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 11
3 Scientific Challenges in Developing Investigational Combination Therapies Participants identified numerous scientific challenges to developing investigational combination therapies, including the need for: • etter animal models and validated preclinical tests; B • etter dosing and treatment schedules to avoid toxicity, yet be B effective; • etter benchmarks, endpoints, and clinical trial designs for combi- B nation therapies; • way to prioritize which combinations to test; A • way to select patients most likely to respond to combinations; A and • ore basic research on the molecular mechanisms that underpin M cancer and how they interact. IMPROVING PRECLINICAL DEVELOPMENT OF INVESTIGATIONAL COMBINATION THERAPIES Standard preclinical development of drugs involves assessing the effects of varying concentrations of experimental compounds in in vitro or animal models and using those results to determine initial doses to test in clinical trials. Such preclinical development presents numerous challenges that may be exacerbated in the development of combination therapies, including cell lines or animal models that do not adequately mimic the tumor, tumor microenvironment, or the propensity to develop 11
OCR for page 12
12 COLLABORATIONS TO DEVELOP COMBINATION CANCER THERAPIES resistance, and a lack of biomarkers for efficacy. In addition, many animal models do not adequately mimic the immune response to tumors, so it is difficult to assess how immunotherapies are working in those models. There are also challenges that are unique to the development of combina - tions. For example, animal models appropriate for one therapeutic class might not be appropriate for another class with which they are being combined. These challenges were discussed at the workshop, as well as ways to address them. Key Suggestions for Improving Preclinical Development of Combinations from Various Workshop Participants • F unds to develop novel animal models that better mimic human cancer • U se of non-cancer animal models (e.g., as autoimmune or infectious disease models) as surrogate efficacy mod- els for anticancer immunotherapies • S trategies to ensure target engagement and inhibition • I nnovative approaches to maximize dose and schedule of combinations • B etter ways to distinguish on-target versus off-target toxicities • G reater use of animal models to identify resistance mechanisms • G reater use of statistical modeling Inadequate Models of Human Tumors and Tumor Microenvironment Dr. Lewis Cantley, professor of medicine at Harvard Medical School and director of the Beth Israel Deaconess Hospital Cancer Center, noted that cell lines do not adequately model the diversity of tumor types, but rather those tumor cells that can grow in a petri dish or under other com - mon laboratory conditions. “Once you establish these cell lines, they have been selected for and evolved to grow on plastic and they are not selected to grow in vivo. They have clearly evolved away from the original tumor from which they arose and do not represent what you see in the disease,” he said. Consequently, he suggested that the most appropriate models in which to test cancer therapies are mouse explant models in which the tumor cells are growing within the animal, which ideally should be
OCR for page 13
13 SCIENTIFIC CHALLENGES IN DEVELOPING COMBINATION THERAPIES a “humanized” mouse.1 Alternatively, he suggested researchers mutate the same targets that are altered in the human cancer in the same tissue at the same time of development in the animal model. “Both of these are more powerful approaches than what we currently do,” Dr. Cantley said. But Dr. Kurt Bachman, Head of Translational Medicine and Biology for the Cancer Metabolism discovery unit at GlaxoSmithKline (GSK), pointed out that multiple tumor explants are needed to capture the diver- sity of tumor types. The tumor cells available for explants may not have the tumor subtype likely to respond to the combination therapy being tested preclinically. “We want to target K-ras mutant lung cancer, but those explants may not have K-ras mutations,” said Dr. Bachman. It is also more expensive to test therapies in explant animal models than in numerous cell lines, Dr. Stern added. Dr. James Zwiebel, chief of the Investigational Drug Branch in the NCI Cancer Therapy Evaluation Program (CTEP), noted that NCI recently launched the Center for Advanced Preclinical Research, which will serve as a national resource for comprehensive pre - clinical testing of anti-tumor efficacy and selectivity, biodistribution, and metabolism in early-stage candidate drugs using genetically engineered mouse models (NCI, 2011a). “That’s an approach that hopefully will gain traction,” he said, and added that Dr. Terry Van Dyke is coordinating this effort and looking for interested parties to participate in it. Dr. Bachman stressed that understanding how the tumor microen - vironment affects growth of the tumor is crucial to improving cancer therapy and pointed out that his laboratory is starting to grow tumor cell lines in different microenvironments to see how that influences the action of inhibitors they’ve developed. He noted that effects seen in cell lines grown in three-dimensional cultures are different from when they are grown in standard culture conditions. “We are doing a lot of experiments to see if our culture conditions shift anything so the cell line looks more like a primary tumor that we can use to better predict what we are going to see in the clinic,” he said. Inadequate Models of Human Immune Responses to Cancer Dr. Nils Lonberg, senior vice president of Biologics Discovery at Bristol-Myers Squibb, noted that immunotherapy combinations cannot be tested in standard tumor models in which tumors are grafted onto immunodeficient mice. Dr. Haleh Saber, supervisory pharmacologist in the FDA Office of Oncology Drug Products, concurred, adding that she 1 Humanized mice have become an important research tool for the in vivo study of hu- man cells and tissues. Humanized mice are immunodeficient mice engrafted with human hematopoietic cells or tissues, or mice that transgenically express human genes.
