Personal Genomics – Why 23&Me Doesn’t Work

by Bradley Miller on November 5, 2009

In PEHub yesterday an article about 23&Me and the financial issues it’s been having.  As an entrepreneur and having had plenty of great ideas poo-poo’ed by investors and industry folks alike, it’s really hard for me to understand why anyone would have invested in 23&Me as a company.  What I don’t understand is why highly skeptical VCs have invested in a business who’s central premise, while certainly desirable, is so far from reality at this point that it’s amazing anyone would invest. It’s certainly an important idea – scanning our individual genetic make-up to discern health risks and prevent them. Who wouldn’t want to understand what preventable diseases they’re prone to? I certainly would (well, to an extent – but that’s for another post).

For the uninitiated, 23&Me is a personalized genomics company that will take a couple drops of your saliva, extract your DNA and screen it for hundreds if not thousands of genetic disease markers.  The company name is derived from the 23 chromosomes humans contain – 23 from mom and 23 matching from dad. But the company over promises and under delivers.  At the end of the day, the fact is that biomedical science isn’t advanced enough yet for us to make meaningful predictions off of the information screened by 23&Me. And, to boot, there are other companies like Navigenics that have a little better model of screening, but they’re still pretty far off mark – as of yet. That said, Navigenics’ science and results are much better than 23&Me, but that’s like saying Peet’s is better than Starbucks – Peet’s may have better beans, but the coffee’s still not all that good.

A gene chip by the company Affymetrix.  That little square on the chip can yield information on 500,000 different genetic variations.

A gene chip by the company Affymetrix. That little square on the chip can yield information on 500,000 different genetic variations.

Let’s dive right to the core of the issue – the biomedical science behind 23&Me. 23&Me (and Navigenics) use “gene chip” technology , which can screen thousands of genes at once and tell you where you have variation (mutations)  that are known to be correlated to disease.  In other words, if the gene chip picks up that you have a variation in a gene that has been correlated to a heart illness, 23&Me argues that you have a higher chance of developing heart disease. While that certainly seems logical – “I have a gene that shows a higher risk of heart disease, I better do something about it” – it remains somewhat misleading.

A closer look at that square on the gene chip - this is what the chip looks like under magnification when it's read by a computer.  The different colors indicate different gene results.

A closer look at that square on the gene chip - this is what the chip looks like under magnification when it's read by a computer. The different colors indicate different gene results.

We have to dive a little deeper here to understand why the findings from these tests don’t correlate to real disease risk. When researchers do genetic studies (the type of studies 23&Me base their tests on), most of the time they find correlations between gene variations and a disease. And I want to stress – these are correlations – and are not purely 100% causative like the genetic testing companies would like you to believe.  Put another way, these genes are found in these diseases, however they are not the root cause of the disease.  Most diseases are due to multiple genetic mutations, which means the underlying causes for these diseases are much more complicated than just one genetic mutation.

We now know of 20 genes that correlate to height, but they only explain 3% of variation in height between people.  Only 3% of the difference in height between these two men!  What about the other 97% of difference?  More research!

We now know of 20 genes that correlate to height, but they only explain 3% of variation in height between people. Only 3% of the difference in height between these two men! What about the other 97% of difference? More research!

For example, a study came out in the New England Journal of Medicine that detailed just how little we know about how our genes and how they become translated in to real world physical traits and disease. This review study illustrated that we know of 20 genes that correlate to the differences in height between people.  While that sounds impressive, turns out that those 20 variations explain only about 3% of the true variations in height. I’m 6’5″ – those 20 genes explain only about 1/3 of an inch of the variation in height between my 5’6″ wife and I! 20 genes!! Why then do we then think that 1 gene will detail risk for heart disease or cancer?  The bottom line is that before we can accurately correlate and make meaningful disease predictions based on genomics, much much more research needs to be done.

So, let’s come back up to the surface. I’ve detailed why genetic research to date isn’t as complete as these companies would have you believe. The personal genetic variations they uncover, while using the most advanced technology and knowledge we have, isn’t sufficient to fully explain disease risks. The companies are selling a service based on scientific misconceptions – people are accepting 23&Me’s marketing, rather than good science. And that lack of scientific and clinical substance is why the medical community hasn’t embraced these tests.

I’m going to get in trouble with these companies because they don’t directly make these claims – but I think it’s implied based on their marketing and how they discuss their product. Doing these tests even just for curiosity’s sake is even a waste of money – they don’t truly tell you anything useful.

As one caveat, Navigenics does have a much better platform than 23&Me and how they correlate gene changes to disease risk is much better than 23&Me. They do take a look at diseases more holistically – let’s say they’re screening for heart disease and for argument’s sake that they screen for 25 different genes correlated to heart disease. They take that information and integrate all the risk factors to give you a more accurate risk analysis based on population statistics. It’s a little better, but I wouldn’t spend the money for it yet.

Don’t get me wrong, these types of products and services are the future of medicine. Maybe not in this direct way, but we will be screening people for disease risk.  No, not for insurance reasons, but rather to attempt to prevent diseases before they take hold. It’s just too early for this type of genomic analysis to be accurate enough to truly act upon. Although I’d like to be an optimist, the technology isn’t there, medical practice hasn’t accepted these tests (and they shouldn’t) and the businesses like 23&Me are floundering as a result. All of this to say, as a consumer patient hold on to your money for now, the scientific community has a long way to go before we really have the information necessary to make strong clinical correlations and to make valid disease predictions.

{ 14 comments… read them below or add one }

1 John Girard December 4, 2011 at 9:39 am

Interesting take — thanks for sharing this!

