Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 49
Improving the Quality of Cancer Clinical Trials: Workshop Summary SCREENING FOR PREDICTIVE MARKERS Biomarkers may possibly be used to predict a number of factors relevant to cancer treatment, including aggressiveness of a tumor and the need for treatment, likelihood of responding to specific treatments, likelihood of developing adverse reactions to treatment, and prognosis. On the second day of the conference, the first session focused on the progress and challenges linked to identifying and validating such predictive markers, as well as applying them in a clinical setting. Drs. Pierre Massion of Vanderbilt Ingram Cancer Center, James Heath of California Institute of Technology, Dan Sullivan of Duke University, and Daniel Von Hoff of Translational Genomics Research Institute addressed these issues in the presentations, and provided answers to specific questions posed by the National Cancer Policy Forum at a panel discussion that followed the presentations. Several presenters stressed the clinical need for predictive biomarkers in oncology. Using lung cancer as an example, Dr. Massion pointed out that such biomarkers are needed for every step in patient management. Markers that can predict a person’s risk of developing lung cancer are needed for people who smoke and might benefit from heightened screening or participating in various chemoprevention trials. Blood, sputum, or other non-invasive biomarker tests are needed to improve diagnosis once physicians detect a suspicious lesion in a patient’s lungs. Currently, diagnosis can only be done reliably with invasive surgery or bronchoscopies. Also needed are biomarkers that can predict the likelihood that a small early lesion in the lungs will progress to a deadly cancer. A suspicious lesion that is less than 2 centimeters cannot be accurately diagnosed as malignant using a PET scan, and may not be accessible via bronchoscopy. A needle biopsy poses the risk of lung collapse, with the only other proactive option—surgical removal—being even more invasive and risky. Physicians can take the “wait-and-see” approach to such lesions, of which 20 percent may be malignant. But with that approach, one may miss a chance of curing an aggressive lung cancer. Because only 30 percent of patients with lung cancer respond to radiation or chemotherapy, and most of them will experience some toxic reactions to those treatments, there is a great need for biomarkers that can predict response to treatment, recurrence, and prognosis. “We are overtreating cancer in general, and in some cases we undertreat the proper subset of patients,” Dr. Massion said, adding that predictive biomarkers will “allow us to eventually provide adjuvant therapy to the appropriate population, narrow down the patient selection, and decrease costs of therapy.”
OCR for page 50
Improving the Quality of Cancer Clinical Trials: Workshop Summary Drs. Massion and Heath gave examples of biomarker tests that are already being used in a clinical setting. These include the ER and HER-2/neu biomarker tumor tests that predict response to various breast cancer treatments, gene signature tumor tests based on the activation patterns of 21 or 70 genes that predict breast cancer recurrence or survival (Paik et al., 2004), and a tumor genetic signature test based on 100 genes that can distinguish between two types of similarly appearing lymphomas, and predict survival following treatment (Dave et al., 2006). Many other predictive biomarkers are in the early stages of clinical testing, including one that uses the patterns of 40 proteins in a blood sample to distinguish between prostate inflammation and prostate cancer,10 an eight-gene signature blood test that predicts response to lung cancer treatment (Taguchi et al., 2007), and those that use the patterns of proteins in the blood, or genes activated in airway epithelial cells to predict lung cancer (Yildiz et al., 2007) (airway epithelial cells can be easily brushed off and collected noninvasively for such diagnostic testing). Initial testing of the blood biomarker test for predicting lung cancer suggests it has a sensitivity of 58 percent and a specificity of 85.7 percent. “Although this beats any biomarkers that are currently available in the blood for patients with lung cancer, it may not have the sensitivity you wish for in the early detection approach and the specificity may not be optimal,” Dr. Massion noted. Despite the need for biomarkers in oncology, and the thousands of cancer-related biomarker publications over the past 10 years or so, only about 20 cancer biomarkers have been approved by the FDA (Figure 13), and many of these are not used routinely in clinical practice. (Ludwig, 2005). “The world of cancer biomarkers is a very humbling one, and not a successful one to my eyes,” Dr. Massion said. Dr. Heath added that “the world of biomarker discoveries has been advancing over the past decades in leaps and bounds, but it is still remarkably immature.” The Challenges of Clinical Validation A major hurdle that needs to be overcome for more predictive biomarker tests to enter the clinic is the validation of the diagnostic accuracy and usefulness of existing candidates, according to Dr. Massion. Such validation should be done via several large, independent studies at multiple 10 Personal Communication, J.R. Heath, E.W. Gilloon Professor of Chemistry, California Institute of Technology, October 5, 2007.
