Hacker News new | past | comments | ask | show | jobs | submit login
Be Careful about 23andMe’s Health Test (nytimes.com)
173 points by tysone 45 days ago | hide | past | web | favorite | 99 comments

I’ve always assumed that my 23amdme results were merely guidelines. If it says I don’t have something, then that means nothing. If it says I do have something, then that means I should be on the lookout for other corroborating data or symptoms.

It actually helped me find an explanation for some weird symptoms I was having. After I told the doctor, “23amdme says I may have X” he said “yeah all those things could be symptoms of X” and we went from there.

Exactly. 23 and me helped find an older half sister, but the health stuff? Eh...

> If it says I do have something, then that means I should be on the lookout for other corroborating data or symptoms.

Unfortunately the power of suggestion is a very real thing, so this isn't nearly as rational as it appears at first glance. There's a reason the medical profession is seriously concerned about outside tests, and it's not just cartel behavior. It's probably the same mechanism as the Baader-Meinhof effect.

The medical profession can only treat a limited number of conditions. They work on a science based approach which requires costly experiments so the range is limited. Even when something is discovered changes can take a long time to filter to you. These type of tests offers insights not available to most through a doctor visit even if the accurancy is way off in some cases.

That seems like the right way to use it. Unfortunately, I think a lot of people will not be so pragmatic.

Did any treatment result from this discovery?

Not the person you asked, but in my case, yes, after additional independent verification. It is not a huge deal, but it did result in a real quality-of-life improvement. I am grateful for what I learned.


Not yet. Just watching to see if becomes serious.

23andMe currently uses a genotyping chip, not gene sequencing. So, this article is misleading at the top (but gets more accurate at the bottom).

Early variant discovery suffers from the winner's curse phenomenon: we overestimate the effect of newly discovered mutations.

As time has gone on, it has become clear that even for many Mendelian diseases, the genetic background plays a significant role in terms of who will develop the disease (or how early they will do so). The estimates of "penetrance" have fallen as we have obtained more data.

This affects clinical genetic interpretation as well as the automated stuff that 23andMe is doing.

This is our biggest gripe with the service. We have a lot of consumers that ask us to perform gene sequencing for them expecting to get something similar to 23andMe. It's always a hurdle to explain the difference between what we do (targeted gene sequencing) and genotyping chip. They're a worlds difference and requires some reading.

FWIW, gene testing may never be able to predict the exact odds, since some of the diseases heavily depend on environmental factors (what you eat, how much you exercise, how much you're exposed to carcinogens and sun, etc). In other words, the real distribution here is _posterior_, whereas the best 23andMe can do is somewhat accurately predict the _prior_. FWIW, my analysis from years ago seems to match the specific health issues my parents are now having in their late 60s. I haven't done the test for them, so it's not like 23andMe just cheated by randomly throwing in the answers from their health questionnaire. I now bring those up with my physician and optometrist and have them check for signs of disease.

What baffles me is why they aren't doing it themselves. A full genome sequencing (!) costs just $1k. I pay more than that for health insurance every month.

> gene testing may never be able to predict the exact odds, since some of the diseases heavily depend on environmental factors (what you eat, how much you exercise, how much you're exposed to carcinogens and sun, etc).

All of those examples are themselves strongly predicted by your genes. The weakest is what you eat, but just knowing what country the customer is based in is probably enough to go forward with.

An extract from the article: “23andMe has said that its health tests can raise awareness about various medical conditions and empower _consumers_ to take charge of their health information.”

It’s a damn NEW DARK AGE way of thinking to refer PEOPLE through one single universal word all the time - A CONSUMER!

Choose any word from x, and carefully observe your thoughts afterwards:

”23andMe has said that its health tests can raise awareness about various medical conditions and empower _x_ to take charge of their health information.”

x = [“customers”, “individuals”, “humans”, “people”, “citizens”, etc]

> It’s a damn NEW DARK AGE

It's a business, and if you give them money that makes you a consumer. What's wrong with that?

I think the issue is mainly that people are multi-dimensional and referring to them as `consumers` boxes them into "thing x that buys our stuff and exists for consumption" instead of framing is as "this is a person that we developed technology because it will improve their lives (and give us money)". Profit is important for a business and helps drive efficiency and measure effectiveness, but I think it is still good use language that doesn't dehumanize the fact that the world is made of people and meeting the needs of people is ultimately the important thing, not simply number in Excel that represent our cashflow.

