Electronic Genetics Computer-assisted risk assessment

by David Orenstein free-lance writer

DCMS member David M. Euhus, MD, associate professor of surgical oncology at University of Texas Southwestern Medical Center at Dallas, fell in love with an Apple Macintosh computer in 1985. The ensuing marriage of his expertise in oncology and passion for programming has produced important new tools for breast cancer counselors and their patients.

A study in the Journal of the National Cancer Institute (2002;94:844-851) describes how Dr Euhus led a team of researchers in determining that a widely used computer model for predicting breast cancer gene mutations is as effective as some of the best genetic counselors. The findings confirm that the software, called BRCAPRO, provides help for genetic counselors who must advise patients about their risk for breast or ovarian cancer risk. “Our message is that the software is a nice adjunct for the experienced genetic counselor and a reasonable mentor for the inexperienced counselor,” Dr Euhus says.

Dr Euhus’ interest in using computers to help fight cancer extends beyond the study. In 1997, using self-taught programming skills, he wrote an interface called CancerGene that makes BRCAPRO and other cancer risk-assessment programs much easier to use. It puts a friendly graphical face on the otherwise arcane, text-based software. BRCAPRO uses family medical history and other factors to make a statistical prediction of whether a patient is likely to have a genetic mutation linked to breast or ovarian cancer. Counselors using CancerGene can draw a patient’s medical history like a family tree. A license to use CancerGene is free, and some 1000 clinics in 40 countries use it.

Dr Euhus became interested in using computer models to predict cancer risk when UT Southwestern hired him in 1996. He wanted to carve a niche for himself. Research in cancer genetics was still emerging, as was the theme of applying sophisticated computation to understanding human genetics. He became attracted to the idea of putting statistical models for cancer risk prediction to better use through computers. The task was to convert the mathematical models into working software.

He did this for a model called the Gail model (which was the subject of another Euhus-led study published in the same journal, J Natl Cancer Inst. 2002;94:858-860), but when he saw BRCAPRO, he knew he had to adapt his programs to that, too. “It was an epiphany,” he says. “There was a beautiful model that accounted for every relative in the family.”

Now he is participating in an NCI study that will compare the cancer risk predictions of 10 models. He is developing software and a database for the 3000-family trial, and is waiting for legal clearance to produce a version of CancerGene and BRCAPRO that could run on the Web. Some Belgian physicians have requested a Web version so the tool can be used by genetic counselors there. Although this generally is a good idea, one must be careful if putting complex medical risk assessment software on the Web because of the chance that the general public might misunderstand it.

Computers have become indispensable tools in medical research and practice, Dr Euhus says. “I don’t think we can move ahead without computers. As we start looking at genes and signaling pathways, you don’t have to look far before things get impossibly complex.” The challenge is not in feeding the computers enough data—there is plenty—but in coming up with the analytical software tools that can make the data understandable.

David Orenstein is a technology and business writer in Silicon Valley. To learn more about a technology topic in Computing Care, e-mail him at davealli@attbi.com.