MADRS

The question readers ask is what counts as a good MADRS score, and in our view that is the wrong question to ask of a trial. An individual score describes a patient. What a trial reports is a difference between two groups, and the useful yardstick is not the 0 to 60 range but the much narrower band the drug-versus-placebo gaps actually occupy. Across the readouts we have covered, those gaps run from 3.6 points to 8.1. So a 4-point separation is not a small number because it is small out of 60; it is a middling number against the results we have reported. That band is our own coverage and not a survey of the field, which is larger than what we have written about, so treat it as a working reference and not a boundary. Read any MADRS headline by asking what it was measured against, then where it sits among the trials you can actually compare it with.

We glossed the scale in Issue 11: "MADRS is the depression scale almost every trial in this field is scored on. It runs from 0 to 60, higher is worse, and the useful thing is the yardstick rather than the range: the drug-versus-placebo gaps we have reported on it run from 3.6 points in Compass's placebo-controlled pivotal to 8.1 in Definium's Emerge. That narrow band is where this field's arguments happen."

The scale comes from Stuart Montgomery and Marie Asberg, whose name is written Asberg here and Åsberg in the original, "A new depression scale designed to be sensitive to change", British Journal of Psychiatry, 1979, volume 134, pages 382 to 389 (doi:10.1192/bjp.134.4.382). The title is the point: it was built to detect change under treatment rather than to describe depression exhaustively, which is why it became the trial instrument.

Response is usually defined as a percentage fall rather than a fixed score, and the threshold is a choice the sponsor makes. Issue 6 records Compass defining "a clinically meaningful response, defined as a 25% or greater MADRS reduction" at Week 6, and our reading of what that implied: "on the trial's own framing fewer than four in ten of the top-dose patients cleared even a soft response bar at Week 6."

We first used the scale in Issue 1, reporting Swiss real-world scores falling from around 31 to around 20, and noting that "a change of that magnitude on the MADRS is usually considered clinically meaningful."