23andMe was always junk science Personalised medicine can’t be trusted

About a decade ago, everyone was very excited about commercial genetic testing.

For a few quid you could get a website with some name like MyGeneHistory dot com to send you a little swab stick, which you would swizz around your gums and send back to them. They would, in turn, do some DNA sequencing on it, and purport to tell you all sorts of exciting things about your ancestors, or your disease risk, or your most suitable diet. There were loads, of which the most famous was probably 23andMe, which recently filed for bankruptcy protection.

One of the funnier ones, the now extinct BritainsDNA, would tell customers stuff about how their ancestors were Germanic tribespeople marauding across Europe and terrorising Romanised town-dwellers in the first millennium, or that they “arrived in the British Isles and Ireland around 3,000 BC”, or brought porridge (seriously) to Britain some time before the birth of Christ.

Obviously that’s all silly, because — bear with me on the maths here — you had two parents, four grandparents, and so on. Assuming 30-year generations (probably too long, but let’s be conservative), 1,000 years ago you would have had around 800 million great^33 grandparents. Since the world population was probably about 300 million at the time, that’s unlikely.

What it actually means is that your ancestors interbred. And that, in turn, means that everyone who was alive in Europe about a thousand years ago was either the ancestor of every European person alive or of no one. If you are a white European, your ancestors, past a certain point, are the same as every other white European’s ancestors.

The science of BritainsDNA wasn’t completely made up. The technology they used — looking at mitochondrial and Y-chromosome DNA, passed down through the female and male lines respectively — really has led to some important scientific breakthroughs, particularly in confirming that humans came out of Africa about 200,000 years ago. But it couldn’t tell you anything very interesting about your ancestors specifically. It made its money (briefly: it shut down in 2017) by taking the real and interesting population-level science and claiming it could say meaningful things about individuals.

This is a pretty standard trick — vitamin companies have done it for years. They note that some mineral or chemical is important in some bodily process, like saying vitamin B helps your body process energy — which is true — and then use that to imply that taking their vitamin B supplements will boost your energy, when in most cases your body has all the vitamin B it needs and will just piss out any extra that you give it. Commercial genetics companies, often, simply follow the same playbook.

23andMe promised a future of precision-guided, individualised medicine, in which a simple genetic test would tell you your risk of various diseases. It was not as overtly silly in its claims as BritainsDNA, but even early on it had problems. In 2013 the US Food and Drug Administration said its scientific claims were unevidenced and ordered it to take down its health test products: it relaunched them two years later for a more limited range of diseases, but it still faced two key issues.

One was that the last two decades have seen the great hopes of personalised genetic medicine not fade away exactly but become more complicated. The framework that dominated genetic research was called the “candidate gene” model: scientists looked for “the gene for” diabetes, or “the gene for” obesity, or schizophrenia, or alcoholism, or heart disease.

But the rise of genome-wide association studies — research which looks at the entirety of human DNA and uses big data techniques to see the correlation between certain genes and phenotypic outcomes — revealed candidate genes, “genes for” things, to be vanishingly rare. They do exist — Huntington’s disease and cystic fibrosis are caused by single-gene mutations — but most health conditions are caused by the interaction of hundreds or even thousands of genes spread across the genome, each of which has only a tiny effect on your risk of depression or gout or whatever. It is not especially useful to know that you have a 3.8% lifetime risk of Parkinson’s rather than 2.7%.

And the other problem is that often the information is of limited use. If you learn you have a raised risk of high blood pressure or melanoma, the advice will be the same as for everyone else: take exercise, wear sunscreen. 23andMe says it could lead to earlier screening and detection, and perhaps there’s something in that, but cancer screening is itself of ambiguous benefit. (To be fair, it is of more use if applied to higher-risk populations.) I don’t think it’s too harsh to say that a lot of the time, these “polygenic risk scores” of diseases will simply scare people by telling them of largely irrelevant increases in their risk which they can do very little about.

It’s easy, perhaps, for me to say that these genetic medicine companies were overhyped. Perhaps a more interesting exercise is predicting which current healthcare fad won’t survive scrutiny.