Even customer is a better word, since it represents a person who comes to you voluntarily for an exchange of value, not just consumption.

Consumer is appropriate here and the word makes a difference. I'm a customer of 23andMe, but I'm not a consumer (purchased for others, have not used myself).

We live in a world where words mean something specific, generally to communicate an idea, not warm fuzzy feelings, and it's important that we use the right words so not to fall into the trap of multi-national businesses pretending their employees are "family" because it sounds nice.

Thank you! :)

You're not the consumer. You're the product. The goal of these companies is to own a giant database of genetic information about the population to then sell that information to others.

If you were the consumer, you'd have to pay far more for a "test": these tests are just a method to get people to enter themselves in the database, and for them to sign their rights away, to 23andMe.


... and to monetize your genetic information infinitely without compensating you

They are salespeople, and they should present us the fairy tale, not the technical lingo, for maximal effect.

It is quite easy to use "consumer" to create a seemingly plausible argument, which is interpreted quite different by many if you replace consumer with any one of the other words suggested.

It is probably because very few has consumer as a part of their self perception, hence many are likely to identify less, and empathize less with a group or person described as such.

It's not wrong in a technical sense, but it can be used to avoid a stronger public reaction to e.g people getting hurt by a business practices.

One might frame it like this: If my business transaction as a consumer ends up with the loss of an arm, I'm still the one who has to live with only one arm.

That makes you a customer!

https://news.ycombinator.com/item?id=19063829 - McKinsey Advised Purdue Pharma How to ‘Turbocharge’ Opioid Sales, Lawsuit Says.

I’m not in any sense offended by the words company chooses, and neither I care about them being polite or not, the heart of the issue is the attitude and the word consumer expresses that very attitude rather accurately.

1. Netflix has consumers

2. Mubi has customers

I think you're being pedantic. "Consumer" is more descriptive than your other words.

Edit: I appreciate your concern about consumerism and the effects of being treated like cattle, I just think more descriptive words create better conversions.

Go on, develop the argument.

Shouldn't that be enough? The authors use more descriptive words to describe the circumstance, which creates better understanding. They could use, say, "humans", but not everyone uses their product and that would be a wasted opportunity for better language.

They're basically saying.... In a producer consumer economic design pattern....

"We're not just threads man!"

This is a topic that is often debated. Consumer has become the more generic term, particularly when speaking about the economy and about legal rights of people. Though it has the negative connotation from "consumerism" which is why the word is controversial and not considered polite.

In this particular case, there are two ways to read the statement from 23andMe. You could interpret "consumers" as the people who have actually bought a product from 23andMe. Or, you could interpret it as the overall population of people who have health information and may buy the product. With the first interpretation, I think "customers" would definitely be the better word and would fit with typical business etiquette. Though I think the second interpretation is what 23andMe actually meant, in which case "consumers" or "people" would be more appropriate. If I had written the 23andMe blog post, I would have used "people".

Comparing the google search results for "empower consumers" and "empower people" is quite interesting.

"citizens"? "residents", "tourists", "non resident aliens"?

Please tone down the histrionics. You're right, but only accidentally. They probably should have said 'customers' as that is the word that is the most accurate and specific from their pov.

if it said "people" would you not complain that it's not a public good?

You're going get upset about the word consumers and not complain about empower?

Is it me or are there suddenly a lot of anti-DNA testing articles being published lately? Is there a particular lobby that I'm not aware of?

Maybe just because most of these DNA services are based on bad science, and eventually people were bound pay attention to this?

I think people are beginning to think of the "tech industry" as a whole, as being a two edged sword. The honeymoon is over.

Not just "people", but "journalists" who see their own field collapsing and attribute a combination of blame and envy to tech.

Or maybe, just maybe, a lot of the things going on in the tech industry could have ramifications due to severe privacy violations along with a large amount of startups/companies selling bullshit to consumers.

The tech industry is just another industry filled with plenty of dishonest, selfish, sociopathic people and organizations that inculcate groupthink, diffuse responsibility, and achieve a collective level of amoral selfishness beyond that of any of the individuals involved in it. This is true, and this is also not news to anyone in the tech industry.

