By: David Shifrin, PhD Science Writer, Filament Life Science Communications
Last week, we talked about the updated ASCO policy on genetic testing in oncology. With that out of the way, I wanted to take a quick look at two issues that were only mentioned in the accompanying editorial but worth another pass.
Authors Paul Yu, Julie Vose and Daniel Hayes mention the extreme heterogeneity across human populations. This variability complicates the analysis of genetic information, making it difficult to determine with high confidence whether a specific variant is a disease risk factor.
Two additional factors compound this problem. One is ethnic diversity. Genetic variants discovered in a study cohort may indeed turn out to be relevant to disease. However, they may be unique (or relatively so) to the specific subpopulation from which the cohort was drawn. As Yu et al point out, any given variant found to be common in that group may be related to disease…or it may not. Just finding a variant means little.
Figuring out how to approach ethnic diversity as it relates to molecular genetics has long been a question. In 2004, Richard Cooper wrote a book chapter titled Critical Perspectives on Racial and Ethnic Differences in Health in Late Life. Most of the content is beyond the scope of this blog. However, Cooper says
“At stake is whether or not we can move beyond […] the generalizations built on estimation of genetic distance that have preoccupied population geneticists and anthropologists. Specifically, it is now possible to ask a set of testable questions: Can the global variation in the human genome be aggregated into subunits, and do those units correspond to the categories we call race? Can we assess the relative magnitude of shared and nonshared genetic material among population groups? Is there variation in causal genetic polymorphisms that is associated with important differences in chronic disease risk? Is it possible to conceptualize the collective human genome as a whole, and express that concept in quantitative terms?”
Certainly, recent advances in technology and ever-expanding datasets have provided some (partial) answers to Cooper’s questions over the past ten years. Even without the advent of molecular biology, disease risk could often be seen to be higher in certain populations.
Now though, the “easy” stuff has been covered (sickle cell anemia, Jakob Creutzfeldt disease), and we are left with the complex task of using bioinformatics and molecular tools to find disease-causing variants. Or, more accurately, combinations of variants that together increase disease risk.
On the other hand, as Cooper points out, the idea of “race” is in many ways archaic and doesn’t take into account the realities of genetics. Thus, it becomes a fine line between binning subpopulations on various criteria – whether genetic, geographic or environmental – to delineate relevant disease-causing factors, and creating convenient but potentially irrelevant or distracting categories.
The other issue mentioned by Yu et al is with studies where the decision to investigate an individual is based largely on family history. If history of disease is used as an inclusion criterion, the pool of participants will, tautologically, be narrowed and the results biased. “However,” says Yu, “when evaluations are applied to general patient populations that have no known family history of the cancer in question, subsequent studies often detect a less potent gene effect.”
The point then is not necessarily to avoid these early, more selective studies. It’s impractical to do so; we have to start somewhere and high-risk populations (e.g. individuals with a family history) are the logical place. Rather, these points are valuable reminders that no matter how much technology advances, inherent limitations to study design will remain based on evolution and geography. We have to be clear about these limitations and carefully temper any conclusions with the knowledge that the next study of a different cohort could look very different.