Any such comparison wouldn't work for just an N=1. You'd need many more people before you could start to draw a conclusion about the Theranos test versus a conventional test. Just an individual doing this wouldn't be anything more than anecdotal at best.
I'm assuming they've done this study, but have refused to publish it for some reason. But this is the real question from the article - where is this study? You don't have to disclose the full technique for the test in order to publish the comparison.
Agreed, a n=1 test wouldn't reveal the most insidious kind of failure: guessing to cover up crummy data (err, I mean "leveraging machine learning techniques to maximize accuracy"). I doubt they're that evil, but it would be nice to know for sure.
> I'm assuming they've done this study, but have refused to publish it for some reason.
Because they're a bog-standard vertical integration play and if that fact comes to light before they are done using the hype train to raise money and grab market share then their competitive advantage will evaporate overnight. If they only release test results to interested parties under NDA it prevents public sector science from rebutting their claims to disruptive rather than incremental innovation, allowing the hype train to steam ahead unimpeded.
I'm assuming they've done this study, but have refused to publish it for some reason. But this is the real question from the article - where is this study? You don't have to disclose the full technique for the test in order to publish the comparison.