4 ms·

Here's my view on this: This doesn't test the full gene sequence, only a few SNPs. Say someone does the test, and it comes back positive. They then call all th

by Gatsky 9y ago

Here's my view on this: This doesn't test the full gene sequence, only a few SNPs. Say someone does the test, and it comes back positive. They then call all their relatives. Those relatives have the choice of getting the same test done with 23andMe, or seeing their doctor. Usually proper BRCA gene tests are only reimbursed under certain circumstances, so mostly the relative will have to pay at least $500 to get tested properly. Most of them will just go to 23andMe. The growth in testing because of this effect is going to be significant. All the hard work of doing this kind of testing eg counselling, detailed family pedigrees, contacting family members, discussion of pros/cons of testing etc is being put back onto the individual and their healthcare providers. Good for 23andMe I guess, especially since their value seems tied to how much data they have. Whether this is net beneficial to society I'm not so sure.

ac29 9y ago
"23andMe demonstrated a high level of accuracy (greater than 99% concordance to Sanger sequencing) and precision (demonstrated by studies yielding greater than 99% reproducibility and repeatability)" Sanger sequencing is pretty accurate, on the order of 99.9%. So if 23andMe is matching those at over 99% accuracy and precision, a full sequence may not be necessary. The FDA took a long time to approve this, so I'd imagine it passes their standards as well.
dalbasal 9y ago
If you are thinking of macro-societal impact... I think you have to ignore the specifics of this particular test and think of the impact of these sorts of test on medicine, culturally. For example, there's a sort of Principe in medicine that you shouldn't order tests unless you will action treatments based on results. For example, if you detect a risk of breast cancer that can only be avoided by mamosectamy, only do the test if you're willing to to the procedure. Otherwise, all you have is a more stressed out patient. Data gathering is totally different in medicine for such reasons. The big, relevant exception is cancer, where a statistical, risk estimating approach is commonplace. In any case, genetic testing (and other ambient data gathering) may give us good reasons to change this approach. It's a transition and taking a step from toy to medicine (this is what 23andme is doing) is en route to such a change. What could emerge downstream is personalized cancer screening protocols. Maybe you should start getting colonoscopies at age 20 and I should start at 60.
- Gatsky 9y ago
  (Disclaimer: I'm an oncologist) There is definitely a science of designing and performing useful diagnostic tests. I have seen many cases of illogical or misinformed testing that has harmed patients physically, psychologically and financially. I ultimately believe that more information is better, however, so far the transition to more direct consumer testing is failing miserably (eg Theranos, or all the companies claiming SNPs can help design dverything ftom skin care to exercise regimens). Then there are just crackpots who coerce vulnerable patients into expensive and totally unproven, never-to-be validated tests that cost thousands of dollars.
hyperbovine 9y ago
Nearby variants are highly correlated due to their spatial proximity, a phenomenon known as linkage disequilibrium. Therefore a dense array of SNP genotypes is often sufficient to make statistically accurate inferences about the portion of the genome you do not directly assay. I myself have used this strategy to estimate my risk for possessing the ApoE variants from my 23andme callset. Is this 100% accurate? Of course not; no medical diagnostic is. But it is accurate enough to be used as a screening procedure in a lot of cases.
- Gatsky 9y ago
  If you start off with an inaccurate screening test, you waste a lot of resources with follow up testing, not to mention the psychological impact with BRCA tests.
  - hyperbovine 9y ago
    Define inaccurate. 23andme is arguing that their diagnostic is to within a fraction of a percent as accurate as the gold standard.
    - Gatsky 9y ago
      ...for the few SNPs that they actually test. A negative 23 test does not exclude BRCA mutation. To apply link disequilibrium as you brought up doesn't change that.
      - hyperbovine 9y ago
        The FNR for this test is comparable with Sanger. I guess I don't really see your point.
        Gatsky 9y ago
        Your 1st point is something like "these tests are useful for screening, you can get information about unassayed regions using statistical genetics approaches". My point is that this is not useful, since screening tests need to have certain properties not provided by SNP tests. The biggest issue is that 23andMe do not test the whole gene. This is like only doing a mammogram on one tenth of only the left breast, obviously a futile screening endeavour even if the FNR for that 10th is gold standard level. But most of all, unlike other screening tests which are designed to help the patient as much as possible (within societal resource constraints), 23andMe are offering a very limited test because that is what suits them, that is all that is possible with the SNP arrays around which they have built their business model. My original post was highlighting that unlike what is suggested on the gushing press release, this is not actually a great moment for direct to consumer genetic testing. If they were really interested in helping screen for BRCA mutations, 23andMe could swap out some of the other genes which tell you trivial stuff like how fast you metabolise caffeine, or some rare drug metabolism effect and include more BRCA SNPs. But they don't, because it is all about how many tests they sell and how useful the information is to pharmaceutical companies.
a13n 9y ago
I've heard that one of Helix's differentiators is that they sequence your whole genome rather than a small subset. Maybe worth checking out. https://www.helix.com https://www.helix.com (I have no affiliation)
xiphias 9y ago
I agree with you that 23andme should test for all known cancer mutations, but after getting a test from them and exporting the data to Prometheus, I got so much more knowledge about myself than just the cancer screening.

