Sunday, December 01, 2013

FDA Ordered Genetic Testing Co. 23andMe To Stop Selling Home Test Kit

But 23andMe has continued to sell its $99 "Saliva Collection Kit and Personal Genome Service (PGS)" in the face of this order:

Here's an excerpt from FDA's Warning Letter to 23andMe:

FDA Warning Letter to 23andMe, Inc. 11/22/13
"However, even after these many interactions with 23andMe, we still do not have any assurance that the firm has analytically or clinically validated the PGS for its intended uses, which have expanded from the uses that the firm identified in its submissions. In your letter dated January 9, 2013, you stated that the firm is “completing the additional analytical and clinical validations for the tests that have been submitted” and is “planning extensive labeling studies that will take several months to complete.” Thus, months after you submitted your 510(k)s and more than 5 years after you began marketing, you still had not completed some of the studies and had not even started other studies necessary to support a marketing submission for the PGS.

It is now eleven months later, and you have yet to provide FDA with any new information about these tests. You have not worked with us toward de novo classification, did not provide the additional information we requested necessary to complete review of your 510(k)s, and FDA has not received any communication from 23andMe since May. Instead, we have become aware that you have initiated new marketing campaigns, including television commercials that, together with an increasing list of indications, show that you plan to expand the PGS’s uses and consumer base without obtaining marketing authorization from FDA.

Therefore, 23andMe must immediately discontinue marketing the PGS until such time as it receives FDA marketing authorization for the device."
They have until December 6 to respond or:
"Failure to take adequate corrective action may result in regulatory action being initiated by the Food and Drug Administration without further notice. These actions include, but are not limited to, seizure, injunction, and civil money penalties." 
Lots of questions here.  If 23andMe's genome service can, as they say, determine risk for disease:

The disease risks produced by 23andMe's test are based on correlations.  Many people claim that science based on correlation is junk science.  I don't think epidemiology is junk science.  As I've said often, epidemiology generates hypotheses which provide the foundation for clinical research. The question is ... how actionable is science based on correlation?

Not only are the results based on correlations, but the population used for comparison is limited ... those people who had the resources to have their genome analyzed, and who voluntarily offered their results for study.  This is not a broad cross-section of the world's population.  It's an elite group, characteristics of whom remain proprietary.

The above addresses external validity, that is, how generalizable results are to other people, other populations.  But there is also a problem with internal validity ... how accurate is the test?  The consumer spits in a tube and mails it back.  Can you think of an instance where the saliva in one's mouth might not contain only the DNA of the person spitting?  Can the sample degrade on its trip through our postal system?  The FDA also has a problem with its validity: "we still do not have any assurance that the firm has analytically or clinically validated the PGS for its intended uses."

Genes do not predict risk in a vacuum.  They operate in the real world.  They interact with what we eat, what drugs we take, what environmental chemicals and radiation we are exposed to, how much or how little sleep and stress and exercise we experience.  Love, laughter, grief, loss ... so many things affect the relationship between genes and expression of genes.  Assigning risk based on genes alone is fraught.  Bernard Munos at Forbes said as much:

23andMe: A Fumbling Gene In Its Corporate DNA?
"Traditional clinical research faces severe limitations and is ill-suited to help us inderstand the relationship between genotype and phenotype. ... Many flawed genes can produce the same phenotype, and disentangling what causes what is a challenge of staggering complexity."
Finally, how private are the results?  Insurance companies and employers have discriminated based on genome reports, even though they're not supposed to (GINA).  As we now know, electronic communication is monitored in this country ... even when the communication is encrypted, even when there are laws (HIPAA) protecting medical privacy.

So many questions.


AB said...

You ask if people will seek testing for diseases they don't have. They already do. The US has primed its citizens to get radiated, scoped and tested for everything under the sun. Of course they will continue to do so. They love to run to the doctors. But at least with genetics, they might not be running for every cancer under the sun, maybe we can reduce it to the ones they are high risk for. Even better would be educate the populace on real prevention, which is, e.g. don't eat red meat if you are high risk for colorectal cancer.

You ask if an extra 25 billion will improve health care. It can, but it depends how it is used. Genetics are a road map, and if we can stop an endless array of middlemen from draining off their slice of the pie, there's a chance some real pathways live here.

Bix said...

I agree with a lot you said, especially "The US has primed its citizens to get radiated, scoped and tested for everything under the sun." There's a lot of money to be made there, and I'm sorry to say a lot of it is being made at the consumers' expense. The expense of their health as well as their wallet.

If you are going to market a tool for risk assessment, validate it. This tool has not been validated for all of its uses. The FDA should allow it to be marketed for uses that can be validated, such as eye color. It should not be sold to assess risk for, say, a chronic disease like diabetes unless and until the risk value is substantiated.

For example ... At one time, a person's LDL cholesterol was thought to be a good risk assessment tool for predicting heart attack. This is no longer the case, evident from the AHA's and ACC's recent report. (More than half of hospital admissions for heart disease have LDL less than 100 mg/dl, considered "low risk." So ... LDL levels to assess risk were not sufficiently validated before they were acted upon ... before, e.g., they were used as justification for statin prescriptions.

A genome test should be sufficiently validated for its use in assessing risk (for each and every of the 100s of diseases 23andMe says it's useful for!) before it is used as justification for any additional test, any surgery, any further expense.