Theranos Has Struggled with Blood Tests

It's not necessarily bad business or bad practice to launch a company before the technology is ready, and to use a more conventional, cheaper substitute in the interim. I remember reading about a company that offered automated translation, but started out with human translators to validate the increased market size at the lowered price point. They used the validation to raise enough money to continue growing a user base at a per-unit-loss as well as develop the machine translation software to the point of marketability.

Here's a nice profile from the New Yorker in late 2014:

http://www.newyorker.com/magazine/2014/12/15/blood-simpler

I mean "nice" as in, a good read...it's a flattering profile of Holmes -- she comes off well, as a genuinely passionate person, and even if she had stayed in school, it sounds like she would've still made a great impact.

But the article also raises concerns that seem to be corroborated by the OP:

> Clarke argues that finger-stick blood tests aren’t reliable for clinical diagnostic tests; because the blood isn’t drawn from a vein, the sample can be contaminated by lanced capillaries or damaged tissue. Holmes strongly disagrees: “We have data that show you can get a perfect correlation between a finger stick and a venipuncture for every test that we run.” When I asked for evidence, I was sent a document by Daniel P. Edlin, Theranos’s senior product manager, titled “Select Data.” It purported to show favorable results from numerous comparison tests. I asked Edlin if the tests had been conducted by an independent third party. He replied by e-mail: “The clinical tests were conducted by a combination of Theranos and external labs,” but he wouldn’t say which ones.

Um, OK. I don't know much about this testing process...but...what trade secrets are being protected by hiding the methodology and source of the comparative test results?

Sounds like the accuracy testing should be blind but isn't. Why not?

It seems like disclosing that a sample is a test is more likely to negatively affect the accuracy of the test than improve it. The need for double blind trials is normally well understood. This is not an experiment, but as a test it seems like it would benefit from at least single blind: lab is unaware - accuracy tests are mixed randomly into the genuine population of testing requests. Kind of like how the TSA is occasionally tested by inspectors who bring all sorts of weapons through.

Anecdata: I'm a physician, and get tests done for myself and my family at the Theranos place in Palo Alto, for around $5/test (vs $50 for other labs), and I have no concern about the accuracy of the tests. This abstract discusses some of the possible reasons for variation in test results: http://www.ncbi.nlm.nih.gov/pubmed/10887452 Of course, none of this proves that the tests are reliable (consistent) and accurate, just that there are several explanations for unexpected and unrepeatable values in a given patient.

I believe parts are a scam and parts are good The pathology diagnostics field is ripe for disruption with easy access and visible pricing and even if they do this well they will probably disrupt the big players. In terms of new technology I very much doubt they have anything substantial.