by Gatsky 9y ago

ac29 9y ago

"23andMe demonstrated a high level of accuracy (greater than 99% concordance to Sanger sequencing) and precision (demonstrated by studies yielding greater than 99% reproducibility and repeatability)" Sanger sequencing is pretty accurate, on the order of 99.9%. So if 23andMe is matching those at over 99% accuracy and precision, a full sequence may not be necessary. The FDA took a long time to approve this, so I'd imagine it passes their standards as well.

dalbasal 9y ago

If you are thinking of macro-societal impact... I think you have to ignore the specifics of this particular test and think of the impact of these sorts of test on medicine, culturally. For example, there's a sort of Principe in medicine that you shouldn't order tests unless you will action treatments based on results. For example, if you detect a risk of breast cancer that can only be avoided by mamosectamy, only do the test if you're willing to to the procedure. Otherwise, all you have is a more stressed out patient. Data gathering is totally different in medicine for such reasons. The big, relevant exception is cancer, where a statistical, risk estimating approach is commonplace. In any case, genetic testing (and other ambient data gathering) may give us good reasons to change this approach. It's a transition and taking a step from toy to medicine (this is what 23andme is doing) is en route to such a change. What could emerge downstream is personalized cancer screening protocols. Maybe you should start getting colonoscopies at age 20 and I should start at 60.

Gatsky 9y ago
(Disclaimer: I'm an oncologist) There is definitely a science of designing and performing useful diagnostic tests. I have seen many cases of illogical or misinformed testing that has harmed patients physically, psychologically and financially. I ultimately believe that more information is better, however, so far the transition to more direct consumer testing is failing miserably (eg Theranos, or all the companies claiming SNPs can help design dverything ftom skin care to exercise regimens). Then there are just crackpots who coerce vulnerable patients into expensive and totally unproven, never-to-be validated tests that cost thousands of dollars.

hyperbovine 9y ago

Nearby variants are highly correlated due to their spatial proximity, a phenomenon known as linkage disequilibrium. Therefore a dense array of SNP genotypes is often sufficient to make statistically accurate inferences about the portion of the genome you do not directly assay. I myself have used this strategy to estimate my risk for possessing the ApoE variants from my 23andme callset. Is this 100% accurate? Of course not; no medical diagnostic is. But it is accurate enough to be used as a screening procedure in a lot of cases.

Gatsky 9y ago
If you start off with an inaccurate screening test, you waste a lot of resources with follow up testing, not to mention the psychological impact with BRCA tests.
- hyperbovine 9y ago
  Define inaccurate. 23andme is arguing that their diagnostic is to within a fraction of a percent as accurate as the gold standard.
  - Gatsky 9y ago
    ...for the few SNPs that they actually test. A negative 23 test does not exclude BRCA mutation. To apply link disequilibrium as you brought up doesn't change that.
    - hyperbovine 9y ago
      The FNR for this test is comparable with Sanger. I guess I don't really see your point.
      - Gatsky 9y ago
        Your 1st point is something like "these tests are useful for screening, you can get information about unassayed regions using statistical genetics approaches". My point is that this is not useful, since screening tests need to have certain properties not provided by SNP tests. The biggest issue is that 23andMe do not test the whole gene. This is like only doing a mammogram on one tenth of only the left breast, obviously a futile screening endeavour even if the FNR for that 10th is gold standard level. But most of all, unlike other screening tests which are designed to help the patient as much as possible (within societal resource constraints), 23andMe are offering a very limited test because that is what suits them, that is all that is possible with the SNP arrays around which they have built their business model. My original post was highlighting that unlike what is suggested on the gushing press release, this is not actually a great moment for direct to consumer genetic testing. If they were really interested in helping screen for BRCA mutations, 23andMe could swap out some of the other genes which tell you trivial stuff like how fast you metabolise caffeine, or some rare drug metabolism effect and include more BRCA SNPs. But they don't, because it is all about how many tests they sell and how useful the information is to pharmaceutical companies.

a13n 9y ago

I've heard that one of Helix's differentiators is that they sequence your whole genome rather than a small subset. Maybe worth checking out. https://www.helix.com https://www.helix.com (I have no affiliation)

xiphias 9y ago

I agree with you that 23andme should test for all known cancer mutations, but after getting a test from them and exporting the data to Prometheus, I got so much more knowledge about myself than just the cancer screening.