One of the key issues in the dust-up over the FDA’s insistence on regulating 23andMe’s service is the question of how 23andMe’s health-risk results differ from other forms of health-risk information. Today, geneticist Joe Pickrell offers a sharp post that unpacks this a bit. He asks Should the FDA regulate the interpretation of traditional epidemiology? It’s a damned good question.
Many online sites, Pickrell notes, including government sites, offer online “risk calculators” where you enter things like your age, weight, and small bits of medical history and then get a readout of your risk of things like heart attack, stroke, or cancer. These are precisely the sort of things that 23andMe offers assessments of, to the FDA’s alarm. It took me just 15 seconds, for instance, to learn I supposedly stand a 6.3% chance of having a stroke in the next 10 years. It took me another 15 seconds, pretending I was a woman of my demographic profile and with a family-health history that my siblings shere, to be told by the National Cancer Institute that I have a 3.5% risk of getting breast cancer over the next 5 years and a 22.4% lifetime risk.
Okay. A 1-in-5 chance of breast cancer. Worrisome. I’m done having kids and not really using these for anything anymore. Maybe I should get a mastectomy?
This is precisely the nightmare scenario that the FDA raises in its stern letter to 23andMe. It fears 23andMe’s risk assessment will raise alarms that lead people to radical action. But as far as I can see here, the only real difference here is that 23andMe’s assessment is based on a few snippets of genetic information, while the Nattional Cancer Institute’s assessment is based on a few snippets — and I mean a very few — of demographic, health, and family history information. Yet one is regulated because it uses a machine that parses spit, and the other is not, because it just asks the consumer some questions and then plugs those remembered or improvised responses into its risk formulas.
I fully recognize that for the FDA, which is apparently charged with regulating any physical device that takes medically relevant materials from your body, is obligated to look more closely at a DNA spit test than an online risk calculator. I’m not blaming them for doing this job. But from a wider, not-strictly-legal standpoint, one nonlegal, this seems in many ways an arbitrary distinction. The spit test is just a way of obtaining what amounts to demographic, health, and family information that happens to come from chemicals in your spit instead of from (unreliable) memories in your brain. Yet the law (and many people) insist this is fundamentally different. Am I missing something here? Sincere question.
Consider this hypothetical: Let’s say 23andMe wants to deliver its service but avoid the whole device-plus-evidence-based-medical-risk-information-equals-medical-advice dilemma that is bringing the company such grief. It wants to lose what the FDA seems to see as thing requiring regulation: a fairly direct connection between a physical device that analyzes spit and spit-derived risk information that 23andMe then gives the customer.
So the company sets the service up differently: It genotypes your spit, but instead of calculating and listing the health risks automatically, its sends you a sheet of paper or a file with the raw genotype data in it. (You can already download that raw data, which is all Gs, Cs, Ts, and As.) Then they direct you to an online form, much like ones where I just learned my stroke and hypothetical breast-cancer risk, and allows you to enter in those forms specific lines of the raw data to learn, say, your statistical risk of breast cancer. They can separate them further by charging a separate fee — say, $5 — for using the form.
Would that be okay? Do the service remain a medical device, or does it become the more benign thing represented by a information-distributing educational website like the National Cancer Institute’s? You can easily argue it makes them more like the NCI. So would the FDA need to regulate them? If so, how can they regulate them — but not some service that uses an online form to ask you for bits of remembered or improvised information?
I hope these are recognized as rhetorical questions. I offer them because I suspect that to large extent this 23andMe problem is not about medical diagnoses. It’s a problem about how to update regulatory policy and practice in the face of newly available data that is of a sort that was once viewed as the medical profession’s, but is now rightly being viewed as the individual’s.