OCR for page 14
14 COLLABORATIONS TO DEVELOP COMBINATION CANCER THERAPIES wanted to test a treatment for leukemia that combined genetically modi - fied human immune cells with a small molecule that was designed to activate a particular gene in the cells. “There was no in vivo model so we couldn’t do the animal pharmacology and toxicology studies,” she noted. In addition, models appropriate for one type of therapy—a vaccine, for example—might not work for another type, such as a small molecule, noted Dr. Ramzi Dagher, vice president for Worldwide Regulatory Strat - egy and regulatory head for the Oncology Business Unit at Pfizer, Inc. Thus, finding the appropriate model in which to test their combination can be challenging. Dr. Lonberg added that animal cancer models tend to be limited in how well they mimic the full spectrum of interactions between the host and the tumor that are key to assessing how well combination immu - notherapies are working. He suggested using surrogate efficacy models, such as autoimmune or infectious disease models, to assess the effects of combinations of agents in immunotherapy. For example, the NOD 2 autoimmune mouse model can show synergy between immune system molecules by revealing a heightened autoimmune response, such as diabetes, when both molecules are combined compared to when they are given singly. Some researchers, such as Dr. Rafi Ahmed at Emory Uni - versity, have also used chronic viral infection models, particularly the LCMV3 mouse model, to reveal interaction between various components in the host immune system and the effects of that interaction on the viral load of infected cells (Kim and Ahmed, 2010). Alternatively, researchers, such as Dr. James Allison at Memorial Sloan Kettering Cancer Center, have tested combination immunotherapies preclinically by creating ani- mal versions of the human antibodies or other immunotherapies that have been developed, and testing those in animals with intact immune systems. But even these animal models may not fully mimic how the human immune system interacts with the tumor, according to Dr. Lonberg. He pointed out that the initial immune response to a tumor is an elimina- tion phase in which the host immune system attacks the tumor. But then an equilibrium ensues. During this equilibrium phase, tumor cells express immunoevasion molecules that enable them to survive in equilib - rium with the host immune system, with occasional tumor cells escaping immune defenses. 2 Non-obese diabetic (NOD) mice exhibit a susceptibility to spontaneous development of autoimmune insulin-dependent diabetes mellitus. 3 Lymphocytic choriomeningitis virus (LCMV): This mouse model has been useful for ex - amining mechanisms of viral persistence and the basic concepts of virus-induced immunity and immunopathology.
OCR for page 15
15 SCIENTIFIC CHALLENGES IN DEVELOPING COMBINATION THERAPIES Dr. Allison’s animal model only mimics the initial elimination phase of an immune response. “You don’t have time in a tumor model like that to look at equilibrium and escape,” Dr. Lonberg said. So any advan- tages or disadvantages a combination immunotherapy might have in that regard cannot be predicted in preclinical testing in such animal models, he said. One participant stressed that it is critical that the therapeutic mecha - nism targeted by a treatment is present in the animal model in which it is tested, and is relevant to human disease. For example, immunotherapy that acts as a CTLA-44 blockade does not work in a lot of animal models, he said, although ipilimumab, a monoclonal antibody targeting CTLA-4, has recently been approved by the FDA5 for patients with advanced mela- noma in first- and second-line treatment. Dr. Cantley suggested using well-designed mouse models in which researchers can verify that each drug had adequately hit its target and had the desired downstream effects, that is, blocked the pathways that foster tumor growth. Evidence of those blocked pathways can then be gathered from the repeat biopsies taken from patients being clinically tested with the drug combination. Given current deficiencies, Dr. Stern suggested that there be better access to animal models for combination therapies or funds to develop them. “For wet bench investigators, the bottleneck is often moving from cell biology to animals,” he said. Combined Toxicity “Sometimes [drug] synergy is going to take us in the direction of enhanced toxicity,” Dr. Flaherty noted. For example, Dr. Engelman described a combination therapy that was highly effective when tested in vitro, but when he gave the maximum tolerated dose of each of those drugs to mice simultaneously, they killed every mouse tested. “You want to shut down these pathways [in tumors], but these are very important pathways for lots of cellular processes. It was only when we started play - ing with different schedules and doses that we were able to find the sweet spot where the mice lived and the tumors shrank,” he said. Dr. Engelman suggested being more creative and innovative in how combination therapies are scheduled and dosed. “Lots of these thera- 4 CTLA-4 (Cytotoxic T-Lymphocyte Antigen 4) is a protein that plays an important regula- tory role in the immune system. It is a member of the immunoglobulin superfamily, which is expressed on the surface of Helper T cells and transmits an inhibitory signal to T cells. 5 See http://www.fda.gov/newsevents/newsroom/pressannouncements/ucm1193237. htm (accessed December 14, 2011).
