Federal Agency Roles in Cancer Drug Development from Preclinical Research to New Drug Approval: The National Cancer Institute and the Food and Drug Administration (2005)

The National Cancer Institute (NCI) has been involved in the discovery and development of many of the anticancer agents currently in use. At $50 billion in public funding over the past 30 years, it has been the largest such public investment in the world and one that is without parallel in any other therapeutic area. A guiding principle of NCI activities related to drug development has been that they must complement industrial efforts, and the goal has been to expedite the best molecules for cancer treatment from discovery into clinical trials (Monks et al, 1997).

NCI was the early leader in discovery and development of cancer drugs (short of production and marketing), but in the past few decades, the private sector has invested substantially in bringing new drugs to market, although it is still highly dependent on discoveries from NCI-supported research. NCI’s challenge is to complete the transition from the dominant or sole force to a complementary role that leverages the public resources that it commands to best advantage, as well as remaining a leader in new approaches to, and the underlying basic science for, drug development.

In 1944, the Public Health Service Act made NCI a division of the Public Health Service’s National Institute of Health (NIH) with the intent of encouraging NCI research. The American Society for the Control of Cancer also reorganized itself in 1944, becoming the American Cancer Society, and continuing to work in cooperation with NCI on research and educational activities (American Cancer Society, 2004).

When the NIH clinical center opened on the Bethesda, Maryland campus in 1953, clinical research projects in cancer were transferred to the center. However, grants for extramural research continued, particularly in the growing area of chemotherapy. In 1955, uncertain about industry interest or academic capacity and dissatisfied with NCI’s lack of direct involvement in drug discovery (Sausville and Feigal, 1994; Zubrod, 1984), Congress appropriated funds to NCI for a national drug development effort, the NCI National Chemotherapy Program.

Accordingly, the Cancer Chemotherapy National Service Center (CCNSC) was formed, primarily as an extramural program, with all of the functions of a pharmaceutical house run by NCI and operations dispersed in industry and academic institutions (Zubrod, 1984). With a 1955 budget of $5 million, initial contracts were let to four screening centers operating through the CCNSC (Zubrod et al., 1977). By the mid 1960s, the CCNSC had combined its intramural and extramural activities in drug development.

Also during this time, with another $5 million from Congress, NCI initiated the cooperative group model for testing chemotherapeutic agents in clinical studies. By 1958,

17 groups had been formed. The Eastern Cooperative Group, the first of the cooperative groups, carried out a trial comparing two drugs in breast cancer, Hodgkin’s disease, and melanoma that was designed by an NCI Clinical Panel. The Panel also assessed the early principles of clinical trials in cancer chemotherapy and reviewed trial data coming from an emerging number of cooperative groups around the country (Zubrod, 1984). This was the first active, prospective involvement of NCI in large-scale cancer clinical trials.

In 1976, the CCNSC was incorporated into NCFs Developmental Therapeutics Program (DTP), the current locus of federally funded preclinical drug discovery and development. The heart of DTP was to become a large-scale in vivo drug screening operation that, by the late 1970s, tested up to 40,000 compounds per year in a variety of mouse leukemia models (Chabner, 1990). More than 500,000 chemicals were tested in laboratory animals in this program, and several hundred were tested in clinical trials. By the late 1970s, some 45 chemicals had been found effective in various cancers. Overall, DTP has had a role in the discovery or development of approximately 40 percent of the current U.S.-licensed chemotherapeutic agents (Sausville and Feigal, 1999), with the rest coming directly from the domestic or international pharmaceutical industry.

Preclinical models used by NCI to select new drugs for cancer clinical trials have evolved over time due to improved understanding of the biologic factors that affect the success of treatment, such as the relationship of tumor cell growth kinetics to drug responsiveness, to retrospective analyses of correlations between clinical and preclinical efficacy, and to the development of the NCI Drug Information System, a computer inventory of compound structure and activity in the mouse models, that limits the screening of analogues and directs the focus to novel structures (Schabel, 1969; Skipper et al., 1970. Venditti, 1981, Venditti et al., 1984).

In the 1940s, the S37 tumor was used to screen 300 chemicals and plant extracts, but by 1955 mouse leukemia models were selected as the initial systems for large-scale screening because they were relatively inexpensive and allowed for high throughput of compounds (Monks et al., 1997; Vendetti, 1981, Venditti et al., 1984, Zubrod, 1984). From its inception, therefore, until the mid 1970s the mouse screen would be used to process more than 400,000 compounds (Khleif and Curt, 2000). There are examples where this appeared to predict well for the clinic, but there was concern that screening against animal leukemia may have created a bias toward drugs that were active only against rapidly growing tumors. While treatment of human leukemias and lymphomas had improved during this period, lesser improvements in the chemotherapy of most human solid tumors were achieved (Venditti, 1981, Venditti et al., 1984).

In the 1970s, the availability of the athymic nude mouse, a model with a defective immune system that accepted transplanted foreign (non-mouse) tumors (Giovanella et al., 1974), permitted inclusion of human tumor grafts in a screening panel to identify agents in addition to those selected by mouse tumor screens. Candidate agents were screened

against matched pairs of animal and human tumors from the same organ site to determine if tumor origin was a factor in drug selection (Venditti et al, 1984). To limit the number of compounds tested through this more complex approach, NCI used a relatively sensitive and cost-efficient in vivo prescreen—a specific mouse leukemia that was sensitive to most classes of clinically effective drugs—and the criteria for activity in this model were set low (Venditti et al., 1984). The panel of tumors against which selected materials was tested included mouse breast, colon, and lung tumors, human tumor grafts of the same types, and an additional mouse leukemia and melanoma.