Diet is always a safe bet. At least one of the commercial genetics companies has had a go at it: DnaNudge, which I wrote about back in 2020, tests customers’ DNA and claims to tell you which foods you should avoid. A geneticist at the time pointed out to me that this was close to being “snake oil”. Again, the genetic variants involved might tell you some statistically significant population-level thing, like 125 out of every 100,000 people with a particular gene might have wheat intolerance, instead of 103 in the background population. But it’s of very little use to individuals. (That company got given £161 million by the government for COVID testing, incidentally.)

These days people are very excited about the ZOE app, which again came to prominence in the pandemic. Users’ blood glucose is continuously monitored by a personalised health widget as they eat different foods, and — as Deborah Cohen and Margaret McCartney wrote in UnHerd two years ago — they also have to poo into a “hammock” over the loo bowl and post the resulting sample off for testing to tell you about your “gut microbiome”, the collection of bacteria and other organisms in your intestine. Then ZOE provides a personalised nutritional programme of foods to eat or avoid, which it says can help you lose weight, sleep better, avoid health risks, and so on.

But again this is pretty suspect. For one thing, while there are some extraordinary claims about the gut microbiome and its links to various health conditions — people have said it causes Alzheimer’s, autism, depression, and a hundred other things — the evidence is very thin. The long and short of it is that the research is plagued by small sample sizes, bad statistical techniques, and overextrapolation from tiny animal studies. It seems pretty unlikely that you can give an individual any useful information.

And as Cohen and McCartney pointed out in 2023, while continuous glucose monitoring is important in diabetics whose bodies can’t control their blood sugar levels, healthy people can, and it’s not clear that a short-term glucose spike brought on by eating a banana is particularly informative. And the same food seemed to have different effects in the same person at different times, or even when monitored by two different devices at the same time, suggesting that the tests are unreliable.

So how about longevity? Anti-ageing science is undoubtedly promising. There are hints that a particular chemical change to one of the four “letters” of our DNA tracks, and possibly influences, our biological age; and there are encouraging therapies which could reverse that chemical change and which, just possibly, could slow or stop aging. Other interesting avenues of research include telomeres, the tips of our chromosomes, which grow shorter as we age, allowing the chromosomes to fray. Reversing that process could influence aging. There are strong hints that severe calorie restriction can extend life in mice.

But the science here is uncertain and complicated. Surely no one would try to sell it yet? Well, obviously they do. The tech entrepreneur Bryan Johnson has for the last several years has been on an extreme anti-ageing regime called Project Blueprint: he claims now to be “the healthiest person alive” (“If health were an Olympic sport, I’d be the gold medalist”). And that’s great! I hope it works out for him and that he lives a long time. I also hope that longevity science improves, because ageing has many very unpleasant side effects (ask my knees) and spending more time on the planet with your loved ones is good.

What Johnson also does, though, is sell “Longevity Mix” supplements on his website for £55 a go, or you could treat yourself to the “Blueprint Metabolic Stack”, which “streamlines your wellness journey with a simple format comprising a supplement drink, protein meal, seven daily pills, and premium Extra Virgin Olive Oil”. A mere £386 for a month’s supply, and it comes with a free hoodie!

Johnson is not alone. As The Wall Street Journal noted recently, anti-ageing is big business. Not just anti-wrinkle face creams and things, although they’re silly enough, I suspect: actual medicines which purport to slow bodily ageing. One scientist locates the ageing process in a family of proteins called sirtuins, which are involved in metabolism, DNA repair, inflammation, and the function of your metabolism and mitochondria. There was some evidence that boosting the expression of the gene that creates sirtuins can indeed extend life — in yeast, and possibly worms and flies. But the pharma industry analyst Derek Lowe argued that even that research is hopeless, “a steaming pile of artifactual results, overinterpretation, and publication bias”. And yet you can buy sirtuin supplements for $720 for an annual subscription.

My fairly confident prediction is that people who buy these supplements will not live significantly longer, on average, than other, similarly health-conscious members of their social class. Come back and ask me in 60 years.

23andMe was a product of its time, a period in which gene sequencing was the exciting science of the moment. But it was also somehow timeless: people have always sought to turn cutting-edge science into snake oil, at least as far back as 1918, when the discovery of radiation led to radium-infused “toothpaste, hair cream, cosmetics and even suppositories”. This overhyping of science and selling it for a monthly subscription fee is a business model destined to outlive even Bryan Johnson.

---

Tom Chivers is a science writer. His second book, How to Read Numbers, is out now.

TomChivers