Thank you. Missing 30% of high grade dysplasia is (or ought to be) a deal killer. It’ll certainly be a patient killer.
A side question for which there’s probably no data is whether the misses are a result of genetic diversity in the lesions or a result of just non-uniform distribution of lesion markers in fecal samples . IOW …
Imagine we have a subject who we knew via colonoscopy had significant lesions. Now we give them the test once and it comes back clean. Would rerunning the test, say, weekly keep missing the problem every time since the screen isn’t sensitive to whatever markers that particular patient+lesion is throwing? Or would we see hit or miss detection based on the luck of whether there’s enough lesion marker present in the sample to be detected?
There’s an implicit assumption in most medical testing that the Principle of Mediocrity applies: samples are uniform (enough). Pretty good bet for a CBC. But are they in this case?
At $600 a pop it’s obviously not going to be repeated often in the clinic. The point of the experiment above would be for the manufacturer to learn how it’s failing and therefore how it ought to be improved.
In principal the test’s sensitivity can be turned up, at the cost of creating even more false positives. But if there’s huge variation in actual factual signal from sample to sample then that’s not the right solution. The right solution for that would be more and larger samples homogenized before processing.
Again not that I expect you (or anyone) can explicitly answer the question. Just musing on the issues inherent in this kind of testing.