OCR for page 16
16 COLLABORATIONS TO DEVELOP COMBINATION CANCER THERAPIES pies are going to require three or four drugs, and a patient cannot be on all of them ad infinitum. They have to be pulsed or sequenced—we can’t just give them everything every day and only dose-reduce when they experience too much toxicity.” One regimen he suggested testing was giving monotherapy with periodic pulses of an additional treatment aimed at killing off those tumor cells that have become resistant to the monotherapy. Divergent Effects Depending on Dose or Sequence Dr. Patricia LoRusso, professor of internal medicine at Wayne State University Medical School and director of the Center for Experimen- tal Therapeutics at Karmanos Cancer Institute, and Dr. Lutzker gave an example of the extensive preclinical testing of combination targeted cancer therapies done by Genentech. This preclinical testing of an MEK inhibitor combined with a PI3K inhibitor, which took about a year, not only assessed additivity versus synergy in various genetically diverse cancer cell lines, but also tested a wide range of daily dosing versus intermittent dosing in animal models that aided subsequent clinical trial design. “Genentech did an excellent job in trying to figure out how best to dose escalate. In a first-in-patient study [Shapiro et al., 2011], we were able to conduct multiple arms simultaneously so that we could more efficiently define the combination of each of the drugs leading the pack, which has helped us in the final outcome of this study,” Dr. LoRusso said. “It is important when you get in the clinic to make sure you have drugs that can actually achieve the types of pharmacodynamic effects that you want or you hope to see in patients,” Dr. Lutzker added. For combinations that include immunotherapies, dose scheduling is key, Dr. Schlom pointed out. He noted that studies have found that tumor vaccines given after chemotherapy regimens are not as effective as those given prior to chemotherapy. Dr. Lonberg added that in one of his studies of two immunotherapy drugs, he found that when the drugs were given sequentially, there was a much more modest effect than when they were given together. Finding the appropriate dose of an immunotherapy is also critical, Dr. Schlom added, because many immune modulators have dual func- tionality, depending on dose, including many immune stimulants that have no effect at high doses. He noted that these potential therapeutics have been shelved merely because they showed no effects and were toxic at the maximum tolerated dose in Phase I studies, but they might have some useful effects at lower doses and in combination with other treat - ments. “It’s not only a matter of drug interaction, but it’s also a matter of
OCR for page 17
17 SCIENTIFIC CHALLENGES IN DEVELOPING COMBINATION THERAPIES what biologically makes sense in terms of sequencing and combining,” summed up Dr. Canetta. Differing Pharmacokinetics and Pharmacodynamics Between Mice and People Dr. Cantley stressed that the pharmacokinetics and pharmacodynam- ics of drugs are dramatically different in mice and in humans. “We need to get beyond this fear that if a combination kills mice it’s therefore going to stop a clinical trial, because those mice data don’t mean anything,” he asserted. Dr. Saber added that all oncology drugs are toxic and “the ques- tion is, can you monitor those toxicities? Most of the time we can and we adjust the dose so you’re good to go.” But Dr. Engelman noted that although pharmacokinetic and pharma - codynamic data do not directly transfer from the mouse to human, they can suggest a framework for how to reduce dosing to counter toxicity. For example, researchers can use the mouse to test the effects of reducing both drugs on both targeting and toxicity versus reducing the dose of just one drug, or keeping the dose of both drugs, but increasing the duration between doses. Dr. Donald Berry, professor of biostatistics at the MD Anderson Cancer Center, suggested going from bench to bedside and back to the bench by doing Bayesian statistical modeling of mouse preclinical test results the same way one would do for a clinical trial. “If there is no relationship between the mouse results and the human results, then we will just focus on the clinical aspects,” he said, “but there is a tremen - dous opportunity to augment the one with the other. You can do it with a statistical model.” Dr. LoRusso stressed that “we need to have much higher standards as to what we are considering effective combinations preclinically. I don’t know that the models have failed us. I think the way we are interpreting the models is what’s really failing us.” Dr. Engelman added that seeing a treatment response that is greater than a control response in preclini- cal tests does not necessarily mean that the treatment will cause clinical responses, but rather that the treatment has a biological effect. It is more important that the treatment causes significant tumor shrinkage in pre - clinical tests, he said. “If we can’t see tumor regressions in a simple 200 mm cubed tumor—which is the most homogeneous sensitive model— than what’s the likelihood that Mr. Jones, who has a huge amount of cancer that is heterogeneous, is going to benefit?” he said. But Dr. Lutzker countered that most human tumors do not grow as fast as tumors in mice. “I’m not prepared to give up on a combination just
OCR for page 18
18 COLLABORATIONS TO DEVELOP COMBINATION CANCER THERAPIES because in a xenograft6 it didn’t shrink the tumor,” he said. Dr. Engelman agreed and said he would view a lack of tumor growth in a xenograft model as a positive indicator, but added that “maybe as a community, we have been too accepting of seeing a biological effect and assuming that would translate into a therapeutic benefit in the clinic.” On-Target Versus Off-Target Effects Dr. Cantley suggested determining whether the limiting toxicity seen in animal models stems from how the combination affects the target, or alternatively whether it is due to how one or both drugs affect something other than the target. “If it’s on-target combined toxicity, then you have done the best you can. If it’s off-target, that means you try another com - bination, another PI3K or MEK inhibitor, for example. Fortunately, we have 18 of one and 7 of the other, so the probability that all combinations are going to have the same toxicity is unlikely,” he said. He noted that often doses of the combination hit the targets hard enough before toxic- ity is seen either in the mouse or the human. “Sometimes you don’t need to reach the mean toxicity because the toxicity is not on target,” he said. Dr. Saber suggested basing dose selection on data, when available, from Phase I clinical trials with the single agents that researchers plan to use in combination. Often sponsors will test a few doses of the single agents in people before combining them. But Dr. Roy Herbst, professor of medicine and chief of the Medical Oncology Section at Yale Comprehen - sive Cancer Center, said it is possible that lower doses of the two drugs combined might be more effective than the same dose of either agent used singly. Identifying Resistance Mechanisms Dr. Cantley described in his presentation how he often goes back and forth from bench to bedside. He uses animal models to determine what causes resistance to targeted treatments, and thus what treatments should be combined. He does this by doing a mutational analysis of the tumors removed from drug-resistant mice. For example, through this procedure he has discovered that resistance to a PI3K inhibitor can occur through amplification of MET. Armed with this information, he biopsies patients who have not responded to a PI3K inhibitor to see if their tumors also have MET amplification or produce high levels of MET protein. If that is 6 A xenograft is a surgical graft of tissue from one species (in this case, a human) to an unlike species (in this case, a mouse).
