Football fans who check popular apps like ESPN for updates on games in progress are accustomed to seeing, in addition to the current score, a number indicating “win probability.” By dynamically crunching data based on a set of constantly changing conditions—down and distance, time remaining in the game—and comparing those scenarios with outcomes in thousands of other games, statisticians developed a means of predicting which team is likely to prevail.
Now, researchers at the School of Medicine have borrowed this concept to develop a predictive model that offers cancer patients more insight into their chances of recovery and may improve treatment decisions. Coined CIRI, for Continuous Individualized Risk Index, the algorithm aggregates all data available as treatment proceeds rather than relying on data at one particular moment to determine prognosis. For example, CIRI might assess not only how treatments have affected the size of a tumor, but also levels of cancer DNA in the patient’s blood over time and how these compare with other patients’ levels at similar stages of therapy. These and other data sets offer a way to continually update a patient’s probable outcomes and give doctors better information when considering potential treatments. “It might tell us, ‘You’re going down the wrong path with this therapy, and this other therapy might be better,’ ” says Stanford oncologist Ash Alizadeh, MD ’98, PhD ’03. “Now we have a mathematical model that might help us identify subsets of patients who are unlikely to do well with standard treatments.” Alizadeh, an associate professor of medicine, led the research team along with associate professor of radiation oncology Maximilian Diehn, MD ’01, PhD ’04.
Research so far indicates that CIRI outperformed standard prognostic predictions in breast cancer and two blood cancers. CIRI may also improve patients’ confidence in managing their diseases. Should they schedule a vacation for next year or focus on putting their affairs in order? “We are trying to come up with a better way to predict at any point during a patient’s course of treatment what their outcome is likely to be,” says Alizadeh.
Kevin Cool is the executive editor of Stanford. Email him at email@example.com.