But this is the human condition. Every organization is prone to these kinds of things. And to some extent, every organization has to be self-interested and prone to value their own interests over those of anyone else.

Which isn't a statement of defeatism--it's worthwhile to curb this tendency, and to identify particularly bad instances of it. It's just that the media themselves are also subject to the human condition, and that despite pretending to be disinterested public servants, they have incentives of their own. And those incentives are only stronger and stronger due to the sheer desperation that the news industry finds itself in.

There's an old saying that you should never pick a fight with a man who buys ink by the barrel. The tech industry didn't just pick a fight with one of these men, it threatened the livelihood of all of them at once. So maybe, just maybe, we should price that in a little bit whenever we see yet another hit piece directed at the tech industry.

> attribute a combination of blame and envy to tech.

Well, it doesn't help when techbros from 4chan and other 'techie'-linked spaces routinely troll these same journalists by suggesting that they should "learn to code", or something. I'm not saying that their envy is justified, but it seems that there's more than one side at fault here!

It isn’t ‘techbro’ trolls that are engaging in the learn to code trolling, it’s people who believe that the news media gleefully printed hundreds of articles about how blue collar workers needed to quit complaining about the new global economy and learn to code, and that they are being hypocritical now that the (supposedly) same forces are impacting their industry.

I’m not endorsing the campaign which is hurtful at a time when someone has just lost their job, but to interpret it as techbros holding coding up as the way of the future is a mistake. Most of these people are also skeptical of coding as a lasting career.

who pays any attention to what someone on 4chan says?

As someone who works in tech, it feels like people are finally waking up (just a little bit) to how much bullshit is in the tech industry. Personally, I would credit (blame?) "blockchain". The whole blockchain industry has been such a roller-coaster of fail that perhaps it awakened some fresh skepticism.

It's the hype backlash. There were a lot of over the top pro DNA testing articles in years past so now the hype cycle has moved on to the next stage.

This is interesting. Is there actual research on “hype cycles?” I want to read it if so.

I just checked and indeed there is research.

The most popular model that we all know is called the Gartner Hype Cycle. https://www.gartner.com/en/research/methodologies/gartner-hy...

Arxiv has a few articles on hype cycles for example:

"A simple interpretation of the growth of scientific/technological research impact leading to hype-type evolution curves", Campani, 2014. https://arxiv.org/abs/1410.8685

"Hype Cycle Dynamics: Microscopic Modeling and Detection", Hashemi, 2018. https://arxiv.org/abs/1809.08939

You can also type "gartner hype cycle" in Scholar or sci-hub and there are many hits.

Not that I'm aware of. It's more a comment on how at some point the press overhypes some tech which then leads to an inevitable correction. Hopefully it's true value is appreciated afterwards but that's not a given.

not sure why this is downvoted, it's a fair question. Conspiracy theories are fun but I think the apparent trend reflects two complementary forces: 1. More media attention to genetic testing in general (because of 23andme, other companies entering the market) 2. A great deal of concern from the medical genetics community that these tests, or at least some of them, are way ahead of the science. Geneticists who are drawn to the field out of a desire to impact health take exception to seeing their work used to take advantage of vulnerable groups. (This can go to the extreme: some geneticists would prefer that the science remain 'pure' and will probably never think it's ready for prime time.) Disclosure: lapsed geneticist

One company screws up, people start to look more carefully at others.

But I can neither confirm nor deny the existence of such a lobby that you speak of, which is funded by.. umm.. crime industry?

Or you know, privacy or something?

If a family member signs over their genes, and they share a lot of genetic material with me, my privacy is compromised.

I like to think I should have some form of agency over the intrinsic things that make me who I am, but this industry couldn't care less about that. That squicks me out.

Would you send me a scan of your fingerprints? Or do your crime industry connections suggest that you shouldn't do that?

You are right. I was trying to think who might have deep pockets to fund lobbying against genetic testing, that's all. Privacy is definitely a valid concern that I forgot about!

>I was trying to think who might have deep pockets to fund lobbying against genetic testing, that's all.

I think in this case the deep pockets would mainly be on the side of selling more genetic testing, as opposed to selling less. After all the pocket holders themselves aren't being tested.

"my privacy is compromised"

As well as the privacy of your ( possibly still unborn ) children.

