How much evidence
before it may be sold
Phase I only says it can be tolerated. Everything after that is these twenty lessons — choosing a dose in phase II, calculating how many people phase III needs, nailing down the primary endpoint in advance, spending a little alpha at each interim look, and finally stacking the dossier and standing in front of a regulator. The material is the trial programme of two drugs that actually reached approval in pulmonary fibrosis. Python runs right inside this page.
Start lesson 1 →How you'll learn
How many people phase III needs is not taste, it is arithmetic. Put in effect size, variance, alpha and power and the number comes out. You turn the sliders yourself.
Why does a regulator want the same trial twice? And what happens when it succeeds only once — that really happened, and the drug was approved anyway.
Primary endpoint, testing order and interim timing are fixed before anyone sees the data. You run the simulation that shows what collapses if they are not.
When the trials end they become paper. The five CTD modules, the review clock, the advisory committee, and which trial each sentence of the label actually came from.