesity or overweight. It states that in patients without type 2 diabetes, average blood concentrations from the tablet and the injection were similar — but that in patients with type 2 diabetes, average concentrations from the tablet were lower than from the injection, possibly through reduced absolute bioavailability. And it states that given that difference, together with higher variability in blood concentrations from the tablet across all patient populations, some patients with type 2 diabetes taking the tablet may reach concentrations below the therapeutic range.

That is one label, one molecule, two delivery forms, and an explicit statement that they are not interchangeable for one group of patients. It is a good reminder that even within a brand, the evidence is form-specific.

What to ask when you see a comparison

Ask whether the two things being compared were ever in the same trial. If they were not, the comparison is arithmetic performed on two separate experiments, and the honest way to present it is with both trial designs beside it.

If they were in the same trial, ask three questions. What was the primary endpoint — because a weight figure inside a blood sugar trial is a secondary measure. Was it blinded, because several of these were open-label. And what was the comparator being given for, since a drug compared at its diabetes amount is not that drug at the amount studied for weight.

Then ask the question that settles most of it: who was in the trial. Every head-to-head comparison supporting any of these labels was run in adults with type 2 diabetes. If you do not have type 2 diabetes, there is no trial on any of these labels that compared these drugs in people like you.