This approach identified new active agents that would have been missed by the mouse tumor models. However, many key disease events, such as metastasis, did not occur in the graft models raising concerns they might miss effective drugs and limiting their predictive utility. In the 1990s, an NCI investigator tested 12 known anticancer agents against 48 human cancer lines transplanted into mice and found that 30 of the tumors did not show a significant response (Plowman et al., 1997). Subsequent changes in screening (see below) improved relationships between models and actual clinical results, but correspondence between experimental and clinical data has never been complete (Johnson et al., 2001).

While these changes took the NCI screen from a compound-oriented approach to a tumor-specific approach, the high cost required a two stage system in which the mouse leukemia model was the prescreen. Moreover, the bias against selection of drugs specifically active in solid tumors remained a problem, and demonstration of significant preclinical activity in human breast, colon, or lung cancer graft models by a given drug did not necessarily predict for clinical activity in patients with those diseases (Khleif and Curt, 2000).

By 1982, 2,164 compounds had been assigned to the tumor panel, of which 1,084 were tested against the complete panel. Although the NCI attempted to introduce more rational criteria into the selection of screening candidates, particularly novelty of structure and known biological activity, the majority of compounds tested continued to be random chemical entities submitted by chemical and drug companies and academic laboratories.

In the mid 1980s, NCI was questioning the value of the program given its slowing yield of promising drugs and failure to produce new solid tumor agents, but decided not to cancel out of concern that the private sector would not fill the void. New molecular screening targets, for example, growth-factor inhibitors, oncogene products, and protein kinases, were considered, but NCI decided to develop a cell-line-based screen representing the major classes of solid tumors because that would allow relatively inexpensive and rapid testing against broad panels of human tumors and would be adaptive to the needs of natural product screening (Chabner, 1990). The screen was designed so that for each compound tested, both the absolute and relative sensitivities of individual cell lines comprising the screen were sufficiently reproducible that a characteristic profile or fingerprint of cellular response was generated.

Since 1990, the program has used a primary in vitro screen followed by evaluation in hollow fiber (see discussion below) and tumor graft models (Sausville and Feigal, 1999). The decision to abandon large-scale in vivo screening disappointed many in the pharmaceutical industry (DTP Program Review Group Report, 1998). NCI maintained that the in vitro screen would provide a practical means for the selection of compounds of interest for in vivo testing, thereby reducing the randomness and cost of less discriminating screening methods and allowing universities (which lacked the resources to submit the larger quantities of agents needed for in vivo testing) to submit more compounds for initial testing (Grever et al, 1992).

The infrastructure for a large-scale cell-line screen was built at NCI’s Frederick Cancer Research and Development Center, initially equipped for 10,000 compounds and currently accommodating 20,000 compounds per year. The screen uses 60 human tumor cell lines, representing leukemia, melanoma and cancers of the lung, colon, brain, ovary, breast, prostate, and kidney. Natural products collected in an NCI repository are also a major source of chemical entities screened (see Box 1)

The aim of the screen is to prioritize the further evaluation of synthetic or natural compounds that demonstrate selective growth inhibition or cell killing of particular tumor cell lines. A 60 cell line dose response data set produced by a given compound results in a biological response pattern that can be utilized in pattern recognition algorithms to assign a putative mechanism of action to a test compound, or to determine that the response pattern is unique. Although the particular growth inhibitory response of a single cell line might be relatively uninformative, the pattern of response of the cell lines as a group can be used to rank a compound according to the likelihood of sharing common mechanisms. The COMPARE computer algorithm quantifies this pattern (Monks et al., 1997), and COMPARE searches the database of screened agents to compile a list of the compounds that are most similar (Paull et al., 1989).

Following characterization of various cellular molecular targets in the 60 cell lines, it may be possible to select compounds most likely to interact with a specific molecular target. Successes in use of the screen to identify the underlying mechanisms of the multi-drug-resistant form of cancer have led NCI to develop collaborations to go beyond growth inhibition and cell killing to characterizing mechanisms of action through the expression of molecular targets in the 60 cell lines (Monks et al., 1997). Examples include cancer suppressor gene status (Weinstein et al., 1997) or the presence of an oncogene (Koo et al., 1996).

In early 1995, NCI, in reviewing screen data, observed that many agents were completely inactive under the conditions of the assay. A protocol for a three-cell line prescreen was developed in collaboration with the Information Technology Branch of DTP that could eliminate approximately 50 percent of drugs from 60 cell line testing

without a significant decrease in ability to identify active agents and thus increase the throughput and efficiency of the main cancer screen with limited loss of information.

In the mid 1990s, for agents identified by the COMPARE algorithm, NCI began incorporating assays using semi-permeable hollow-fibers implanted in animals in the in vivo phase of drug development. These fibers allow tumor cells to grow in contact with each other, and more than one tumor can then be implanted into a single animal providing greater efficiencies than would be obtained through a single in vivo experiment (Khleif and Curt, 2000). A standard panel of 12 tumor cell lines is used for the routine hollow fiber screening of compounds with in vitro activity, and alternate lines can be used for specialized testing on a non-routine basis. The premise of this technique is that advancing potential anticancer agents identified in an in vitro screen to preclinical development requires a demonstration of in vivo efficacy in one or more animal models (Hollingshead et al, 1995), and hollow fiber screens appear to correlate well with clinical results (Johnson et al., 2001).

In 1997, the Director of NCI formed the Developmental Therapeutics Program Review Group (called the Horwitz Committee after its chair, Susan Horwitz of Albert Einstein College of Medicine) and charged it with the task of making recommendations to enhance NCI’s ability to discover new and useful antitumor drugs, particularly those with unique and novel mechanisms of action rather than simply antiproliferative effects.

The Horwitz Committee was enthusiastic about enhancing DTP decisions concerning its resource allocations through an oversight group established to continuously monitor the discovery and development process for any drug target or drug candidate by bringing together and coordinating distinct proposals from different laboratories or institutions nationwide.

