Hacker News Reader

Comment by tptacek | Hacker News Reader

Now consider that this is a marketing message that wants to have it both ways: it alerts clients to genetic "risks", which are very likely to be subjected to secondary testing, but does not want to be liable for the accuracy of its negative results, which are very unlikely to be challenged.

It would be better if 23AM presented medical genomic information in a neutral way, with external links to descriptions of genomic variations but no assertion of diagnostic significance at all, or even a mention of how to obtain a diagnostically significant result.

Unfortunately, the way 23AM is packaged (see the website), they are incentivized to do the opposite; positive results are recognized by their clients as valuable, and, whether 23AM likes it or not, so are the negative results.

View on HN

Here is a company who is trying to make genetics more accessible, approachable, and comprehensible.

> It would be better if 23AM presented medical genomic information in a neutral way, with external links to descriptions of genomic variations but no assertion of diagnostic significance at all, or even a mention of how to obtain a diagnostically significant result.

What you're suggesting would effectively revert the consumer experience about half a step above "go to lab X to sequence your DNA, then hop on over to SNPedia.com to figure out what it all means."

> Unfortunately, the way 23AM is packaged (see the website), they are incentivized to do the opposite...

Perhaps this is more subjective than worth talking about, but that's not the impression I had when using and working on the product. All the more reason they need to be careful, for sure.

Disclaimer: I worked at 23andMe's frontend team briefly, a few years ago. While I have mixed feelings about their engineering practices and other things, I do believe in and support their mission.

It is impossible to read 23AM's current website and not come away with an impression of the tool's clinical ambitions. I strongly buy into the idea that everyone working on the product is doing so with the best intentions and I'm thrilled to be living in a time when products like it are available at its price point. But their marketing is incautious and might need to be tuned up.

> But their marketing is incautious and might need to be tuned up.

You're probably right. Unfortunately it's too easy to create a gap between the public-facing marketing messaging and the real spirit of the product. I can think of many companies who struggle with this.

If you do become a customer, it's evident that they perform an admirable effort to make the results appear as neutrally educational as possible. Every data interpretation is heavily cited and explained, all the while urging further clarification from a medical professional. (Let me know if you'd like some screenshots.)

Personally, this is above and beyond of what I would expect, and there are competitors (especially foreign) who do far less. I can see why the FDA would prefer to have everything that could be interpreted as medical advice to be regulated and controlled, but I'm not convinced that it would yield a net benefit to consumers in this case.

What would be wrong with offering a product that sequences your DNA, then highlights the most interesting bits with links to external sources? A customer would still get the benefit of knowing high-risk/important things to check out, and 23AM would be clear. The website could wrap it up to make it a smooth experience.

Probably nothing.

"Here is a company who is trying to make genetics more accessible, approachable, and comprehensible"

No, it really isn't. It's trying to make a mass-market product that it can sell to unsophisticated end-users who don't have a prayer of understanding what they're seeing. Accuracy and precision are secondary concerns, and without those you're not actually doing genetics, but pseudo-science.

I have a number of friends who have received 23's "analysis", and come to me to interpret all/part of their results. They're garbage. You shouldn't take any of it seriously, in any way. To properly interpret even the simplest parts of their report, you need a background in statistics and genetics that only highly trained people have. And even then, there's usually not enough information -- you need to do tons of digging into the literature to find out that some relative risk increase is highly conditional on population or that the original experiment was flawed in some subtle but important way.

As they say: things should be made as simple as possible, but no simpler. There are things that are complicated for a reason, and you can't just discard the inherent complexity and claim it as a virtue. It's not disruption to be irresponsible.

Well, sure. That's because dispositive results are valuable to their customers. It's not a marketing problem: they have someone (some) people want. Saying that it would be "better" that they not provide this service seems to be an odd usage of "better".

The liability concern is real: they aren't perfect, they'll goof up tests, and in some measure (though IMHO largely overstated -- the real risk here is a little fear and a needless diagnostic procedure) people will suffer when they do. But that doesn't seem to me to be insoluble. If no one insures that kind of thing right now then maybe they could be induced to, or 23AndMe could work up something on their own via investors.

