Biotech Values

[biomaven0]

Interesting analysis from a biostatistician. His main complaint here seems to be comparing change from baseline of some measure in intervention vs. control groups. One issue is that degree of change is related to initial baseline as well as any impact from the intervention. Thus, for example, if you are measuring on say a 5-point scale, then someone that starts out at 5 can only go down, and vice versa from someone starting at zero. So, among other things, regression to the mean distorts your outcomes.

http://www.fharrell.com/2017/04/statistical-errors-in-medical-literature.html#glucpct

Note discussion of recent NBIX/Abbvie elagolix trial.

Instead he proposes some sort of ANCOVA analysis - but I can't say I recall seeing that used much at all in practice.

Comments from our resident stats gurus?

Peter