Know the questions a skeptical FDA reviewer will probe — before you're in the room.
The last post argued that how you randomize is a design decision most biostatisticians treat as settled before the interesting work begins. REMAP-CAP showed what happens when that decision is taken seriously. The carat package showed that even standard covariate-adaptive randomization has inferential consequences most of us aren't
A simulation study on why historical control borrowing reduces sample size, and what happens when the history is wrong
Last week I published what I submitted to the FDA on their January 2026 Bayesian guidance. Comment 3 argued that when you compose Bayesian trial components (priors, borrowing, sequential monitoring), the system's operating characteristics can diverge from what component-level analysis predicts. This post shows you the math. The
