Scientists have been searching for cancer biomarkers for many years, but the methods of discovery have changed as new technologies have been developed. Traditionally, scientists have relied on conventional laboratory research tools, such as gel electrophoresis and immunohistochemistry, to identify altered genes and changes in mRNA and protein expression (Ross et al., 2004). Progress in this work has been slow because researchers could examine only one or a small number of candidate markers at a time, and the methods required some prior knowledge and experience with the potential markers of interest. More recently, many novel high-throughput technologies (Kiviat and Critchlow, 2002; Fan et al., 2004; Aebersold et al., 2005; Weckwerth and Morgenthal, 2005; De Bortoli and Biglia, 2006; IOM, 2006a), especially in the fields of genomics and proteomics, have made it easier to interrogate hundreds or even thousands of potential biomarkers at once, without prior knowledge of the underlying biology or pathophysiology of the system being studied (Table 2-1). As a result, there has been a flood of new data and renewed interest in discovering novel biomarkers for use in drug development and patient care.

The goal of these discovery methods is to identify genetic variations or mutations as well as changes in gene or protein expression or activity that can be linked to a disease state or a response to a medical intervention.

Analysis of these large datasets requires sophisticated algorithms and bioinformatics to identify individual markers of interest or to derive signatures or patterns of many markers (reviewed by Cristoni and Bernardi, 2004; Englbrecht and Facius, 2005; Tinker et al., 2006). Although these methods are continually evolving and being improved, there is still a great need for novel approaches to data analysis, especially with regard to network oriented models that can incorporate many different types of data to fully integrate the vast complexity of biology in health and disease. However, identifying biomarker patterns or specific changes in genes or the products of gene expression in tumors is only the beginning of the process to develop cancer biomarkers.

Before a candidate biomarker can be put into use, it must undergo several stages of confirmation, validation, and qualification for use (Wagner, 2002; Feng et al., 2004; Ransohoff, 2004, 2005; Simon, 2005; De Bortoli and Biglia, 2006). Analytical validation is the process of assessing the assay or measurement performance characteristics, while qualification is the evidentiary process of linking a biomarker with the biology and clinical end-

points (that is, clinical validity and utility). Both are intended to ensure that a biomarker is fit for a specified purpose, but it is often not clear how best to prove the performance characteristics of a biomarker-based test, especially for those that employ newer technologies, since many lack a gold standard for comparison (IOM, 2006a). Ultimately, a test used to make clinical decisions must, in combination with an intervention, lead to a beneficial impact on patient outcomes. Thus, use as a clinical diagnostic also involves evaluations of benefit, harm, cost, and effort (Ransohoff, 2004).

Once an appropriate method has been selected for measuring the biomarker or pattern of markers, the technical parameters of the test must be well defined to establish sensitivity, specificity, reproducibility, and reliability of the measurements. However, different technology platforms may be needed at different stages of biomarker development. Platforms for biomarker discovery generally need to process many biomarkers simultaneously, but they can be low throughput in terms of the samples processed. Platforms for clinical research need to be high throughput in terms of specimen processing, but they usually focus on a smaller number of markers. Platforms for clinical practice need to be inexpensive and robust, and ideally results should be easily and objectively quantifiable.

The validity of the biomarker as an indicator of a biological, pathological, or pharmacological process must also be confirmed in carefully designed studies. Validation is necessary for each potential use of a biomarker, and the level of evidence needed to implement a biomarker varies with the intended use. For example, in the drug development process, biomarkers can play a role at many different stages, from early, exploratory research to surrogacy for a clinical endpoint in large clinical trials, and the required degree of validation increases along that continuum (Table 2-2; see also Table 1-2). In the drug development process, the highest level of evidence is required if the biomarker is to be used as a surrogate endpoint—it must be qualified for that specific use by clearly demonstrating in clinical studies that the marker accurately predicts the clinical endpoint of interest. Correlation and plausibility is not sufficient (Srivastava and Wagner, 2002; Wagner, 2002; Fleming, 2005). For instance, tumor shrinkage might seem to be a plausible surrogate for treatment efficacy, but in fact, tumor shrinkage in response to a drug does not necessarily lead to improved patient survival (Norton, 1997; Citron, 2004; Hudis, 2005). Similarly, inhibition of preinvasive abnormalities is widely thought to predict a reduction in invasive cancer (Kelloff et al., 2006), but the relationship has not been fully validated in clinical studies, and a recent study even showed that the antiestrogen

raloxifene can reduce the risk of invasive breast cancer without significantly reducing the incidence of ductal carcinoma in situ, a preinvasive lesion with potential to develop into invasive cancer (NSABP, 2006).