What about your face? It's used more often.

I see it the other way. I've noticed a very marked increase in promotion of these tests over the last year or so (including people on YouTube being sponsored to take them). I think we're now seeing some of the backlash to that promotional onslaught.

I think you're right but for different reasons. Over the past few years, we've seen the toxicity that can come from the management of healthcare companies like they are the same as software companies. The easy example here is Theranos, which was engaged in fragrant fraud until it was uncovered by the WSJ. The company focused its time and resources on the design and aesthetic, but was from a scientific perspective, 100% BS.

That says to me that if you don't take an aggressive regulatory supervision of the healthcare market, (literal) snake oil salesmen will come in and promise the world. It's one thing to let juicero fail because there's no critical health concern that happens when the product ultimately fails. When it's a device or service that impacts patient care, we have a moral obligation to ensure we are doing best to ensure we are positively impacting care.

Because become very popular and are being hyped, and so the press has decided to shone some light on them?

Adulterers would have a reason to be opposed to such testing but I doubt they have a lobby.

I don't think it is some lobby. All these concerns are legitimate. Besides, people are scrutinizing tech companies more anyway. I have always felt uneasy when friends happily send their DNA to some company just to get some stats of doubtful accuracy without realizing how their data can be used. So I am happy about the increasing awareness and scrutiny. I hope these companies get regulated better. Just like with GDPR, I should have control of exactly what data from my DNA they can use, and how they can use it.


By that you mean the original articles promoting 23 and Me were paid advertising and this is the nytimes' attempt to be less biased?

No. Obviously both are paid advertising.

So this seems to suggest there may be plenty of false negatives, but I think the key factors in this article are at the end: Family history and genetic contributions to disease.

I personally decided to throw our 23andme tests out, for a variety of reasons (health privacy being one), but one of them is health anxiety. It feels unnecessarily anxiety inducing to provide this info, in either case. If i have a mutation, I may live with some elevated fear of something inescapable that never materializes. If I get negative results, I may wonder “what about the ones they didn’t check?” Either way, I end up with less actionable information.

Kind of a psychological phenomenon here - more data in this case doesn’t necessarily result in more information, because interpretation matters so much in this arena.

I,OTOH, which I had had my 23&me results years ago since what they tell me now might have prevented a costly experience back then. I totally get the "health privacy" issue, what with Google looming in the background along many dimensions, but for me, that was not a big issue. I also think that there is an aspect of 23&me that is being ignored in the NYTimes piece, the social network growing around real phenotypes in the context of genotypes.

The data can be analyzed by other parties for additional insights. I'm a big fan of Dr. Rhonda Patrick -- her team allows you to submit your 23andMe data for additional tests: https://www.foundmyfitness.com/genetics

I've not yet done this myself but am planning to do so later this year.

I'm still not sure about privacy side of the 23andMe. Whole testing procedure could be designed to be completely anonymous, yet upfront I have to provide name, address, birthday and e-mail.

Why do people willingly provide DNA samples together with personal information?

Do they verify the information, couldn't someone use a pseudonym? (Though then your relatives who take the test might be surprised by the new relative they discover..)

I bought a test and had to use my own name for shipping and payment (because of the delivery and payment services involved). Someone else (cough, cough) registered the test for their own sample and the results.

For 23andme, the discovery of relatives is optional. A user can choose to not receive any information about potential relatives, or to disclose minimal information, to allow messaging without names, or to disclose a name. I may be misremembering some of the granularity, but names do not have to be automatically disclosed.

even if they didn't, they need postal name/address to send the kit

Not really. I can buy the kit from a reseller or even OTC by cash and send it back without providing a return address.

Apart from that, I shouldn't need to register on their website to access my results. There could be unique, randomly generated URL on a scratch card in the kit, where the results will be put as soon as they are done with sequencing.

this is a bad fucking idea to allow your DNA sequences to be shared with the government.

but seems like a trend now, just to blindly copy what celebrities endorse so

more power to you folks who submitted your DNA, you can never take it back, and they have it forever.

Any crime your descendants or you commit in the future or the propensity to commit crime under the right psychosexual circumstances carefully driven by subliminal messaging, are fucked.

California has been stealing DNA from babies and guests of the criminal justice system for over 30 years.


that is quite chilling but not surprising from the US.

