Blood test that finds 50 types of cancer is accurate enough to be rolled out

Re:Blood test that finds 50 types of cancer is accurate enou... by AlexCoventry on Friday June 25, 2021 @12:51PM (#27633308)

If we optimistically assume the claimed 0.5% false positive rate is accurate, and all of the US got tested annually, that's 1.64M false positives per year. Cancer.gov is telling me that approximately 1.8M USians were expected to get cancer last year. That's a positive predictive value of 52%. That still seems highly informative, to me; much higher than the PPV for mammography according to this (admittedly old) study: [1].

Assuming the 0.5% FP rate holds (again, I know that's optimistic), would you still regard universal testing with this method to be harmful?

[1] https://jamanetwork.com/journals/jama/article-abstract/40935...

[ Parent | on HN ]

Re:Blood test that finds 50 types of cancer is accurate enou... by amelius on Friday June 25, 2021 @01:21PM (#27633719)

Isn't this exactly why the headline reads "blood test good enough to be rolled out"?

I.e., higher FP rate, and it wouldn't be rolled out.

[ Parent | on HN ]

Re:Blood test that finds 50 types of cancer is accurate enou... by AlexCoventry on Friday June 25, 2021 @01:56PM (#27634152)

IIRC, the actual article says they're going to do a large trial, with 150k people, and someone in the article expresses skepticism that the estimated FP rate of 0.5% is accurate, because so far it was tested in people where there's some evidence of cancer.

[ Parent | on HN ]

Re:Blood test that finds 50 types of cancer is accurate enou... by dan-robertson on Friday June 25, 2021 @06:32PM (#27637076)

Is that actually good though? It feels to me that there’s a big difference between explaining the implications of a positive result to someone who understands bayes theorem and someone who doesn’t.

Another aspect of cancer screening is that detecting cancer earlier can improve key statistics like 5 year survival without affecting the actual disease in any way. Which can make screening sound more effective than it is.

[ Parent | on HN ]

Re:Blood test that finds 50 types of cancer is accurate enou... by TaupeRanger on Friday June 25, 2021 @12:57PM (#27633373)

Good question. The proof is in the pudding. Run a well powered RCT and if the intervention helps patients live longer and/or better, then we should consider using it! It will be hard to design a study that includes all possible detected cancers, but there are ways around this - perhaps we begin by studying those cancers that are most common and lethal, and go from there.

[ Parent | on HN ]