What a Weight-Loss Percentage in a GLP-1 Trial Is Actually Measured From

Answer first: it is a group average, measured against a starting weight the trial chose

A number like a twenty percent weight reduction is not a description of what one person lost. It is the average percent change in body weight, across everyone assigned to one arm of one trial, from a weight recorded at a specific moment the protocol defined, to a weight recorded on a specific week the protocol defined. Change any one of those three things and the number changes.

Three things travel with that percentage and almost never travel with the headline. The trials ran for roughly a year and a half. Everyone in them, including the people getting placebo injections, was also enrolled in a diet and exercise program. And the average sits on top of a spread wide enough that a large minority of people in the same arm did substantially better or worse than it.

None of that makes the figures wrong. It makes them narrower than they sound. What follows is what each part of the number actually refers to, taken from the Clinical Studies sections of the labels themselves.

The clock runs for about a year and a half, not a season

The tirzepatide weight-management trials assessed weight reduction after 72 weeks. The semaglutide injection trials ran 64 to 72 weeks depending on the study, and the semaglutide tablet trial ran 64 weeks. The orforglipron trials also ran 72 weeks. Seventy-two weeks is nearly seventeen months.

That length matters twice. First, the early part of each trial was spent building up to the amount being tested, so the headline percentage is not describing the first few weeks of anything — the labels note that the tirzepatide trials assessed weight at 72 weeks with at least 52 of those weeks spent at the amount under study. Second, a percentage earned over seventeen months is not a rate you can divide by four and call a monthly expectation. The curves in these trials flatten. The number is an endpoint, not a slope.

So when a figure appears next to a program that bills monthly, the useful question is how many months of continuous treatment that figure represents. In these trials, the answer is well over a year.

Everyone in the trial got a diet and exercise program — including the placebo group

This is the part that surprises people most. In the tirzepatide weight trials, every patient received a standard lifestyle intervention: instruction on a reduced-calorie diet, physical activity counseling with a recommended minimum of 150 minutes a week, and counseling on behavior-modification strategies, beginning with the first dose and continuing throughout. The semaglutide and orforglipron weight trials describe the same arrangement.

The placebo groups lost weight on that alone. In the main tirzepatide trial in adults without type 2 diabetes, the placebo group's mean change was a 3.1 percent reduction. In the corresponding semaglutide trial, the placebo group's mean was 2.4 percent. In the two newer semaglutide trials, placebo means were 3.9 and 3.8 percent.

One semaglutide trial went further and put everyone on an intensive program: an eight-week low-calorie diet, then a reduced-calorie diet for the remaining sixty weeks, with physical activity starting at 100 minutes a week and rising toward 200. The placebo group in that trial lost 5.7 percent — more than double the placebo result in the trial with the standard program. Same molecule, same class, different background program, different placebo number.

That is why the label reports two figures for every arm and not one. The mean change from baseline includes whatever the lifestyle program contributed. The difference from placebo is the part attributable to the drug. In the main tirzepatide trial the highest dosage studied produced a mean reduction of 20.9 percent and a difference from placebo of 17.8 percentage points. In the main semaglutide injection trial the figures were 14.9 percent and 12.4 points. The smaller number in each pair is the one that isolates the medication.

Two of these trials did not start the clock at the patient's original weight

In a tirzepatide trial designed around lifestyle therapy, patients first went through a twelve-week intensive lifestyle lead-in before any randomization. Only those who had already lost at least 5 percent of their body weight during that lead-in were allowed into the randomized period. Of the 806 patients initially enrolled, 227 left during the lead-in, 141 of them specifically because they did not reach that 5 percent threshold. The 579 who continued had already lost an average of 6.9 percent.

The headline result from that trial — an 18.4 percent mean reduction — is measured from the weight at randomization, after the lead-in, in a group pre-selected for having already responded to a diet program. It is a real result about maintenance and further reduction in that group. It is not the same quantity as a percentage measured from an untreated starting weight.