OCR for page 19
19 SCIENTIFIC CHALLENGES IN DEVELOPING COMBINATION THERAPIES the case, he will consider entering these patients into a trial that tests a combination of a MET inhibitor with a PI3K inhibitor. PRIORITIZING COMBINATIONS TO TEST Strategies for Prioritizing Which Combinations to Test in the Clinic as Suggested by Various Workshop Participants • U sing stricter preclinical benchmarks for effectiveness, such as tumor shrinkage, and demonstrating consistent effects in multiple animal models • D emonstrating adequate pharmacokinetics and evidence of target activity at clinically relevant doses • D oing high-throughput in vitro screening of drug combina- tions to detect synergy • S ubprioritizing so there is testing of the best drugs of each class • U sing genetic analyses and response biomarkers • T esting combinations that optimize the benefit of already approved drugs The growing number of targeted therapies that could be tested in combination, as well as the limited government and industry resources for such testing and the finite number of patients in whom combinations can be tested, suggests the need for a better way to prioritize which combinations get tested in clinical trials, several participants pointed out. Such prioritizing is key to developing a focus for patient advocates, federal agencies, and pharmaceutical chief executive officers (CEOs), said Dr. Michaele Christian, former NCI CTEP director, so everyone knows what the high-priority combinations are. But such prioritization can be challenging. As Dr. June noted, even in restricting combination therapy to combinations of immunotherapies, there is “a menu that is much too large to test in a combinatorial approach without some way of prioritization.” Dr. Engelman added, “We are going to have more combinations than we have patients.” Dr. Engelman suggested using stricter preclinical benchmarks for effectiveness when deciding which combinations to test in the clinic. One of those benchmarks should be seeing tumors shrink in animals, as opposed to blocking tumors from forming or from growing. “Lots of times we get excited about a biological effect, yet the tumor still grows slowly or the cells still grow slowly and that does not predict for efficacy in the clinic,” he said. Dr. Engelman added that he would prefer to see
OCR for page 20
20 COLLABORATIONS TO DEVELOP COMBINATION CANCER THERAPIES synergy versus additivity in preclinical tests, but the most important effect is seeing the tumor shrink, regardless of whether it is caused by synergy or additivity. He also noted that he has been impressed with a web-based system7 that Drs. William Pao and Mia Levy at Vanderbilt University have built to disseminate information on patients’ tumor mutations and responses to various therapies to enable a genetically-informed approach to cancer medicine. The My Cancer Genome website is an international collaboration of contributing physicians and physician scientists that com- piles information on the mutations influencing cancer progression and growth, potential therapies that may be effective against specific muta- tions, and available clinical trials that target specific mutations. These data can be used to prioritize which combinations of targeted cancer therapies should be tested in the clinic, and can inform clinicians at the point of care about tumor mutations and possible targeted therapies. Dr. Bachman said that when testing combinations in cell lines, he also aims for finding synergistic, not just additive, results. Dr. James Doroshow, deputy director for Clinical and Translational Research at NCI, suggested that combinations be tested clinically only if they work in at least three xenografts, and that they be based on a biologi- cal mechanism for which there is an assay. Before testing a combination clinically, Dr. Helen Chen, associate branch chief of the Investigational Drug Branch at NCI, said that the agents in the combination should have already demonstrated adequate pharmacokinetics and some evidence of activity or target engagement at clinically relevant doses and exposures. Ideally the individual targets should be validated, and priority should be given for combinations that have shown a higher degree of efficacy, such as those that have converted growth inhibition to a tumor cell kill, she said. Dr. Cantley pointed out that sometimes agents used singly do not have a significant effect because they are not tested at high enough doses, and subsequent combination therapies using those same agents do show an effect. “We have learned that you really have to hit these targets hard,” he said. In addition, some immunotherapeutic agents only work in com - bination and not singly, several experts in this field pointed out. Dr. Chen also suggested assessing whether the synergism of the com - bination is seen consistently across all preclinical models, and if not, whether a predictive marker can be identified to choose those patients likely to respond to the synergistic interaction. Dr. Lutzker added that “in order to do small clinical trials, it is critical to try to understand which patient you want to test the combination in,” and suggested not doing any clinical testing unless there is a biomarker test that can be done simulta - 7 See www.mycancergenome.org (accessed December 14, 2011).
OCR for page 21
21 SCIENTIFIC CHALLENGES IN DEVELOPING COMBINATION THERAPIES neously to assess which types of patients respond or do not respond to the treatment. To systematically assess which combinations should be tested clini - cally, Dr. Stern is collaborating with Dr. Marcus Bosenberg to conduct high-throughput screening of 40 compounds at 3 concentrations on 30 tumor cell lines that model common human combinations of mutations. The tested drugs were heavily weighted toward drugs that target tumor cell growth signaling, but included conventional cytotoxic therapies as well. This research has revealed numerous additive or synergistic inter- actions in various combinations for specific tumor genotypes, some of which revealed novel pathway interactions. Drs. Stern and Bosenberg are currently linking these functional results to phosphoproteomic data as well as exome sequencing so researchers can use it to predict combina- tion drug sensitivity according to genotype, such as by BRAF status, ras 8 status, and other genetic aspects of a tumor. Dr. Barrett suggested a more personalized approach to determining which combinations of therapeutics should be tested in patients. Such an approach can be taken by assessing the genomes of patients’ tumor, iden- tifying which genetic mutations are driving the growth of those tumors, and then devising combinations that block those drivers. To do this, Dr. Barrett uses a small amount of tumor tissue that can be obtained from standard or needle biopsies. Then he uses flow cytometry to separate tumor cells based on the duplications or deletions of chromosomes or other characteristics that can be measured by examining individual cells with a laser. He then uses comparative genomic hybridization to geneti - cally profile these subpopulations of tumor cells, noting that multiple populations of tumor cells can be present in a single biopsy. From this pro- filing he said it is possible to detect more than 100 chromosomal aberra- tions, which are then quantified and ranked according to how likely they are to be influencing the development or growth of the patient’s tumor. Based on this information, statistical and bioinformatics techniques are then used to depict what he calls a “wiring diagram” of the activated pathways that are fueling the tumor. This is then used to determine the most appropriate combination therapy. “There is lots of heterogeneity, but we find all the populations and can purify them out and often find convergence on these pathways,” Dr. Barrett noted. “What we need to do is identify the concurrent aberrations and mutations in each tumor cell population if genomics is really going to help advance the development of these targeted therapies, particularly combination therapies.” As Dr. LoRusso noted, there are a lot of drugs of similar class. For 8 The ras family of genes code for proteins involved in cell signaling, cell growth, and apoptosis. Mutations in ras genes can lead to cancer.