Similarly, the criteria for validation vary among the many possible clinical uses of biomarkers (see Table 1-1). For example, validation of a biomarker for screening, which entails the systematic testing of an asymptomatic population to identify evidence of particular type of cancer, requires proof that the biomarker detects the disease with a high degree of sensitivity, specificity, and positive predictive value. Ultimately, the clinical value of a screening test will also also depend on whether the routine use of the test, combined with appropriate interventions, reduces the morbidity and mortality due to that disease. In contrast, the clinical validation criteria for a diagnostic biomarker, which is used to definitively determine the presence or absence of cancer in patients with symptoms or a known abnormality, are less onerous.

The vast majority of candidate biomarkers never progress past the initial discovery phase, and very few become qualified as surrogate endpoints or useful clinical tests, in part because further evaluation is expensive and time-consuming, with uncertain outcome (and thus a risky endeavor)

(Hayes et al., 1998; Wagner, 2002; Srivastava and Wagner, 2002; Feng et al., 2004; Altman and Riley, 2005; Fleming, 2005; Simon, 2005; IOM, 2006a, 2006b). However, it has also been argued that the rules of evidence to assess the validity of biomarkers are both underdeveloped and not routinely applied (Ransohoff, 2004; Altman and Riley, 2005; LaBaer, 2005). The mechanisms of disease and pharmacologic response are complex and challenging to discern, and lack of appropriate study designs and analytic methods exacerbates the challenge. Perhaps as a result, both genomic and proteomic studies of the same cancer types have often identified discordant biomarker candidates and patterns (reviewed by Diamandis, 2004; Dalton and Friend, 2006; Quackenbush, 2006).

Because of the enormous number of genes and proteins analyzed in genomic and proteomic studies, many false findings can be expected unless appropriate statistical methods are used (Simon, 2005). For example, in the discovery setting, overfitting can lead to erroneous identification of markers or patterns in association with disease. This occurs when multivariate analysis is used to assess associations between large numbers of possible predictors and an outcome—that is, a pattern is found that fits perfectly, but by chance (Ransohoff, 2004, 2005; Simon, 2005, 2006). Overfitting can be easily identified by checking reproducibility in a separate, independent group of individuals, but most published studies do not report this essential assessment of reproducibility (Ransohoff, 2004, 2005).

Sample bias can also render conclusions drawn from a biomarker study invalid (Ransohoff, 2005). Bias can be defined broadly as the systematic but unintentional erroneous association of some characteristic with a group in a way that distorts a comparison with another group. The design, conduct, and interpretation of randomized clinical trials to assess medical interventions place high importance on ensuring that the treated and untreated patient populations are similar in every respect except for the treatment to avoid biases that could affect the outcome and thus the conclusions drawn from the results. However, most research on molecular markers for diagnosis or prognosis entails observational studies, in which it is difficult or impossible to ensure or even fully assess the similarity of the comparison groups, and which are much more likely to result in biased conclusions as a result. In fact, Ransohoff has suggested that bias can be so powerful in non-experimental observational research that a study should be presumed biased until proven otherwise (2005). He notes that a single bias might produce errors sufficiently large to invalidate results. Thus, great care must be taken in the design, conduct, interpretation, and reporting of such research.

The validity of biomarker research also depends on the generalizability of the results. Generalizability concerns how broadly the results can be applied and depends on the characteristics of study participants and how they were selected (Ransohoff, 2005). Initial studies often establish proof of principle, but they have limited generalizability. Subsequent larger studies then aim to assess broader generalizability.

However, the developmental sequence for biomarker development is much less well defined than for drug development. In drug development, the phases of research are clearly delineated in a step-wise fashion to examine several key issues, including generalizability. Phase I studies aim to establish appropriate doses and to identify potential side effects, while phase II studies begin to address biological activity and adverse events. Phase III studies are larger and seek more definitive conclusions on efficacy and safety. Patients in phase I and II studies have often failed all available treatment options and have diverse characteristics, whereas phase III trials select patients who are more representative of how the drug will actually be used in clinical practice (Ransohoff, 2005).

