Improving Survival Predictions From Adjuvant Therapy

March 14, 2011
Lynne Peeples

The growing popularity of personalized cancer care has increased interest in tools that can guide physicians to the best-tailored therapies for their patients.

The growing popularity of personalized cancer care has increased interest in tools that can guide physicians to the best-tailored therapies for their patients.

In his talk at the Miami Breast Cancer Conference, William F. Symmans, MD, professor of pathology at the University of Texas MD Anderson Cancer Center in Houston described some approaches to this complex task of predicting survival from adjuvant treatment.

Until now, prognostic tests have offered little more predictive power than basic pathologic stage or tumor phenotype, noted Dr. Symmans.

"I think we're at a crossing point," he said. "We need to reconsider the idea that one assay-or one composite result-is going to tell us everything we need to know about each and every treatment decision, as well as natural history."

"Rather, we should move to multiplex platforms," he added.

To that end, one potential strategy is to compile a series of predicted outcomes based on different combinations of treatments. Assembling results from various clinical trials could, in effect, create a menu of prognostic tests from which a patient and doctor might choose, suggested Dr. Symmans.

The popular recurrence score, used to identify patients that could benefit from adjuvant chemotherapy, is based on this idea. But it has limitations.

For example, a recent trial (Southwest Oncology Group 8814) found that patients who received anthracycline-based chemotherapy followed by tamoxifen had similar survival probability in all recurrence score groups.

It may be that patients with low recurrence scores benefited from the endocrine therapy and not the chemotherapy, while the reverse may have been true for patients with high scores, explained Dr. Symmans.

A more modular approach to tailoring regimens might also work, with a treatment-specific predictive model built from individual components, such as sensitivity to chemotherapy, 70-gene prognostic score, and nodal status. This could be designed to incorporate concurrent or sequential therapies.

Dr. Symmans' team recently investigated one potential component: the sensitivity to endocrine therapy (SET) index, a measure of genes related to endocrine receptors (doi: 10.1200/JCO.2010.28.4273). While this index was not prognostic for patients with node-negative and endocrine receptor-positive cancer who did not receive adjuvant treatment, it did predict survival for patients on tamoxifen with node-positive disease. The measure offered little predictive power for node-positive disease.

"The extent of disease matters, even within biologic subgroups," Dr. Symmans noted.

In the end, different approaches may yield different insights, he said. Putting them together could result in even better tests and more optimally tailored regimens.

Meanwhile, it may also be important to consider whether a predictive test should be guided by the relative or absolute survival benefit.

"It's the latter that patients sitting on the other side of your desk are interested in," said Dr. Symmans. "They want their absolute probability of survival high."