The Committee recommended to DTP: a focused screening program for active compounds using assays for which it has developed expertise and capacity; providing public access to its repository of compounds, research tools, and information databases; working with the government, academic, and industrial communities to develop, evaluate, and deploy new assays in both the internal and external scientific communities; and fostering a more collaborative approach to screening by serving as a matchmaker between chemists and biologists for the analysis of novel agents. Along these lines, the Committee recommended that DTP assume a leadership position in informatics, facilitate developing cancer therapeutics through expansion of its current operations, and increase access of extramural investigators to natural product repositories, select chemical libraries, engineered cell lines, chip and microarray technology, standardized reagents for cancer immunotherapy, and information databases. It further recommended that extramural funding be directed to support cooperative groups in different parts of the country and facilitate alliances between government, academic, and industrial resources (http://deainfo.nci.nih.gov/advisory/bsa/bsa_program/bscdevtherprgmin.htm).

As a result of the Horwitz Committee’s recommendations, the algorithm for the drug discovery pathway was changed. The committee recommended that the 60 cell line screen be preceded by a 3 cell line screen to first identify lead antiproliferative agents. A subset of compounds identified in this manner could then be analyzed in the 60 cell line screen to gain insight into a compound’s mechanisms of action. The next step would be to define the mechanism of action or identify a novel compound as affecting a defined biochemical, cell, or tissue physiological endpoint with the intent of advancing to development only after an effort to establish a molecular endpoint of a compound’s action. Further characterization of a compound could then be conducted in an appropriate animal model or via the hollow fiber assay.

The new algorithm relies less on activity in tumor graft models and more on a priori definition of a mechanism of action or molecular target in defining a strategy for subsequent development (Sausville and Feigal, 1999). In essence, an effect on an important molecular target becomes a constant signal after which pharmacologic, scheduling, and toxicologic studies follow. This would also allow incorporation of target or molecular endpoints in early clinical trials to follow logically from the preclinical experience. More formal safety testing would be undertaken for suitable compounds, proceeding to phase I clinical trials using the biological, pharmacological, and toxicological properties to define optimal dose and schedule conditions for human studies.

NCI is also reclassifying the cells in the panel according to the types of genetic defects the cells carry. That way, if drugs that address the specific defects or targets can be identified, they could theoretically be matched to a patient’s tumor cell makeup. DISCOVERY is a computer program that uses a clustering algorithm to group

compounds by cellular response patterns. The data collected from the primary in vitro screen is analyzed with the help of algorithms and assembled into unique patterns of activity that cut across cell types. Correlations of compound activity are made with mechanisms of action for particular molecular targets and then used to generate hypotheses that relate to the potential targets, and the molecular targets of unknown drugs can be deciphered by analyzing their cell line screen data and comparing it with the activity of the agents in the database. These analyses and correlations have recently been enhanced through the use of publicly available data on gene expression patterns for thousands of expressed genes from NCI’s collaboration with Brown and Botstein (Ross et al., 2000) as a means of aligning response patterns with patterns of gene expression (Sausville and Johnson, 2000).

The Biological Resources Branch (BRB) is one of the extramural arms of DTP. The program supports preclinical and early clinical studies of biological response modifiers (BRMs) through a program of grants and contracts. These studies assess the effects of novel biological agents and explore relationships of biological responses with antitumor activity. An NCI Preclinical Repository distributes selected agents for peer-reviewed intra- and extramural preclinical studies. Other contracts support the production and in vivo evaluation of monoclonal antibodies, immunoconjugates and other biologicals.¹

BRB staff also provide oversight of the Biopharmaceutical Development Program (BDP) at the Frederick Cancer Research and Development Center, which produces a variety of biopharmaceuticals under current Good Manufacturing Practices (cGMP) for Phase I and II human clinical trials or advanced preclinical animal testing. NIH grant holders and other peer-reviewed scientists are encouraged to contact the BRB for discussions at an early stage regarding use of this resource. NCI program funding can also be provided through the Rapid Access to Intervention Development Program (RAID) mechanism (see further discussion below) or through DTP.

There are three types of production services: small quantities of interesting new proteins that may be implicated in disease processes and would also support testing in small animal models; larger quantities of well-characterized material for preclinical testing to support evaluations of efficacy and toxicity in appropriate animal models of disease; and clinical-grade material to support final preclinical work, IND submission, and Phase I and II clinical trials.

The development group deals with incoming projects at initial stages including evaluation, yield, efficiency, and characterization of products. This group also focuses on process development after initial evaluation for supporting cGMP production in BDP. The laboratory has the capacity to construct recombinant bacterial strains and cell lines, has expertise in a variety of fermentation processes including bacterial, yeast, fungal,

insect, and mammalian cell culture, and focuses on process optimization² with implementation of highly efficient bioreactor production utilizing mammalian and microbial fermentation pilot plants with the capacity for product recovery.

Recent advances in molecular biology have dramatically improved understanding of how cancer cells work. Specific molecules have been identified that are responsible for the initiation and progression of tumors. This provides the opportunity to move beyond empiric screening of agents by their effects on tumor cell growth to detection targeted to a particular molecule of biologic importance in cancer development or progression. In the late 1990s, NCI’s efforts to revamp its preclinical drug discovery work were refocused in this direction (Sausville and Feigal, 1999). The tools for this approach include sequence information that defines the primary structure of relevant target molecules; expression vectors for large scale target production; physical and computational techniques to allow routine elucidation of three-dimensional structure; advances in screening technology to promote increased efficiency in assessing large numbers of candidate lead structures; and synthesis approaches to generate large numbers of test compounds (Sausville and Feigal, 1999).

Currently, methods for target identification are relatively slow and unreliable. The NCI multiple cell line screen, when employed in conjunction with cell-based assays, provides a useful empirical method for identifying molecules that are likely to target DNA or proteins or may be novel. But for small molecules with novel activities, this screen cannot suggest a likely target (Horwitz Committee, 1998). Researchers have tended to rely on methods in which the small molecule is used to purify its target from a cell extract, usually by affinity chromatography, although this method tends to be tedious. Through investments in genomics and related technologies, a more radical approach to target identification is now becoming available using new tools (called hybridization technology) that allow the study of effects on a whole genome, not just an isolated portion.

