One DNA test said he was likely to get Alzheimer’s, the other said he wasn’t
nytimes.com
nytimes.com
> Word of caution to those with data from Ancestry.com: in our experience, based on data in OpenSNP and from Promethease users since 2006, Ancestry data always reports rs429358 as (T;T), even for people who's data from other sources indicates they are (C;T).
Loads. Are you seriously telling me that you can't think of anything useful to do with this knowledge? (What about if he wants to have kids?)
> The doctor referred Mr. Fender to a geneticist, but it turned out he did not see patients under 50 who were not symptomatic and had no family history of the disease.
> Mr. Fender then tracked down Jill Goldman, a genetic counselor specializing in dementia at the Taub Institute at Columbia University Medical Center, who described a multistep process of counseling and confirmatory testing that’s been the standard of care for 25 years. She typically serves people at high risk of inheriting a disease, and insurance usually covers both the consultations and the tests. But it was unlikely to cover the costs in the absence of family history. “It was like a chicken-and-egg thing,” Mr. Fender observes. “I needed a medical test to prove to them that it was real, but I couldn’t get a medical test until I could prove to them that it was real.”
> Meanwhile, he happened to see a holiday special — $69 — for Ancestry’s genetic risk test.
There are problems here, but they're not with 23andMe or Ancestry.com, I don't think.
> 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...)
'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
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 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/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-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.
Obviously, there are perfectly accurate and FDA approved tests for BRCA1 and BRCA2. The point is 23andMe don't do these tests. Therefore, how can you argue that their testing for lowly penetrant poorly characterised ultra-rare risk loci (which means most of the positive tests will be errors, as in this case) is somehow a great benefit to humanity, if they don't offer by far the most common, clinically significant and life saving genetic test there is?
As I said, the answer is of course that 23andMe are acting in their own interests. Anyone paying for the privilege to give away their genetic information to a private company should realise this.
> 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 testing for ultra rare lowly penetrant AD loci is more useful?
Your entirely tangential example from DeCODE makes no sense. DeCODE was not a user pays self-referred program. It was a research project.
If you read my original comment, I did not state in past tense that 23andMe made billions (I presume you are merely careless in reproducing what I said, rather than disingenuous). My comment specifically was 'Meanwhile 23andMe makes billions selling the data to pharma?', indicating that there is potential for them to make billions or that they are in the process, not yet complete, of making billions. Their valuation by VC funding is entirely supportive of that potential and process. The CEO has come out many times stating their business plan is to sell population genetic data to pharma.
How can one thing be good if another thing is even gooder?
> As I said, the answer is of course that 23andMe are acting in their own interests. Anyone paying for the privilege to give away their genetic information to a private company should realise this.
It is not a negative-sum game. It is from the self-interest of the butcher we are able to buy dinner, not his benevolent regard for humanity. 23andMe can do a lot of good while seeking to do well.
> And testing for ultra rare lowly penetrant AD loci is more useful?
Apparently. Illumina and/or 23andMe chose to put it on the chip, so they apparently felt it was worthwhile.
> Your entirely tangential example from DeCODE makes no sense. DeCODE was not a user pays self-referred program. It was a research project.
My point there was a comparison of the regulatory regimes that genetics companies (DeCODE is not a 'research project', is a former NASDAQ-traded (DCGN)corporation which is now private post-bankruptcy, which did in fact use 'user pays self-referred programs' in addition to the national database they were hired for) are under such that even when they know with high accuracy about very serious mutations like BRCA, which you brought up as a huge dereliction, they may still not be allowed to tell people, which illustrates the incoherence and harm of the medical system in regards to genetic data; given the question about whether they would even be allowed to tell people (like you think they should not be allowed to tell people about AD loci), why would they spend resources testing for it?
