6 ms·

Really? You don't see a problem with these companies selling tests to people and then dumping the cost of confirming the result and dealing with the implication

by Gatsky 8y ago

Really? You don't see a problem with these companies selling tests to people and then dumping the cost of confirming the result and dealing with the implications on the already strained healthcare system? Meanwhile 23andMe makes billions selling the data to pharma?

gwern 8y ago
I don't because 23andMe didn't flag it in the first place, disconfirming it was an order of magnitude cheaper than his interactions with the supposed adults turned out to be and didn't involve any catch-22s or bullshit reasoning (a geneticist should know that family histories can be wrong and de novo mutations exist, and anyone with 2 brain cells to wire together should realize the incredible value of learning you have early-onset AD rather than paternalistically dismiss it as useless data), and the DTC companies are approaching 10 million customers with few major issues - not that this could be called a major issue in the first place. > implications on the already strained healthcare system? The healthcare system didn't do squat. That's kind of the problem here. In fact, if we're going to discuss the system as a whole, I would point out that not sequencing everyone is immensely costly and a major failing of the healthcare system, as testing would pay for itself just in terms of better dosing of drugs like warfarin. There shouldn't even be a question of learning whether you have early-onset AD, it should already be known. > Meanwhile 23andMe makes billions selling the data to pharma? They don't. The deals are in the low millions as far as is publicly known (and if you're referring to GSK's recent investment in 23andMe, that wasn't a sale). And if they are, it's not clear whether that makes 23andMe even net profitable (they're at >$1b in total VC), and even if they have revenue, that doesn't mean their investors have made 'billions', and finally, if they did, good for them, because that means they created value by subsidizing DTC testing for almost a decade, cut through the regulation, and created a large useful database on par with the UK Biobank, which wouldn't've happened otherwise. (If you don't believe me, ask someone how useful the Million Veteran Project has been compared to 23andMe or UKBB; or just do a site search on biorxiv.org...)
- Gatsky 8y ago
  You haven't stated the population risk of having a pathogenic de novo PSEN1 mutation. This would be an essential preface to whatever argument you are trying to make about the case described in the parent article. 'few major issues' is a liberal interpretation of being shut down by the FDA. Besides, nobody is measuring the impact of the test, so how do you know there are 'few major issues'? > not sequencing everyone is immensely costly and a major failing of the healthcare system I do cancer genomics in my day job. I would not agree with this comment on any day of the week at the present time. The problem with 23andMe and their ilk is it just isn't that useful. Warfarin use is being phased out anyway, because of the many non-genetic causes of variability and the availability of drugs with more reliable pharmacokinetics. This same principle applies to the vast majority of serious medical conditions - lifestyle factors greatly outweigh any genetic component. But the biggest problem with acting like 23andMe is a useful test is they don't actually do a proper test for the most common genetic condition that leads to a serious condition with high penetrance - BRCA1 and BRCA2 mutations and breast and ovarian cancer. They don't do a proper damn test, they just test 3 out of the thousands of variants in the gene. Why do you think that is? Why don't they do a proper version of the most useful and important test, which can actually be used to prevent the patient getting cancer? The related question is - who benefits from 23andMe testing? The company or the 'Me'? Re the 'billions', which of the following is more likely: - VCs have given 23andMe >$1b in total because they think it can make multiple billions by selling the data - Any other reason
  - gwern 8y ago
    > You haven't stated the population risk of having a pathogenic de novo PSEN1 mutation. The population prevalence of any early onset AD is, and always will be, extremely low because of its pathogenicity, just like any other gene which is terminal with ~100% penetrance. The prior probability will be low, even without checking, and the posterior won't be large either for the usual Bayesian reasons. I shouldn't need to state this about the population risk, it should be obvious, especially to a cancer genomicist. Unless you want to argue that the mutation in question is very common...? > 'few major issues' is a liberal interpretation of being shut down by the FDA. Being nearly shut down by the FDA is no one's problem but 23andMe's. I was actually referring to unwanted discoveries about family secrets, which are the major downside of 23andMe testing. (Infidelity is a good deal more common than early-onset Alzheimers.) > Warfarin use is being phased out anyway, because of the many non-genetic causes of variability and the availability of drugs with more reliable pharmacokinetics. This same principle applies to the vast majority of serious medical conditions - lifestyle factors greatly outweigh any genetic component. 'Warfarin is being phased out' is another way of saying 'warfarin is still used', and the uses go well beyond just warfarin (https://www.biorxiv.org/content/early/2016/07/23/065540 https://www.biorxiv.org/content/early/2016/07/23/065540 covers a number of other drugs, 2 years ago), the uses only increase with time, and of course there's a chicken-and-egg thing here: drug response can't be predicted without large datasets, and large-scale sequencing supply will induce its own demand. (Just like how the UK Biobank's results induced even more interest in genetics.) Many other things can be usefully predicted and PGSes are increasingly of clinical utility (https://www.gwern.net/docs/genetics/heritable/2018-khera.pdf https://www.gwern.net/docs/genetics/heritable/2018-khera.pdf https://www.gwern.net/docs/genetics/selection/2018-torkamani.pdf https://www.gwern.net/docs/genetics/selection/2018-torkamani...) and they add onto 'lifestyle factors' prediction, considerably, there is no reason to use only one. > They don't do a proper damn test, they just test 3 out of the thousands of variants in the gene. Why do you think that is? Presumably because validating the other thousands of variants to the degree it took to satisfy the FDA for the 3 variants they do look at would be extremely expensive, and I suspect there may be other reasons as well (why would their Illumina chips look for those variants in the first place given that they are rare and they can't use them?). And knowing isn't that useful. DeCODE knows the Icelanders with BRCAs, but isn't allowed to tell them: http://pulitzercenter.org/reporting/right-not-know-when-ignorance-bliss-deadly http://pulitzercenter.org/reporting/right-not-know-when-igno... > Re the 'billions', which of the following is more likely: You're backpeddling furiously. First it was '23andMe made billions' off poor helpless customers, now it's merely they might make billions. So now you're damning them for doing what they haven't even done yet. Wow. It's not a negative-sum game, you know.