In 2000, as a first step to refocus its efforts in this area, NCI announced the availability of Molecular Target Drug Discovery grants. Program announcements for this initiative stated that “Rather than depending on in vitro and in vivo screens for antiproliferative activity, investigators can now focus on new molecular targets and pathways essential for the development and maintenance of the cancer phenotype.”³ The NCI announced that it was reorganizing its drug development programs from early drug discovery to the conduct of clinical trials in order to bring forward new types of agents based on strong rationales. Clinical evaluation of new agents would include appropriate measurements to verify target modulation. Investigators funded through the program were asked to identify a novel molecular target, to validate the target as a basis for cancer drug discovery, and to develop an assay for the target. Molecular target laboratories are

funded to produce libraries of potential anticancer compounds for distribution, to develop screening assays, and to confirm a drug’s initial ability to alter its target.

As of 2004, 76 research groups were being supported via this initiative—investigating an aberrant protein that enables cancer cells to evade programmed cell death, a “stress response” protein that is overexpressed in tumors and may play an important role in cancer growth, or how DNA changes (methylation) can lead to cancer.⁴

The Cancer Genome Anatomy Project (CGAP) is an interdisciplinary program established and administered by NCI to achieve a comprehensive genetic description of normal, precancerous, and malignant cells so as to determine the changes that occur when a normal cell is transformed into a cancer cell, and then to apply that knowledge to the prevention, detection, and management of cancer.⁵

Since its inception in 1996, the program has had four primary initiatives. The Human Tumor Gene Index identifies genes expressed during the development of human tumors. The Cancer Chromosome Aberration Project characterizes the chromosomal alterations that are associated with malignant transformation. The Genetic Annotation Index identifies and characterizes the different genetic DNA sequences associated with cancer, and the Mouse Tumor Gene Index identifies genes expressed during the development of mouse tumors.

CGAP supports the production of serial analysis of gene expression (SAGE) libraries and their sequencing, while the National Center for Biotechnology Information (NCBI) has created a repository for the sequence data, has developed data processing algorithms, and has developed and maintained the SAGEmap website at NCBI.⁶ Everything is shared openly with the research community. The CGAP web page contains databases of a wide array of human and mouse genomic data and provides information on new experimental methods. Biological reagents developed through the program are available to researchers at cost. In the future, the program may move into the development of functional genomics and proteomics databases.

According to NCI, researchers have started mining the CGAP databases and are already discovering new, potentially cancer-causing genes, identifying candidates for molecular targeting research, and helping to build assays for cancer cell signature research. Investigators who are funded through the program are required to agree not to patent the sequences they acquire. And because NCI has obtained a declaration of exceptional circumstances under the Bayh-Dole act, sequencing project contractors do

not retain title to inventions developed with the federal funds, and NCI can require immediate disclosure of all such data.

The Laboratory of Molecular Technology (LMT)⁷ in Frederick, Maryland, is an integrated molecular biology laboratory focusing on high-throughput gene discovery and analysis including advanced sequencing, genetics, genomics, and proteomics technologies, together with associated bioinformatics and information management. General access to these advanced technologies is provided to the NCI community through the LMT core service laboratories.

The lack of adequate animal models has led to the development of transgenic mouse models to be used as preclinical assays to determine the likelihood of success for novel agents being considered for clinical studies. The Mouse Models of Human Cancers Consortium⁸ involves 20 groups of academic researchers who have created and are making available to researchers mice with defined genetic alterations that predispose the animals to certain types of cancer that could serve as a basis for testing new molecular targeting treatment and prevention strategies. Academic members of the consortium are developing ties to pharmaceutical industry sponsors to facilitate the testing and evaluation of new compounds in these mouse strains.

The high cost of X-ray and nuclear magnetic resonance equipment and the specialized expertise required to run experiments on these instruments exclude large numbers of scientists who are interested in pursuing molecular structural studies using these technologies. Many of the most interesting structural problems will involve complex systems, such as those involved in multistep intracellular processes. Multicomponent complexes are difficult to crystallize, and, frequently, usable structural data can be obtained only by using a scarce (synchrotron beam) technology. NCI is collaborating with the National Institute for General Medical Sciences on the National Beam Program to provide both this technology to quickly identify the structure of important molecular targets in cancer cells and efficient computer modeling to identify potential anti-cancer agents suited to hit the targets based on these structures.

Several NCI initiatives are aimed at creating libraries of synthetic, biological, and natural compounds for testing against validated molecular targets. Samples of these are provided at no cost to investigators. NCI has made available more than 140,000 synthetic chemicals, 80,000 natural products extracted from plants and marine organisms, and a variety of biological agents for use in studying the compounds.

Specific assays are used to identify those that hit defined target molecules. Approaches used to achieve these goals include: the National Cooperative Drug Discovery Group Program (NCDDG), funded as cooperative agreements in response to a Request for Applications (RFA)⁹ that was established in 1983 and supports 13 multi-

disciplinary groups involved in the discovery of new, synthetic or natural-source derived anticancer drugs; and the Biology-Chemistry Centers, funded as program project grants¹⁰ that support six multi-disciplinary research centers that bring together approaches to the generation of structural diversity and novel, specific assays directed at molecular events or targets important in the cancer process and thus suitable for cancer drug discovery.

A number of preclinical mouse tumor models have been developed. Unfortunately, no one model serves as a satisfactory predictor of human cancer, but the experience gained over the past three decades by using these models for studying the effects of various experimental therapeutics can serve as a baseline for assessing new targeted agents. Four major types of in vivo models are available to assess the efficacy of potential new therapeutics in cancer. Each model, as described below, has advantages and disadvantages.

