Analytic

A single fixed approval threshold misprices risk in both directions. Calibrating the costs of wrong approvals and wrong rejections with US burden-of-disease data, Montazerhodjat and Lo find the false-approval rate a conventional trial should accept is about 27.9% for pancreatic cancer, a terminal illness with no existing therapy, and about 1.2% for prostate cancer, against the standard 2.5% used for both. On this accounting the institution is too cautious for people with lethal untreated disease and too permissive for people with milder disease.

Health and flourishing