Gene Signatures Predict Breast Cancer Treatment Response in Clinical Study

BORDEAUX, France, Nov. 16 -- Gene-expression signatures in breast tumor samples can accurately predict which patients will not respond to two common chemotherapy regimens, researchers here found.
Action Points
Explain to interested patients that these are preliminary results that need to be confirmed in larger prospective studies.
Explain that the results apply only to the specific treatments and types of patients included in the study.
The signatures had a negative predictive value of 90% in forecasting patients' responses to the regimens, according to Hervé Bonnefoi, M.D., of the Bergonié Institute, and colleagues in the online edition of The Lancet Oncology.
Their findings were part of a phase III clinical trial sponsored by the European Organization for Research and Treatment of Cancer (EORTC) comparing the two regimens: one, dubbed FEC, combines 5-fluorouracil, epirubicin and cyclophosphamide for six cycles, and the other, known as TET, consists of docetaxel for three cycles followed by three cycles of docetaxel and epirubicin.
Positive predictive values for the signatures were lower. But if only those patients identified by the signatures as likely good responders had received the treatments, the researchers said, the rate of pathological complete response would have been about 70%.
As it was, 44% of patients in the study achieved pathological complete responses.
"This is such a big improvement that we need to do another study to make sure it is really true," said the researchers.
The signatures reflect expression levels of dozens of different genes. They were developed in 2006 by a group at Duke University under Joseph Nevins, Ph.D., who participated in the current study.
Dr. Nevins's research had indicated that the signatures could predict response to individual chemotherapies in estrogen receptor-negative breast cancers.
Based on those studies, Dr. Bonnefoi's group included the predictive signatures in EORTC's 10994/BIG 00-01 trial, testing neoadjuvant FEC and TET in a randomized, head-to-head comparison. The genomic-signature tests were conducted on frozen biopsy samples taken before chemotherapy began.
The group applied the tests to 125 patients in the trial, 66 in the FEC group and 59 receiving TET.
Among 40 cases in the FEC group predicted to respond well to FEC, 27 obtained a pathological complete response. In the TET group, 25 pathological complete responses occurred in the 35 cases predicted to have good responses.
If the findings are confirmed in a follow-up study, said the authors, they recommend that the genomic-signature tests be used in clinical practice to select patients for these regimens.
Those patients likely to have poor outcomes to these treatments would therefore be spared the expense and toxicity and could receive more effective treatment more quickly.
They also identified an additional benefit to using predictive genomic testing in clinical trials of novel drugs. By selecting for inclusion those patients unlikely to respond to conventional drugs, the researchers said, "it should be possible to do much smaller trials."
Data from such trials "would apply only to a narrowly defined group, albeit one with a poor outcome with conventional treatment," they acknowledged. "Whether it is right to proceed in this way is a matter of legitimate debate, but clearly we think it is."
Another drawback identified in the study was that the genomic analysis could not be conducted in more than 25% of eligible patients. There were 212 patients in the clinical trial with estrogen-receptor negative breast tumors, the main inclusion criterion for the expression analysis.
Of those, 55 had to be excluded because their samples contained too few tumor cells or had insufficient genetic material to allow the gene-expression analysis.
Also, the analysis was performed on frozen biopsy samples. The standard protocol at many clinical centers is to preserve samples with formalin and paraffin.
Dr. Bonnefoi's group said the gene-chip platform used for the analysis was designed for paraffin-embedded samples, "so only small modifications to the protocol might be needed to test paraffin-embedded material." But the group did not test this.
Funding for the study came from EORTC and more than a dozen nonprofit foundations and government agencies. The authors declared no conflicts of interest.Primary source: Lancet OncologySource reference: Bonnefoi H, et al "Validation of gene signatures that predict the response of breast cancer to neoadjuvant chemotherapy: a substudy of the EORTC 10994/BIG 00-01 clinical trial"Lancet Oncology 2007; DOI: 10.1016/S1470-2045(07)70345-5.