Combining gene signatures for breast cancer with clinical factors such as patient age and tumor size can improve predictions about the risk of recurrence in women with early-stage disease, new research suggests.

This strategy may also help physicians select the most appropriate chemotherapy regimens for women who undergo additional (adjuvant) therapy to prevent recurrences, the researchers report in the April 2 Journal of the American Medical Association.

Gene signatures are characteristic patterns of gene activity in cells that may reflect the underlying disease biology. A number of breast cancer signatures have been developed to predict clinical outcomes and several are being tested in trials such as TAILORx and MINDACT.

Building on this work, researchers at the Duke Institute for Genome Sciences and Policy and their colleagues have now tested the hypothesis that biological information in breast cancer gene signatures is independent of clinical risk factors and that integrating these two sources of information can improve risk assessment beyond traditional methods.

"We can predict who might respond to chemotherapy, but more than that, we can predict which type of chemotherapy may benefit a particular person," says lead investigator Dr. Anil Potti of Duke. "This is truly, in my mind, the next step in personalized medicine."

His team retrospectively analyzed nearly a thousand breast tumor samples for which there was clinical and pathological information. Using this information, they assigned nearly 600 samples to three risk groups: low, intermediate, and high.

The researchers then used gene signatures to further stratify the probability of recurrence among each risk group. The signatures include well-characterized genes involved in cell communication and other molecular pathways that are deregulated in the disease. The result was "clusters" of patients with a range of clinical outcomes, and the clusters were confirmed in 391 additional samples.

The researchers plan to validate the strategy in several prospective trials. The current results reflect the limitations of a large, retrospective study, which included tumor samples from different studies and in some cases lacked complete clinical data.

For instance, the status of the HER2 gene, which can affect prognosis, was not always known. Another limitation was the small number of patients within certain clusters, which hindered statistical comparisons.

It is only through very large studies that the true genetic diversity of breast cancer will be revealed, notes Dr. Chiang-Ching Huang of Northwestern University, who coauthored an accompanying editorial.

"Cancer is heterogeneous, and you cannot see the subtypes or the differences in the risk of relapse and responses to therapy in small numbers of cancers," says Huang. "This study shows that if we pool our efforts we can get large enough samples to see if there are real signatures with prognostic power."

With so many experimental gene signatures for breast cancer, what the field needs now are studies to assess these classifiers in concert, says Dr. Lisa Anne Carey of the University of North Carolina Lineberger Comprehensive Cancer Center, who was not involved in the research.

Such studies may further document genetic alterations in breast cancer subtypes and eventually lead to clinical tools.

"This is intriguing work that begs to be taken further," says Dr. Carey.

—Edward R. Winstead