by Gatsky 8y ago

gwern 8y ago

I don't because 23andMe didn't flag it in the first place, disconfirming it was an order of magnitude cheaper than his interactions with the supposed adults turned out to be and didn't involve any catch-22s or bullshit reasoning (a geneticist should know that family histories can be wrong and de novo mutations exist, and anyone with 2 brain cells to wire together should realize the incredible value of learning you have early-onset AD rather than paternalistically dismiss it as useless data), and the DTC companies are approaching 10 million customers with few major issues - not that this could be called a major issue in the first place. > implications on the already strained healthcare system? The healthcare system didn't do squat. That's kind of the problem here. In fact, if we're going to discuss the system as a whole, I would point out that not sequencing everyone is immensely costly and a major failing of the healthcare system, as testing would pay for itself just in terms of better dosing of drugs like warfarin. There shouldn't even be a question of learning whether you have early-onset AD, it should already be known. > Meanwhile 23andMe makes billions selling the data to pharma? They don't. The deals are in the low millions as far as is publicly known (and if you're referring to GSK's recent investment in 23andMe, that wasn't a sale). And if they are, it's not clear whether that makes 23andMe even net profitable (they're at >$1b in total VC), and even if they have revenue, that doesn't mean their investors have made 'billions', and finally, if they did, good for them, because that means they created value by subsidizing DTC testing for almost a decade, cut through the regulation, and created a large useful database on par with the UK Biobank, which wouldn't've happened otherwise. (If you don't believe me, ask someone how useful the Million Veteran Project has been compared to 23andMe or UKBB; or just do a site search on biorxiv.org...)

