The same genes also predicted the recurrence of prostate and lung cancer and medulloblastoma, the most common form of childhood brain cancer, according to the study published in the New England Journal of Medicine.
"It seems to work across tumors," said Dr. Michael Clarke, professor of cancer biology at Stanford University, who performed the research with scientists at the University of Michigan.
Since Clarke isolated the first cancer stem cell in a solid tumor in 2003, cancer stem cell research has held out promise
for cancer treatments. Cancer stem cells are believed to be the cells that fuel cancer's growth.
In this study, Clarke and his colleagues isolated a minority of breast cancer stem cells that were feeding tumors. They then generated a 186-gene "invasiveness" gene signature, used to measure overall aggressiveness of patients' cancer.
The test classified patients into either good or poor prognosis with notable accuracy. Of those who the test classified as having a good prognosis, 81 percent were cancer-free 10 years later, compared with 57 percent given a poor prognosis. The researchers used databases of 600 early high-risk breast cancer patients from the Netherlands.
When the gene signature was paired with a previously identified group of genes, accuracy went up. In that test, less than 50 percent of those given a poor prognosis were cancer-free after 10 years, Clarke said.
The findings were replicated when predicting outcomes for prostate cancer, lung cancer and medulloblastoma.
Eventually, a commercial gene test could help individualize cancer treatment. For instance, someone with a low probability of cancer recurrence would get less invasive surgery while another person with high probability of recurrence would get more aggressive treatment, the researchers said.
"It's not a test that is going to be available tomorrow," Clarke said. "But I think it lays the foundation to bring it to the clinic."
Over time, the researchers plan to refine the list of genes needed to home in on a patient's risk of recurrence, which could make a genetic profiling test more cost-effective.
Joan Massague, chair of the Cancer Biology and Genetics Program at Memorial Sloan-Kettering Cancer Center in New York, wrote in an accompanying editorial that the findings are "exciting."
But he cautioned that much more work needs to be done before such a test could be used in the clinic.
"Predictions of a good prognosis have been proposed as criteria for limiting the use of necessary adjunct therapy," Massague wrote. "It remains to be seen whether such criteria would prove a sufficient incentive for the widespread use of prognostic signatures."
Stanford plans to incorporate the findings into its cancer care program, said Dr. Irving Weissman, director of the Stanford Institute for Stem Cell Biology and Medicine.