How to Read a GLP-1 Study Headline in Five Minutes

To read a GLP-1 study headline in five minutes, ask five questions in order: where did this come from, who was studied, what was measured, compared with what, and what number is missing. Set a timer, spend about a minute on each, and you will catch most of the spin before it reaches your group chat.

You do not need a statistics degree. You need a routine, a little suspicion, and the patience to click one link past the headline. Below is that routine, tried out on real GLP-1 trials.

Why do GLP-1 headlines need a routine?

GLP-1 news moves fast, and most of it is technically true. The trouble is what gets left out. "Weight-loss drug cuts heart attack risk by 20%" and "Weight-loss drug: 1.5 fewer heart events per 100 people over about three years" can describe the same trial. Both are accurate. They just feel very different when you read them over breakfast.

A routine keeps you from reacting to the feeling and helps you find the facts. Here it is.

Minute 1: Where did this actually come from?

Scroll to the bottom of the article, or search the drug name plus "trial", and find out what kind of source sits under the headline:

• A company press release. Often the first public news, and usually written before any outside review.
• A conference presentation or abstract. Real data, but short, and not fully peer reviewed.
• A preprint. A full paper posted before peer review. The medRxiv server says preprints "should not be relied on to guide clinical practice."
• A peer-reviewed journal article. The fullest version, with methods and limitations.
• An FDA label or safety communication. The official word on what a medicine is approved for and what it warns about.

Real example: On August 7, 2025, Lilly's press release on the ATTAIN-1 trial of oral orforglipron reported an average 12.4% weight loss on the top dose. The peer-reviewed paper in the New England Journal of Medicine appeared on September 16, 2025, and its primary result was 11.2%. Nobody lied. The two documents used different ways of handling people who stopped the pill, which we explain in our post on trial dropouts and missing data. But a headline built on one number would not match the other.

Minute 2: Who was studied?

Results belong to the people who were in the study. Look for three things: health conditions, size and length.

• SELECT enrolled 17,604 adults aged 45 or older who already had cardiovascular disease, a BMI of 27 or more, and no diabetes. Its heart results describe that group, not every person on semaglutide.
• STEP 1 enrolled 1,961 adults with obesity (or overweight plus a weight-related condition) and no diabetes.
• SURMOUNT-1 enrolled 2,539 adults, also without diabetes.

If you have type 2 diabetes, are under 45, or have never had a heart problem, the headline may still be interesting. It just was not measured in people like you. Trial eligibility rules decide who those people were, so they're worth a glance.

Minute 3: What was measured, and for how long?

"Works" can mean very different things:

Headline word	What the study may actually have measured
"Weight loss"	Average % change in body weight at week 68 or 72
"Protects the heart"	Fewer heart attacks, strokes or cardiovascular deaths
"Lowers inflammation"	A blood marker such as hsCRP, not a health event
"Helps cravings"	A questionnaire score

A change in a lab marker is not the same as fewer hospital visits. Also check the clock: STEP 1 lasted 68 weeks, SURMOUNT-1 72 weeks, and SELECT followed people for a mean of 39.8 months. A 12-week result cannot tell you about year three.

Minute 4: Compared with what?

Every result is a comparison, so name the comparison:

• Placebo: a dummy treatment. In STEP 1, both groups also got lifestyle support, and the placebo group lost 2.4% on average versus 14.9% with semaglutide.
• Another drug: SURMOUNT-5 compared tirzepatide with semaglutide head to head (−20.2% vs −13.7% at 72 weeks).
• Nothing randomized at all: health-record studies compare people who chose different treatments, which is a different and weaker kind of comparison for proving cause and effect.

A headline that says "Drug A beats Drug B" based on two separate placebo trials is comparing apples with apples grown in different orchards.

Minute 5: What number is missing?

This is where the fun is. Big percentages love to travel alone. Ask for their partners:

• A relative number needs an absolute one. SELECT's 20% lower risk came from 8.0% of the placebo group vs 6.5% of the semaglutide group having a major heart event, a difference of 1.5 percentage points. Our relative vs absolute risk guide does the full math.
• An average needs a spread. STEP 1's average was −14.9%, yet only about half (50.5%) reached 15% or more. More on that in average weight loss vs responder rates.
• A benefit needs its harms. In SELECT, 16.6% of people on semaglutide stopped the study drug because of adverse events, compared with 8.2% on placebo.

Try it: spot the trick

1. "New pill helps people lose 12% of their weight!"

Show answer

Check which number this is. ATTAIN-1's 12.4% came from the press release's "efficacy estimand" (roughly, people who kept taking the pill). The journal's primary result, counting everyone who was randomized, was 11.2%. Both are real numbers, but they answer different questions.

2. "GLP-1 drug cuts heart risk by a fifth for people with obesity."

Show answer

Two catches. First, SELECT studied people with existing heart disease, not everyone with obesity. Second, "a fifth" is relative: the absolute difference was 1.5 percentage points over about 3.3 years.

3. "Study links GLP-1 drugs to [scary thing]."

Show answer

"Links" usually signals an observational study or adverse-event reports, which show associations, not proof. Look for randomized data or an FDA statement. For example, after reviewing trials and a large observational study, FDA asked in January 2026 for the suicidal-ideation warning to be removed from GLP-1 weight-loss labels. Our post on why adverse-event reports aren't proof explains the difference.

Scoring: 3 out of 3 means you are ready for the GLP-1 headline decoder quiz. Fewer than that is fine too. Each trick only fools you once.

The five-minute card (copy this)

Minute	Question	Where to look
1	What is the source?	Bottom of the article, journal site, ClinicalTrials.gov number
2	Who was studied?	Methods / "participants" paragraph
3	What was measured, for how long?	"Primary end point" in the abstract
4	Compared with what?	"Randomly assigned to…" sentence
5	What number is missing?	Results table: absolute rates, spread, harms

What should you do with a headline after five minutes?

If it passes, it is worth bringing up at your next appointment, especially if it concerns a medicine you take. If it fails, you have saved yourself a worry spiral. Either way, nothing in a headline is a reason to change your dose or stop a medicine on your own. Talk to your prescriber or pharmacist first.

If you track your own treatment in GLPily, you can jot the headline in your journal and bring it, with your own logs, to that conversation.

Common questions

What is the fastest way to find the original GLP-1 study?

Search the trial name (STEP, SURMOUNT, SELECT, ATTAIN) or the drug name plus "NEJM" or "Lancet", or look for a ClinicalTrials.gov number starting with "NCT" in the article. That number leads to the registry entry, which links to publications.

Is a press release about a GLP-1 trial reliable?

It can be accurate, but it is written by the sponsor and usually arrives before peer review. Treat it as a preview and check the full paper when it comes out.

Why do two articles about the same GLP-1 trial quote different numbers?

Often because they report different analyses (for example, everyone randomized vs people who stayed on treatment), different doses, or different time points. Check which group and which week each number refers to.

Should I change my GLP-1 treatment because of a news story?

No. Bring the story to your prescriber or pharmacist. They can tell you whether it applies to your medicine, dose and health history.