Direct-to-consumer DNA testing company 23andMe faced another round of criticism last week—this time the scrutiny was familiar. The New York Times ran a story by a writer who submitted her genetic material to several testing companies, including Anne Wojcicki's Google-funded service, and got wildly varying results.
Back in 2010 government officials from the Government Accountability Office pulled the same routine. Their stinging report was brought before a congressional committee questioning the accuracy and legitimacy of genetics testing companies like 23andMe who operated without FDA approval. The GAO report revealed that one DNA sample was sent to 23andMe and three other genetics testing companies and each provided a different assessment of the fictitious consumer’s risk for prostate cancer.
"We all tested for slightly different markers so we included seven genetic data points and other people included like 10 genetic data points," Wojcicki explained to me over the summer when I asked about accuracy. "Because we were looking at different data we got different numbers. But there was nothing wrong with the accuracy."
Independent genetic counselor Laura Hercher added, "I think on science, 23andMe is very strong. It's accurate. But my exception is, okay ... take an example like diabetes. They say you have this and this genetic variant, and therefore your risk of diabetes is increased 10% over the general population." But a 10% risk over the general population is useless, Hercher says, since factors such as lifestyle, diet, exercise, and weight play a much higher relative role. "It's like if you had a hundred stocks, and you looked at three of them and said these three went up 10%, so you're up 10% for the year. Well, yeah, except for the other 97 stocks."
But Wojcicki agreed that our understanding of genetics is rudimentary and growing. Unless one gets her entire DNA sequenced—a costly test that looks at all 3 billion of an individual’s nucleotides—any interpretation of results is somewhat theoretical. Genomic testing like 23andMe examines but a fraction of our DNA segments. "Nobody can quantify for you what’s the impact of eating fiber every day, for instance," says Wojcicki. "We can say we think it’s good. But some people might say ‘Oh, it reduces your risk of colon cancer by 20%, some people might say it reduces your risk by 25%.’ We just don’t know yet. We’re still getting data to understand exactly what the risk means for certain genetic variants in your DNA. For something like prostrate cancer, understanding exactly what each A, C, G, and T means is still being understood." (It should be noted that 23andMe does test for more gene-predictive conditions such as Parkinson’s, BRCA1 and 2 mutations, and Alzheimer’s as well.)
What eats at Wojcicki, though, is any public skepticism that wants to write off her company as practicing bunk science. But she says the results they come up with are 99.9% reproducible. "There’s almost no other tests that are out there that are going to give you that kind of reproducibility. There’s still a lot of work to do on all the interpretation but that’s part of what’s evolving. And that’s part of the reason why we’re a different kind of test, why we keep in touch with our customers, and we update you every time we know something new. Because we realize that the data is evolving."
What now remains to be seen is how Wojcicki, who didn’t comment for this latest piece, will weather recent blows to 23andMe’s public image. Or how in 2014 23andMe’s mission and business model will have to evolve to keep growing, let alone surviving. Most importantly: Do consumers see the kind of results 23andMe offers as definitive enough to trust?