8 ms·

I think the line of reasoning makes sense: people will make medical decisions based on this information, but 23andme has never given us any reason to (or not to

I think the line of reasoning makes sense: people will make medical decisions based on this information, but 23andme has never given us any reason to (or not to) believe their tests are accurate. Without someone reviewing their process, their error rate is unknown. That can be dangerous. This is similar to the problem with many alternative medicine claims. 'Alternative medicine' may be very helpful to an individual, but it's important not to overstate its effectiveness or reliability.

dnautics 13y ago
how would you define 'error rate'? Do you mean "lab handling error rate"? Or "information presenting" error rate? The way 23andMe works, is they run your DNA and they show "percentages" based on already done studies. Then they throw in caveats (like, if you're not of the correct race relative to what the study was done in). So what would you suggest? That the service track all of their users over a long-term and then see whether the percentage of their users that get condition X matches the percentage they predicted? And then do some sort of statistical analysis on a possibly moving number (because 23andMe updates their information if academic science improves) with a population that might turn out to have differing biases...? And then prove this information over what, 100000 SNPs on the chip? What you are basically arguing is that 23andMe should not be allowed to offer its service.
- epistasis 13y ago
  Ultimately both error rates matter, and are measurable. It's not necessary that these error rates are zero, it's necessary that they are measured, controlled, and acknowledged, which is what 23andMe is not doing. Even the FDA isn't arguing that 23andMe should not offer it's service, just that it shouldn't market it. And really, given that they don't seem to be following the basic quality assurance practices of the industry, they should not be marketing at all.
  - dragonwriter 13y ago
    > Even the FDA isn't arguing that 23andMe should not offer it's service, just that it shouldn't market it. They aren't even really arguing that either; they are more arguing that if they are going to market it, then they need to provide studies supporting the accuracy and take other steps to get it approved to be marketed for the particular uses it is being marketed. They are saying in the Warning Letter that 23andMe must stop marketing the product because 23andMe has abandoned the approval process (and apparently was not preventing 23andMe from marketing the product as long as they were working through the approval process), not because the product is inherently unmarketable.

dnautics 13y ago

how would you define 'error rate'? Do you mean "lab handling error rate"? Or "information presenting" error rate? The way 23andMe works, is they run your DNA and they show "percentages" based on already done studies. Then they throw in caveats (like, if you're not of the correct race relative to what the study was done in). So what would you suggest? That the service track all of their users over a long-term and then see whether the percentage of their users that get condition X matches the percentage they predicted? And then do some sort of statistical analysis on a possibly moving number (because 23andMe updates their information if academic science improves) with a population that might turn out to have differing biases...? And then prove this information over what, 100000 SNPs on the chip? What you are basically arguing is that 23andMe should not be allowed to offer its service.

epistasis 13y ago
Ultimately both error rates matter, and are measurable. It's not necessary that these error rates are zero, it's necessary that they are measured, controlled, and acknowledged, which is what 23andMe is not doing. Even the FDA isn't arguing that 23andMe should not offer it's service, just that it shouldn't market it. And really, given that they don't seem to be following the basic quality assurance practices of the industry, they should not be marketing at all.
- dragonwriter 13y ago
  > Even the FDA isn't arguing that 23andMe should not offer it's service, just that it shouldn't market it. They aren't even really arguing that either; they are more arguing that if they are going to market it, then they need to provide studies supporting the accuracy and take other steps to get it approved to be marketed for the particular uses it is being marketed. They are saying in the Warning Letter that 23andMe must stop marketing the product because 23andMe has abandoned the approval process (and apparently was not preventing 23andMe from marketing the product as long as they were working through the approval process), not because the product is inherently unmarketable.