> My comment specifically was 'Meanwhile 23andMe makes billions selling the data to pharma?', indicating that there is potential for them to make billions or that they are in the process, not yet complete, of making billions
Oh, so the rest of that paragraph where you were contrasting how 'Meanwhile 23andme makes billions' to things that had already happened or were happening were also just 'potential' too? I see. How very careless (not disingenuous) of me.
Why bother to be flippant this far down a thread on HN?
> It is not a negative-sum game.
I remain surprised that you should be quite so sanguine about the mutual benefit of giving away data to tech companies. 23andMe and others recently announced they would ask for explicit consent before sharing DNA information with other parties. How a genomics company could go on for this long without guaranteeing the privacy of its users is baffling.
> Apparently. Illumina and/or 23andMe chose to put it on the chip, so they apparently felt it was worthwhile.
'Worthwhile' and 'marketable' to a consumer audience are not equivalent, as I have repeatedly pointed out with BRCA testing.
> My point there was a comparison of the regulatory regimes that genetics companies...
You referenced the sequencing of Icelanders, which was indeed a research project (you've read the papers surely) and not a self-referred program, regardless of what else DeCODE does or is. You can't make a point about regulatory conditions in that setting and transfer it to consumer testing. But are you trying to say that it isn't worth testing BRCA because DCT companies might not be allowed to tell people the result, even though 23andMe are currently offering BRCA results?
> Oh, so the rest of that paragraph...
I would suggest that your ongoing efforts to tell me what I meant in my comment are better spent elsewhere. Particularly since as a privately held company, we don't know 23andMe's revenue, what deals have been made or what their future plans and projections are. What we do know is how much VC funding they have received, and the return on investment that VC's expect.
"But scientists are discovering that — to a surprising degree — we contain genetic multitudes. Not long ago, researchers had thought it was rare for the cells in a single healthy person to differ genetically in a significant way. But scientists are finding that it’s quite common for an individual to have multiple genomes. Some people, for example, have groups of cells with mutations that are not found in the rest of the body. Some have genomes that came from other people."
The DNA placed on the chip will be a heterogeneous mixture of whatever kind of cells provided. How they’d handle mosaicism, whether somatic or otherwise, I don’t know.
if you're lucky it won't tell you much you don't know, but learning about all the possible things to look out for (heightened risk of diabetes and adverse reactions to a medication, for me), sorted by magnitude, is exactly the kind of thing you'd hope to learn from personal genotyping. it's much more granular than what is offered by eg 23andme's health service, and is updated with new research. it's more than worth the $5.
https://promethease.com/privacy
> After 45 days Promethease deletes your Promethease report.
> At no time is your DNA data shared - or sold - to any external party, period. We also do not sell any products like vitamins or supplements.
[...]
> We do not claim any ownership rights in your genome and it belongs to you only.
There is no ethical excuse for withholding information from someone about their own biology.
It kinda falls under "well, I didn't actually lie... I just didn't say more than absolutely necessary. It isn't my fault if they assumed wrongly."
In the example of early-onset alzheimer's, complete information would come with a recommendation of retesting, given the incidence of errors, a likelihood of the disease (was given), and whatever the accepted medical advice to follow afterwards is. It would probably be best to include warning signs of when to see the doctor about it if most doctors and insurances won't cover anything without symptoms.
You can have all the legal disclaimers on this as well.
They were more about ethics counseling than interpretation of results at all.
This person was not a very helpful consultant, but a layer of bureaucracy. I would have loved someone to help me interpret results. In my case it was someone who I said “I want this exact test on this exact gene.” And they said “Are you sure, if you know here’s what it can mean.” And I said “Yes” so the lab ran the test and gave me my results.
Curiosity killed the cat.
Oh, you wouldn't tell the insurer? That'll probably be grounds for invalidating the policy, if and when it ever comes to light.
The more we are able to know about the likelihood of future medical issues on an individual level, the more dysfunctional the "insurance" model will become.
It’s a bit like being put onto death row.
Also, feelings matter. So “it’ll just hurt your feelings” is dismissive claptrap.