Many physicians are terrible at interpreting results of screening tests:

http://www.bmj.com/content/333/7562/284

For screening tests where the problem being screened for is rare, most results will be false positives, but the result is not intuitive.

Doctors being lousy Bayesians is (if anything) an argument in favor of services like 23andme!

Right. IF 23 And Me can show that its test are accurate. But they can't (or have refused to do so). That's the whole point.

23 And Me should only have to prove that it is better than the alternative, not that it is accurate (to determine whether the service is of net benefit to the public). Depending on the counter-factual, 23AM could provide a relatively inaccurate test that would still be better than no test.

Unfortunately for us, the FDA is on a quest for "safety"; the agency does not try to help the citizenry live longer or happier.

Safety isn't part of a long and happy life?

> 23 And Me should only have to prove that it is better than the alternative,

23andMe isn't even doing that.

You seem to believe the average doctor is well-versed in the mechanics of genetic testing, or (perhaps equivalently) that the average doctor isn't wholly dependent on the mechanisms of FDA approval for product selection.

Any doctor educated in the past two decades is well versed in genetic testing, and any doctor specializing in areas where genetic testing is a frequently used aspect of planning/diagnosis (such as OB/GYN) will be highly familiar even if trained long ago.

The 23andme product explains explicitly how it should be used by a physician, and there are all kinds of warnings and checkboxes before the user is shown any data that falls into the area of overlap.

Perhaps any one of several trained medical professionals on HN could chime in here to explain (a) how suitable 23AM's current product is for clinical use and (b) how competent a typical physician is at evaluating clinical genomics products without FDA's assistance. 'carbocation? 'niels_olson?

('carbocation has already expressed some tacit support for FDA's action, but I'm not qualified to interpret it in the context of this particular subthread).

> The 23andme product explains explicitly how it should be used by a physician

Generally, for diagnostic tools, drugs, etc., it is not enough that the product labelling explains how it should be used, but it is required to demonstrate to the FDA that there is evidence that it works as marketed when used according to those directions. The FDA Warning Letter specifically calls out 23andMe's failure to either identify existing approved products on the market to which the product is equivalent (which wouldn't reduce the requirements for proving this) or to meet the requirements for demonstrating the effectiveness of their offering as a new product without existing, approved equivalents.

How do you think 23andme should answer that question?

There are at least three different products in the space and to claim similarity would be to likely invite over-regulation:

- "full scan" companies, that do a full genetic scan and do not limit the scan to SNPs

- personal genomics companies that focus on SNPs

- medial diagnostics like the MSAFP test

> How do you think 23andme should answer that question?

I don't know enough about 23andMe's operations and those of what appear from the outside to be generally similar services to answer that. Assuming that they really are an innovative service, they would need to conduct and provide the studies supporting the validity of their innovative service.

  Any doctor educated in the past two decades
  is well versed in genetic testing

I'm not sure what definition of "well versed in genetic testing" you had in mind but this is untrue for most or all reasonable definitions.

  any doctor specializing in areas where genetic testing
  is a frequently used aspect of planning/diagnosis
  (such as OB/GYN) will be highly familiar even if trained long ago.

Also an overstatement. The trouble is that "genetic testing" is a large and complex field. Are you referring to a single-gene test for a well-known and well-characterized disease? Or whole exome sequencing? Because in the coming years doctors will come into contact with both. In general these tests will require specialist interpretation; not general physicians or OB/GYNs.

Regardless of whether physicians are skilled or not at helping their patients interpret the results of a proprietary test, the point I was making is that if the test produces an unacceptably high number of false positives then this would result in an unnecessary increase in the number of times physicians have to actually have those conversations.

If physicians are forced to waste huge amounts of time and effort to console an influx of patients who have been wrongly told they have some terrible disease, this is going to reduce the amount of time they have available to work with patients with actual terrible diseases.

If other medical device/diagnostic makers have to comply with the law, so should 23andme.

What if it results in some accurate positives and saved lives? You think the FDA should limit access to information curious people can obtain for $100 about their own genome that might help them behave proactively toward their own health just in case it results in more patients behaving proactively?

I don't know a single person who thinks Elvis is alive...