Toshiba device tests cancer types from a drop of blood (opens in new tab)

CA-125 test for ovarian cancer is similar. "Best we've got", not "what we'd like to have".

No, not necessarily. Check out these numbers:

Sensitivity is what fraction of the affected people are actually found. Here: 90%, so 10% are missed.

Specificity is what fraction of the unaffected people are detected as such. Here: 95%, so 5% wrongly detected ("false positives").

In Europe, there are 60 cases of lung cancer per 100 000 people.

That makes 54 correctly detected per 100 000, missing 6 cases. That also means 5000 people incorrectly suspected of lung cancer (5% of 100 000).

Update: using the accuracy from the article itself, we would still get a total of 1000 of false negatives (affected but not detected) and false positives (unaffected but suspected). Incidence is still 60/100 000.

Screening techniques are a real problem if their false positive rate is high. Any significant increase in the work up rate will be a real burden on the health system (so you had better see a corresponding benefit). Worse, gold standard for follow up is going to vary by disease and location, but if it is say a biopsy and you are screening a large population you are often going to start killing people without the disease. Not many, of course, but you have to balance that against the putative improvements.

No, because you'll be telling 1 out of 20 healthy people they tested positive for cancer. Or, if it's 95% selectivity per test, which it probably is, 13 out of 20 healthy people will test positive.

I'd suggest reading these to get a better idea of the potential harms of screening tests in the asymptomatic population:

https://sciencebasedmedicine.org/a-skeptical-look-at-screeni...

https://skepticalinquirer.org/exclusive/the_screening_test_t...

https://fivethirtyeight.com/features/the-case-against-early-...

90% sensitivity = 10% false negatives. When the patient has the condition, it is detected 90/100 of the time.

95% specificity = 5% false positives. When the patient doesn't have the condition, it is correctly not detected 95/100 of the time.

Can you please post links to the research that you mention?

Not exactly. What Theranos has done is to make clear in its aftermath why the problem of making a diagnosis from a single drop of blood is difficult. In short, the sample size is too small and may not contain biomarkers of the diseases of interest even if they are present in the body. Unless this advancement in its application can show how and why this is not a problem, there will be resistance towards accepting what is being put out without further substantiating evidence.

At early stages you can simply blast it with chemo/radiation before it manifests into something intreatable though.

The problem w/ Theranos was that at first glance everything about it was physically too small. A single drop of blood is too little, the cost of running the tests was too low - to do all of the separate tests it was supposed to be able to do and still be within the realm of the possible.

The Toshiba machine doing one type of test from one drop of blood at a cost of ~$184 is well within the range of what should be possible in the real world.

The signal to noise ratio is still devastatingly bad.

Most cancers aren't throwing large amounts of detectable crap into your bloodstream. If they were, people would be doing routine cancer tests when you do things like cholesterol screening.

The fact that nobody can do this with vials makes me suspicious of being able to do this with drops.

That's a good start. If they'd be able to do 13 tests from a drop of blood from each of my fingers that'd be 130 tests!

Most likely some form of Windows CE, but close enough.