Delineating the phases for biomarker development is more difficult, in part because biomarker tests can be used for many different purposes, and thus research to assess the usefulness of a test must be designed to examine specific applications. The variability in technologies used to identify biomarkers further complicates the situation. As a result, proposals to establish developmental phases for biomarkers have focused on specific uses, such as early detection or surrogate endpoints, or specific methods (Pepe et al., 2001; Baker et al., 2004; Zolg and Langen, 2004). Perhaps because of this variability in the development pathway, the level of assessment and oversight for biomarkers is also more variable, and usually less stringent, than for drugs. The process of developing and implementing biomarkers differs from that of drugs in other ways as well, including economically. These issues are covered in more detail in Chapters 3 and 4.

If the full potential of cancer biomarker-based tools in early detection, treatment, and drug development is to be realized, it will be important to optimize efforts to discover and validate putative biomarkers. Progress in biomarker discovery and development is directly dependent on the capacities of the technologies available. Initially, high-throughput methods to discover biomarkers and expression patterns focused on nucleic acids,

because the methods were more advanced and fully characterized than for other cellular components. The Human Genome Project¹ spurred interest in the development and application of methods to access nucleic acids, and although there is still a need for standardization of reagents, platforms, and analyses, considerable progress has been made in the field. In addition, ongoing work to sequence the genomes of individual tumors as well as other organisms continues to spur the development of new technologies. For example, single-molecule sequencing² is likely to lower the cost of sequencing significantly, and should reduce the problems that arise from normal cell contamination of tumor samples (IOM, 2006a).

However, there are limitations to what can be learned from genomics approaches to assess DNA and RNA. Although RNA assays can detect dynamic changes in gene expression as well as identify upstream DNA-level abnormalities in cancers, it is more costly and difficult to work with than the comparatively stable DNA. It can be argued that proteins, which perform many essential functions in cells, may provide more meaningful biomarkers than either type of nucleic acid because changes in DNA and RNA are not always directly linked to altered protein expression, modification, or function. But progress in the identification of protein biomarkers has lagged, in part because proteins are more numerous and far more subject to quantitative and post-translational structural changes than genes, and in part because of the limitations of current technologies (Tyers and Mann, 2003; Aebersold et al., 2005; Hartwell, 2005; Cottingham, 2006; ). Technologies used to examine other types of biomarkers, such as metabolomics, are even less developed and characterized. Metabolomics entails the study of metabolic responses to drugs, environmental changes, and diseases via identification of small-molecule metabolite profiles; that is, it attempts to measure the metabolic consequences of altered genes and protein expression.

Proteomics research aims to interrogate extremely complex protein mixtures in blood and tissues. It has been estimated that blood contains more than 100,000 different protein forms with abundances that span 10–12 orders of magnitude (Anderson and Anderson, 2002; Jacobs et al., 2005). The leading high-throughput proteomics technology, mass spectrometry (MS), has limited ability to identify and quantify proteins in complex mixtures. Many known biomarkers occur at very low abundance and would not be identified by current technologies (Aebersold et al., 2005; Jacobs et al., 2005; Kolch et al., 2005). New fractionation methods (for depletion or enrichment) prior to MS could improve the process, as currently available methods are tedious and expensive, and are not amenable to high-throughput analysis. Furthermore, identification of the various peptides or proteins detected by the technology remains a difficult challenge. Antibodies raised against specific protein biomarkers could fill this gap, but currently antibodies are not available for most of the proteins that could be disease biomarkers (Anderson and Anderson, 2002; Aebersold et al., 2005; Hartwell, 2005; Jacobs et al., 2005; Cottingham, 2006).

