Reference ranges, prediction intervals [GxP / QC / QA]
❝ Thinking very hard about it I have not ever seen any CRO use the prediction interval approach to define ref. ranges.
I can't remember ever seeing a path lab defining their own reference ranges: whether CRO, central lab involved in Phase III trials, or the lab next door where I had my cholesterol level measured once some years ago (I never went back: what you don't know won't hurt you). Whenever I asked, the answer was that they were taking the ranges recommended by the vendor of the machine and reagents. Which may have been determined in a rather different population on another continent.
Actually I worked for some time at a hospital lab which was involved in such an exercise on behalf of the manufacturer, to establish the normal ranges for thyroid hormones on a new lab machine. They even used a sample of my own blood (talk about freely given consent). Not sure there was much statistics involved then.
❝ After WHO published their rule about aberrant hemoglobin a few years back, some CROs revised their reference ranges in funny ways so that anyone on the brink of keeling over due to anemia could still be enrolled. [...] But I digress, this latter aspect is more a question of low ethical standards than a question of reference ranges.