I wonder if this had impact on the drop in crime rates during the 90s.

Imagine a carefully targeted campaign against genetically disposed to violence and allergy to education. You can sterilize entire 'hoods bye putting guns, drugs and alcohol there and just wait for the environment to infect all others.

Like considering the overwhelming prison population in the US is black or latino, you can't help but wonder at some grand conspiracy that's aimed at sorting out the unfit from the normal.

Relevant fine print from FDA decision:

i A prominent and conspicuous limiting statement that the test provides only a preliminary test result that needs to be confirmed using an independent pharmacogenetic test without such a limitation prior to making any medical decisions. Alternatively, appropriate design verification and validation activities, including the generation of robust analytical data demonstrating appropriate analytical accuracy and reliability of test results for each genetic variant included in the test report, must be performed that demonstrate that the test can be used to make well-informed clinical decisions.

In other words, any test results (at least related to drug metabolism) require... confirmation with another test?

Two things: 1. 23 and Me has a history of overhyping their science when it comes to medical claims. FDA back in 2013 threatened regulatory action due to their wafarin (blood thinner) metabolism claim.


2. FDA regulates marketing and claims that you can make for products including drugs/devices (but that whole regulatory framework is a story for another day). You need to have certain regulatory approval for formalized medical testing/reimbursement (remember that 23 and Me is targeting consumers, not traditional patients).

They are just saying don't take medical action based on our results and hold us liable.

Why doesn’t 23 and Me test the whole gene? I can’t imagine it’s because they are ignorant. Is it cost or do they not have all the gene info due to how they acquire the sample or something else?

23andme only searches for known snips (SNPs, single nucleotide polymorphisms), they don't scan the genome, that is still a fairly costly and slow procedure (although it's radically getting cheaper and cheaper!).


> that is still a fairly costly and slow procedure

How many zeros are we talking about? Does anybody offer the service?

Lots of people offer whole genome sequencing. Dante Labs is currently offering it at EUR 259 (no affiliation: I got their mailshot because I bought their lower still Black Friday offer). Typical lab costs are thought to be more in the $1k range, with the assumption that some of these offers are made with the intent of rapidly acquiring a large corpus of anonymised full genome sequences which they hope most of their customers won't opt out of being analysed for commercial scientific research.


Cost, but they keep some of the PCRed DNA around and could do it for all their existing client base in the future, for free if selling what they can of the data begins to outweigh the cost.

According to https://en.wikipedia.org/wiki/$1,000_genome, the cost for whole genome sequencing is now under $1000, with some companies now coming very close to 23andMe's $100-$200 price. I haven't found any info on whether 23andMe is moving in this direction instead of their current methodology.

Veritas Genetics offered 30x coverage full genome sequencing in the lead up to Xmas for $199 to the first 1000 customers.

They sold out in about 6 hours.

Veritas offered $999 full genome sequencing since 2016.

My guess is that we should see a new regular/lower price later this year.

23andme used to give health reports for years and these issues were known. After all , i doubt its clients consider it a health service, it is at best a novelty. Plus, its health reports are completely unavailable in europe. Still, if someone is curious to check their disease risks, there is promethease

I realize this is just my personal experience. That being said, I was an early customer of 23andme, and their health checks actually alerted me to two different things that I have since taken appropriate measures for. They were legitimate issues that I hadn't sussed out previously, but further testing confirmed them. I'm grateful for what I learned.

i have the old reports too, and even though i did not learn something i didn't know, i wish i could have new ones, with the added knowledge from ~5 years of research.

Same. There's always https://promethease.com/ , but 23andme did/does a better job at data presentation.

I whole heartily disagree. There is significant patient risk when it comes to bad testing information. Women may attempt to elect to get invasive procedures (such as double mastectomy) based on their perceived BRCA status. This may lead to the rare infection and potential death, not including the emotional and physical suffering. And as the article alluded to, 23 and Me did "cross the line" when it started reporting warfarin related variants, when the science wasn't well understood [1].

Americans have already agreed that when it comes to healthcare markets, more rather than less regulation is necessary for making clairvoyant decisions (ie Federal Food, Drug, and Cosmetic Act which created the modern FDA). I think there is clearly potential for policy reform, but the complete lazzie-faire economic model just simply doesn't apply to the healtcare market.

