Research Grants Awarded
Prospective Evaluation Of Molecular Triaging With Pharmacogenomic Tests To Select Neoadjuvant Treatment
SCIENTIFIC RATIONALE: Adjuvant treatment of breast cancer provides the best opportunity for long-term cure, but each current treatment regimen can only cure a minority of patients, and the number and complexity of effective treatments is increasing rapidly with new agents and new combinations. It is a vital challenge to extend the repertoire of molecular diagnostic tests that guide treatment selection, particularly to identify those who are expected to benefit from a standard treatment, and to intelligently combine molecular agents with chemotherapy regimens for those who actually need such combination to achieve complete benefit. We have developed microarray-based pharmacogenomic predictors of pathologic response, and resistance, to a current standard chemotherapy regimen (T/FAC: paclitaxel, followed by 5-FU, doxorubicin, and cyclophosphamide) and these predictions were associated with survival. In addition, a measurement of estrogen receptor-related transcription predicts survival benefit from adjuvant hormonal therapy, and a threshold of HER2 gene expression that defines pathologic HER2 status and potentially converts some patients to HER2-positive status.
HYPOTHESES: Our principal hypothesis is that use of pharmacogenomic tests to select each patient?s treatment based on predicted response will improve the overall benefit from treatment of a patient population. Secondly, that pharmacogenomic predictions of benefit from a specific treatment regimen are based on tumor biology and pharmacology, so can be generalized to predict benefit from other similar treatment regimens. Thirdly, that improved understanding of the biology of primary breast cancers that have a poor response to neoadjuvant treatment can lead to improved prediction and treatment strategies.
SPECIFIC AIMS: 1) To conduct a prospective single-center trial to validate the use of individual predictors and the molecular triaging algorithm to assign neoadjuvant treatment, and to test whether addition of a novel targeted agent improves the pathologic response rates in patients who were not selected for a standard treatment, 2) to test whether predicted benefit from T/FAC chemotherapy can be generalized to predict benefit from similar chemotherapy regimens that sequentially combine anthracycline-based and paclitaxel treatments, 3) to conduct a multicenter, randomized, prospective clinical trial to evaluate the impact on pathologic response rate when microarray-based pharmacogenomic tests are used as the basis for neoadjuvant treatment selection in broader clinical practice, 4) to identify and characterize promising molecular targets of primary or acquired resistance to treatment, and 5) to develop automated methods that would enable pharmacogenomic analyses to be reliably performed in external reference laboratories.
STUDY DESIGN: During the first two years, we will test our microarray-based pharmacogenomic predictors in a proof of concept, single-center, prospective trial (N=350 patients) with primary objectives to: 1) confirm the positive predictive value of the T/FAC chemotherapy predictor, 2) identify an increase in pathologic response from concurrent bevacizumab plus T/FAC chemotherapy for patients who are not predicted to respond to T/FAC, and 3) obtain preliminary results for overall pathologic response rate, compared to matched historical controls. Concurrently, microarray-based response predictions from samples contributed by five centers that used four similar regimens that sequentially combine paclitaxel and anthracycline-based chemotherapy (N=196 for each regimen) will be independently compared with distant relapse-free survival. This will identify which chemotherapy regimens could be selected using the pharmacogenomic test. During the last three years we will conduct a multicenter, prospective trial (N=600 patients) that will randomize according to whether or not the results of microarray-based molecular triaging results are reported to the treating oncologist. Each center will select standard and investigational treatments to be used in the study and control arms, so that only the method of treatment selection is different. The trial has 80% power to detect a 10% increase in pathologic response rate as significant between the study and control arms. Laboratory studies will evaluate pre-treatment tumor biopsies and residual cancer after treatment to identify and then characterize molecular targets of resistance, using array-based gene expression and gene copy number, reverse-phase protein lysate arrays, and in vitro functional genomic and cell biological techniques. These results will be included in multi-stage computer models with clinical, radiologic, and pathologic variables to predict tumor shrinkage. We will also develop automated methods for reporting prediction results from microarray data for future use by other clinical reference laboratories.
IMPACT: The prospective clinical trials will establish internal validity of the molecular triaging concept, and then external validity of its use in a broader clinical setting employing additional treatments. The correlative studies will expand our understanding of molecular targets of resistance to current adjuvant treatments. The main impact will be the eventual integration of pharmacogenomic testing into diagnostic use to increase the overall rate of long-term cure by molecular triaging of each patient to the specific adjuvant treatment plan that is expected to be most effective, to avoid the treatments that won?t be effective, and to know in advance who stands to benefit from participation in a clinical trial with investigational treatment.
