Over the last few months, the lab-testing startup Theranos has endured a relentless stretch of bad publicity, morphing from a health tech darling to a Silicon Valley cautionary tale. It started last October, when The Wall Street Journal published an investigation of the company’s much-hyped technology, which promised it could run a host of lab tests (including those for diabetes, high cholesterol and prostate cancer) from just a few drops of blood taken from a finger. Its proprietary technology and lower prices, the company said, would democratize blood testing and encourage its use for routine health monitoring. But the Journal reported that the proprietary device wasn’t all it was cracked up to be, and that the company was running most of its tests on conventional lab machines. Since then there has been news of dissolved business partnerships, problems with regulators, criminal and civil probes and just this week, personnel changes. Founder and CEO Elizabeth Holmes, the Stanford dropout who was the subject of much positive media coverage before the WSJ’s report, could be banned from the industry for two years.
Much of the debate over the last few months has been about whether Theranos’s technology can actually do what the company has said it can. But there’s a dubious assumption at the heart of Theranos, arguably just as damning as the questions about its technology. Theranos wants us to believe that ostensibly healthy people can get healthier by having more tests. But the science suggests that is far from true.
To see why, let’s go deep into Holmes’s pitch for why Theranos was so important. Let’s use her 2014 TEDMED talk, which has been viewed more than 380,000 times, as a proxy for her elevator pitch:
We define diagnosis today as the determination of the presence of disease from its signs and symptoms. Yet diseases often begin so much earlier than when symptoms first appear. We see a world in which every person has access to actionable health information at the time it matters. A world in which no one ever has to say, “If only I’d known sooner.”
It seems obvious: Finding a disease as early as possible would lead to fewer deaths than finding it late. But as more and more research is showing, it’s not as simple as going exploring in healthy people to see what turns up, then treating what you find. “The reservoir of abnormalities is bottomless,” said Dr. Brenda Sirovich, co-director of the Outcomes Group at the VA Medical Center in White River Junction, Vt. “And we don’t know anything about what those things mean in an asymptomatic vacuum.”
For example, levels of prostate-specific antigen in the blood are elevated in men with prostate cancer, so a logical leap suggests that we might be able to detect nascent disease by looking for it in seemingly healthy men. But the PSA test (available from Theranos on its own or as part of a set of routine men’s health tests) is questionable as a way to screen for prostate cancer. Screening with PSA provides at most a small benefit in preventing mortality from prostate cancer, and some studies suggest it harms more men than it helps through the side effects of treatment, which can include impotence and incontinence.1
The effectiveness of screening for a given disease before signs and symptoms appear depends on a host of conditions, said Dr. John Ioannidis, director of the Stanford Prevention Research Center — including whether there’s an effective treatment for whatever is found, and whether people will fare better if they get treatment without waiting for symptoms. He added that a successful screening test needs to be accurate, without too many false positives or false negatives.
But depending on the prevalence of a disease or condition in a given population, false positives can outnumber true positives, even for a test that’s quite accurate.
The wider the pool of people being tested, the greater the chance of false positives, which is why screening guidelines generally limit the population to be screened. The more independent tests you do at once, each with its own chance of error, the larger the chance that at least one of those tests produces an incorrect result, said Rebecca Goldin, director of STATS.org and a professor of mathematical sciences at George Mason University.
Screening may also find disease that would never have caused a problem if it had been left alone — what researchers call “overdiagnosis.”2 “There are a zillion things that are a disaster to screen for,” said Ioannidis. Last year, he was an author of a study that reviewed the evidence behind screening for 19 potentially deadly diseases3 and found it was “uncommon” for targeted screenings to lead to reductions in death from the disease being screened.
Holmes’s talk also advocates that individuals should have more power to order blood tests on their own:
And today I can go buy a deadly exotic animal, a venomous viper, a military truck or armored vehicle … but I can’t order a blood-based pregnancy test or an allergy test because that could be dangerous. God forbid I stop eating peanuts.
She elaborates on this point in a 2014 New Yorker profile, noting that Theranos’s direct-to-consumer vision would require changes to state laws and regulations, some of which prohibit blood tests without a prescription.
Holmes doesn’t get into details about the full range of tests that the company would like to see available without a prescription or expert guidance. (Theranos didn’t respond to emails requesting comment.) Other companies are also pursuing this direct-to-consumer vision. But there are reasons people generally can’t order tests themselves. For one thing, some screening tests merit a discussion ahead of time about whether the potential benefits outweigh the potential harms.4
Tests are also hard for consumers to interpret. Few lab results are inherently binary. “Very few blood tests give you a ‘Yes, you have this condition,’ or ‘No, you don’t,’” said Dr. Kenny Lin, an associate professor of family medicine at Georgetown University Medical Center. Context — symptoms, personal and family medical history — is essential to interpret those results. Dr. Eleftherios Diamandis, division head of clinical biochemistry at Mount Sinai Hospital in Toronto, wrote in a 2015 opinion paper about Theranos that a layperson whose PSA value was 20 micrograms per liter “will assume, based on statistics, that he would have a more than 50 percent chance of harboring prostate cancer; and ask for a biopsy.” But if his PSA was 1 microgram per liter a few days earlier, he likely has benign prostatitis, not cancer.
Even something as seemingly straightforward as the allergy test Holmes mentions is fraught. A 2015 study found blood-based allergy tests in kids usually resulted in misdiagnosis and overly restrictive diets. There are real consequences to telling people they are allergic to things they aren’t.
One other idea is central to Holmes’s talk: that giving people more information about their risk of disease will cause them to make meaningful and beneficial changes.
When individuals have access to the information about their bodies, they can begin to change outcomes. Type 2 diabetes alone, which drives 20 percent of our health care costs, can be reversed through changes in lifestyle, in diet, in exercise. Yet today there’s 80 million Americans who are prediabetic and 90 percent of them don’t know that they are.
Certain people probably are motivated to make positive lifestyle changes merely by getting more comprehensive information about their health. But in the aggregate, data alone doesn’t seem to fuel better decisions. For example, almost 35 percent of American adults are obese, which is likely quite apparent to them, yet many have great difficulty losing weight. A review of studies published in March concluded that communicating genetic risk for diseases does not change people’s health behaviors, including smoking, diet, exercise, alcohol use and sun protection. And a study published last year found that screening for Type 2 diabetes didn’t change total physical activity, smoking habits or alcohol consumption.
As for prediabetes, while there is an awareness campaign to identify those who have it (about 86 million, according to the Centers for Disease Control and Prevention), there’s little evidence that that doing so will change outcomes. And identifying someone as “sick” can change self-image and cause stress, even though most people with prediabetes will not actually develop diabetes, said Gabriela Spencer Bonilla, who conducts health services research at the Mayo Clinic. It will also funnel millions more people into lifestyle management programs that are already strained by the needs of people with full-blown diabetes.
Not all of Theranos’s vision is misguided. Making lab tests cheaper, expanding access to tests used for diagnosis, and reducing the pain of traditional blood tests are “clearly worthy goals,” Ioannidis wrote in a critical look at Theranos published last year. But the vision of giving everyone widespread access to more blood tests in the belief that more information is always better is naïve at best, and harmful at worst.