Forecasting the path of breast cancer in a patient

Chances of survival depend on which organs breast cancer tumors colonize first

Contact: Zen Vuong at (213) 300-1381 or zvuong@usc.edu

USC researchers have developed a mathematical model to forecast metastatic breast cancer survival rates using techniques usually reserved for weather prediction, financial forecasting and surfing the Web.

For decades, medical schools have taught doctors that the best way to treat cancer and metastatic progression is to memorize a list of tumors and their typical migration patterns. Metastasis is the development of malignant tumor growths elsewhere from the primary site of cancer.

“This is akin to back in the days when weather reporting depended solely on a barometer and experience,” said Jorge Nieva, an associate professor of clinical medicine at the Keck School of Medicine of USC and co-author of a new study. “Medical students are taught very fundamental cancer progression patterns. What the modeling does is it brings the sort of complexity of modern-day weather forecasting to trying to understand where tumors go, when they go and how they get to that location. This type of mathematical modeling is wholeheartedly different from what most medical students learn today.”

The study, published online Oct. 21 in the journal npj Breast Cancer, a Nature Partner Journal, looked at 25 years of data regarding 446 breast cancer patients at Memorial Sloan Kettering Cancer Center. It focused on a subgroup of women who were diagnosed with localized disease but later relapsed with metastatic disease.

The model shows that cancer metastasis is neither random nor unpredictable. Survival depends significantly on the location of the first metastatic site or “spatiotemporal patterns.” In other words, USC researchers uncovered a framework to explain how tumor cells circulate through a patient’s bloodstream over time to settle in various organs. The path that varies depending on tumor makeup and treatment decisions.

“There’s nothing like this in the cancer world; there’s nothing really like this in the disease progression community even though the techniques are well-developed in other contexts,” said Paul Newton, lead author of the study and an aerospace and mechanical engineering professor in the USC Viterbi School of Engineering. “Our long-term goal is to build comprehensive predictive computational simulations of metastatic cancer. Ultimately what we want to do is tailor those models to individual patients using their individual characteristics.”

The concept

The framework USC researchers built combines scattered data points doctors are already collecting in order to produce an understandable, comprehensive cancer map. The system design is comparable to information Google collects to predict Web surfing patterns and to determine PageRank.

“If somebody is reading about breast cancer on Wikipedia, the likelihood that she is going to jump to a lung cancer page or a bone cancer page is much higher than the likelihood of her jumping to the Costco website,” said Newton, who is also a professor at the Norris Comprehensive Cancer Center in the Keck School of Medicine of USC as well as professor of mathematics. “These probabilities of jumping from one page to another are not all equal. Where you jump to next depends strongly on where you currently are. This observation lies at the heart of our model.”