Semaglutide weight loss cannot be predicted from the first visit

Design

Prespecified secondary analysis of a randomised, double-blind trial

Population

7,594 adults with obesity and atherosclerotic disease, without diabetes, on semaglutide for at least 1 year

Primary outcome

Peak percentage body weight loss and its baseline predictors

Effect

Median peak loss 12.5% (IQR 7.9–17.9); baseline model explained under 10% of variability

This prespecified analysis of the SELECT trial looked at 7,594 adults with obesity and established atherosclerotic disease, without diabetes, who stayed on semaglutide for at least a year. The researchers tested whether routine baseline information — demographics, history, laboratory values and measurements — could predict how much weight each person would lose at their lowest point.

It could not. Median peak weight loss was 12.5% (IQR 7.9–17.9), reached after a median of 19 months. The full model explained less than 10% of the variation between individuals, and most of that came from sex: women lost a median 16.6% against 12.4% in men. People whose maximum dose was 2.4 mg lost a median 13.3%, compared with 6.0% for those who stayed at 0.24 mg.

In clinic this means there is no reliable way to tell a patient in advance whether they will be a large or modest responder. The honest counselling is a range, a timeline of more than a year to the lowest weight, and a plan to reassess on the patient's own trajectory rather than a predicted one.

• Quote a range when counselling: median peak loss was about 12.5%, with a quarter losing under 8%.
• Tell patients the lowest weight typically comes after about 19 months, not in the first few months.
• Expect larger losses in women on average, but do not use sex to predict an individual's response.
• Judge response on the person's own weight trajectory at full tolerated dose, not on baseline features.
• These were people without diabetes; weight loss with semaglutide is usually smaller in type 2 diabetes.

Why it matters

The hope of picking likely responders before prescribing is not supported by the routine data clinicians actually have.

Don't overread it

The dose comparison is not randomised — people who stayed on low doses may differ in tolerance and adherence.

The statistics, in plain English

An interquartile range of 7.9–17.9% means the middle half of patients lost between those amounts; a quarter lost less and a quarter more. A model that explains under 10% of variability leaves more than 90% of the difference between people unexplained, which is why it cannot predict an individual's result.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.