Not to shift this too far off topic (as if it's been on topic for a few pages anyway), but there are 2 different test types that you're asking about here.
One is mutational testing of circulating tumor DNA (ctDNA) in a person with a known or suspected tumor in lieu of an actual biopsy. These can be faster than a biopsy or somewhat useful when a biopsy is difficult/impossible due to patient or system specific factors. Commercial tests like Guardant 360 and Foundation Liquid CDx are in this category. I personally find them more useful than just guessing what it is based on a CT scan but much less useful than an actual biopsy with tissue based NGS testing. My experience has been that the liquid biopsies don't pick up as many mutations as are present in the actual tumor (when compared side by side) and except for the most clinically obvious tumors, their ability to diagnose the primary site when you only have evidence of metastases is poor. In short, better than nothing, but there are better things.
If you're asking about blood-based multi-cancer early detection such as the Grail Galleri test, it's exciting technology, but you don't really have to read between the lines in
their own manuscript to realize that it's not ready for prime time. In their big screening test of >6500 people, They had 92 positive tests, of which only 35 (a little more than 1/3) actually had cancer. This works out to a true positive test rate of 0.86%. Unfortunately, the false negative rate (negative blood test who were diagnosed with cancer within 12 months of the blood test) was 1.32%, 1.5x higher. Of the 35 true positives, 15 of them were breast, lung, colorectal and prostate, all of which have some reasonable level of evidence for standard screening procedures. So if you want to be really stringent in your definition, this test identified a real cancer in 0.3% of people in the study that would not have been picked up by standard screening when the actual cancer rate in the study was 1.8%, so it caught a little over 15% of the actual cancers. The author's own conclusion is "This study supports the feasibility of MCED screening for cancer and underscores the need for further research evaluating its clinical utility." Which translates to, "our test kind of works, but it mostly doesn't".
To be transparent, I accrued roughly 50 people to this study on my own (my institution at the time was the highest accruing to the study and the 2nd author who now works at Grail used to be at that institution). I think this technology has a lot of promise, and I imagine that there are some AI informed analysis models and testing strategies that are going to make this type of testing a standard in the next decade or so. But the currently available products just aren't there yet.
Now...aren't you sorry you asked?