A study published yesterday in Science Translational Medicine from a group at Johns Hopkins University set out to determine the best-case power of genetics to provide clinically meaningful information about risk of common diseases. The authors did this using data from identical twin pairs: by seeing how often twins with identical genomes develop the same diseases, we can judge the extent to which those diseases are determined by genetics, and the extent to which they might be predictable if we completely understood their genetics.
The study’s key findings were that most people could obtain a result that they are at significantly elevated risk of at least one disease. But for most diseases, they would learn that their genetic risk places them not far from the average risk across the population. In a few cases, the authors found that a genetic test could potentially identify most individuals who eventually will develop a disease: this was true for thyroid autoimmune disease, type 1 diabetes, Alzheimer’s disease, and coronary heart disease in men. The authors conclude that genetic testing will not replace conventional preventive medicine.
We agree with the authors that it is important to set reasonable expectations for what genetic testing can and cannot do. We think the positive finding that genetic testing can have some clinical utility for risk assessment for common disease is encouraging. We also think it is very encouraging that there are a few diseases where genetic testing could be particularly powerful — Alzheimer’s disease and coronary heart disease are not small potatoes and early alerts of increased risk of these conditions could have substantial public health benefits.
Many of our more than 125,000 23andMe customers receive a risk report for a common disease that indicates that they are at substantially increased or decreased risk, using criteria similar to the ones used in this study. In addition to disease risks, 23andMe customers also receive information about their carrier status for inherited diseases and possible drug responses based on genetics, both of which already make an impact in personalized medicine today.
On a more technical note, the study’s finding that negative genetic test results are mostly uninformative is largely an artifact of the authors’ mathematical modeling procedure and the fact that risk predictions for each disease were binned into just two buckets for “negative” and “positive” results. The authors pre-specify a baseline risk that for most diseases is not much smaller than the population average risk, and disease risk associated with different genotype classes is constrained to fall between the baseline value and 100%. Thus, it is impossible for anyone to have disease risk much less than the population average, and most individuals with “negative” results are pegged at the baseline risk level. Models that place some individuals at substantially lower risk would also be consistent with the observed twin data, but this study is not effectively exploring the potential utility of these negative findings.
Do we believe that genetic testing could ever substitute for conventional preventive medicine? No, and we don’t think it should. Instead, it is more useful to think about genetic testing as one of many sources of information — along with family history, lifestyle, and conventional clinical testing — that can inform disease risk assessment and preventive medicine. Genetics will not be equally informative for everyone, and will not replace these other approaches, but this is actually quite compatible with personalized medicine — that we should find and leverage the information that has the biggest impact on care at an individual level.
