23andMe and the Future of Consumer Genomics

by Reihan Salam

The FDA has ordered 23andMe, a leading consumer genomics start-up, to cease marketing its genetic testing service. Founded in 2007, 23andMe has managed to drive down the price point of its core offering to $99, for which you get a report on 254 diseases and conditions. The FDA contends that 23andMe’s collection method is a medical device, and as such it must be approved by the FDA. Christina Farr of VentureBeat reports on the FDA’s concerns:

Regulators are concerned that a false positive could result in unnecessary surgery, and false negatives could lead to patients being less aggressive about screening for various health risks. The agency is most concerned that 23andMe is testing various mutations of genes that indicate a woman might have a high risk of getting breast or ovarian cancer. The FDA believes a false positive on that test could cause a woman to undergo a needless mastectomy.

The FDA, a government agency that is responsible for food and patient safety, has “serious concerns” if the results of a 23andMe test are not “adequately understood by patients” or if it returns incorrect test results.

This certainly looks like an example of a regulator that is overstepping its bounds. 23andMe’s saliva testing kit is not dangerous, and it offers all many of disclaimers regarding how its findings ought to be used. If the results of a 23andMe test are not “adequately understood,” patients nevetheless need to go through a gatekeeper — a primary care professional of some kind — to undertake a risky medical intervention. One gets the impression that the FDA is now seeking to protect us from hypochonrdia, in which case it will have to do more than regulate harmless saliva testing kits. Perhaps the FDA should redefine internet-enabled devices as medical devices, as these devices are used to access WebMD and Yahoo Answers, where all kinds of information is shared about medical conditions, not all of which is relevant or reliable.

Recently, Razib Khan and David Mittelman weighed in on the future of consumer genomics. They describe 23andMe as part of an important first step for the sector — by using older genomic techniques, 23andMe was able to give its customers fairly detailed information concerning their health and ancestry. But as the market for genetic genealogy matures and as the technology arounds genomics advances, we will start to see direct-to-consumer start-ups that offer far more useful information:

As analysis of ancestry and genealogy saturate over the next decade, it seems likely that a natural segue will be made to medical genomics, as massive sample sizes can finally be leveraged into fine-grained prediction. Genetic genealogy is an applied enterprise of phylogenetics, but adding phenotypic information to the rich implicit pedigrees may finally allow for both true and surprising outcomes in regards to health risk prediction and the general interpretation of genome variants. Eventually the personal information ecology in which we are embedded will be rich, tagged and amenable to being synthesized and reduced down to actionable units. Genetic genealogy complements written records, but its power is such that it can yield information as a standalone tool. Not so when it comes to medical and assorted trait predictions. Most genetic traits of interest are complex, with multiple, sometimes obscure, lines of causality. Modern studies that use hundreds of individuals are notoriously underpowered, and are rapidly being superseded by pooled data sets of hundreds of thousands. Yet this is nothing compared to what will no doubt be the norm in the next decade, as diverse strands of information continuously thread together, and generate a perpetual stream of novel insightful predictions. Imagine that biomarkers are constantly recording your feedback to a particular diet or activity, and this information is intersected with millions with varying levels of genetic relatedness and lifestyle. The computational requirements are immense, but the past few decades have shown us that we can always be surprised how far Moore’s Law can take us.

Getting to this point, however, will require that norms surrounding privacy will have to change, a development that Khan and Mittelman believe will come to pass as Google and Facebook nudge the public towards accepting greater openness and as universal coverage obviates concerns about uninsurability, one of the chief anxieties (though certainly not the only one) behind the desire for medical privacy.