Home / Research
Study teardowns
Each teardown takes one study and reads it closely. We look at what was actually tested, what the comparator was, who funded and analyzed it, how dropout and missing data were handled, absolute versus relative effects, and who was enrolled. We also cover what the study robustly shows.
Every number comes from a source we opened and cite, with the date it was checked. Charts are rebuilt from the published tables, and figures we couldn't verify are left out. Educational only, not medical advice.
Latest
GLP-1 · Semaglutide 2.4 mg · NEJM 2021
STEP 1: what the semaglutide obesity trial shows, and what it doesn't
−14.9% vs −2.4% at 68 weeks, but the placebo arm got a lifestyle program, the sponsor ran the analysis, and people regained about two-thirds of the lost weight within a year of stopping.
How we read a study
- Design: randomization, blinding, registration, and prespecified endpoints.
- Comparator: placebo, usual care, or an active drug, and what both arms actually received.
- Funding and roles: who designed, analyzed, and wrote it.
- Dropout and missing data: how many finished, and how the analysis handled the rest.
- Effect size: absolute differences next to relative ones.
- Population: who was enrolled and who was excluded.
- Durability and replication: what happened later and in other studies.