Is 23andMe in Trouble? Should They Be?

The direct-to-consumer (DTC) gene testing company 23andMe seems to be in a strange position. The Food and Drug Administration (FDA) is gradually loosening the legal limits it placed on the company’s operation, and the industry as a whole does seem still to be growing.

However, 23andMe, along with other direct-to-consumer gene testing companies, faces at least two other problems: increasing criticism of the value of their tests; and the rapid approach of whole-genome testing, which could make their technology obselete.

All that might explain the decision of 23andMe founder and CEO Ann Wojcicki to write a piece for STAT boosting her product.

The headline was “Consumers don’t need experts to interpret 23andMe genetic risk reports,” which was the essential message of the essay. She opened and closed it with a comparison to the introduction of the first at-home pregnancy tests, which “some physicians” warned against, on the grounds that women couldn’t handle the information. In between she focused on breast cancer and the company’s new tests for (as she said) “three variants among the thousands of mutations that contribute to [breast] cancer risk,” which are “found mostly in people of Ashkenazi Jewish ancestry.”

There are a number of flaws in the piece. For a start, the comparison with pregnancy tests is not good. When they were first introduced to the U.S. in the early 1970s, some doctors did indeed complain, perhaps because of a breach of their monopoly interest. But when one wrote in complaint to the American Journal of Public Health, in November 1976, the editor appended a short note, which memorably concluded, “Not everyone needs carpenters to hammer in their nails.” Most people soon agreed.

In contrast, 23andMe launched the direct-to-consumer gene testing business 12 years ago, and it is still very controversial among academics, physicians, and public interest advocates in health fields. Breast Cancer Action, for example, examined the FDA’s decision to allow 23andMe to offer results about the three genetic variants associated with breast cancer, and concluded that the test “may do more harm than good.”

Tellingly, Wojcicki’s article undermined her own argument for bypassing expertise. Though she presumably had access to professional editorial guidance both at her own company and at STAT, she messed up a statistic related to breast cancer, confusing total lifetime risk with increased lifetime risk. Why does that error matter? As Hank Greely pointed out, “if she, casually, makes a mistake in relative risk v absolute risk (one that cuts against her own argument), how often are her customers going to make that mistake, or others?”

Wojcicki’s article overall comes across as an infomercial. Yes, it is labeled as “Opinion” and the author clearly says, in the second paragraph, “My company.” But it reads like a sales pitch. For instance, it ignores or downplays the facts that: only about 5–10% of all breast cancers are linked to gene mutations; those that are genetic are not necessarily linked to the BRCA1 or BRCA2 variants; and women with those variants have risks worth watching but may well never develop symptoms.

But many readers would find it easy to gloss over phrases such as “could mean” and focus on the much more dramatic “potentially lifesaving.” This is also exactly the problem with the company’s presentation of test data: The caveats are included, but it’s only human to jump to the “conclusion” — which is an apparently precise statement of the consumer’s risk level.

More than that: A similar murkiness is inherent to the company’s business model. In all of 23andMe’s ubiquitous ads, the company is presented as one that sells spit kits and delivers genetic information to consumers. In fact, its bottom line relies on aggregating the data those tests provide and then re-selling it to pharmaceutical companies. This, too, is (now) mentioned on the company’s website; but many people are surprised to hear it.

The scientific context for Wojcicki’s article is challenging for the company. Back in January, Carl Zimmer wrote a good piece in the New York Times (based on research published in Science) about the difficulty of predicting outcomes based solely on genes. Indeed, many phenotypic variants now seem to be not just polygenic but omnigenic — completely interconnected. The week before the article appeared, HuffPost published a substantial article about error rates in DTC tests.

Writing in Gizmodo about Wojcicki’s comparison of pregnancy and gene tests, Kristen V. Brown asserts that “it’s not that simple.” There are good reasons to encourage patient participation in their own healthcare, and  some indications that “in general, people acted rationally in response to their results: If there was something concerning, they followed up with a doctor.” (Of course, this does not apply to false reports that the customer is clear.) Brown sensibly concludes that “we need a lot more research.”

Finally, in February, Smithsonian magazine considered at length, without resolving it, the question of whether parents should genetically test their child. In closing, they quote Timothy Caulfield, research director of the University of Alberta’s Health Law Institute, who thinks the science is not yet ready:

“The reality is that if you have a generally healthy kid, there aren’t a lot of specific interventions that you can do. For most risks, it comes back to the same things: Exercise. Eat well. Wear sunblock. Get a good night’s sleep. You don’t need genetic tests to tell you that.”

What kind of business model would that be?