Very nice article John. The first study is a poster child for what is wrong with so many clinical trials: they are designed so there is very little chance that a conclusion of any kind will be made at the end. Combine sample size samba (gaming the effect size to be more than the minimal clinically important difference) and having an unre…
Very nice article John. The first study is a poster child for what is wrong with so many clinical trials: they are designed so there is very little chance that a conclusion of any kind will be made at the end. Combine sample size samba (gaming the effect size to be more than the minimal clinically important difference) and having an unrealistically high estimate of the event rate is a recipe for futility, wasted resources, and raising ethical concerns about needlessly exposing patients in a setting where learning from them is unlikely. Contrast that with the design diagrammed here: https://hbiostat.org/bayes/bet/design
Very nice article John. The first study is a poster child for what is wrong with so many clinical trials: they are designed so there is very little chance that a conclusion of any kind will be made at the end. Combine sample size samba (gaming the effect size to be more than the minimal clinically important difference) and having an unrealistically high estimate of the event rate is a recipe for futility, wasted resources, and raising ethical concerns about needlessly exposing patients in a setting where learning from them is unlikely. Contrast that with the design diagrammed here: https://hbiostat.org/bayes/bet/design