Average Weight Loss vs Responder Rates: Two Different Trial Results

An average weight loss tells you what happened to the group as a whole. A responder rate tells you what share of people crossed a set line, such as losing at least 5%, 10% or 15% of their starting weight. GLP-1 trials report both because they answer different questions, and reading only one gives you half the picture.

Think of a garden where every seedling got the same water and sunlight. The "average height" might be 15 cm. The seedlings themselves range from 4 cm to 30 cm. Trial results work the same way.

What does "average weight loss" in a GLP-1 trial mean?

Most GLP-1 obesity trials list one main headline number: the mean (average) percentage change in body weight from the start to the final week.

• STEP 1 (semaglutide 2.4 mg, 1,961 adults without diabetes, 68 weeks): −14.9% on semaglutide vs −2.4% on placebo.
• SURMOUNT-1 (tirzepatide, 2,539 adults without diabetes, 72 weeks): −15.0% (5 mg), −19.5% (10 mg), −20.9% (15 mg) vs −3.1% on placebo.
• ATTAIN-1 (oral orforglipron, 3,127 adults without diabetes, 72 weeks): −11.2% on the 36 mg dose vs −2.1% on placebo.

An average is good for comparing groups within the same trial. It's not good at telling you what you would experience, because almost nobody lands exactly on it.

What is a responder rate?

A responder rate is the percentage of people who reached a chosen threshold. Trials usually pick round numbers like 5%, 10%, 15% or 20%. Here's STEP 1 laid out both ways:

Lost at least…	Semaglutide 2.4 mg	Placebo
5%	86.4%	31.5%
10%	69.1%	12.0%
15%	50.5%	4.9%

Read that table slowly. Three things jump out:

1. About half did not reach 15%, even though the average was −14.9%.
2. About 14 in 100 didn't reach 5% on semaglutide.
3. About 1 in 3 people on placebo lost 5% or more. Placebo groups got lifestyle support too, and that alone works for some people.

What do the other big trials show?

SURMOUNT-1 reported that 85%, 89% and 91% of people on the three tirzepatide doses lost at least 5%, vs 35% on placebo. At the 10 mg and 15 mg doses, 50% and 57% lost 20% or more, compared with 3% on placebo.

ATTAIN-1 on the 36 mg dose: 54.6% lost at least 10%, 36.0% lost at least 15%, and 18.4% lost at least 20%, compared with 12.9%, 5.9% and 2.8% on placebo.

That last one is a nice example of why you need both numbers. The average was about 11%, yet nearly 1 in 5 people on the top dose lost a fifth of their body weight. Meanwhile, many others lost less than the average. One number can't hold all of that.

Spot the trick

1. "The average person on this drug loses 15%."

Show answer

The trial's average was about 15%. That's not the same as "the average person." In STEP 1, roughly half of people on semaglutide lost less than 15%. Trial averages also depend on how researchers handled people who stopped treatment.

2. "86% of people lost weight on semaglutide!"

Show answer

Check the threshold and the placebo group. The 86.4% figure is the share who lost at least 5%. On placebo, 31.5% hit that same mark. The fair comparison is the difference between groups, not 86% on its own.

3. "Drug X: average 21%. Drug Y: average 15%. X is clearly better."

Show answer

Only if they were tested in the same trial. Separate trials differ in who they enrolled, how long they ran, and how they handled dropouts. A head-to-head trial like SURMOUNT-5 (tirzepatide −20.2% vs semaglutide −13.7% at 72 weeks) is a much fairer comparison than two separate averages.

Scoring: 3 out of 3 and you can officially roll your eyes at "average person" headlines.

Why do trials choose 5%, 10% and 15% as thresholds?

The lines are study definitions, not laws of nature. "At least 5%" is a long-standing benchmark in obesity drug trials: STEP 1 and SURMOUNT-1 both made it a co-primary endpoint alongside the average change. Higher thresholds show how many people got larger results, and those are the numbers that tend to end up in headlines.

What counts as a good result for one person depends on their health goals, starting point and other conditions. We cover that in what counts as a response to weight-loss medication.

Can an average and a responder rate disagree?

They can tell different stories. Imagine two made-up trials, each with a 12% average:

	Trial A	Trial B
Average loss	12%	12%
Lost ≥10%	70%	45%
Lost ≥20%	5%	25%

In Trial A, most people landed near the middle. In Trial B, results were spread out: more big losers, more small losers. Same average, very different experiences. (These numbers are invented to show the idea, not taken from any study.)

When a paper includes a "waterfall plot" (one bar per person, sorted from most loss to most gain), look at it. It's the most honest picture of spread you'll find.

What should you take from this for your own treatment?

If your results don't match a trial average, you are not broken, and the medicine has not necessarily "failed." Trials show a wide range of results, and some people respond less for reasons researchers are still studying.

What helps is having your own trend line instead of a stranger's average. Logging weight and measurements over weeks (in a notebook or an app like GLPily) gives you and your prescriber something real to look at. Any decision about dose or switching medicines belongs in that conversation, not in a comparison with a trial table.

For other kinds of headline math, see relative vs absolute risk, and for how dropouts shape the averages above, read why GLP-1 trials report dropouts.

Common questions

What percentage of people lose weight on semaglutide?

In STEP 1, 86.4% of people on semaglutide 2.4 mg lost at least 5% of their body weight over 68 weeks, 69.1% lost at least 10%, and 50.5% lost at least 15%. Results in everyday care can differ.

What percentage of people lose 20% on tirzepatide?

In SURMOUNT-1, 50% of people on 10 mg and 57% on 15 mg lost at least 20% of their body weight at 72 weeks, compared with 3% on placebo.

Why does the placebo group lose weight in GLP-1 trials?

Placebo participants usually get the same diet and activity counseling as the medicine group, and that support helps some people lose weight. That's why trials compare the two groups instead of looking at one alone.

Is the average or the responder rate more useful?

Both. The average compares groups; the responder rate shows how results were spread across people. Look at them together, along with the placebo numbers.