A variety of native tumor types in immunocompetent mice and rats exist, and these models are reliable, available in numbers adequate for good statistics, low cost, and with well established responses to current anticancer therapies. The disadvantages are that the systems are fully rodent and that these fast growing tumors may not accurately model human tumors.

Genetically engineered mice that will develop tumors are generally immunocompetent, and the tumors are genetically mouse and localized in the usual sites. The disadvantages of these models are the requirements for breeding (and frequently, licensing) the animals that make these models high cost. The tumors usually develop late in the animal’s lifespan, so these models are slow in developing. Moreover, there are few tumor types available, and it is difficult to obtain enough animals to establish reliable statistics. Importantly, very few of these models have been validated as representative of human disease through molecular markers and response to current anticancer therapies.

Human tumor grafts in mice have the advantage of human malignant cells of a wide variety of types and, in many cases, reliable tumor growth. The response of many of these tumor models to current anticancer therapies is well-established. The disadvantages are that the hosts are immunodeficient, the tumors are generally slow-growing, the tissue supporting the cancer cells is of mouse origin, and the animals are costly and require special housing. Due to the cost of the studies, fewer animals may be used than are actually needed for reliable statistical analysis. Unfortunately, all animal models suffer from inconsistency in predicting responses of human cancers to therapy.

There are other models that are used in studying chemoprevention and metastasis. Carcinogen-induced tumor models have been used in chemoprevention research but have been difficult to apply to therapeutic research. Models of metastatic disease use rats or mice that have naturally metastatic cancers, or involve injecting tumor cells intravenously to produce lung metastases, intrasplenically for liver metastases, intracardially or

intratibially for bone metastases, or into the internal carotid artery for brain metastases. A number of experimental cancers have now been developed that produce fluorescent markers that, with the necessary instrumentation, allow detection of metastases by fluorescence. These genetically engineered models are costly, and natural or engineered models have highlighted the problem that tumor response to therapy markedly depends upon the anatomical location of the disease.

The lack of uniform standards for determination of tumor volume doubling time, tumor growth delay, log₁₀ cell-kill, or other endpoints is a weakness in the field. Experiments are frequently terminated before any endpoint is reached. Guidelines for the conduct and analysis of in vivo studies of new therapies, based on results from previous testing, could help developers weed out less active compounds earlier in the process. Guidelines to establish uniform criteria for antitumor activity and for in vivo data analysis would allow comparisons to be made before new therapies are taken to clinical trial. For targeted therapies, tumor expression of the target should be confirmed. Activity in several tumor models establishes a stronger case for a new agent than activity in a single model. Use of established methods for the determination of additivity/synergy for combination regimens should be the standard. An agent that survives higher hurdles during preclinical testing would presumably have a greater chance of clinical success, and guidelines for uniform data gathering and data analysis should allow improved selection of agents for clinical trials.

Once a compound of interest is identified, animal models are critical to assess preclinical toxicology. In the 1970s, NCI used only dogs and monkeys in its preclinical toxicology protocols (Khleif and Curt, 2000). In 1979, NCI and the FDA reviewed existing data and concluded that toxicity studies performed in mice could in most cases replace the more costly and time-consuming large animal studies. Some believe that the NCI toxicology protocol has performed well in predicting safe initial doses for clinical trials (Khleif and Curt, 2000). Others believe that it is costly and non-productive to generate large amounts of animal toxicology data without some guidelines or assessment of whether the data are of any use (Schein, 2001). Clearly, when it comes to qualitative or organ-specific forms of toxicity, the role of preclinical toxicology studies in animal models has been ambiguous.

The development of drugs within NCI has evolved into two stages at present, each of which requires a toxicologic evaluation. The first involves a preliminary assessment of toxicity (range-finding studies) usually in two preclinical animal models (rodent and nonrodent) with the determination of maximum tolerated dose and drug behavior in the animal (pharmacokinetics) in both species. If the drug meets the program criteria for full-scale development, a more complete toxicity evaluation (IND- directed studies using the proposed clinical route and schedule) is performed that leads to the filing of an IND (Tomaszewski and Smith, 1997). FDA has defined the battery of preclinical studies that are considered important for assessing the safety of oncology drugs prior to filing an IND. In the past decade, NCI has conducted agent-directed drug development projects, including toxicologic evaluations on eight drugs and concluded that their evaluations

allowed an impressive prediction of maximum tolerated doses and dose-limiting toxicities.

Early drug development work is more typical of traditional academic research, and downstream preclinical and clinical testing has traditionally been the domain of industry. While it is often the case that a company will partner with, or acquire, promising research at the transitional stage, academic researchers often wish to carry forward research that originated in their institutions. To aid them, NCI has initiated several programs aimed at reducing or removing the rate-limiting barriers that typically delay clinical validation, for example, scale-up of production, development of suitable drug formulations, development of analytic methods, stability testing, animal toxicology, and planning for clinical trials (Schein, 2001)—processing challenges that are routinely undertaken by industry.

NCI’s Rapid Access to NCI Discovery Resources (RAND) program was initiated to make available to academic institutions on a competitive basis the discovery and early preclinical development contract resources of DTP so as to provide a broad range of early preclinical assistance for anticancer therapeutic discoveries in academic laboratories. The program is intended to remove the most common barriers between basic research and the development of new molecular entity therapeutics. It assists academic investigators in the discovery of small molecules, biologics, or natural products through mechanisms such as the development of high-throughput screening assays, computer modeling, recombinant target protein production and characterization, and chemical library generation. It also assists in the development of analytical methods for pharmacokinetic and metabolic studies, and in vivo pharmacokinetic, toxicity, and efficacy studies. Compound samples accepted after DTP review can be screened for the academic originator using the NCI’s three cell line, one day prescreen, and a follow-up of actives in the NCI 60 cell line screen.

