FDA Cites Crazy Stats, Misuses Stats, on PPV or FP FN of "Probabilistic tests"

 The Genomic Health Oncotype test gives out ranges mostly from 5% cancer reccurence risk to 40% cancer recurrence risk.   So if you get the former, and recur, is that a False Negative?  If you get the latter and do NOT recur, is that 60% FP?   I don't think so, it's simply statistics.

FDA seems to deliberately misunderstand this, logging Oncotype Dx and other probalbilistic calls as being either right or wrong and thus either FN or FP.  

I mean to explain in a blog.  here are links.

###

2. FDA LDT Proposal: Now that this is under review at OMB,  

 

2. There’s a point of view that, given the large number of substantive comments received and the relatively quick turnaround from the agency, it’s unlikely that FDA would have made significant revisions from the proposal.

 

2. Given the workload this would entail for FDA, however, it would seem that something would need to be grandfathered in, which leaves the key question being which tests / type FDA views as posing the greatest risk (e.g., prenatal screening, cancer treatment decisions) and is therefore less likely to grandfather in.

 

2. To that point, I managed to find this Sept. 22, 2023 FDA memorandum on specific testing examples “that raise public health concerns,” which was listed as a footnote citation in that joint FDA / CMS statement from a few weeks ago

The "few weeks ago" is FDA CMS press release.

THe second is this memo which I discussed above briefly

https://www.regulations.gov/document/FDA-2023-N-2177-0076

I've filed hard copy as 

2023 FDA 1003 Shuck Goofy Memo on FP FN on Probabilistic LDTs 14p.pdf