How Useful is a Report from a Company Like 23andme to Your Health? The Promise and Peril of Personal Genomics
How do we separate hype from reality as personal genomics companies ramp up ads, social media, and celebrity influence campaigns that directly target consumers? In February, we sat down with two University of Pittsburgh experts—Jeremy Berg, Associate Senior Vice Chancellor for Science Strategy and Planning, Health Sciences, and Mylynda Massart, Assistant Professor of Family Medicine—to discuss how the heavy consumer pitch can cloud medical practice, science, understanding, and the road ahead for medical education as personal genomics becomes increasingly relevant in the clinic. The discussion was taped live from the Sci-Mic Podcast Stage at the 2019 meeting for the American Association for the Advancement of Science in Washington, D.C. The following has been edited for brevity and clarity. For the full discussion, listen to the Pitt Medcast episode here:
Today, millions of people are raising their hands for genetics testing through direct-to-consumer personal genomics companies like 23andMe. And even though there are disclaimers that the reports they’re getting back are not for diagnostic or medical decision-making purposes, for many consumers the reports raise questions that you, Mylynda and Jeremy, would find familiar given your expertise in the emerging field known as precision medicine, also known as personalized medicine.
Let’s start there. What is precision medicine? What’s your elevator speech for laypeople?
Jeremy Berg: Precision medicine is using genomic information in addition to other factors to try to customize diagnosis and potentially treatment of a wide range of diseases.
Mylynda Massart: When I explain precision medicine to my patients or to my family, it’s really about taking all of the factors that lead up to our individual health. So, looking at our genetics and all of our environmental factors, such as our nutrition, our exposures in our local environments, even our traumas.
There are some compelling stories and videos out there from these companies: origin stories, reunions for long lost family members. There are also stories along the lines of, for example, a woman who didn’t know about her Ashkenazi Jewish heritage, and the mail-order test results told her that she had an 80 percent chance of developing a deadly cancer. So she followed up with her doctor, had preventative surgery, and, lo and behold, every woman in the family has this genetic variant, and this is life changing, life saving.
But what kind of information are we—the consumers—really getting from these companies, and what do we need to be thinking about as we try to process this information?
JB: In terms of the technology, most of the direct-to-consumer tests are not whole-genome sequencing, but are a sampling of known variations across the human genome. So it’s a very incomplete picture of your whole genetic background.
But more important than that is: All of this is really complicated. In some cases, there are variations that are so-called highly penetrant, [meaning they] are very predictive. If you have this [gene] variant, then you are certain or very likely to have a particular disease. The archetype of that is sickle cell anemia. You have two copies of each gene. If you have two copies [of a particular variation in a hemoglobin gene], you have sickle cell disease; if you have one copy, you have sickle cell trait.
But for most other diseases or other traits, things are much more complicated, where there are lots of variations, and they all contribute a little bit [in addition to] other factors like environment and exposures and experience.
And that complexity tends to get lost in the sort of cartoon version of the reports that come back. And one concern that I have is that consumers take it much more seriously or uncritically than they should for their own benefit.
MM: I think that when the direct-to-consumer testing market first opened up, it was really fun, and that’s really what the key term was. It was engaging, the general society was learning more about genetics and genetics terminology, and they were having fun doing it. Now, the level of results are increasing in clinical utility. There are results coming out now on BRCA mutations [for breast cancer], genes that may contribute to risk of Alzheimer’s disease or Parkinson’s disease, and a lot about pharmacogenomics—how genetics affect the metabolism of everyday medications (which can result in adverse reactions, or toxicity levels, or drugs simply not working or being therapeutically effective at all).
And to me, the part about that that’s so important is that while it’s still fun and interesting and engaging, these results now need to be brought into the clinical world and interpreted from a clinical perspective. And the doctors out there really need to recognize the limitations of these tests: How much is actually accurate? What is left that perhaps didn’t get tested? And what we call residual risk. How do you explain to a patient and the general population the limits of the capacity of this testing to really be informative?
And then on the flip side, a lot of these tests are done in [approved] labs and are probably very accurate. But the FDA has passed some pretty strict regulations that these tests, to be used clinically, need to be repeated and confirmed [in order to determine how to apply a result to] an individual’s care plan.
Regarding these reports about “risk,” can you clear up that term?
JB: One very fundamental aspect is the difference between absolute risk and relative risk. So if you have a test result, and it says you have a 10 times higher risk than that of the general population based on your genetic variation, it depends what the baseline risk is. If the absolute risk for the population is 1 percent, and you have a tenfold higher risk, it means you have a 10 percent absolute risk, which may be worrisome or not, depending on what the condition is and so on.
But tenfold sounds really scary! Ten percent, not so much.
The other [aspect is that] the uncertainty in the models that leads to these risk predictions is pretty substantial. The models are based on some outrageously gross simplification of everything, because that’s the best we can do. Conveying the uncertainty in the risk is something that I think really gets lost in the shuffle.
MM: Even with BRCA, having one copy of that mutation does not guarantee that someone will ever develop breast cancer in their lifetime or ovarian cancer in their lifetime. It simply is, again, about risk, and having a general-population understanding of risk and where these risks fit in, what they mean, and what are the preventative interventions that could be put in place to reduce their risk.
JB: The tests are never going to be deterministic. Genetic determinism—meaning, if you knew your genome sequence, and we understood everything, you could predict your whole future life—is just absolutely false. And the studies that [disprove genetic determinism] go back decades and decades, looking at identical twins, where their genomes, for all intents and purposes, are identical. Some traits are closely shared, but a lot of other traits, including getting different diseases, aren’t.
MM: Right. And I think that’s where this greater concept of precision medicine comes into play. Taking the genetics component and recognizing what it does contribute as well as its limitations. And then really starting to understand, and using computer-assisted technology for all the other types of exposures and data and family histories, and then being able to better refine that predictive model. Even so, it’s unclear whether that will ever reach 100 percent, but it will increase in its predictive aspect as technology develops.
What is needed to bring us further along in precision medicine and bring all of its promise into reality?
JB: Well, one of the obvious things is more data, and Mylynda can talk more about that. Then, the challenge is still substantial—to do the computer analysis to try to develop better risk models.
A lot of genetic data originally were focused on people who had reason to believe that they have a risk for a particular disease. [For example], if you test people who think they have a [higher] risk for breast cancer because they have a family history of breast cancer, and you identify a gene. So you’ll identify particular variations in that population of patients. [But] then the question is: What’s the prevalence of that same genetic variation in a general population that doesn’t have any risk for breast cancer? And that’s the sort of thing that can now be done with these larger studies.
MM: The All of Us study is about putting together that million-person research cohort to gather a vast amount of data, to bring it together and organize it in a way that researchers can finally start looking at all those different layers that contribute to precision medicine. And there were a lot of things that had to line up to finally be able to do a study of this size. We had to have the computing technology, the ability to store [and analyze] all that data. And the cost of these analyses needed to come down. Now, someone can have their genome sequenced for $500 to $1,000. All of that had to line up to create this large project that’s being funded through the National Institutes of Health.
Also, one of the large emphases of the All of Us study is to collect that data from a very diverse population. We’re trying to understand the components that lead to strong health and to longevity, as well as the components that lead to disease and illness. We need to look at a vast population that’s very diverse to apply that to unique communities and have something that’s meaningful.
Photo of TVs: Getty Images. Portraits: Jeremy Berg By Joshua Franzos, Mylynda Massart By John Altdorfer. Photo Illustration: Elena Gialamas Cerri