How Precision Medicine Can Save More Lives and Waste Less Money (Part 1 of 2)

We all have by now seen the hype around the Obama Administration’s high-profile Precision Medicine Initiative and the related Cancer Moonshot, both of which plan to cull behavioral and genomic data on huge numbers of people in a secure manner for health research. Major companies have rushed to take advantage of the funds and spotlight what these initiatives offer. I think they’re a good idea so long as they focus on behavioral and environmental factors. (Scandalously, the Moonshot avoids environmental factors, which are probably the strongest contributors to cancer) . What I see is an unadvised over-emphasis on the genetic aspect of health analytics. This can be seen in announcements health IT vendors, incubators, and the trade press.

I can see why the big analytics firms are excited about increasing the health care field’s reliance on genomics: that’s where the big bucks are. Sequencing (especially full sequencing) is still expensive, despite dramatic cost reductions over the past decade. And after sequencing, analysis requires highly specialized expertise that relatively few firms possess. I wouldn’t say that genomics is the F-35 of health care, but is definitely an expensive path to our ultimate goals: reducing the incidence of disease and improving life quality.

Genomics offer incredible promise, but we’re still waiting to see just how it will help us. The problems that testing turns up, such as Huntington’s, usually lack solutions. One study states, “Despite the success of genome-wide association and whole-exome and whole-genome sequencing (WES/WGS) studies in revealing the DNA variants that underlie the genetic basis of disease, the development of effective treatments for most diseases has remained a challenge.” Another says, “Despite much progress in defining the genetic basis of asthma and atopy [predisposition to getting asthma] in the last decade, further research is required.”

When we think about the value of knowing a gene or a genetic deviation, we are asking: “How much does this help predict the likelihood that I’ll get the disease, or that a particular treatment will work on me?” The most impressive “yes” is probably in this regard to the famous BRCA1 and BRCA2 genes. If you are unlucky enough to have certain mutations of these gene, you have a 70% lifetime risk for developing breast or ovarian cancer. This is why testing for the gene is so popular (as well as contentious from an intellectual property standpoint), and why so may women act on the results.

However–this is my key point–only a small percentage of women who get these cancers have these genetic mutations. Most are not helped by testing for the genes, and a negative result on such a test gives them only a slight extra feeling of relief that they might not get cancer. Still, because the incidence of cancer is so high among the unfortunate women with the mutations, testing is worthwhile. Most of the time, though, testing is not worth much, because the genetic component of the disease is small in relation to lifestyle choices, environmental factors, or other things we might know nothing about.

So, although it’s hard enough already to say with any assurance that a particular gene or combination of genes is associated with a disease, it’s even harder to say that testing will make a big difference. Maybe, as with breast or ovarian cancer, a lot of people will get the disease for reasons unrelated to the gene.

In short, several factors go into determining the value of testing: how often a positive test guarantees a result, how often a negative test guarantees a result, how common the disease is, and more. Is there some way to wrap all these factors up into a single number? Yes, there is: it’s called the odds ratio. The higher an odds ratio, the more helpful (using all the criteria I mentioned) an association is between gene and disease, or gene and treatment. For instance, one study found that certain genes have a significant association with asthma. But the odds ratios were modest: 3.203 and 5.328. One would want something an order of magnitude higher to show running a test for the genes would have a really strong value.

This reality check can explain why doctors don’t tend to recommend genetic testing. Many sense that the tests can’t help or aren’t good at predicting most things.

The next section of this article will turn to behavioral and environmental factors.

About the author

Andy Oram

Andy Oram

Andy Oram writes and edits documents about many aspects of computing, ranging in size from blog postings to full-length books. Topics cover a wide range of computer technologies: data science and machine learning, programming languages, Web performance, Internet of Things, databases, free and open source software, and more. My editorial output at O'Reilly Media included the first books ever published commercially in the United States on Linux, the 2001 title Peer-to-Peer (frequently cited in connection with those technologies), and the 2007 title Beautiful Code. He is a regular correspondent on health IT and health policy for He also contributes to other publications about policy issues related to the Internet and about trends affecting technical innovation and its effects on society. Print publications where his work has appeared include The Economist, Communications of the ACM, Copyright World, the Journal of Information Technology & Politics, Vanguardia Dossier, and Internet Law and Business.