Many patients decide to receive an adjuvant treatment (shortly after diagnosis) because this intervention provides the best chance of long-term cure. However, it is important to recognize that there are several distinctly different treatment regimens (often combining or sequencing different drugs for clinically proven benefit) that are each capable of curing a minority of patients. Unfortunately there is currently no test available that can be relied on for selection of the most effective adjuvant treatment for an individual. There are prognostic tests that provide information about the inherent biological risk, and that can also be generally informative about the inverse relationship between sensitivity to chemotherapy, versus hormonal therapy. However, we will need new tests to reach the next level of understanding where we can select the best chemotherapy regimen and/or hormonal treatment for an individual, identify who needs to try new treatment possibilities, and recognize when partial benefit from chemotherapy then hormonal therapy add to a cure. Each test should predict the expected benefit from a specific treatment regimen, but still be generalizable to accommodate the subtle variations in the way that treatments are prescribed. Finally, we must insist on clinical trials of new approaches to customize adjuvant treatment, because demonstrated accuracy of tests to select treatments is as important as the effectiveness of the treatments themselves.
We have developed pharmacogenomic (gene expression) signatures to accurately predict response and resistance to our current standard chemotherapy regimen (T/FAC, i.e. paclitaxel (T) followed by 5-FU, doxorubicin, and cyclophosphamide (FAC)). Those predictions also appear to be strongly associated with distant relapse-free survival in the same patients. We have defined a signature to measure the level of gene transcriptional activity from estrogen receptor (the target of hormonal therapy), and that predicts patients? survival benefit from adjuvant hormonal therapy. We have defined a threshold of HER2 gene expression that accurately identifies HER2 status of breast cancer, and have identified a subset of patients whose tumor would be converted to HER2-positive status by HER2 gene expression levels, and might benefit from trastuzumab. All of these predictive signatures are obtained from a standard needle biopsy of the patient?s breast cancer, are determined from analysis of commercially available gene expression microarrays, and could feasibly be translated into routine clinical diagnostics to provide the prediction results before any treatment is selected.
We are now ready to conduct a proof of concept prospective clinical trial (specific aim 1) to investigate whether the use of these predictive signatures to select a treatment will increase the response rate in the selected patients, whether patients who are not selected for a standard therapy will benefit from the addition of a new investigational treatment, and whether the overall approach to triage each patient to a specific treatment could improve treatment response overall, as compared to a carefully matched control group of unselected patients. At the same time, we have assembled a team of collaborators to test our T/FAC chemotherapy predictor on samples from patients who received one of three similar chemotherapy regimens in common use. We will provide our chemotherapy response predictions for an independent comparison with survival after treatment (specific aim 2).
Results from the first two specific aims will lead to a prospective clinical trial that is designed to test the impact on treatment response when molecular triaging is used (versus not used) to select treatment in different centers, allowing any of the chemotherapy regimens for which the response predictor has been validated, and allowing each center to design their own clinical trial treatment for patients who are not selected to respond to a standard treatment. Concurrent laboratory studies will identify molecular targets of resistance from the number of gene copies (DNA), gene expression levels (RNA), and protein signaling in the breast cancer, comparing these before and after treatment, and focusing specifically on patients whose cancer was resistant to the pre-operative treatment. The most promising gene targets will undergo detailed laboratory characterization of their cellular functions and role in response to individual drugs. We will also use the clinical, pathologic, radiologic, and all molecular information (DNA and RNA microarrays, and protein signaling) from the trial participants to train a multi-scale computer model to predict tumor shrinkage. Concurrently, we will develop automated analysis and reporting methods that should facilitate the transition of pharmacogenomic tests for molecular triaging into clinical diagnostic laboratories.
Within five years we will have completed the prospective proof of concept and multicenter trials that are needed to demonstrate the impact of molecular triaging to individually recommend adjuvant treatment, identified and characterized genomic targets of resistance for further development, and developed automated analysis and reporting methods that can be applied in clinical diagnostic laboratories. Herein lies a strong opportunity to increase the overall rate of long-term cure by molecular triaging of each patient to the specific adjuvant treatment plan that is expected to be most effective, to avoid the treatments that won?t be effective, and to know in advance who stands to benefit from participation in a clinical trial with investigational treatment.