In addition to improving the sensitivity, specificity, and dynamic range of these technologies, it will be important to process the resultant data efficiently and effectively, by developing new software packages, algorithms, and statistical and computation models, including those that can integrate data from multiple inputs, such as proteomic and genomic data from the same samples (Cristoni and Bernardi, 2004; Englbrecht and Facius, 2005; Tinker et al., 2006). These approaches will also necessitate novel sample preparation procedures, from a variety of sources such as blood, plasma, tissues, and cells, and for a variety of analytic technologies, including metabolomics, proteomics, and genomics. Finally it will be necessary to develop new assays to translate discovery into viable clinical tests, and to develop novel approaches for real-time in vivo detection, via imaging and nanomaterials. The Human Proteome Organisation (HUPO), an international consortium of national proteomics research associations, government researchers, academic institutions, and industry partners, has begun to examine some of these issues in a pilot phase of its Plasma Proteome Project (Omenn et al., 2005), and progress is being made on several fronts (de Hoog and Mann, 2004; Ong and Mann, 2005), but much work remains to be done. There is a significant need for new and improved technologies for biomarker discovery and development, particularly in the field of proteomics. Such new technologies also will yield dividends in improved

capabilities for understanding fundamental cellular processes in cancers and systems biology in general.

Although technology development and directed discovery have not traditionally been a primary focus of National Institutes of Health (NIH) funding, and the NIH peer review process generally does not favor high-risk projects (IOM, 2003), that has been changing in recent years and there are some precedents for funding initiatives that focus on large-scale discovery projects and also catalyze the development of improved technologies for biomedical research. As noted above, the Human Genome Project drove not only the development of new technologies, but also improvements in the existing technologies through automation, data standards, and quality control (Aebersold et al., 2005; Hartwell, 2005). More recently, the NIH Roadmap proposed a framework for the priorities NIH as a whole must address in order to optimize its entire research portfolio, laying out a vision for a more efficient and productive system of medical research (NIH 2006b). The NIH Roadmap identified opportunities in three main areas: new pathways to discovery, research teams of the future, and re-engineering the clinical research enterprise. The first main area, pathways to discovery, aims to advance the quantitative understanding of complex biological systems by deciphering the many interconnected networks of molecules that comprise cells and tissues, their interactions, and their regulation. In addition the program aims to increase access in the research community to new and better technologies, databases and other scientific resources that are more sensitive, more robust, and more easily adaptable to evolving needs.

The Protein Structure Initiative (PSI), a $600 million, 10-year venture funded by the National Center for Research Resources of the National Institute of General Medical Sciences, is a recent example of an NIH program that explicitly funded technology development in the initial phase of the project. The overarching goal of the initiative is to determine the structure and function of thousands of proteins by 2010, with the final product serving as an inventory of all the protein structure families in nature. In the first phase of the initiative, PSI funded nine research centers that focused on developing novel and innovative approaches and technologies, such as robotic instruments, to determine protein 3-D structures from knowledge of their amino acid sequences (NIGMS, 2006). Technological innovations were developed for each step of the process, from the initial target selection, to the final poststructural analysis. According to NIH leadership, PSI succeeded in reducing costs four-fold from the initial year, and the techno-

logical improvements are likely to have a broad impact on protein structure research, beyond the PSI-funded work (Norvell and Berg, 2005).

NCI currently also sponsors development of novel nanotechnologies as tools to accelerate advances in biomarker research. Clinical applications of nanotechnologies include measurement and analysis of biomarkers in vivo for early cancer detection, prevention and monitoring of treatment response. NCI programs such as the Innovative Molecular Analysis Technologies Program provide funding for research projects to develop new and emerging technologies, including nanotechnology methods and tools, and for refining existing technologies through to development and commercialization. In September 2004, NCI announced a 5-year $144.3 million initiative to develop nanotechnologies to be used in cancer research (NCI, 2004). The goals of this initiative, the NCI Alliance for Nanotechnology in Cancer, include the development of research tools to identify new biological targets, agents to monitor predictive molecular changes and prevent precancerous cells from becoming malignant, imaging agents and diagnostics to detect cancers early, and systems to provide real-time assessments of therapeutic and surgical efficacy.³ Emerging nanotechnologies include quantum dots, gold nanoparticles, and cantilevers. Quantum dots and magnetic nanoparticles can be used for barcoding of specific analytes, and gold and magnetic nanoparticles are components of a possible alternative to PCR known as the bio-barcode assay. Nanotechnologies can be used to genotype at high-throughput, and some researchers believe that they have the potential to reduce cost for many diagnostic applications (Azzazy et al., 2006). In addition, the size of nanoparticles makes them compatible with in vivo molecular manipulation and measurement (Yezhelyev et al., 2006). Nanodiagnostic assays have already been used to detect Alzheimer biomarkers in cerebrospinal fluid (Azzazy et al., 2006).

The National Cancer Institute (NCI) has also recently launched new funding initiatives for proteomics research, noting that “current proteomic technology approaches are insufficient to reliably and reproducibly discover, identify, and quantify peptides and proteins of clinical significance for cancer” from complex patient samples. The NCI’s Clinical Proteomic Technologies Initiative for Cancer program is a 5-year $104 million program that includes two funding opportunities for proteomics technology development: Advanced Proteomic Platforms and Computational Sciences (DHHS, 2005) and Clinical Proteomic Technology Assessment for

Cancer (CPTAC) (DHHS, 2006b). The goal of the former is to improve technology for protein/peptide detection, recognition, measurement, and characterization in biological fluids and to develop computational, statistical, and mathematical approaches for the analysis, processing, and exchange of large proteomic datasets. The goal of the CPTAC is more specifically to improve measurements of proteins and peptides with mass spectrometry and affinity-based proteomics platforms. The CPTAC Request for Application notes that

While these NIH initiatives are to be commended, a review of projects funded through them suggests that many projects focus primarily on improving existing technologies, rather than on the development of completely novel technologies. To achieve the latter, it might be better to undertake a highly directed contract-based program. An examination of the Defense Advanced Research Projects Agency (DARPA) might prove instructive in that regard. DARPA is the central research and development organization for the Department of Defense, and it has focused on research projects that are high risk but also have potential for high payoff if successful (Box 2-1; IOM, 2003). As such, this approach is particularly amenable to technology development, and past leaders of NIH and NCI have expressed interest in adopting some aspects of the DARPA model to spark technological innovation. In fact, under the leadership of former NCI director Richard Klausner, NCI launched a pilot program that was modeled in part after DARPA, as well as other agencies, including the National Aerospace and Space Administration. Established in 1999, the Unconventional Innovations Program (UIP) focused on the development of novel, long-range technologies to support cancer research. The Program funded research through contracts instead of grants, allowing for enforcement of deadlines for specific milestones along the research track in order for researchers to

continue to receive funds. UIP management actively recruited the interest and involvement of investigators from disciplines that have not traditionally received support from NCI and assembled interdisciplinary research teams focused on cancer detection technologies, including nanotechnologies (IOM, 2003; NCI, 1999).

Regardless of which funding model is used to foster innovative technology development, it will be essential to take an organized, comprehensive approach to the problem and to include a broad array of extramural experts in all aspects of program planning, execution, and oversight. For example, program managers with current expertise in the field could be recruited to direct the projects, similar to the DARPA approach. In addition, panels of experts should provide an oversight role in establishing goals and reviewing progress and performance of the program and individual contracts. Con-

tinuation of contracts should be highly dependent on reaching pre-defined milestones and deliverables.

The involvement of multiple federal agencies is important, as no single agency is likely to have the needed expertise to address all issues, but it will be important for one agency to take the lead in organizing intra-institutional efforts. Given NCI’s current funding level and recent initiatives and interest in biomarker discovery and development, it may seem an obvious choice for the lead agency for this endeavor, but to date it has not yet developed an adequate overarching leadership strategy. The NCI programs described above are relatively narrow in focus, and there is very little coordination or communication among them, with no unifying strategy or oversight. Without appropriate organization and funding, researchers will be unable to muster the resources and knowledge to achieve broad gains, and progress

is likely to continue to be slow and piecemeal. Finally, as noted in the last section of the chapter, providing academic scientists with appropriate incentives and resources will be critical if they are to undertake successfully work that has not been traditionally viewed as an academic pursuit.

Analysis of human tissues is essential for biomarker discovery and validation. Human tissue has been collected and stored in biorepositories for more than 100 years in the United States, and it is estimated that there are more than 300 million tissue specimens from more than 175 million cases stored in the United States, with new specimens accumulating at a rate of more than 20 million per year (reviewed by Eiseman and Haga, 1999). These specimens are collected by a broad array of institutions, both federal and private, but most patient samples were originally collected for diagnostic and therapeutic purposes, so the vast majority are not used in research. NIH is the largest single funding source for tissue repositories, and NCI in particular supports many different tissue repositories for research (Box 2-2). However, despite the common funding source, these biobanks

vary a great deal in their design, methods, standards, data collection, and informed consent, making it difficult to compare data from different studies or to combine data from different repositories. Indeed, there has been a lack of nationally agreed-on quality control and standard operating procedures, which limits the usefulness of existing collections (Eiseman et al., 2003). In addition, genomic and proteomic analyses require tissue preservation methods that differ from those most commonly used in diagnostic pathology.