A semaglutide withdrawal trial has the same structure. Patients went through a twenty-week run-in on the drug; 99 of 902 were permanently off the product before randomization even happened, most commonly because of adverse reactions. Mean weight at enrollment was 106.8 kg; mean weight at randomization, twenty weeks later, was around 96. The reported percentage runs from that later, lower number.

The label does not leave this to inference. It states plainly that because patients who discontinued during titration and those who did not reach the target amount were not eligible for the randomized period, the results may not reflect the experience of patients in the general population who are first starting the drug. That sentence is printed in the prescribing information, and it is the single most useful thing on the page for anyone comparing figures across trials.

The average is not the outcome, and it is not a ceiling either

Every one of these labels reports the distribution alongside the mean, because the mean conceals it. In the main tirzepatide trial, at the highest dosage studied, 90.9 percent of patients lost at least 5 percent of their body weight — but 56.7 percent lost at least 20 percent. In the main semaglutide injection trial, 83.5 percent reached at least 5 percent, while 30.2 percent reached at least 20 percent.

Read those pairs carefully and two things follow. A phrase like up to twenty percent is not the average and not what most people got; it is a threshold that a subset crossed. And the average is not a cap: in every one of these trials, some people lost considerably more than the mean and some lost little or nothing, which is exactly why the labels print four thresholds instead of one.

The distributions in these trials are wide and they overlap between arms. Nothing published tells you in advance where in that distribution any individual will land.

A meaningful share of the final numbers are imputed, and the placebo arms are missing more

At the 72-week mark in the main tirzepatide trial, body weight was missing for 21.6 percent of the patients assigned to placebo, and for roughly 10 percent of those assigned to the drug. In the newest semaglutide trial in patients with obesity, the figures were 15.0 percent missing for placebo and around 5 to 6 percent for the treated groups. In the orforglipron trial in adults without diabetes, 24 percent of the placebo group was missing a Week 72 weight.

Those gaps are filled statistically. The labels describe the methods — multiple imputation drawing on patients who dropped out and were retrieved, or on observed data within the same treatment group. This is standard, disclosed, and reasonable practice. It is also a real feature of the numbers: a portion of every headline figure is estimated rather than weighed.

The asymmetry is the interesting part. More placebo patients went missing because more of them quit. In the main tirzepatide trial, 26.4 percent of the placebo group discontinued study drug versus 14.3 to 16.4 percent of the treated groups. In the orforglipron trial in adults without diabetes, 30 percent of the placebo group discontinued versus 22 to 24 percent of the treated groups. A comparison against a group that is leaving faster than yours is a comparison worth understanding before quoting the gap between them.

Not every percentage in this class measures body weight at all

The semaglutide trial in patients aged 12 and older used percent change in BMI as its primary endpoint, not percent change in body weight. The reported 16.1 percent figure is a BMI reduction. The body-weight change in the same trial, 14.7 percent, was explicitly not included in the prespecified hierarchical testing.

BMI and body weight are different quantities, and in a group still growing they can move differently. A percentage lifted out of an adolescent trial and set beside an adult body-weight percentage is comparing two different measurements that happen to share a percent sign.

The same caution applies elsewhere in these labels. The obstructive sleep apnea trials measured an apnea index as the primary endpoint and reported weight as a secondary result. The cardiovascular trial measured time to a first cardiac event. Percentages in this class are attached to whatever the trial set out to measure, and that is worth checking before treating two of them as comparable.

How to read one of these figures in the wild

When a percentage appears in an advertisement, a program page, or a headline, four questions get you most of the way to what it means. Which drug and which trial. Measured from when, and to when. Compared against what, and what did that comparison group get. And is it a mean, or a threshold that some proportion of people crossed.

If the answer to any of those is unavailable, the number is not necessarily false. It is simply not yet interpretable, and it is not a forecast for you. The trials establish that these drugs produce weight reduction well beyond a diet and exercise program alone, over a period longer than a year, in populations selected the way trial populations are selected. That is a strong finding. It is a different statement from a promise about a particular person over a particular quarter.