I would argue, though, that 23 & me has a bigger marketing problem than scientific problem. In fact, I don’t think that the typical consumer of 23 & me’s products cares at all about the nuances of the science — what the big difference is between causation and correlation, how immature the research is at this point, etc. From the average consumer perspective, the data is “good enough” (note: I don’t believe this to be true myself — I just think that the typical consumer won’t make the distinction). Especially when you see luminaries like Sergei Brin making significant life choices and investments based on these very same markers (http://www.wired.com/magazine/2010/06/ff_sergeys_search/all/1), I think the science issue is relatively less important.

Put another way, my guess is that very few people see the service, get interested, and then say “wait, I’m not going to spend $5/month — the science is bad.” Rather, I think that there are not enough people getting to that purchase point to begin with – and that’s a market-size / marketing problem.

If I were 23 & me, I would change up my marketing — focusing on the heritage and ancestry issues more than the disease markers.

2 Jan May 19, 2012 at 3:34 pm

I actually believe that most who go to 23 7 Me do it for exactly the last reason you mentioned above – Ancestry issues and heritage. The more people involved in any genome project for any reason the better.

3 peter blumenauer April 18, 2013 at 7:42 am

This company will be successful for one reason: Curiosity.
People like to know their ancestry, etc.. But this technology, although in it’s infancy, gives them a peek into the unknown. AND, it’s only a hundred bucks.. Accurate or not, it’s like astrology..
Just saying…

4 Kim May 9, 2013 at 8:50 pm

Thanks for the information. I am not a scientist. Not even close. After listening to a radio ad, I was interested in 23 & Me. Kind of compelling. At least the way its advertised. Besides, I’ve spent 100 bucks on worse things. But after reading your piece. Nah. I’ll save my money.
Thanks.

5 Tim Collins August 27, 2013 at 6:19 am

You briefly mentioned insurance. If 23 & Me and the like, are as important as the sellers would have us believe, is the test covered by insurance? If not, why not?, too new, not really conclusive to predict, diagnose or treat disease? Wouldn’t they have a vested interest in that? Besides, who wants to find out their related to Justin Bieber, Miley Cyrus, Flava Flave or OMG, Al Sharpton, or worse!

6 Sharon September 1, 2013 at 9:10 am

So I am thinking about this – because I have Hashimoto’s and my mother died from something (ending in kidney failure) when I was young. If there was a greater chance of me having something else related to this (I have other symptoms), I would ask my doctor to test me for this. I also have two children who may have inherited whatever this crap is. Still thinking…

7 jack lawson September 30, 2013 at 2:26 pm

just looked at my cousins profile it said .03 chanceof native american ral accurate great grandma was 100% native american

8 Jeff Smith October 11, 2013 at 11:27 am

jack lawson: maybe great grandma was 100% native America; maybe she wasn’t.

But one thing I do know 100%: there has been inter-breeding between native Americans and Europeans since virtually the day the Europeans arrived 500+ years ago…

Native Americas have also been taking European women prisoner and integrating them into their tribes since the beginning as well. Well known examples such as Cynthia Ann Parker, mother of the great Quanah Parker abound. If your grandfather was Quanah Parker, your great grandmother would be 100% (or close) European.

Further, it is rather difficult to trace native America genealogy since there are virtually no written records. Oral traditions, yes, but anyone who has ever played “telephone” knows that is hardly reliable.

9 Lulu November 26, 2013 at 10:23 am

My husband and I ordered our kits after he was diagnosed with a incurable disease, out of curiosity. The blessing in disguise came after one of my close relatives contacted me, he was adopted and was hoping we could help him find his father. We did. They have been reunited and very close now. It was a God Send. As far as the medical conditions, etc., it is fun reading. But in no way should this replace seeing your own doctor, as 23andme states.

10 Bradley Miller November 29, 2013 at 7:14 am

Cindy – thank you for the comment. For your close relative – did he find his father through 23andme? I’m super curious about those events. For your husband – I hope that you two are doing OK – did 23andme’s kit help you in any way? I wrote the post almost 4 years ago now, and while things have changed regarding genomics, I’d have to weigh in that the more we find, the less we know. That’s not a bad thing, per se, but we’ve got a long way to go until we have enough data to drive genomic discoveries. It’s an exciting time, though.

11 John Sullivan MD November 29, 2013 at 1:37 pm

This is a misleading article and demonstrates how far behind the author is in genomics and genetics. 23 and me uses the same state of the art technology and base risk guides published in the scientific peer reviewed literature. No other medical genetics testing facility performs testing differently. Not only is the FDA wrong but the review of the genetic risk issue is totally misrepresented by the blog author

12 Bradley Miller November 29, 2013 at 1:59 pm

Wow, strong words John. Clearly, SNP profiling really is cutting edge in the era of whole genome sequencing. When you start using 23andme for diagnostics, you let me know. That said, we are in an era when we should own our data and sequence, so I applaud them there. I have said that I think 23andme is doing a great job engaging and driving mainstream acceptance of genetic testing. http://www.medicinethink.com/23andme-genetic-testing-consumer-engagement/

13 charlotte ann dehgan January 21, 2014 at 3:18 pm

I did the test because I wanted to know my Alzheimer ‘s Disease risk. Both parents were diagnosed on autopsy. In addition to that, I found out I had 2 positive markers out of 4 markers for Chronic Leukemia. Since my Mom had two first cousins with this disease I felt this was important. I shared the results with my Doc and he felt that it was important enough to enter those results into my electronic medical record.

14 David March 8, 2014 at 12:24 pm

So now it’s 2014. Is it still too early for this to be accurate?

Leave a Comment

{ 2 trackbacks }

Previous post:

Next post: