How to read a GLP-1 clinical trial

Who took part in the trial

Look for diabetes status, age, starting BMI, related conditions and major exclusions. A study in adults without diabetes does not predict results for people with diabetes or for adolescents. Trial participants also get a set level of follow-up and support, which often differs from routine care.

Selection also happens after treatment starts. In a withdrawal trial, people who reach a later randomisation point may already have shown they tolerate the first phase of treatment. Results from that phase do not describe what every new user goes through.

What a percentage means

Percentage weight change usually means change from a stated baseline. A 15% reduction from 100 kg is 15 kg. From 80 kg it is 12 kg. The sum helps you read a result, but it will not predict yours.

Tell average weight loss apart from the share of people who reached a threshold. Tell percentage points apart from relative percentages too. If two averages are 10% and 15%, the gap is 5 percentage points, and the second figure is 50% larger than the first. Both statements are true and give quite different impressions.

How treatment changes and missing data were handled

Some analyses estimate outcomes whether or not people stayed on treatment. Others estimate an outcome assuming people kept taking the medicine as planned. These answer different questions and can give different headline numbers from the same trial.

Check how many people stopped, why they stopped and how missing measurements were handled. A strong result among people who finished treatment may not describe everyone who started. Safety findings need context as well: how events were defined, how long the trial ran and how many people took the medicine all affect what the numbers mean.

Separate trials do not make a league table

Trials differ in population, dose, length, support programme and analysis method. Setting their headline averages side by side gives some context. A well-designed head-to-head study answers the comparison question far better. Even a direct comparison applies most clearly to the products, doses and people studied.

Find the full paper, protocol and trial registration where you can, and tell peer-reviewed findings apart from a press release. Ask whether a claim is about weight, glucose, cardiovascular events or another outcome. A good summary gives the result along with its limits. A selected average does not show one medicine will suit everyone best.

Check who funded and analysed the work

Funding and author affiliations should be easy to find. Industry funding does not make a trial invalid by default, and an independent author does not guarantee sound methods. Read the design, the outcomes set in advance, how missing data was handled and the limitations. Check the full paper says what the promotional headline says. If a result comes from an exploratory subgroup or an analysis run after the main trial, the paper should state this. The distinction affects how much weight the finding deserves in a broad treatment claim.