Combining Predictive & Prognostic Signatures Provides Personalized Care Recommendation
Rabiya S. Tuma · Oncology Times · 2007
SAN FRANCISCO and BARCELONA—Clinicians already have access to gene signatures that provide prognostic information for individual breast cancer patients. Now, researchers have identified several gene signatures that predict an individual patient's response to chemotherapy or endocrine treatment, researchers reported at scientific meetings this fall. One research group has combined prognostic and predictive signatures into a diagnostic panel that can guide individual treatment. The panel also identifies patients who are least likely to respond to existing therapies and should go onto a clinical trial. “So far the prognostic profiles were the most robust, and the predictive profiles were still starting to become mature. Now that you have both in a more robust way, you can start to combine them,” said Laura Van't Veer, PhD, Head of Molecular Pathology at the Netherlands Cancer Institute, who developed the original 70-gene prognostic signature (now called MammaPrint). “I think in the end you will have one complete assay where you have a prognostic profile, several predictive profiles, maybe some single gene tests, and that will give you a full picture of the patient—what the best tailored treatment is.” Predicting Response to Endocrine Therapy Although physicians regularly test for estrogen receptor and progesterone receptor status and use the results to guide treatment, less than 60% of the women whose tumors express hormone receptors respond to endocrine therapy. With that issue in mind, W. Fraser Symmans, MD, Associate Professor of Pathology at the University of Texas M. D. Anderson Cancer Center, and colleagues developed a 200-gene signature that predicts response to tamoxifen therapy in an independent validation patient cohort, according to data he presented at the Breast Cancer Symposium in San Francisco. The team hypothesized that as the estrogen receptor (ER) works by activating other genes, tumors that had a high expression of estrogen-responsive genes would be more dependent on hormone stimulation, and thus more sensitive to endocrine therapy. With that in mind, the team identified 200 genes that appeared to predict estrogen dependence. They refer to the signature as the sensitivity to endocrine therapy (SET) score. During the development of the signature, Dr. Symmans and his colleagues found that if they split women into three groups based on their SET score, there was a significant difference in disease-free survival between the women with high, intermediate, and low scores. Those with high and intermediate scores appeared to have increased survival following tamoxifen treatment compared with those with a low score. In the team's current work, they tested the SET signature on an independent group of 250 women who received tamoxifen. The data are available for 176 of these women thus far. When the women were grouped into high, intermediate, and low based on their SET score, there was a statistically significant difference in their disease-free survival. The 10-year survival rates were 88%, 89%, and 71%, respectively, for the high, intermediate, and low SET groups. The low SET group accounted for 64% of the population.Figure: W. Fraser Symmans, MD (left) and Lajos Pusztai, MD, DPhil, in a photo that appeared earlier this year in an M. D. Anderson publication.Disease-free survival was not significantly different between women with high, intermediate, and low SET scores in a validation cohort of 298 women who had ER-positive tumors but who did not receive tamoxifen therapy. “We conclude that the SET index is an independently validated predictor of benefit from endocrine therapy for patients with ER-positive breast cancer,” Dr. Symmans said. “The SET index carries no prognostic information in the absence of endocrine therapy.” Building a Personalized Medicine Panel In addition to the SET index, the M. D. Anderson team developed a 30-gene signature that predicts pathological complete response to neoadjuvant chemotherapy in ER-positive breast cancers and has now combined these two tests with the 76-gene prognostic signature from Rotterdam into a single diagnostic panel, which can guide an individual patient's treatment. Lajos Pusztai, MD, PhD, Associate Professor of Breast Medical Oncology, reported the findings during a plenary talk at the American Association for Cancer Research-National Cancer Institute-European Organization for Research and Treatment of Cancer (EORTC) International Conference on Molecular Targets and Cancer Therapeutics. Using messenger RNA isolated from a single biopsy, clinicians could run all five tests—and come up with a very good idea as to how the patient should be treated. Dr. Pusztai and colleagues tested the approach using molecular and clinical data that were collected prospectively in previous trials. Out of these 198 women with node-negative ER-positive breast cancer, the team found that 55 had a good prognosis and 143 had a poor prognosis. Of those with the good prognosis, 21 were expected to respond to chemotherapy, and 34 were not. Nineteen of the 21 expected to respond to chemotherapy had a low likelihood of response to endocrine therapy, one had an intermediate likelihood and one had a high likelihood. Of the 34 unlikely to respond to chemotherapy, 20 had a low SET score and were thus unlikely to respond to endocrine therapy, six had an intermediate score, and eight had a high score. Thus, only one patient out of 198 had a good prognosis, and was predicted to respond to both treatment regimens. Of the 143 with a poor prognosis, 79 were predicted to respond to chemotherapy, while 64 were not. Of the 79 likely chemotherapy responders, 71 had a low SET score, four had an intermediate score, and four had a high score. Of the 64 patients unlikely to respond to chemotherapy, 38 also had a low SET score, indicating that they are unlikely to benefit from endocrine therapy, while 10 and 16 had intermediate and high scores, respectively. Based on these three tests, 38 (19%) patients were predicted not to respond to either therapy available, and to have a poor prognosis. “This is good information to know,” Dr. Pusztai said. “Probably this could motivate people to participate in research studies. These individuals really need new drugs, as opposed to the other ones who can be cured with existing treatment modalities.” Prospective Study Being Reviewed by FDA The current experiment, which Dr. Pusztai referred to as “in silico,” was performed by reanalyzing previously collected microarray information that was stored in a computer database from a past trial and linked to clinical outcomes. Although all of the samples