All output from a RAND project is returned to the originator of the project as synthesized or isolated materials, high-throughput screening or pharmacokinetic methods, informatics output, or in vivo screening results, among others. NCI will not acquire intellectual property rights to inventions made by its employees with research materials under RAND (or the RAID program, described next), unless the originating investigator and NCI mutually agree that it is in the investigator’s best interest. If an NCI contractor is in a position to file an invention report and elects to retain rights under the Bayh-Dole Act, the contractor will, as provided under the contract, offer the principal investigator a first option to negotiate a license to the invention.

Once the optimal compound is selected, the Rapid Access to Intervention Development (RAID) program facilitates further preclinical development and generation of evidence that a new molecule or approach is a viable candidate for expanded clinical evaluation. RAID provides academic or non-profit institutions NCI resources for the pre-

clinical development of drugs and biologics on a competitive basis. Resources include chemical synthesis and good manufacturing practices, formulation research, clinical dose form manufacturing, bulk manufacture of monoclonal antibodies, and formulation for production of recombinant proteins, among others.

RAID is intended to remove the most common barriers between laboratory discoveries of new molecular entities and clinical trials. The program, in December 2004, provided resources to academic institutions in 52 projects ¹¹, but it is not intended to be a pipeline for materials for NCI-held INDs. Most of the products in the RAID program will be studied clinically under investigator-held INDs at the originating (or a collaborating) institution. RAID resources are limited, and it is not an unconditional commitment to develop a particular compound for the clinic, nor is it meant to assist industry in the absence of an academic partner.

RAID is also not a grant program to a particular laboratory. It is expected that the great majority of resources committed through RAID will be through use of NCI new-agent development contracts and of NCI staff expertise in service of meritorious academic projects. Some steps in the process may best be carried out in the originating laboratory, in which case NCI will initially attempt to provide necessary support through existing suitable funding vehicles, but this pathway for support may not be the ultimate avenue used.

RAID is designed to accomplish the tasks that are rate-limiting in bringing discoveries from the laboratory to the clinic. Division of Cancer Treatment and Diagnosis contractors perform the work in direct consultation with the originating laboratory. In some cases, RAID supports only the one or two key missing steps necessary to bring a compound to the clinic; in other cases it supplies the entire portfolio of development tasks needed for an IND.

NCI has pursued several mechanisms to partner with the private sector in drug development. These mechanisms are in addition to the typical industrial relationships allowed by law and NIH policy, for example, Cooperative Research and Development Agreements (created by the Federal Technology Transfer Act of 1986 to enable cooperative research and development relationships and technology transfer between NCI or other federal agencies and industry) and Small Business Innovation Research grants (authorized in the Small Business Innovation Act of 1982 which, through a set aside of agency resources, provides grants for up to two and a half years to support research and development leading to commercialization in small-business-qualified companies.

The Drug Development Group provides support for academic and corporate-derived compounds when NCI is responsible for conducting and monitoring the drug’s clinical development. A number of promising agents have been developed through this program. A novel compound (PS 341) presented to NCI by Millennium Pharmaceuticals, was the

first in a new class of agents that take aim at a new cellular target, the proteasome enzymes. These enzymes play an important role in the breakdown of proteins that regulate the cell cycle, and inhibition of the breakdown of these proteins by the new agent can lead to cancer cell suicide. Histone deacetylase (HCAC) was another agent pursued through this program. Certain cancer-causing genes induce cancer when they block the normal expression of healthy genes. HCAC inhibitors relieve this suppression. In cooperation with extramural organizations, NCI has studied the anti-tumor effects of several such inhibitors, including pyroxamide, an inhibitor identified by an NCI-supported cancer center that considerably reduced tumor growth in animals.

The Flexible System to Advance Innovative Research (FLAIR) provides funds to small businesses to develop cancer therapeutic and prevention agents from basic discovery to clinical trials. As of December 2004, 20 active FLAIR grants and phase I clinical trials were supported by the FLAIR program.

The Radiation Modifier Evaluation Module (RAMEM) program serves individual investigators and industry in the design and development of treatment programs for the use of novel molecular, biologic, and cytotoxic agents in conjunction with radiation therapy. Integration of molecular imaging, molecular signatures, and molecular therapeutics with radiation therapy is a high priority of NCI’s Intramural Program because new anti-cancer agents may ultimately be used in combination with radiation therapy.

Preclinical and clinical research with novel agents for cancer treatment and prevention requires usable tools to determine that the intended molecular target producing or associated with cancer has been affected by the agent. The Interdisciplinary Research Teams for Molecular Target Assessment is a new program that enables interdisciplinary teams of scientists to develop such tools. The teams will define the molecular basis for these research tools and develop and validate novel biochemical, pathological, pharmacologic, immunologic, molecular, or imaging methods and reagents to measure the effect of new target-directed drugs in proof-of-principle laboratory models and clinical trials. These methods and reagents must, therefore, be suitable for in vivo use in animal models and in human beings. The first set of applications for this program was funded in early FY 2001.¹²

In 1955, the National Cancer Institute began to organize clinical trials of the first effective anticancer agents in leukemia patients. Since then, NCI has become the largest U.S. network for clinical trials of any type, with support through a variety of mechanisms, including the largest, the Cooperative Group Program and the Cancer Centers Program, as well as grants and contracts to individual investigators and institutions. NCI-supported trials are carried out in diverse settings, including cancer centers, academic medical centers and community hospitals. Early trials (phase I and

many phase II) are often carried out by single investigators or small groups at of a few institutions. Phase III trials are generally associated with the Cooperative Groups.

As of March 2004, NCI was sponsoring about 1200 clinical trials. About 1000 are treatment trials, and the rest are supportive care, screening, prevention, diagnostic and genetic. Most are phase I and II trials and about 200 are large, phase III trials, the mainstay of the Cooperative Groups (Table 1).

As noted earlier, the Clinical Trials Cooperative Group Program was established by NCI with a congressional appropriation for this purpose in the 1950s when most new anticancer agents were being developed through NCI funding. Over time, through consolidations and other changes the Program reached its current status of 12 NCI-supported Cooperative Groups, made up of academic institutions from the United States, Canada (National Cancer Institute of Canada, Clinical Trials Group), and Europe (European Organization for Research and Treatment of Cancer), emphasizing phase III studies, 11 for adult and the 12^th, the Children’s Oncology Group, for childhood cancer. Some groups (Children’s Oncology Group) consist of investigators with a particular specialty, others (Radiation Therapy Oncology Group) study a specific therapy, and still others (Gynecologic Oncology Group) focus on a group of related cancers. The 2004 Cooperative Clinical Research budget of $180 million supported the accrual of about 28,000 patients that were contributed from more than 1,700 institutions to group

coordinated trials at about 1000 sites, although 85 percent of patients are enrolled at about one-quarter of the sites, mainly large cancer centers. The Program is explicitly designed to promote and support clinical trials of new cancer treatments, explore methods of cancer prevention and early detection, and study quality of life and rehabilitation issues during and after treatment (Schilsky, 2002).

Cooperative Group investigators may work on non-Cooperative Group trials, such as early phase and phase III NCI or industry trials at their institutions or phase III trials cosponsored by industry and carried out by Cooperative Groups. The latter trials using investigational agents proprietary to a pharmaceutical or biotechnology company are subject to specific cooperative research and development or clinical trials agreements (CRADA or CTA)¹³ dealing with intellectual property, confidentiality, and other uses of the investigational agent (http://ctep.cancer,gov/industry/industry.html). All the cooperative groups are subject to on-site monitoring under NCI guidelines.

The Cooperative Groups and the NCI-designated cancer centers collectively conduct a larger number of trials and enroll more patients than other entities, but other groups now compete in this territory. For example, U.S. Oncology, a for-profit network of cancer providers that treats about 16 percent of U.S. cancer patients, has accrued over 16,000 cancer patients into clinical trials and played a significant role in FDA approval of eleven anticancer drugs that have entered the market over the last ten years (http://www.usoncology.com/OurServices/USONResearch.asp), and there are other entities for organizing trials, such as the clinical trials network of the National Comprehensive Cancer Network (NCCN), as well as private sector contract research organizations Some of these groups extend the reach of clinical trials to patients who would otherwise be outside the membership of the Cooperative Groups, others may draw from the same pool of patients, for example, NCCN members are 19 large U.S. cancer centers that are also part of the Cooperative Group system.

As productive as the Cooperative Groups have been, the conduct of phase III trials has been seen as problematic for many years. A major issue has been, first of all, the slow and difficult procedure for developing, reviewing, and initiating protocols. Secondly, patient accrual can continue over many years before the goals are met. As a result, by the time trials are completed, the questions being addressed may no longer be relevant. Furthermore, the 28,000 patients entered into Cooperative Group trials annually comprise only about 2 percent of the 1.3 million adults diagnosed with cancer each year (ACS, 2002), or about 2 percent of the new cases and a much smaller percentage of all people living with cancer.

Finally, while phase III trials are, with rare exceptions, required for final FDA new drug approval, most of the trials that in practice define the uses of a cancer drug take place after FDA approval, the importance and scope of such trials in cancer being far greater than for drugs for other conditions. A whole range of tumor types other than those for which the drug is approved, and in combinations with a wide variety of other

drugs, are very often studied once a drug becomes available, and the Cooperative Groups play a leading role in this work. This complicates the federal role in developing and regulating cancer therapies.

If, as is anticipated, the number of new cancer drugs in development increases—possibly dramatically—the NCI clinical trial network will be pressed to meet new challenges, including: greater involvement with pre-approval trials; more rapid development of protocols for post-approval trials; improving accrual rates; greater numbers of clinical questions likely to be generated by the larger numbers of agents (many of which should have known molecular targets); among others

In addition to increasing enrollment in trials, efforts to select patients based on matching cancer molecular profiles with drug targets might identify the most appropriate patients for trials of particular agents. Industry is clearly moving in this direction, and NCI has the tools to contribute to the process, and could facilitate collaborations in this area (Canetta, R., personal communication to the National Cancer Policy Board, 10/11/01). A similar concept may move phase I dose-finding trials from the goal of the maximum tolerated dose to a focus on a biologically effective dose, that is, matching the dose of a targeted agent to levels of the target measured in blood and/or tumor tissue. Monitoring the activity of new agents in this more precise way could provide an anticancer effect through saturation of activity against the tumor molecular target before general clinical toxicity appears in the patient, but will require tests of appropriate markers in tumor and blood specimens from each patient studied in phase I trials.

Facilitating drug development needs of this sort could be accomplished by good use of existing (some relatively new) programs. The NCI’s Early Detection Research Network (EDRN) already funds molecular marker laboratories at various institutions around the country. These laboratories are focused on discovery of markers for cancer diagnosis, but they are also aware of the value of many of the markers as targets for cancer therapy or prevention. Such markers can be picked up by industry, developed in collaboration with industry, or used in trials by any of the entities described earlier. Core laboratories funded by NCI’s Specialized Program of Research Excellence (SPORE) could serve the same function. As described in the Critical Path section of Chapter 3 of this background paper, FDA is also focusing on the development of markers and their use in improving evaluation of new medical products (drugs, biologics, and devices) in the recently announced Critical Path Initiative and Critical Path Opportunities List (http://www.connectlive.com/events/fdacriticalpath/—12/3/04)

Potential administrative difficulties stem from the fact that that the EDRN is overseen by the Division of Cancer Prevention, the Centers and SPOREs by the Office of the Director and the Division of Cancer Biology, Diagnosis, and Centers, and the cooperative groups by the Division of Cancer Treatment and Diagnosis, and each office and division has its own priorities and responsibilities. However, reasonable communication, coordination, and the normal process of publication and dissemination of scientific advances should help to make these resources available in drug development,

and NCI is aware of these coordination, academic-industry relationship, managerial, and other challenges (Klausner, R., personal communication to the National Cancer Policy Board, 4/10/01).

In 1996, in response to growing concerns that its clinical trials portfolio was increasingly inefficient and unresponsive to changing needs—including the Cooperative Groups Program as a centerpiece of its clinical trials efforts—NCI commissioned an external review of its clinical trials program in its entirety (DTP Program Review Report, 1997). In the resulting document, called the Armitage Report after its chair, James Armitage, the Review Group wrote that the clinical trials system is “an intricate and large research laboratory without walls. This complexity has bred inefficiencies and eroded the ability of the system to generate new ideas to reduce the cancer burden.” The Review Group considered recruiting and retention of clinical scientists, recruiting of participants to clinical trials, improving clinical trials methodology, increasing collaboration and cooperation in clinical trials, and NCI’s organizational framework and structure for implementation of clinical trials.

The Group recommended that NCI should: increase funding for cooperative groups to fully recommended levels; reduce and limit data collection to study endpoints and patient safety and fund some large simple trials in common cancers to establish treatment differences; enforce uniformity of data collection; enlist advocates, industry, and the FDA to develop uniform standards and reporting requirements; provide cooperative groups and cancer centers with the means to access all relevant electronic databases and test the new NCI informatics system; and develop strategies to convince payers that clinical trials represent a better standard of care and ultimately result in decreased costs.

As a result of this review, NCI created an internal implementation committee that developed constructive responses to these recommendations, among others. Several of these initiatives are highlighted or outlined more specifically in NCI’s FY 2005 Bypass Budget (NCI, Nation’s Investment in Cancer Research, 2004). For example, continued support of the Cancer Trials Support Network centralizes the common administrative, financial, and data collection activities of the cooperative groups. Since 2002, physicians outside NCI cooperative groups have been enabled to enroll patients into NCI-sponsored clinical trials. In 2003, NCI and FDA signed a multi-part Interagency Agreement to share knowledge and resources in order to enhance the efficiency of clinical research and the scientific evaluation of new cancer medications (see FDA chapter below). Other new initiatives strengthen the scientific planning for large trials and aim to double the rate at which Phase III trials are completed.

A persistent concern has been how well cooperative group efforts are coordinated, if at all, with clinical trials underway at NCI-designated cancer centers (P30) and SPORE (P50) programs. In 2003, a National Cancer Advisory Board (NCAB) review of the P30 and P50 programs recommended more coordination and standardization among these

entities and the cooperative groups.¹⁴ The review group recommended that NCI develop a plan for improved coordination of all clinical research mechanisms, including cooperative groups, phase I and II contracts, SPOREs, and cancer centers, and convert the funding mechanism for cooperative groups and Phase I and II studies from a contract to an assistance mechanism. As of 2004, no action had been taken in response to the recommendations. However, in January 2004, the NCI announced the formation of a clinical trials working group to advise the Director and NCAB on the “development, conduct, infrastructure, and support necessary for the optimal coordination and future progress of the entire range of intramural and extramural clinical research trials” (The Cancer Letter, p. 4, 1/9/04). In early 2005, this working group announced recommendations addressing better coordination, standardization of research tools, forms, contracts, and databases, and essential data collection, speeding of patient accrual, NCI certification of clinical oncologists, expansion of community-based regional IRBs, reduction of regulatory burdens, joint participation of FDA and NCI in meetings with industry, and establishment of a database on federally funded trials and of an external Investigational Drug Working Group, among others (The Cancer Letter, p. 1, 2/25/05). Steps to implement some of these recommendations will await future NCI action.

The Cooperative Group Program and the clinical trials programs more broadly continue to be scrutinized within and outside of NCI. These reviews have focused most closely on structural, administrative, and resource issues, and on the difficulty in accruing patients. They have not, by and large, examined the specific clinical questions addressed in the trials themselves to determine whether the best use is being made of this resource, although the most recent review suggested prioritization of trials so this may change. The possibility that large numbers of irrelevant trials are being carried out is remote, given the mechanisms in place to assure that each major trial is reviewed for clinical importance and soundness. However, the predicted change in the numbers and types of new cancer drugs emerging from the R&D pipeline suggests that a review of the types and mix of questions being addressed, particularly in phase III trials, could be useful.

In 1955, Congress recognized the need for public investment in the discovery and development of anticancer agents. Since then, NCI has established a Developmental Therapeutics Program to support the programs necessary to its pivotal role in the worldwide effort to develop new agents to treat cancer. The DTP supports the preclinical development of novel therapeutic modalities for cancer, the acquisition, synthesis, and definition of activity in both in vitro and in vivo models of cancer, and the advancement of active agents in preclinical models toward clinical evaluation.

In recent years, the advent of genomics, combinatorial chemistry, and high-throughput screening for the identification of potential lead compounds has markedly expanded the number of candidate drugs in the pipeline (although, so far, not the number

of effective drugs emerging). Progress in identifying the genetic and molecular bases for cancer has also intensified efforts to identify more selective and efficacious anticancer compounds. This predicted increase in the number of candidate drugs has already begun to alter the preclinical models and methods used in many therapeutic areas. The changes needed in clinical trials, however, have not kept pace.

A clear role for NCI, in collaboration with other NIH institutes and the outside scientific community, is in continuing to support large-scale science projects, as it has with the Human Genome Project and several current projects, such as, the Cancer Genome Anatomy Project and the Early Detection Research Network, among others. NCI has been very successful in carrying out, and assisting in, the process of developing new drugs for cancer. The challenge for the Institute is to continually reassess where its advantages lie, to consider its historical involvements in all phases of cancer drug development—to focus on basic science and discovery of interesting molecules against novel targets, or on preclinical and clinical development of academic or industry agents in coordination with the FDA in deployment, or both—and to create opportunities for moving forward. As of late 2004, NCI was considering a review to carry out just such a reassessment (Doroshow, J., personal communication to the National Cancer Policy Board, 2004).