But arguing that the only viable solution is to shut down a comparatively cheap and (full disclosure: I'm a customer) really quite entertaining service seems wrong to me. It's a net good; some people (hi!) like it.

Well, I still like it, too, in a way. It's just a bit unnerving sometimes. I have this feeling that the technology is still a bit in its infancy.

That's exactly right. There is, at this time, no such thing as "medical genomic information", of the sort which 23andme purports to provide. There is a great deal of data, but that's not the same thing, and distilling that into information is an entire field of study -- complex, subtle, enormously recondite, and extremely refractory to the layman.

Nobody is arguing that it should be "shut down"; even the FDA doesn't appear to be doing that, and is demanding that they cease marketing only after years of effort trying to alter the product's marketing message.

> But arguing that the only viable solution is to shut down a comparatively cheap and (full disclosure: I'm a customer) really quite entertaining service seems wrong to me. It's a net good; some people (hi!) like it.

That's not what's happening. You're not stupid, and I'm surprised you see it this way.

23andMe are being asked not to market their product in the way they are marketing. Or, if they want to continue to market like they are they have to have the same level of regulation as all other medical device suppliers.

Neither of those stop you from using their product.

OK, the language was inflammatory and I apologize for that bit. But the point stands: it's not a "marketing" problem that they are providing pseudo-diagnostic results. That's what the market wants, because they can't get it from the existing products under the existing regulatory structure.

One way to see this is yours and the FDA's: it's a fairness issue, and they need to play by the same rules. The other is the liberstartuptarian line that the regulatory structure is the thing actually at fault here. In this case, I tend to agree with the latter.

I mean, I had fun with this service, and I wasn't fooled by its marketing: they told me in essence that they'd give me their best guess at my traits and disease risks and drug susceptibilities, etc... And I bought it. A product that just gave me a bunch of genes and markers I could look up on wikipedia wouldn't have gotten my money.

So what's the answer here? Where's the regulatory answer for those of us who want to spit in a tube and get a fun HTML5 experience out of it? It doesn't seem like the FDA has one.

They could slap on a disclaimer saying 'Results for entertainment purposes only. This service has not been evaluated or approved by the FDA.' I'm no expert on regulatory law but I suspect this would immunize them from most oversight...but I also suspect it would have a deleterious effect on sales.

I think they would have to remove much of the product too. The FDA is involved because they consider the product to be "intended for use in the diagnosis of disease." I doubt they could say, "our data shows you have disease X but this is for entertainment purposes only."

You're right. They're not going to be able to say "our data shows you have disease X". They probably have no business saying anything of the sort.

> That's what the market wants, because they can't get it from the existing products under the existing regulatory structure.

The market can have that, all 23andMe has to do is stop selling their product as a diagnostic test.

They sell it as being able to predict your risk of some diseases. They're not a tiny little startup battling against the industry, they have some pretty big money.

They need to either sell it as a novelty ("Not designed to treat or diagnose any medical condition") or step up and get regulated.

Unfortunately we're in a weird situation where 23andMe get regulated, but wingnuts like Sarah Myhill (in the UK) don't. http://drmyhill.co.uk/wiki/Nutritional_Supplements_-_what_ev...

So I definitely agree that something's wrong.

> The market can have that, all 23andMe has to do is stop selling their product as a diagnostic test.

Or, alternatively, perform and submit the studies to show that, used as directed, it works as a diagnostic test as marketed.

I've used 23andMe and I feel that this is how the information IS presented.

When I click through my profile, it shows me strength of assumptions based on the number of conclusive or inconclusive studies that have been reported. You can easily click through each result and see the studies.

I never once felt that they were making a diagnoses. I don't know where people are getting this from.

23AM just got a smackdown letter from the FDA about that:

http://www.fda.gov/ICECI/EnforcementActions/WarningLetters/2...

That letter was linked from the article itself, and is what prompted the fact that this discussion is happening now.

Compare it with Carfax and similar "car history" reporting websites. A positive result is nearly always meaningful, but a negative result means nothing more than a lack of hits in a database that is well known to be incomplete.

To take it one step further, I don't necessarily trust studies linking genomic variants with disease significance. A subset derive correlations using oversimplified models:

http://www.ncbi.nlm.nih.gov/pubmed/23770567

Reply on news.ycombinator.com