Gatsky 8y ago
You haven't stated the population risk of having a pathogenic de novo PSEN1 mutation. This would be an essential preface to whatever argument you are trying to make about the case described in the parent article. 'few major issues' is a liberal interpretation of being shut down by the FDA. Besides, nobody is measuring the impact of the test, so how do you know there are 'few major issues'? > not sequencing everyone is immensely costly and a major failing of the healthcare system I do cancer genomics in my day job. I would not agree with this comment on any day of the week at the present time. The problem with 23andMe and their ilk is it just isn't that useful. Warfarin use is being phased out anyway, because of the many non-genetic causes of variability and the availability of drugs with more reliable pharmacokinetics. This same principle applies to the vast majority of serious medical conditions - lifestyle factors greatly outweigh any genetic component. But the biggest problem with acting like 23andMe is a useful test is they don't actually do a proper test for the most common genetic condition that leads to a serious condition with high penetrance - BRCA1 and BRCA2 mutations and breast and ovarian cancer. They don't do a proper damn test, they just test 3 out of the thousands of variants in the gene. Why do you think that is? Why don't they do a proper version of the most useful and important test, which can actually be used to prevent the patient getting cancer? The related question is - who benefits from 23andMe testing? The company or the 'Me'? Re the 'billions', which of the following is more likely: - VCs have given 23andMe >$1b in total because they think it can make multiple billions by selling the data - Any other reason
- gwern 8y ago
  > You haven't stated the population risk of having a pathogenic de novo PSEN1 mutation. The population prevalence of any early onset AD is, and always will be, extremely low because of its pathogenicity, just like any other gene which is terminal with ~100% penetrance. The prior probability will be low, even without checking, and the posterior won't be large either for the usual Bayesian reasons. I shouldn't need to state this about the population risk, it should be obvious, especially to a cancer genomicist. Unless you want to argue that the mutation in question is very common...? > 'few major issues' is a liberal interpretation of being shut down by the FDA. Being nearly shut down by the FDA is no one's problem but 23andMe's. I was actually referring to unwanted discoveries about family secrets, which are the major downside of 23andMe testing. (Infidelity is a good deal more common than early-onset Alzheimers.) > Warfarin use is being phased out anyway, because of the many non-genetic causes of variability and the availability of drugs with more reliable pharmacokinetics. This same principle applies to the vast majority of serious medical conditions - lifestyle factors greatly outweigh any genetic component. 'Warfarin is being phased out' is another way of saying 'warfarin is still used', and the uses go well beyond just warfarin (https://www.biorxiv.org/content/early/2016/07/23/065540 https://www.biorxiv.org/content/early/2016/07/23/065540 covers a number of other drugs, 2 years ago), the uses only increase with time, and of course there's a chicken-and-egg thing here: drug response can't be predicted without large datasets, and large-scale sequencing supply will induce its own demand. (Just like how the UK Biobank's results induced even more interest in genetics.) Many other things can be usefully predicted and PGSes are increasingly of clinical utility (https://www.gwern.net/docs/genetics/heritable/2018-khera.pdf https://www.gwern.net/docs/genetics/heritable/2018-khera.pdf https://www.gwern.net/docs/genetics/selection/2018-torkamani.pdf https://www.gwern.net/docs/genetics/selection/2018-torkamani...) and they add onto 'lifestyle factors' prediction, considerably, there is no reason to use only one. > They don't do a proper damn test, they just test 3 out of the thousands of variants in the gene. Why do you think that is? Presumably because validating the other thousands of variants to the degree it took to satisfy the FDA for the 3 variants they do look at would be extremely expensive, and I suspect there may be other reasons as well (why would their Illumina chips look for those variants in the first place given that they are rare and they can't use them?). And knowing isn't that useful. DeCODE knows the Icelanders with BRCAs, but isn't allowed to tell them: http://pulitzercenter.org/reporting/right-not-know-when-ignorance-bliss-deadly http://pulitzercenter.org/reporting/right-not-know-when-igno... > Re the 'billions', which of the following is more likely: You're backpeddling furiously. First it was '23andMe made billions' off poor helpless customers, now it's merely they might make billions. So now you're damning them for doing what they haven't even done yet. Wow. It's not a negative-sum game, you know.