OCR for page 51
Improving the Quality of Cancer Clinical Trials: Workshop Summary FIGURE 13 Publications and FDA-approved biomarkers. Despite the increasing rates of publications on biomarkers, the number of FDA-approved plasma-protein tests is decreasing. Triangles and the associated trend line represent the number of FDA-approved plasma-protein markers per year. Squares and circles indicate publications under the Medline medical subject heading “biomarker” and text word “biomarker,” respectively. SOURCES: Massion presentation (October 5, 2007) and Ludwig (2005). Reprinted by permission from Macmillan Publishers Ltd: Nature Reviews Cancer 5(11):845-856, Copyright 2005. institutions that use different testing platforms and well-annotated patient samples. This validation remains challenging and needs to keep pace with the rapid progress in assay development. “You see a myriad of publications looking at biomarker discovery and first-phase validation, but very few are putting them within a clinical context. This is where we need to go—what we need to do,” said Dr. Massion. Under the auspices of the NCI and the Specialized Programs of Research Excellence (SPORE) program, he and other researchers have created a group called the Lung Cancer Biomarkers Group, which aims to provide several academic institutions with access to four different sets of patient sample materials held at an NCI repository for the purpose of testing the accuracy of lung cancer biomarker tests. Such testing will not only access accuracy of the tests, but also their reproducibility within and between institutions and how they can provide clinically useful information. Later in the discussion, Dr. Massion added that “we need to establish repositories that allow us to validate biomarkers within institutions and across institutions and across platforms. This is unfortunately very time consuming and also requires centralized repositories of prospectively acquired samples and a great deal of collaboration between institutions.”
OCR for page 52
Improving the Quality of Cancer Clinical Trials: Workshop Summary Dr. Massion gave several examples of clinical trial designs for studies aimed at assessing the clinical utility of predictive biomarker test (Figures 14-16). The simplest study design is to compare outcomes of patients who test positive for the biomarker with those of historical controls (Figure 14). Another design that is much more costly to run is to randomize patients as to whether they undergo the biomarker test or not (Figure 15). Of those tested for the biomarker, patients with positive results receive the intervention the marker indicates is warranted, while those with negative test results receive standard care. This study is designed to determine whether the predictive test improves patient outcomes when compared with unselected patient treatment. This study requires a large number of patients. Another study design does not address the quality or value of the biomarker itself, but rather compares outcomes for two different interventions in those who test positive for the biomarker and as well as those who test negative for FIGURE 14 Clinical utility of predictive markers, study design 1: A single-arm validation study using historical controls for comparison. In this study, all patients receive the biomarker test and outcomes of patients who test positive for the biomarker are compared with those of historical controls. SOURCE: Massion presentation (October 5, 2007).
OCR for page 53
Improving the Quality of Cancer Clinical Trials: Workshop Summary FIGURE 15 Clinical utility of predictive markers, study design 2: Randomization to receive or not receive the biomarker test. This design determines whether the predictive test improves patient outcomes when compared with unselected patient management. In this study, patients are randomized, and one group is given the biomarker test, the results of which influence the therapies received by the patients. The therapeutic outcomes of the tested and untested groups are compared. SOURCE: Massion presentation (October 5, 2007). FIGURE 16 Clinical utility of predictive markers, study design 3: Randomization of treatment irrespective of biomarker test results. This design compares two interventions in both marker-positive and marker-negative groups. In this study, all patients receive the biomarker test, and are randomized into either of the two treatment arms. SOURCE: Massion presentation (October 5, 2007).
OCR for page 54
Improving the Quality of Cancer Clinical Trials: Workshop Summary the same biomarker (Figure 16). This study design randomizes between the two interventions. Dr. Heath noted that often biomarker tests show promise when initially tested on a population of interest, such as men with prostate cancer, versus a healthy control population. “But as soon as you look at them across the general population, they tend to fall apart, and that is one of the reasons why the approval of biomarkers has been relatively slow. What you would like to be able to do is take your top 1,000 candidates and just measure them broadly across all population bases and do it cheaply and rapidly,” he said. One reason Dr. Massion gave for the lack of clinically successful biomarker tests in cancer are their lack of competitiveness in terms of costs and reimbursement. Dr. Heath expanded on this in his presentation on lowering the cost of in vitro diagnostics. He offered several suggestions for making biomarker tests more rapid, inexpensive, sensitive, and clinically relevant; the first and foremost is to ensure the tests are based on accurate and appropriate biology. “The first thing is to get the biology right because obviously if you are measuring the wrong thing, who cares,” he said. The reagents and materials used in the test are also critical. “You will see a lot of interesting devices these days to measure proteins or messenger RNAs or whatever. Many of these tests look exotic. But if it looks exotic, it is probably not going to be something you are going to use in the clinic,” he said, adding that the test should use inexpensive and scalable technology, as well as inexpensive reagents and equipment, such as glass and plastic. “This is a huge issue, and it is probably the limiting issue in antibodies,” Dr. Heath said. Biomarker tests should also require very small amounts of tissue or blood, such as a finger prick of blood, yet be highly sensitive and quantitative because many of the compounds of interest are present in exquisitely minute amounts in patient samples. To improve the sensitivity and quantitativeness of such tests, Dr. Heath suggested using fluorescent markers and a scattering microscope, which has an aperture in front of a light microscope that enables automated counting of trace compounds of interest. It is about 10,000 times more sensitive than standard protein assays and can detect compounds at 100 attomolar concentrations. Given the complexity of the molecular pathway networks that underlie various cancers, biomarker tests should be multiparameter tests that can be automated and done rapidly because, as Dr. Heath pointed out, time equals money. Time can be decreased by not making tests diffusion dependent, as are standard ELISA antibody-based assays. These tests require a few hours
OCR for page 55
Improving the Quality of Cancer Clinical Trials: Workshop Summary for proteins of interest to diffuse to the labeled antibodies with which they bind. Tests can be based on the kinetics of proteins binding to antibodies without requiring such diffusion, according to Dr. Heath, and be completed in 5 to 30 minutes with a cost of 5 to 20 cents per measurement, depending how many measurements are made simultaneously. “The idea of this kind of technology is to put every single possible biomarker you could imagine on the chip so we can capture the diurnal and dietary variation, and all the other fluctuations that tend to foul up the validation of a biomarker, but then also correlating it with traditional pathology and disease,” Dr. Heath said. He also gave examples of what he called “PET on a chip,” which is a microarray test developed by researchers at the University of California, Los Angeles, that partitions cells from a patient’s tumor biopsy into 100 different wells that contain metabolic markers for response to drugs and can indicate—within an hour of when the patient was sampled—to which drug regimen the patient is likely to respond. Bioimaging Predictive Markers Following Dr. Heath’s presentation, Dr. Sullivan gave examples of how bioimaging can be used to predict clinically relevant variables in oncology; the advantages and disadvantages of using imaging biomarkers; and the technical and regulatory challenges of making those biomarkers clinically useful. As previous speakers noted, bioimaging can predict where a cancer drug will concentrate in the body, and various physiologic states such as a lack of oxygen (hypoxia), or diffusivity that can affect drug response. Two small studies suggest DCE MRI might be useful as a predictor of survival following treatment for osteosarcoma or renal cell cancer (Reddick et al., 2001; Flaherty et al., 2008). Two clinical studies found PET imaging of hypoxia predictive of response to drug treatment, and larger multi-institutional trials have been planned to assess the effectiveness of such bioimaging (Rischin et al., 2006; Dehdashti et al., 2003). Two studies also indicate that PET imaging of labeled estradiol predicts response to hormonal therapy in advanced breast cancer (Linden et al., 2006; Mortimer et al., 2001). Diffusion imaging was found to predict response to treatment in brain cancer (Hamstra et al., 2005), and encouraging results from a study of magnetic resonance spectroscopy in non–Hodgkin lymphoma patients have led to a multisite trial to test prospectively whether such imaging can identify patients who would respond to conventional therapy versus patients who should receive more aggressive treatment, such as a bone marrow trans-
OCR for page 56
Improving the Quality of Cancer Clinical Trials: Workshop Summary plant (Arias-Mendoza et al., 2004). Researchers are also starting to conduct clinical studies of the usefulness of assessing multiple imaging biomarkers to predict indolent disease (Shukla-Dave et al., 2007). Although bioimaging is well suited for revealing physiologic measures such as hypoxia, diffusion, or tumor metabolism, such measurements may be strongly influenced by other interfering systemic conditions and result in a misleading reading, Dr. Sullivan pointed out. For example, for FDG imaging of tumor metabolism, “there can be other things going on in the body and the brain, the heart and the skeletal muscle that could suck up all the glucose and give a spuriously low value in the tumor, so these have to be controlled for,” Dr. Sullivan said. Although imaging biomarkers lack the diversity and large number of parameters that can be simultaneously discerned compared to genomic or proteomic tests, he said, imaging biomarkers do provide the spatial and temporal context for the markers, unlike in vitro tests. “Given that cancer is a heterogeneous disorder and a systemic disorder, that contextual information might be important or useful in making predictions of response to therapy,” he noted. For example, drugs such as tirapazamine (SR-4233) are most active in tissue lacking oxygen. PET scanning with a marker for oxygenation could be an especially useful predictor of which patients will respond to this drug. Bioimaging can also preserve some physiological information, such as pH or oxygenation status, that might otherwise become disrupted or lost in the sample preparation involved in the in vitro tests, he added. Such imaging is more likely to accurately reflect the true physiological state of normal or cancer cells. Another advantage of molecular imaging over some genetic or proteomic biomarker tests is that it can be less invasive because it does not require tumor samples. But several technical challenges are involved in developing imaging biomarkers, particularly if they use radiolabels and small molecules. Although it is relatively easy to label antibodies with a radioactive probe for detection in an imaging system, it often is difficult to label small molecules with the radioactive carbon, oxygen, fluorine, or nitrogen atoms used in PET imaging. For example, although the cancer drug gemcitabine has a few fluorines as well as carbon atoms in its structure, several attempts to replace these with radioactive fluorine or carbon atoms for PET imaging failed after much trial and error. “So this has been years in development and they are still not there yet. It is not a straightforward process in many cases,” Dr. Sullivan said. One problem in this regard is that the drug kinetics may not be appropriate for imaging of a labeled drug. Imaging agents are usually better if they
OCR for page 57
Improving the Quality of Cancer Clinical Trials: Workshop Summary have irreversible binding, but some drugs have reversible binding. One also has to develop a rapid synthetic pathway for the radiolabeled drug that can be accomplished within the time constraints of the half-life of the isotope used to label it, which may only be a few hours. It can also be challenging to produce a compound that is sterile, doesn’t induce a fever, and can be immediately injected into a patient. “It can take years to get these bioimaging probes ready for use in patients, and meanwhile the drug development is moving along,” Dr. Sullivan said. An alternative to labeling the drug for bioimaging is to label another ligand for the target of interest, such as a drug analog or a growth factor that binds to the same receptor as the drug. However, although this may enable easier labeling synthesis, the labeled ligand may not reveal drug localization as accurately as the labeled drug and requires more validation. The time course for development of such labeled ligands can be as long or longer than that for labeled drugs, so that also may be out of synch with corresponding drug development. In addition to the technical challenges of developing bioimaging biomarkers, there are substantial regulatory hurdles. Currently the regulatory process for these markers is the same as that for drugs: Studies must show their clinical benefit and safety for patients. “Right now there is no commercial pull for the industry to develop these agents and there is a lack of resources being devoted to their development partially because of this regulatory process,” Dr. Sullivan said. “Many people believe that for these imaging tracers that have no pharmacologic effect, the target or benchmark for efficacy should be the same as it is for devices approved by the FDA; that is, the agent should provide the information which the producer or vendor claims that it provides, and that it would not necessarily provide clinical benefit,” he added. A final challenge that Dr. Sullivan discussed is the validation of imaging biomarkers. He noted many sources of variability in bioimaging that need to be considered and controlled for, including such physical sources of variability as scanner calibration, machine variation, different image acquisition parameters, and different algorithms for data processing. There are also physiological sources of variability, including intra- and interpatient variation and reader variability. Performing the necessary repeatability tests for imaging methods is especially difficult and costly because they are performed on people, not specimens. If imaging is to become a reliable in vivo assay, Dr. Sullivan said, several factors must be in place: uniformity of instrumentation, protocol-specified
OCR for page 58
Improving the Quality of Cancer Clinical Trials: Workshop Summary acquisition of images, independent quality control, reliability and independence of reader interpretation and provenance, auditability, and accessible storage of results. In a presentation in the following session, Dr. Gwen Fyfe of Genentech also noted some of the potential pitfalls of bioimaging biomarkers and the need to properly validate them. “Expensive techniques save time and improve outcome only if carefully validated by clinical outcomes,” she said. She gave an example of a small bioimaging study of a VEGF receptor inhibitor that showed that colorectal tumor vascularity and permeability decreased rapidly following treatment with the inhibitor and seemed to correlate with clinical benefit (Morgan et al., 2003, and Steward et al., 2004). But when a larger randomized study was done, the drug had no impact on overall progression (Hecht et al., 2005). She noted that a bioimaging study that predicts response after only one cycle of treatment might not predict a more durable response. “A biomarker probably has a good negative predictive value—if you don’t see an impact it is unlikely to be useful—but the positive predictive value is really subject to interpretation,” Dr. Fyfe said. “We need to validate these biomarkers very carefully with clinical outcomes before we assume that we can go from Phase I to Phase III based on an imaging result.” But it is debatable what the appropriate validation is, Dr. Fyfe added. “Is it response, [is it] durable response, or is it progression-free survival?” she asked. Not only does there have to be validation of technique, such as reproducibility that considers site and patient variability and timing of analyses, but there needs to be validation of patient benefit. Such validation may be disease-, pathway-, or drug-specific. Even within a pathway, agents may differ significantly in their mechanism of action, she pointed out. Follow-up exploratory trials that assess the relationship of the biomarker effect to clinical outcome are needed to avoid large negative trials. Dr. Fyfe concluded by noting the need for information sharing among academia and pharmaceutical and biotechnology companies to validate biomarkers. The Biomarker Consortium11 is a good start, she said, but 11 The Biomarkers Consortium is a public–private biomedical research partnership managed by the Foundation for the National Institutes of Health that includes government, industry, patient advocacy groups, and other non-profit, private-sector organizations. In addition to the Foundation for NIH, founding members include the NIH, the FDA, and PhRMA. Other partners in the consortium include CMS and the Biotechnology Industry Organization. The Consortium aims to “rapidly identify and qualify biomarkers to support basic and translational research, guide clinical practice and, ultimately, support the develop
OCR for page 59
Improving the Quality of Cancer Clinical Trials: Workshop Summary she called for more government investment in such validation efforts and stressed that all biomarker validation results should be published, both negative and positive. “Each company has its own interests so for us to really work on this there needs to be an NCI-directed effort because I think pharma and biotech will help, but fundamentally that is going to be a very splintered effort, and we are going to learn best by doing careful studies that are in the public domain,” Dr. Fyfe said. In the discussion that followed, Dr. Steven Larson added that the FDA and the U.S. Pharmacopeia, in addition to the NCI, should do more to aid efforts at validating and standardizing imaging biomarkers. “The FDA could develop a path which would allow for qualification of these individual biomarkers for the specific biochemical or pathway purpose for which they are intended,” he said. Clinical Translation Given the complexity of the molecular mechanisms that underlie specific cancers and the challenges involved in developing and validating biomarker tests predictive of those mechanisms, translating the research findings on predictive biomarkers into tests with clinical usefulness can appear to be an especially difficult hurdle to overcome. But Dr. Von Hoff, the last speaker in this session, described a simplified approach to such translation that has been used at the Translational Genomics Research Institute (TGen). “It is an understatement to say that work on biomarkers is complicated and that screening for predictive markers is going to take a while. But we need to help patients who are sitting in front of us right now with refractory cancer. So our clinical research teams are focused on applying what we already know about mutations, translations, and deletions. We feel this is an important policy because there is a lot you can actually do right now,” he said. To apply that knowledge, Dr. Von Hoff proposed that oncologists assess the “clinical and molecular contexts of vulnerability” of their patients’ tumors— which he referred to as a “sixth vital sign”—and use those contexts to help select therapy. As an example of clinical context of vulnerability, he described an 81-year-old patient with lung cancer who smoked for 72 years. Dr. Von Hoff said this patient’s history suggests that his tumor can repair ment of safe and effective medicines and treatments.” The Consortium also plans to “harmonize approaches to identify viable biomarkers, verify their individual value, and formalize their use in research and regulatory approval” (http://www.biomarkersconsortium.org).
OCR for page 60
Improving the Quality of Cancer Clinical Trials: Workshop Summary TABLE 3 Contexts of Vulnerability Tumor Type Vulnerability Agent(s) Ewing sarcoma Growth factor receptor (IGFR1) AMG479; CP751, 871 Ewing sarcoma Phosphoinositide 3-kinase (enzyme) SF1126 Synovial cell sarcoma Gene translocation, Growth factor receptor Iressa/Tarceva Chondrosarcoma Tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) TRAIL interactive agent Alveolar soft part sarcoma Oncogene—gene fusion C-met inhibitor Osteogenic sarcoma C-met (Mesenchymal epithelial transition factor) abnormalities C-met inhibitor Small blue round cell tumors—Ewing, osteosarcoma, neuroblastoma, desmoplastic small round cell, synovial Platelet-derived growth factor receptor Platelet-derived growth factor receptor inhibitor Chronic Myelogenous Leukemia Oncogene—fusion protein (Brc-abl) Gleevec almost any DNA damage, so giving him chemotherapy that induces such damage is not likely to be effective, and other options are warranted. The molecular or genomic context of vulnerability refers to what Dr. Von Hoff called the molecular addiction of the tumors. Many breast cancers, for example, are “addicted” to estrogen and need this growth factor to survive. Over the past few decades, researchers have noted numerous other growth factors, enzymes, and other compounds that tumors need to survive (Table 3). Drugs have already been developed that target these vulnerabilities, Dr. Von Hoff noted, so knowing the tumor addiction can help with treatment selection and improve treatment effectiveness. He has found this approach to be remarkably effective in some cases, even in patients with advanced cancer that has not responded to conventional treatment. For an example, he described metastatic myxoid liposarcoma, a particularly aggressive cancer of the connective tissue whose hall-
OCR for page 61
Improving the Quality of Cancer Clinical Trials: Workshop Summary Tumor Type Vulnerability Agent(s) Acute Lymphoblastic Leukemia Oncogene—fusion protein (Brc-abl) Gleevec Chronic Neutrophilic Leukemia Oncogene—fusion protein (Brc-abl) Gleevec Hypereosinophilic syndrome Growth factor receptor mutations Gleevec Medulloblastoma Growth factor receptor mutations Gleevec, hedgehog Gastrointestinal Stromal Tumor Growth factor receptor mutation Gleevec, sunitinib Prostate cancer Oncogene—fusion protein HDAC inhibitor reversing the phenotype Castleman disease Increased interleukin-6 (growth factor) CNTO 328 NOTE: This table shows the types of vulnerabilities and the therapeutic agents that have been developed to target them in specific tumor types. ACRONYMS: IGFR1 (insulin-like growth factor receptor 1), AMG479 (fully human antibody against IGFR1), CP751,871 (monoclonal human antibody against IGFR1), SF1126 (Vascular Targeted pan-PI3K Inhibitor), TRAIL (Tumor necrosis factor–related apoptosis-inducing ligand), C-met (Mesenchymal epithelial transition factor), HDAC (Histone Deacetylase), CNTO 328 (human-mouse chimeric monoclonal antibody to interleukin-6). SOURCE: Adapted from Von Hoff presentation (October 5, 2007). mark is a specific and relatively simple genetic abnormality (translocation between the 12th and 16th chromosomes). A recent study showed that a drug in development, ET-743, is effective in only 7 percent of all patients with connective tissue cancers, but 97 percent effective in myxoid liposarcomas (Grosso et al., 2007). According to Dr. Von Hoff, a tremendous number of tumors’ genomic contexts of vulnerability are deletions, translocations, or other simple genetic abnormalities. For example, about 64,000 U.S. breast cancer patients have mutations in their BRCA1 or BRCA2 genes, which a recent study showed are likely to respond to drugs known as PARP inhibitors that target the abnormal DNA repair associated with these mutated genes. “Biomarker patterns are great, but they are going to take longer to be useful,” Dr. Von Hoff said. In the later discussion he added that deletions and mutations are easier to measure and are more likely to be reproducibly measured than
OCR for page 62
Improving the Quality of Cancer Clinical Trials: Workshop Summary biomarker pattern assays. Deletion, translocation, and mutation assays are reliable and don’t have the “fudginess” of microarray assays, he said. Dr. Von Hoff reported that researchers at TGen are using small interference RNA techniques to design drugs that target the specific genetic abnormalities that cause tumor addictions. “Instead of treating patients with a drug and then finding some pattern that indicates what the genetic deletion is that makes the drug effective for some patients, we design a drug that targets only that deletion,” Dr. Von Hoff said. “We get the marker for the genetic abnormality and then design the drug to take out cells with the marker.” But determining the addictions of patients’ tumors requires researchers to identify and catalogue the contexts of vulnerability as rapidly as possible. This suggests the need for a centralized clearinghouse for such characterization of patients’ tumors, which would avoid the need to send tumor samples to several different facilities, each with the capacity to evaluate tumors for only one or two genetic abnormalities. To meet that clearinghouse need, Dr. Von Hoff and his colleagues created the Tissue Banking Analysis Center (TBAC). TBAC, located in Phoenix, assays tumors for all molecular targets for which there are therapeutics. TBAC is sponsored by U.S. Oncology and the Molecular Profiling Institute, and is the only one of its kind, according to Dr. Von Hoff. More such clearinghouses like TBAC are needed throughout the world, he said. Patients who have their tumors analyzed at TBAC have the opportunity to participate in Phase I or II clinical trials enriched with patients whose tumors have specific molecular abnormalities. Such clinical trials are models for the approval of a new agent aimed at a specific molecular target in a patient’s tumor rather than designated for a particular histologic type of cancer, Dr. Von Hoff noted. He reported on a new clinical trial design for evaluating an agent against a target rather than against a tumor type. With this design, patients whose tumors have the specific molecular target are treated with an agent aimed at that target. Patients who experience a complete or partial remission continue taking the agent, while patients who progress are taken off the study (Figure 17). Researchers at TGen have instituted five such trials, and using TBAC was key to enabling these trials, Dr. Von Hoff pointed out. To see how commonly researchers would be able to discern a genetic abnormality in patients’ tumors for which there are already agents that target them, Dr. Von Hoff and his colleagues conducted a pilot trial called Target Now (Von Hoff et al., 2006). This study of 112 cancer patients found
OCR for page 63
Improving the Quality of Cancer Clinical Trials: Workshop Summary FIGURE 17 New clinical trial design for evaluating an agent against a target rather than against a tumor type. With this design, all patients receive a test for the FGFR mutation. Those with the mutation receive treatment designed to target the mutation. If the patient experiences a positive response, the therapy is continued. ACRONYMS: FGFR (fibroblast growth factor receptor). SOURCE: Von Hoff presentation (October 5, 2007). that standard immunohistochemistry assays for 13 possible targets found at least one potential target in about three-quarters of the patients, with an average of 1.6 targets per patient for which a conventional therapeutic agent was available. Microarray analyses found an average of 11 targets per patient for which there was a potential therapeutic agent, and virtually all patients had at least one potential target identified with this analysis. The physicians of the patients in this study provided abundant anecdotal evidence that this approach has been remarkably effective in some cases. For example, a patient with advanced ovarian cancer, who progressed on four prior regimens, responded to tamoxifen after estrogen receptors were found as a target. However, he said, there have some dramatic anec-
OCR for page 64
Improving the Quality of Cancer Clinical Trials: Workshop Summary dotal failures of the approach as well. But overall, Dr. Von Hoff estimates that with this targeted approach to treatment, response rates range from 26 to 30 percent, and the response rates are even higher for patients with rare tumors. Meanwhile the average response rate for patients in Phase I clinical trials is about 4 percent. He and his colleagues are currently conducting a prospective clinical trial called the Bisgrove trial to assess this more accurately. The endpoint measured in this trial is time on the therapy selected by molecular profiling, versus time on the therapy the patient had just received prior to the study. The rationale for this endpoint is that the length of time a patient responds to a therapy usually gets progressively shorter with successive therapies as disease progresses. Thus, if the time on therapy increases, it suggests that the profiling-selected therapy has changed the natural history of the disease (Box 1). “If the results from this trial are promising, we will have to rethink whether or not Phase I trials should be done in patients who are not profiled,” Dr. Von Hoff said. “Discovery and use of biomarkers is tough in drug development—it takes a long time,” he concluded. “But we shouldn’t be paralyzed by that and instead should focus on the contexts of vulnerability that are deletions, mutations, or translocations. These are easier to find and will probably foster more dramatic results within smaller clinical trials.” He added, “It is BOX 1 Details on the Endpoint for the Bisgrove Trial Usually the period of time a patient is on successive therapies is progressively shorter. If period B is greater than period A, the profiling-selected therapy has changed the natural history of the patient’s disease. If 30% of patients on this Bisgrove trial have period B longer than period A, then molecular profiling helps. SOURCE: Von Hoff presentation (October 5, 2007).
OCR for page 65
Improving the Quality of Cancer Clinical Trials: Workshop Summary very feasible to molecularly profile nearly all patients’ tumors [for patients] who are candidates for Phase I trials, and there is a great need for clearinghouses where patient tumors can be sent to be assayed for their context of vulnerability.” Panel Discussion In the panel discussion following the presentations, Dr. Massion was asked what is needed to improve biomarker technologies to make them more useful for developing better and more efficient trials. He responded that “our ability to mine these datasets that we obtain from genomics and proteomics is falling behind our ability to generate these data. There is a great need for analytical tools and ways to analyze these data. We are trying to reconstitute a puzzle that is extremely complex.” He added that biomarkers also have to be discovered within specific clinical contexts that address the heterogeneity of cancers rather than testing them within broad populations. “If you address lung cancer as a whole, for example, you are limiting yourself and then you are actually going to face a lot of difficulty in applying biomarkers to specific subgroups. We should consult with our clinicians and epidemiologists to think about biomarker discovery in the specific clinical context and then rapidly test the biomarker in that context, and our preclinical models should mimic that clinical context.” Later in the discussion, Dr. Joe Gray from the University of California, San Francisco, added that microarrays generate an enormous volume of data, with a nearly infinite number of marker combinations that might be predictive. Rather than abstractly analyze these data and search for any pattern that might be predictive, he suggested analyzing the data with the awareness that the data are informative about the underlying biology of specific molecular pathways or networks. “We need to organize the data in the context of the biological process that is deregulated so that any of the following 27 markers actually inform you that it is [for example] the BRCA DNA repair pathway that is actually deregulated and all of these markers ought to point you to that. Until we start thinking about interpreting the data in that context, we are going to be lost in this chaos of marker space,” he said. Dr. Massion also reiterated the need for rigorously validating biomarkers. Such validation can be aided by establishing centralized repositories of prospectively acquired patients’ samples, and by collaboration among institutions so biomarkers can be validated across institutions and platforms.
OCR for page 66
Improving the Quality of Cancer Clinical Trials: Workshop Summary Dr. Heath was asked how the use of predictive markers would affect trial design and implementation. He replied that, although predictive markers would have great value in stratifying patients for clinical trials, the challenge still remains to generate for these trials predictive markers that are noninvasive, such as those found in the blood or another readily accessible body fluid as opposed to the tumor, which requires an invasive biopsy. He pointed out that such noninvasive markers cannot directly detect a translocation or other genetic abnormality, but rather reflect the status of the tumor. Such reflection requires multiple-parameter measurements. Yet physicians and diagnostic companies are more familiar with single-parameter tests such as the prostate-specific antigen (PSA) test, according to Dr. Heath. “They are very uncomfortable looking at a panel of markers that goes through some computation program to give them back an answer, and I think there is a significant amount of physician retraining that has to be done to counter this,” he said. Dr. Von Hoff asserted that Dr. Heath underestimates physicians. “We would love to work with you on this because medicine is really very pattern oriented and physicians are used to putting all those patterns into their decision-making process every day,” he said, adding the example that antibiotic sensitivity testing “is never just black and white”; physicians have to consider many parameters when selecting the appropriate antibiotic for their patients. Dr. Heath noted, however, that “diagnostic companies have been reluctant to go down this pathway, and without commercialization, you can’t get it into people’s hands.” Dr. Mills added that proper validation of predictive biomarkers through clinical trials would be a way “to reassure the practicing physician that the pattern is reproducible and does mean something effectively.” Dr. Mills then asked Dr. Sullivan if the use of predictive markers would increase the number of patients willing to participate in cancer clinical trials. Noting he had no data to back up his opinion, Dr. Sullivan said he suspected that predictive markers would increase the number of such patient volunteers because “if patients had some sense the tests are being used to intelligently sort them out to where there might be some benefit to them and less harm, I suspect that they would find that appealing.” Patient advocate Kathy Meade of the Virginia Prostate Cancer Coalition added that predictive markers would appeal to many of the prostate cancer patients she deals with who are often interested in mechanistic explanations for the tests and treatments they receive and appreciate biomedical thinking that is “outside the box.”