There are several different chemotherapy regimens physicians commonly use to treat patients with newly diagnosed breast cancer. These treatments vary widely in potential toxicity and cost. Depending on the type of preoperative chemotherapy, 15% to 30% of patients will achieve a pathologic complete response, which is defined as no residual invasive cancer in the breast. Patients who achieve pathologic complete response have been shown to have an excellent long-term and disease-free survival.

It is well understood that not all regimens are equally effective for every individual. Some patients achieve cure with short, less toxic, less expensive regimens. Others require longer, more toxic regimens that are more expensive. However, physicians use the currently available treatments without any selection process, because there has been no way to tell which regimen is most likely to benefit a specific individual.

Gene expression profiling may help predict the response of an individual patient’s breast cancer to a specific chemotherapy regimen. This new laboratory technology involves extracting RNA from cancer specimens. Investigators then profile the RNA using a cDNA microarray from Millennium Pharmaceuticals.

The investigators wanted to determine if they could use gene expression profiling to identify patients who will have pathologic complete response to a common preoperative (neoadjuvant) chemotherapy regimen of sequential paclitaxel/FAC (5-FU, doxorubicin and cyclophosphamide).

They collected breast cancer specimens via needle biopsy from newly diagnosed patients with breast cancer prior to institution of chemotherapy. They used the first 24 cases to generate a multi-gene predictor of response outcome. Then, they validated the accuracy of that predictor on 21 new cases. Approximately 80% of patients in both groups had T2-T3 disease. Almost all of the patients had weekly paclitaxel for 12 weeks, followed by 4 courses of FAC. Two patients in each group had paclitaxel every 3 weeks followed by FAC. Three patients in the validation set received trastuzumab concomitantly with paclitaxel.

At ASCO, investigators reported that a specific pattern of expression of 74 genes could predict pathologic response to paclitaxel/FAC chemotherapy with an accuracy of 71%. The positive predictive value was 75%; this means that the test predicted 4 complete responses, and 3 of those 4 cases were in fact pathologic complete responses. Dr. Pusztai said this is proof of principle that this technology can serve as a diagnostic tool.

Investigators reanalyzed the data excluding the 3 patients who received trastuzumab. Studies show trastuzumab can alter sensitivity to paclitaxel and anthracycline therapy. After excluding these patients, the overall prediction accuracy improved to 78%, and the positive predictive value improved to 100%.

However, this first-generation technology is probably not ideal. While the accuracy and positive predictive values are high, the sensitivity of the test is low.

Accordingly, Dr. Pusztai and colleagues have created a second-generation test that uses data from a larger number of patients. It also incorporates a different DNA chip (from the Affymetrix company). The researchers believe this new test will be more accurate than the previous test. The second-generation test is undergoing validation in clinical trials at M.D. Anderson Cancer Center.

Dr. Pusztai said the ultimate goal of this research is to personalize chemotherapy selection according to the genetic characteristics of each patient’s cancer. He believes that it will be possible to create tests to predict complete pathologic response to each of the commonly used preoperative breast cancer chemotherapy regimens. One day, a physician may choose a chemotherapy regimen after running all these tests on the cancer specimen.