OCR for page 26
26 COLLABORATIONS TO DEVELOP COMBINATION CANCER THERAPIES should be,” Dr. Barrett said, adding that “even 2 weeks can be almost a lifetime for some of these patients.” Dr. Perlmutter concurred, saying, “For some patients that 2-week wait is extra scary. We have to not only get better and cheaper in our testing, but we also have to get faster in our testing.” To do more detailed whole-genome sequencing is more time consuming, Drs. Engelman and Barrett noted, and currently is not practi - cal for patient selection, although whole-exome sequencing has led to the discovery of a feasible number of exons—600–800—that could be assessed within 3 weeks and be potentially clinically useful, Dr. Engelman added. Dr. LoRusso gave a positive example of patient selection in a Phase II clinical trial of a MEK inhibitor tested in combination with a BRAF inhibi- tor by Jeff Infante of the Sarah Cannon Research Institute in Nashville, Tennessee. She said Dr. Infante preselected his patients based on the presence of BRAF mutations in their tumors, and preliminary results suggest a better response rate with the combined therapy than what was observed with the single agents (Infante et al., 2011). The trial is still ongoing, with the majority of enrolled patients continuing in the study. Dr. Lutzker added that he preselected patients in a trial that tested a MET monoclonal antibody plus erlotinib. Such preselection was done using an assay for high-level expression of MET by immunohistochemistry. This Phase II study showed strong efficacy in this patient group in terms of progression-free and overall survival, he said. Dr. Cantley noted his team of researchers spent a lot of time discovering and testing biomarkers for early response that were quantitative, predicted clinical outcomes, and worked well across institutions. These biomarkers included those that could be evaluated in positron emission tomography (PET) scans. Dr. LoRusso questioned the relevance of the genetic profiling being done in metastatic tumors to determine appropriate treatment combina - tions to patients with non-metastatic disease, who might be more likely to benefit from combination therapy. “What are the risks that are involved if we are studying these combinations in the wrong patient population at the wrong clinical stage?” she asked, especially if negative findings in a metastatic patient population led to combinations being rejected for further testing in patients with early-stage disease. Dr. Sharon Murphy, scholar in residence at the Institute of Medicine, suggested conducting more combination therapy trials in pediatric cancer populations. She noted that there are extensive tissue banks of pediatric tumors that are clinically well annotated and could serve as valuable resources for investigators. “When we think about combination targeted therapies or targeted treatments, we should think about childhood cancer, which arguably is a better model because genetically it’s simpler than many adult forms of cancer. There are fewer signaling pathways, and children need these drugs too,” she said. She added that investigators
OCR for page 27
27 SCIENTIFIC CHALLENGES IN DEVELOPING COMBINATION THERAPIES should not have to wait to test promising combination therapies in pedi - atric populations until after they show promise in adult Phase I trials. Dr. Chen responded that CTEP has been doing a lot of Phase I and II testing of investigational cancer treatment combinations in the pediatric population. Dr. Samuel Blackman, Director of the Oncology Early Development Unit at GSK, agreed that pediatric populations should be engaged to achieve early proof of concept, and for some subtypes of cancers such testing is easier to do in the pediatric than the adult population because pediatric patients with these tumors tend to be grouped according to the genetic drivers of their tumors and treated in disease-specific programs offered in major pediatric cancer centers. Adaptive Trial Designs According to Dr. Berry, adaptive clinical trial designs are especially suited for answering the numerous questions that combination therapy raises, such as which of several possible drug combinations, patient selec- tion biomarkers, doses, and dosing schedules are the safest and most effective. He said that an adaptive design enables researchers both to answer questions as well as to raise new questions and test new hypoth - eses during the course of a trial. Adaptive trials use Bayesian statistics to model and predict during a trial which option is most likely to be beneficial based on the results to date. Researchers use these predictions, while the trial progresses, to increase the number of patients being tested in the arms showing the most promise, and reduce or drop the number of patients being treated in those arms generating poor results. For example, Dr. Berry designed an adaptive Phase I/II trial for a two-drug combination therapy for leukemia in which the admissible doses expanded or contracted during the trial depending on toxicity and effectiveness. For trials of two agents given separately or together, patients are randomized to each possibility, but “as you are learning, for example, that agent 1 is not doing as well as you might have hoped, you might give it a lower probability [and assign less patients to receive this agent],” he said, adding, “At some point we make a decision as to what is going to be the confirmatory stage [for the agent that is having the best results]” (see Figure 3-1). Another example of an adaptive trial is I-SPY 2 TRIAL (Investiga- tion of Serial studies to Predict Your Therapeutic Response with Imaging And moLecular analysis 2; see Figure 3-2 and Appendix A), which tests various treatments for breast cancer used singly or in combination while simultaneously assessing which patient selection biomarkers are most appropriate for each treatment. This trial has five experimental arms in which new treatments are “plugged in” to be tested once other tested
OCR for page 28
28 COLLABORATIONS TO DEVELOP COMBINATION CANCER THERAPIES FIGURE 3-1 An example of an adaptive trial design that includes several treat- ment arms at the start of the trial (in this case a factorial design: agent 1, agent 2, the combination of agents 1 and 2, and a control arm). As information accrues, the arms that are not performing well can be dropped. Accrual continues with no interruption to carry out interim analyses. Interim analyses may be continued into the confirmatory stage. image Figure 4, xed SOURCE: IOM, 2010b. agents progress to confirmatory trials or fail to show favorable results, Dr. Berry said. It has a factorial design in which single agents plus standard of care are tested against combinations of agents plus standard of care. As the trial progresses, single agents may be dropped because the results are more favorable when they are used in combination, “but meanwhile we have some experience in the single agents and some notion of synergy or additivity,” Dr. Berry pointed out. The I-SPY 2 TRIAL is innovative in that there is an adaptive design with regard to both treatment and the patient selection biomarker for the treatment. Dr. Berry stressed this is critical given that researchers continue to uncover new biomarkers for patient selectivity. “We have to figure out ways that we can update that information and use additional markers to understand who benefits from treatment. The only way to do it is to build it into our clinical trial structure and learn as we go,” he said. Dr. Iannone highlighted that trials can be more efficient if patients in a clinical trial of