[1]. https://www.fda.gov/ICECI/EnforcementActions/WarningLetters/...

people may make bad decision based on their horoscopes, that didnt make them banned

I have Celiac disease. Would it be beneficial for me to take one of their tests to see where my condition originated?

If I were in your shoes, and planned to have children, I would go for it, just in case.

Those videos were totally unnecessary, and could have been still images.

I felt videos 2 and 3 were the most helpful parts of the article. I wish more science/tech articles could rely on graphical/animated explanations. After all, they do speak a thousand words.

If an image says 1000 words, those videos said about 1002 words, and showed up as blank spaces until I right-clicked and selected "play". (Before that, they said 4 words: "This site looks broken".)

Usually pictures (still images) are said to "speak a thousand words", not videos or animations.

Wait, this is from the NYT's editorial board? I love the NYT, but why on Earth would I look to its editors for scientific/medical advice? And why on Earth would they opine?

Well for one is Sarah Jeong. Despite internet conservative white males not understanding irony and sarcasm, she is a thought leader when it comes to the intersection of technology and law. I don't know her writing on science/medicine, but her writing on technology and law are in the top of the field. I trust her (and the NYT brand) to consult with medical professionals who know what they're talking about.

And for what it's worth, this particular NYT article is pretty on point for their analysis as read by a bioinformatics PhD student.

Weird digression! I happen to agree Sarah Jeong's work is worthwhile, but like you said it's not her field, and none of the editorial board list anything about medical focus. Re: consulting experts, sure and agreed - but then why not do it as a reported piece and like, quote the experts?

> this particular NYT article is pretty on point for their analysis as read by a bioinformatics PhD student

Really? With stuff like "23andMe’s breast cancer screen looks at just a handful of places in the gene where mutations are known to appear. That’s like proofreading a document by looking at only a handful of letters"? Bioinformatics research background here, too, and as far as I know all of those screens scan particular sites and not a full sequencing -- because the sites are the only thing we have clinical evidence for anyway.

I don't doubt that there are more comprehensive screens on the market, but arguments like "only 2% of the population has these mutations" are rhetorically misleading without comparative details, like how much of the population has detectable mutations of any kind.

I think using an opinion page level of rigor to make arguments of this kind misleads some people, and leaves everybody else in a place of choosing whether to "trust the NYT brand" as opposed to providing them evidence.

Clinically, we have the option to sequence the BRCA 1/2 genes. It's not uncommon for this to be done if certain clinical criteria are met (such as close family history of brca or being of Ashkenazi Jewish descent.) [1].

Note: Paywall warning. But the uptodate article says the following.

"Direct-to-consumer BRCA testing — In March 2018, the US Food and Drug Administration (FDA) authorized the direct-to-consumer (DTC) company, 23andMe, to include BRCA testing as part of its Personal Genome Service Genetic Health Risk report [60]. It is important to note that the testing includes only the three BRCA Ashkenazi Jewish founder mutations and does not entail analysis of other potential variants in these genes. Pre- and post-test genetic counseling is not required for individuals interested in such DTC testing.

Any individual who receives a positive result from DTC testing should have it confirmed in a clinical lab. Although the likelihood of a false-positive result is very low, it is critical for the result to be confirmed and an individual, signed report should be provided to the patient.

An important concern related to BRCA DTC testing is that women at average risk for breast cancer may be falsely reassured about what the genetic testing results mean for their breast cancer risk and the continued need for routine, age-appropriate mammography screening. On the other hand, for women with a personal or family history of breast cancer who are candidates for genetic testing, testing only for the founder BRCA variants is incomplete, regardless of ancestry. Such women need complete BRCA gene analysis and are likely good candidates for multigene panel testing."

[1]. https://www.uptodate.com/contents/genetic-counseling-and-tes...

Thanks for these details!

Edit: it's still bugging me that this doesn't describe the utility of the additional clinical screen more specifically. Is the point that even if false positives are rare, the treat path is informed by full sequence data? Would love additional insight.

These details are important to me because I think low cost tests can fill a public health niche that "only the most comprehensive test" strategies leave underserved. Any progress in those areas is likely to meet resistance, because it could bite into the margins of folks offering broader packages. So I am really interested in how both scientific and popular articles portray the tradeoffs.

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact