Genomic Profiling Takes Another Step Toward the Clinic
Nevins, Ginsburg and Olson Discuss Progress and Challenge
This article originally appeared in GenomeLife, Issue 11.
As highlighted last year in GenomeLife, a team of Duke researchers has been using microarrays (thousands of DNA molecules arrayed on glass wafers) to assess the gene expression of tumors in order to predict breast cancer patients' risks of recurrence and responses to chemotherapy. More recently, the team (headed by IGSP Center for Applied Genomics & Technology Director Joe Nevins and Mike West, the Arts & Sciences Professor of Statistics) has incorporated additional information into its models, which leads to better predictions. The Duke group now uses six to eight expression patterns totaling hundreds of genes and incorporates all available clinical data into its statistical models. The advantage of such so-called clinico-genomic profiling for oncologists, says Nevins, is that they will no longer have to decide between relying only upon clinical acumen or genome science when deciding how best to treat their patients. "The reality is we don't have to choose. Our approach uses both types of information."
Center for Genomic Medicine Director Geoff Ginsburg hails this development. "What is unique about this work is the marriage of more traditional clinical predictors of outcome with gene expression data from the tumors themselves. The work of others in this field has predominantly kept these two important pieces of prognostic information separate. Joe and Mike's work is at the leading edge of a tsunami of information that will assist clinicians and patients in making more informed decisions about their care."
Science versus Art
Among those hoping to put the work of Nevins, West and their colleagues into practice is Duke surgical oncologist John Olson. Olson, who in March received a Specialized Program of Research Excellence grant to study the improvement of genome-based predictions in breast cancer, sees genomic medicine for the treatment of cancer as inevitable. However, he is the first to admit that we are not there yet; in his view, the obstacles have less to do with assimilating genome science and more to do with changing the art of medicine.
"We need to plan for success," he says, noting that an efficient mechanism for obtaining tumor material for genomic analysis will need to be put in place, as will a systematic approach to educating physicians. "It's a matter of changing the way we think about things and adjusting our practice of medicine accordingly."
Communication Breakdown?
And then there are the patients. Some health care economists have speculated that the more precise risk estimates afforded by clinico-genomic profiling will not be meaningful to many, if not most patients. Thus, if a breast cancer survivor is told that her risk of recurrence is forty percent, will that number be useful to her in making treatment decisions? Olson says that getting patients to understand probabilistic information has always been difficult. "The issue of proper communication with patients deserves at least as much attention as figuring out which genes to use in making predictions."
The key, he says, will be in not throwing out the baby with the bathwater. "The challenge is to reduce the complexity of the information while not discarding its value. If you try to distill everything down to a single number, you risk losing the value of some of the information."
Which brings Olson back to the art of medicine. For him, whatever information he has at hand and whatever choices he can offer his patients must be placed in the context of their daily lives. "How do people make choices about things? Helping patients decide is really an art. How do I communicate all I know in an unbiased way that will help them make a decision? That's hard. We may say to someone, 'This cancer has a 70 percent chance of coming back.' But for that individual it's not 70 percent, it's yes or no. It's 100 percent or zero percent."
Ginsburg says that time and concerted educational efforts will be critical. "Fifteen years ago, few patients understood the meaning of serum cholesterol," he says. "Now even my mother knows hers and with that information is making more informed lifestyle choices. Disseminating that knowledge is incumbent on us as practitioners, but it does take time. One of our goals for the Center for Genomic Medicine is to develop the right studies and educational strategies that will make this information more user-friendly."
Looking Ahead
Using genomic and clinical data together to refine predictions will not be the end of the story. Nevins is energized by the prospect of using the genomic medicine approach to incorporate still other kinds of information. "You can imagine three years from now we'll see protein profiles and genetic variation data that will help predict outcomes in breast cancer. We don't want to be in a situation where we have to choose between data sources. If each of those assays adds prognostic value, then we want to use them together."
In the meantime, the Duke team is trying to apply its current predictive models in a variety of patient populations. Assuming those test runs succeed, the next step will be to move the approach from a research basis to limited clinical applications, for example, to use genomic information to help guide decision-making in selected difficult cases.
Despite acknowledging challenges ahead, Nevins, Ginsburg and Olson are keen to bring as much relevant data into the examining room as possible. Their ultimate goal is to eliminate unnecessary treatments while making sure that those who actually need aggressive therapies are certain to get them. "The aim," says Olson, "is to provide patients all of what they need and none of what they don't."