OCR for page 29
29 SCIENTIFIC CHALLENGES IN DEVELOPING COMBINATION THERAPIES I-SPY-like TRIAL A R D A A N P D T Outcome: O Population I pathCR M V of patients or PFS I E or OS Z L E Y FIGURE 3-2 Design features of an I-SPY-like trial. In this design, patients are adaptively randomized within biomarker subsets to multiple therapies or combi - nation therapies. The bottom four arms in the figure constitute a factorial design New Figure 5 for agents C and D that are being investigated simultaneously with agents A and B. All comparisons are to standard of care. The therapies’ benefits are assessed within patient subsets defined by biomarkers. The adaptive randomization aspect enables study arms that are performing well within a particular patient subset to be assigned with higher probability to patients in that subset. Arms that are per- forming poorly are assigned with smaller probabilities, and if they do sufficiently poorly in a subset, they will no longer be assigned to those patients. If they do sufficiently poorly in all subsets, then they are dropped from the trial. Drugs that have established a sufficiently clear biomarker signature based on their perfor- mance will be graduated from the trial. NOTE: OS = overall survival; pathCR = pathological complete response; PFS = progression-free survival; SOC = standard of care. SOURCE: Berry presentation (June 13, 2011). combination therapies can easily move from one arm to another, based on some early measure of success or failure through a specific pharmaco- dynamic response biomarker. Efficiency is especially improved if there is a high negative predictive value and patients can be quickly assigned to another treatment arm without needing another baseline biopsy. “There is an efficiency for investigators, but there is a huge efficiency and potential upside for patients as well,” he pointed out. Dr. Larry Rubinstein, statistician at the NCI Biometric Research Branch, agreed that the adaptive trial designs Dr. Berry presented were well suited to trials of investigational drug combinations because they avoid the problem of setting the maximum tolerated dose prematurely, which often occurs with standard Phase I trial designs that have a small
OCR for page 30
30 COLLABORATIONS TO DEVELOP COMBINATION CANCER THERAPIES number of patients. Adaptive trial designs also address the ethical imper- ative of aiming to benefit patients by focusing on the dose combinations that are most promising, he said. The randomization of patients to bio- marker tests is also an important way to assess whether the biomarker indicates if a patient is likely to do better when given a particular treat - ment, versus whether a patient is likely to do better regardless of which treatment he or she receives, that is, it enables researchers to distinguish between predictive and prognostic markers. But in addition to assessing the clinical toxicity of varying doses of experimental agents, Dr. Rubinstein suggested introducing in vivo pharmacodynamic assays for efficacy or toxicity, or even using these assays during the course of a clinical trial to assess whether the individual agents are hitting their targets and otherwise working mechanistically as expected. “This means you may end up terminating escalation for an agent, not on the basis of toxicity, but because you appear to have reached the limit of its efficacy,” he said. Such pharmacodynamic assays could be a part of an adaptive clinical trials design, he noted. Dr. Doroshow emphasized the importance of demonstrating a mecha- nism of action early in a clinical trial to validate one’s presumptions in this regard. “We have an enormous number of presumptions going into first-in-human studies, and often those presumptions are wrong,” he said. “It’s critical to get this proof-of-mechanism information for the subse - quent development of which combination to utilize. If you don’t have an assay to demonstrate target inhibition, it’s almost impossible to develop an appropriate schedule, in terms of relating systemic exposure to the targeting effect.” Thus, NCI is currently developing more than 50 assays for evaluating the mechanisms of action of molecularly targeted agents, Dr. Doroshow reported (see Appendix A). Dr. Steven Piantadosi, director of the Samuel Oschin Comprehensive Cancer Institute at Cedars-Sinai Medical Center, stressed that some sort of factorial clinical trial design must be used to investigate the interactions of agents when they are used in combination, and that there be enough sam- pling points in the “two-dimensional dose space” from which researchers can reap adequate information about how the response changes over that two-dimensional surface. Dr. Lutzker added that “modeling the dose– response curve is really critical in that aspect.” Repetitive Tumor Biopsies As a means for assessing whether drug agents are hitting their targets in patients, several conference participants suggested conducting assays on repetitive biopsies of patients’ tumors. Dr. Cantley also suggested examining repeat biopsies from clinical trial patients to assess not only
OCR for page 31
31 SCIENTIFIC CHALLENGES IN DEVELOPING COMBINATION THERAPIES whether each target of a combination therapy has been hit individually, but that “you have hit something that you know should be a consequence of inhibiting both,” he said. In addition to analyzing patient tumor samples prior to a clinical trial, Dr. Engelman noted that his research group biopsies every patient who becomes resistant to a tested targeted therapy. As many as five biopsy cores are taken in one procedure, at no greater risk to patients than a sin- gle biopsy, because these cores are removed through a single transducer needle that makes only one puncture to access tissue specimens from the site. “We could be more aggressive about getting tissue for lots of stud - ies,” he said. Dr. Engelman added that patients are more than willing to have such biopsies performed. Dr. Perlmutter agreed about the general willingness of patients to have needle biopsies performed, but noted some situations in which a biopsy may not be feasible. “I don’t think you would ask a brain can- cer patient to give you a biopsy, but many cancers are biopsied,” she said. “Patients are often more than happy to provide multiple biopsies, but doctors often do not request them,” she added, stressing that a trial should get as much data as possible and patients generally recognize that providing specimens is in their best interests. Dr. Cantley added that the combination of clinicians and patient advocates stressing the importance of the biopsies required in the clinical trials he has been involved with has led to patient willingness to have these biopsies performed and to enroll in