in the database were collected in a prospective manner for the original trial and were a good starting place for initial tests of new approaches, Dr. Pusztai emphasizes that the team still needs to perform a prospective trial to demonstrate conclusively that the approach has clinical utility. The research team has now designed such a study, which is currently under review with the Food and Drug Administration, he said. Women with Stages I to III breast cancer will be eligible for the proposed trial. All participants will undergo fine needle aspiration. The biopsy samples will be analyzed using microarray-based tests for HER-2 gene expression, ER status, chemotherapy sensitivity, and sensitivity to endocrine therapy. Patients who are HER-2 positive will be treated with trastuzumab and chemotherapy. Women who show likely sensitivity to chemotherapy will receive neoadjuvant paclitaxel plus fluorouracil, cyclophosphamide, and doxorubicin. Women whose tumors have a high SET index will receive preoperative aromatase inhibitors. Those patients who are predicted to not respond to existing therapies will be put on new drug trials, with targeted agents selected by the gene expression pattern in the biopsy sample. The goal of the trial is to improve the pathological complete response rate above 25%, which is typically seen with neoadjuvant chemotherapy. Dr. Pusztai said that the choice of a neoadjuvant approach will reduce the wait for trial results, and may make clinicians, patients, and regulators more comfortable with the novel approach. Predicting Response to Traditional Chemotherapy in ER-Negative Tumors So far prognostic and predictive gene signatures have focused on ER-positive tumors. Yet ER-negative cancers are considered to be more aggressive than hormone responsive ones. In his search for signatures that might predict response in ER-negative breast tumors, Hervé Bonnefoi, MD, a medical oncologist at the Institut Bergonié at Bordeaux 2 University in France, turned to colleagues at Duke University led by Joseph R. Nevins, MD, who had already identified signatures that predict response to standard chemotherapy agents in lung cancer cells. The signatures were derived using the NCI-60 panel of cancer cell lines, which represent a variety of tumor types. Thus, Dr. Bonnefoi suspected that the gene signatures may also predict for response in breast cancer patients. To validate the signatures, Dr. Bonnefoi and colleagues used 125 prospectively collected tumor samples from patients who had participated in a previous EORTC clinical trial. That trial randomized patients to either neoadjuvant fluorouracil, epirubicin, and cyclophosphamide (FEC) or neoadjuvant docetaxel and epirubicin (ET). The chemotherapy signatures had strong negative predictive value for both treatment regimens, Dr. Bonnefoi reported during ECCO-14, the European Cancer Conference. Of 66 patients assigned to the FEC arm, the chemotherapy signatures accurately predicted that patients would or would not have a pathological complete response in 79% of patients. Although the positive predictive value of the test was only 68%, with 27 out of 40 responders identified correctly, the negative predictive value was 96%, with the test correctly predicting residual disease in 25 out of 26 non-responders. The microarray tests were similarly effective in the group of patients receiving ET therapy, with an overall accuracy rate of 80%. The positive predictive value was 71%, with 25 of 35 responders correctly identified. The negative predictive value was considerably higher at 92%, with 22 of 24 patients correctly categorized as non-responders. “For clinicians, the most practical information from the test is the negative and positive predictive values,” Dr. Bonnefoi said. “When the test tells you it will not work, it is true in the vast majority of cases.” And thus, like the panel Dr. Pusztai is developing, the gene signatures that predict response to chemotherapy regimens may be most useful in identifying patients who are unlikely to respond to standard therapies and should go onto a clinical trial. When Dr. Bonnefoi's team ran both tests on all of the patients, regardless of what arm the patient was actually treated in, the test predicted that 23 patients (18%) would not respond to either regimen. Only two of those showed a response to their actual therapy. “I think this is very important in terms of a research perspective,” he said. “We know that these patients will not respond very well. I think these patients are ideal candidates for trials with new agents.” If researchers focus on those hard-to-treat patients, they are likely to find regimens or novel agents that work in the patients who most need them, rather than identifying yet more therapies for patients who are likely to have a good response to already existing treatments. This approach could also result in economic savings for health care systems, Dr. Bonnefoi said. Patients who are likely to respond to standard agents can be treated successfully for 10,000 Euro (about $14,400) or less for their entire treatment. On the other hand, patients who are not likely to respond to those agents can be prioritized to receive expensive novel drugs, many of which cost 4,000 Euro (about $5,800) or more per month and must be administered on an ongoing basis. The team is planning to validate the use of the chemotherapy predictive signatures in 180 patients with ER-negative disease. Moving the Tests into the Clinic Any of these molecular tests, of course, have value only if clinicians are willing to use them in the clinic. Just how to encourage such use was the topic of discussion during the question-and-answer period of an AACR-NCI-EORTC plenary session devoted to individualizing therapy. “If the tests were 100% accurate, then it would be a no-brainer,” said one of the researchers, Lajos Pusztai, MD, DPhil, of the University of Texas M. D. Anderson Cancer Center. “The reality is that all of these tests are mediocre in their accuracy. There is room for personal opinion and judgment on how valuable they are.” Therefore, clinicians may feel that they can do just as well without them. Laura Van't Veer, PhD, of the Netherlands Cancer Institute predicted during the panel discussion that as the validation for the tests becomes more robust, the tests will become more widely accepted in clinical practice. When will a panel of molecular diagnostics arrive in the clinic? “It will go in steps,” she said in a subsequent interview. “There are components present already that give you some guidance, for hormonal therapy or lapatinib or trastuzumab therapy. Some things are a reality right now. Others will be added. In five to 10 years, I think we will have a fair amount of tests in one package.”