Thursday, October 22, 2015

Theranos, The Wall Street Journal, and what Everyone is Missing: the Harms of Overtesting

Theranos is a Silicon Valley start-up dedicated to revolutionizing (some would say disrupting) the medical blood testing industry.  The claim to fame of the company was to use a single drop of blood from a finger stick to run hundreds of tests.  This premise led to a $9 Billion valuation and rock-star status for it's founder, Elizabeth Holmes- and was recently threatened by a Wall Street Journal expos√©, which claimed the company's tests were not that accurate and that they used traditional blood testing for most of their work.

There are many good summaries and explanations of the dispute, with most of them siding with the Wall Street Journal over Theranos.  The Wall Street Journal followed up with a story that the FDA had made a surprise inspection (the medical establishment's equivalent of a midnight no-knock SWAT team raid) and ordered Theranos to refrain from using their machines on all but one test- on the grounds that they didn't trust the data Theranos was giving them for approval.

But all of this controversy misses a fundamental aspect of the Theranos story that is far more troubling: Direct-to-Consumer Medical Testing.

In the United States, if a person has a healthcare concern or question, they go to their doctor.  That doctor then determines the appropriate testing (if any) for that person at that time.  These tests are then ordered, the patient gets them drawn, and the results are sent to and reviewed by the physician.  Based on those results, the doctor then contacts the patient if necessary or (more often) does nothing because the lab tests didn't reveal any need for further action.

Right now, a healthy 50 year old man without any symptoms or problems should have exactly 5 blood tests according the US Preventative Services Task Force: HIV, cholesterol, hepatitis B, hepatitis C, and a test for diabetes.   There is no evidence that any other routine testing is helpful, and it may actually be harmful.  How?

Because of the positive predictive value problem.  Almost no tests are 100% sensitive or specific.  Some people who have a disease will test negative, and be falsely reassured.  Far more people will have nothing wrong with them- but still be told that they tested positive, resulting in unnecessary intervention (or at least another test).

Consider this- suppose you take the HIV test, and test positive.  What is your chance of actually having HIV?  If you don't engage in unprotected sex or drug abuse, it's actually lower than you might think.

How does this work?  Let's say you have a test that is 99% sensitive, and 99% specific.  That means that if you test 10,000 people with HIV, 9900 of them will in fact test positive (sensitivity); and if you test 10,000 people without the disease, 9900 of them will test negative (specificity).

Now let's say you live in the United States, where 1.2 million people 13 and older are estimated to have HIV.  The population of the US is around 320 million.

So, if you tested everybody in the United States, of those who have HIV, how many will test positive?  .99 times 1.2 million, or 1,188,000.

How many who don't have HIV will still test positive?  .01 times 320 million (technically 320 million minus 1.2 million), or 3,188,000.

So if you test positive, what's your chance of actually having the disease?  1,188,000/(1,188,000+3,188,000), which comes out to 27%, which is called the Positive Predictive Value (PPV) of the test in this population, or the likelihood that if the test is positive you actually have the disease.

That's fine for HIV- everyone who tests positive on this screening test can then get confirmed by another, more specific test to make sure they really have the disease.  And for that matter, the HIV test happens to be more sensitive and specific than 99%, and there is a big benefit to finding the disease early instead of waiting for symptoms- which is why the scientific consensus is that screening everyone for HIV is a good idea.

But for many, many other conditions, this reality results in many patients getting treatment that they do not need.  There is a serious problem in the US with physicians, nurse-practitioners, and other healthcare providers ordering way too many tests, as documented by Atul Gawande and many others.  In fact, there are many campaigns and efforts to reduce the number of tests doctors order.

The reasons why doctors order too many tests now are complex- patient satisfaction, training, the fact that doctors remember the one disease they missed far more than the countless times they over-tested, medical malpractice issues, and payment incentives to name a few.  But, the problem will likely only get worse if people are told by Theranos that:

When billionaire Mark Cuban publicly advocated getting a bunch of lab tests done every 3 months, just to see what they are, the response across the internet was rapid and near-uniform: this is a terrible idea, though there certainly were some dissenters.  This is called panel profiling, and was all the rage in the 1970s- but was rapidly abandoned in the 1980s because of how many false positives resulted in massive amounts of unnecessary follow-up testing and procedures, as noted in this paper attacking Theranos which preceded the Wall Street Journal piece.

Let's show how this breaks out in practice.  Suppose a patient goes to a lab testing outfit, and asks for "everything".  Among many tests, they may very well get a thyroid test.  Suppose the T4 level comes back a bit high, but the TSH is normal.  The patient feels fine, and has no symptoms.  But, with this abnormal value, what if they push for a thyroid ultrasound to see if anything is wrong with the gland?  It may very well pick up a thyroid nodule, which many people have and which may be a cancer but far more likely is nothing.  Now what?  A biopsy, which can show things like benign tissue or atypia of uncertain significance- a polite way of saying it looks weird.  Or, it can show thyroid cancer- most of which are so slow growing that they will never cause anyone any trouble.  What now?  Well, many patients want it taken out- regardless of how unlikely it is to produce any harm.  And of course, thyroid surgery has significant risks- including vocal cord paralysis (you won't be able to talk) and even death.  In Korea, where thyroid cancer screening went through the roof, there has been a 15 fold increase in the rate of diagnosis of thyroid cancer, but no decrease in mortality from thyroid cancer- meaning surgeons have been overwhelmingly taking out things that never would have bothered anyone.

This is but one example, and problems like these are why the US Preventative Services Task Force recommends very few "routine" blood tests.  In opposition to this, Theranos is trying to have consumers (previously called patients) order and look at their own blood tests- and Lab Corp and Quest Diagnostics (Theranos' competitors) are doing the exact same thing.  Consumer empowerment and caring about one's health is all well and good.  But as shown above in numerous examples, more medical care isn't necessarily a good thing if you're not actually sick.

Ultimately, I must include a caveat here: it is possible that by tracking people's lab values over the course of decades, we will find something revolutionary which will let us predict rare cancers and intervene before it's too late, or something else.  But that is many, many years and perhaps decades off- and will probably never even benefit anyone who buys a blood test now.

We often simply do not have the capability to effectively interpret and use many of the tests that a consumer buys on their own with no medical indication.  And any attempt by Theranos or their competitors to argue that consumers routinely buying labs is a good idea is at best speculative and at worst intentionally misleading.  I am not particularly a fan of the old paternalistic model of medicine, with doctors controlling the flow of all information.  But Theranos and its competitors are trying to replace a system which had its problems with a quite possibly worse one; where companies will have free reign to sell unnecessary tests (and perhaps even devices and drugs) to patients.