protocols in which such biopsies are mandatory. Dr. LoRusso stressed that serial biopsies of patients are the best way to assess “not so much what went right, but more importantly what went wrong” in a clinical trial. She said within the previous year, her research team used full-time technicians to biopsy at least 300 out of 500 patients. “I feel we are still relying too heavily on surrogates and I haven’t found many surrogates that have led me down the appropriate path of tak- ing that drug forward into the appropriate patient subset,” she said. She added that imaging results are also inadequate surrogates. She uses imaging, such as PET or DCE-MRI (dynamic contrast enhanced-magnetic resonance imaging), to assess treatment effectiveness during the course of a clinical trial, but these are very expensive, she said, “for the amount of information that we are not getting because of the variability of multiple factors, including the heterogeneous patient population in a Phase I trial.” Determining Appropriate Dose and Schedule Several participants noted that it can be challenging to determine appropriate dosing because of the variability in how patients respond to different agents. Dr. LoRusso raised the question of whether “all doses of
OCR for page 32
32 COLLABORATIONS TO DEVELOP COMBINATION CANCER THERAPIES combinations are created equal or do we need to personalize the doses of the individual drugs relative to the mutational status and changes in the tumor,” which she said no one has explored yet, but added “it’s not an insignificant issue and I don’t think we can forget it as we are developing these combinations.” Dr. Doroshow suggested that NCI’s toxicogenom - ics program (see Appendix A) should help researchers find correlations between pharmacokinetics and systemic exposure with the genomic pro- files of various tumors according to the class of drug being tested. This information is being made public as it is gathered. Dr. Chen suggested that if the goal of the therapeutic outcome is to achieve a sustained major response or cure, then one should consider giving short but intensive doses that are lethal to the tumors. If that goal is not achievable and continuous therapy exposure is required, she sug- gested that lighter, less intensive therapy may have to be given so it can be tolerated, or that combination therapy be given sequentially rather than concurrently. Dr. Chen added that hundreds of clinical trials testing combinations of these targeted agents reveal they can be quite toxic. Often there is an increase in the severity and frequency of the known toxicity of the single agents used in combination, although sometimes new toxicities arise. In some cases there appeared to be synergistic toxicity, perhaps due to the nonspecific targets of these molecules, and significant dose reductions were required. Some combinations, such as the VEGF (vascular endothe - lial growth factor) inhibitor sunitinib and the mTOR inhibitor temsiroli- mus, had to be abandoned because of their combined toxicity. According to Dr. Chen, combinations of agents that target parallel pathways are less likely to have overlapping toxicity and are better toler- ated, as are agents with more specificity. She added that combinations appear to be less tolerable if they target the more downstream elements of signaling pathways. In all cases, the maximum tolerated dose based on cycles one and two do not appear to predict long-term tolerability. Dr. Chen stressed that what she has learned from all these clinical trials is that full doses of each individual agent are often not tolerable in combination, and that the adverse effects on normal tissues may limit the spectrum and degree of duration of combined target inhibition. This raises numerous questions about the best way to develop a combination dosing strategy to reduce toxicity. These questions are probably best addressed with more intensive preclinical studies to determine the optimal dose and schedule, she said. Such dosing will probably be based on the pharmacodynamics or pharmacokinetics required for synergism, keeping in mind that the dose required for synergism may not be the same as that required for single-agent activity, she said. Dr. LoRusso pointed out that the maximum tolerated doses of the
OCR for page 33
33 SCIENTIFIC CHALLENGES IN DEVELOPING COMBINATION THERAPIES drugs used in a combination are not necessarily meaningful. “There are various trial designs that could actually hurt you sometimes more than help you, depending on which is the most important drug, and how the ratios need to be defined in the clinical scenario,” she said. She added that often toxicities, such as rashes, are expected to be worse in combinations of drugs if each drug has shown such a side effect in Phase I trials. Dr. Herbst noted that about half of the combination targeted cancer therapies currently being tested clinically have shown dose-limiting skin toxicity. However, sometimes the side effect is not seen in the Phase I trial of the combined drugs. Also, additional toxicities not predicted by single-agent studies can surface when the drugs are tested in combination. “What we sometimes can predict or theorize based on preliminary monotherapy data may not actually come true when we do the combination,” Dr. LoRusso said. Dr. June added that the T cells used in cell-based immunotherapies are living and often long lasting and self-replicating, so they have dif- ferent pharmacologic and pharmacokinetic parameters than drugs for which simple clearances can be assessed. Because of this, a clinical trial design quite different from a standard Phase I approach is needed. For example, in a Phase I clinical trial of a cell-based immunotherapy, his research group tries to identify an optimal biologic dose rather than the more standard maximum tolerated dose. Appropriate Endpoints and Other Study Measures Researchers are finding that immunotherapies such as Provenge and various tumor vaccines used to treat cancer often extend survival without delaying time to progression, Dr. Schlom pointed out. These therapies often stabilize rather than diminish the size of tumors and may also extend survival without diminishing the growth rate of metastatic can - cers. These findings suggest that traditional endpoints may not be appro - priate for clinical trials of immunotherapies, according to Dr. Schlom, and that overall survival might be the best indicator of their effectiveness. It is not clear whether this applies only to combination immunotherapies or also to treatments that combine an immunotherapy with standard chemo- therapy or targeted treatments. However, he noted that preliminary data from one study (NCI, 2011b) showed that a tumor vaccine combined with docetaxel did extend time to progression over the docetaxel treatment given singly. Dr. June added that for many immunotherapies, determining the optimal biologic dose is the most appropriate aim of Phase I studies, as opposed to determining the maximum tolerated doses. Especially for immunotherapies that apply live cells, such as modified T cells, the typi - cal dose-escalation Phase I clinical trial design is not appropriate, he said.
OCR for page 34
34 COLLABORATIONS TO DEVELOP COMBINATION CANCER THERAPIES A few participants suggested there should be higher standards for clinical response in trials of combination therapies. “Given the fact that we will be running out of patients and resources [to test combination treatments], we need to be setting our bars way higher than we are,” Dr. LoRusso said. Dr. Wendy Demark-Wahnefried, associate director for Cancer Pre- vention and Control at the University of Alabama at Birmingham Com - prehensive Cancer Center, encouraged researchers to include lifestyle factors, especially measures of energy balance and obesity, when assess- ing the effectiveness of combination therapies because obesity has been shown to affect some of the same molecular pathways targeted by certain cancer drugs. Breast and endometrial cancers, for example, are hormon- ally driven cancers that are affected by obesity, she said, and collecting body mass index data at baseline and at follow-up of patients with these cancers being treated with combination therapies could provide useful information. Dr. Cantley agreed, adding that prostate, colorectal, and pan- creatic cancers are also affected by obesity, presumably through its effects on IGF1, and that he has suggested to industry to modulate those effects by using the diabetes drug metformin in clinical trials. “This is something we’re very much aware of,” he said. Dr. Chen stressed that though many targeted therapies show evidence of being therapeutic in the clinic when used singly, many of those treat - ments fail clinical trials when they are used in combination. For example, VEGF and EGFR inhibitors showed no effect when given with chemo - therapy to treat several cancer types, including colon, pancreatic, kidney, and breast cancers, she pointed out. Combinations that target mTOR have also failed Phase II or III trials. “Do we have a failure of the hypothesis or a failure of the clinical trials because we did not use the right dose or choose the right patients? All these possibilities are possible for different scenarios,” she said. Speeding Up the Collaborative Clinical Trial Process Dr. Vassiliki Papadimitrakopoulou, professor of medicine in the Department of Thoracic/Head and Neck Medical Oncology at MD Anderson Cancer Center, pointed out that multiple steps need to be satis - fied to have several pharmaceutical companies and academic institutions collaborate in combination therapy trials of investigational anticancer agents, including coordinating Institutional Review Board (IRB) reviews, data sharing and analysis, intellectual property agreements, Investiga - tional New Drug (IND) applications, etc. It can take years to accomplish all those steps so that a collaborative, multisite clinical trial of combina - tion therapy can begin. There is concern that during that lengthy start-up
OCR for page 35
35 SCIENTIFIC CHALLENGES IN DEVELOPING COMBINATION THERAPIES time, scientific advances will occur that might indicate that the combina- tions in the trial are no longer the most promising ones to test, she pointed out. “We need to speed things up,” she said. Dr. Cantley added that a major time impediment is acquiring the IRB approvals from multiple institutions. He suggested that presenting a strong trial concept initially to the IRBs can help speed things up, as can having regular face-to-face meetings and teleconferences, and having investigators with clinical trial experience on a research team, in addition to the Principal Investigators, to provide valuable advice and help oth - ers to benefit from their experience. Dr. Papadimitrakopoulou suggested that patient advocates can help speed up the process by putting more pressure on academia to make their IRBs more expedient. Dr. Perlmutter suggested that every multicenter trial have a single IRB12 and noted that for the I-SPY 2 TRIAL there are 15 different versions of informed consent. “It certainly adds expense and confusion that is totally unnecessary,” she said. Dr. John Hohneker, senior vice president and global head of devel- opment of the Integrated Hospital Care Franchise at Novartis Pharma AG, added that many institutions are afraid to commit to and execute an agreement without having their own IRBs approve it. Dr. Flaherty suggested that time and resources could be saved in the long run if there were a precompetitive venue for testing drug combina - tions in a limited number of patients—less than 20—to more rapidly sift out combinations likely to be effective in the clinic. Sponsors would have an incentive to contribute their drugs to such a system because it would be an efficient way of triaging combinations that they do not have the resources and the time to test, according to Dr. Flaherty. “We need to cre - ate some kind of mechanism for cranking through these combinations in relatively small patient numbers in a much more rapid fashion than we currently have the capacity to do,” he said. Drs. Christian and Flaherty also called for strong patient advocacy to support a list of vetted important targets and combinations that should have priority status for clinical tests. “Figuring out a way of having a rolling, ongoing dialogue about prioritization is absolutely critical for the early combinations,” Dr. Flaherty said. Dr. Christian added, “There are all these patient advocacy groups and we just need to figure out how to make them talk to each other about this most important topic.” 12 On July 26, 2011, the U.S. Department of Health and Human Services announced that the federal government is contemplating various ways of enhancing the regulations oversee - ing research on human subjects, as described in an Advance Notice of Proposed Rulemaking (HHS, 2011).
OCR for page 36