In an attempt to address these problems, NCI commissioned a study of 12 U.S. biorepositories (half of which were supported by NIH) to identify best practices, with the goal of developing standard operating procedures. Best practices were identified for all aspects of specimen collection, storage and use, including processing and annotation, storage and distribution, bioinformatics, consumer and user needs, business plan and operations, privacy, ethical and consent issues, intellectual property (IP) and legal issues, and public relations, marketing, and education (Eiseman et al., 2003). NCI also commissioned a blueprint for a National Biospecimen Network (NBN) with the goal of providing a “comprehensive framework for sharing and comparing research results through a robust, flexible, scalable, and secure bioinformatics system that supports the collection, processing, storage, annotation, and distribution of biospecimens and data” (Friede et al., 2003). Primary objectives of the blueprint were to collect biospecimens that were amenable to genomic and proteomic analysis, as well as to ensure uniformity so that data from different studies could be combined or compared. None of the biorepositories examined by Eiseman et al. (2003) had all the characteristics identified as necessary in the NBN report.

The NBN blueprint report has been criticized by some, most notably for its projected costs (Goldberg, 2003). Nonetheless, in 2005, NCI established the Office of Biorepositories and Biospecimen Research with the objective of improving and standardizing biobanking activities and to facilitate the establishment of a National Biospecimen Network (NCI, 2006b). In April 2006, the office put forth first-generation guidelines for all NCI-supported biorepositories (NCI, 2006a), addressing common best practices for research biorepositories, quality assurance, and quality control programs, informatics systems, ways to address ethical, legal, and policy issues (e.g., informed consent, privacy, data security protections, Institutional Review Board oversight, ownership of and access to biospecimens and data), standardized reporting mechanisms, and administration and management structure. Second-generation guidelines, currently being developed in collaboration with the American College of Pathologists and

other relevant extramural groups, will propose evidence-based standard operating procedures. A pilot test of the proposed NBN to evaluate the use of best practices for collecting specimens from prostate cancer patients for biomarker research is also ongoing (NCI, 2006c).

The committee supports the development of practice guidelines and standards as well as the harmonization of ethical, legal, and policy issues, and emphasizes the importance of developing strategies to maximize the quality and usefulness of biorepositories while also protecting patient rights. For example, it is important to develop consensus on common data elements for collecting patient information and to make this information and samples easily accessible to researchers. NCI’s Early Detection Research Network has made considerable progress in this regard and provides a good model for how to proceed with other biospecimen collections (NCI, 2005, Figure 2-1). Supporting and encouraging the use of electronic patient records would facilitate this work as well.

It is also critical to develop strategies to ensure the confidentiality of identifiable patient health information under the Privacy Rule of the Health Insurance Portability and Accountability Act (HIPAA), without

FIGURE 2-1 EDRN informatics infrastructure.

SOURCE: EDRN website, 2006.

impeding research. This may require a reassessment of the privacy regulations established under HIPAA, as well as efforts to promote uniformity in interpretation across states and institutions (Bledsoe, 2004; IOM, 2006b). Promoting interagency harmonization of informed consent is also likely to facilitate research on biomarkers. The committee notes that some privately established biorepositories have been quite successful in dealing with some of the issues that NCI is grappling with, and much could be learned from their example. The Multiple Myeloma Research Consortium Tissue and Data Bank is a prime example (Box 2-3).

However, some challenges remain unaddressed despite these recent activities undertaken by NCI. First, most biomarkers are developed using archived tumor specimens that were collected for other purposes, and many of the discordant findings in the biomarker literature may be due to this retrospective approach (Simon, 2006). Clinical trials to test drugs are usually prospective, with hypotheses, patient selection criteria, analysis plans, and primary endpoints clearly defined in advance of the study. In contrast, biomarker studies are usually performed without a prespecified written protocol defining the hypothesis, eligibility requirements, primary endpoints, or analysis plan. In addition, the specimen population is often very heterogeneous, representing different cancer stages and treatments. This often leads to multiple subset analyses, which increases the chance of false-positive conclusions (Simon, 2005). Many biomarker studies also perform analyses for multiple candidate biomarkers and multiple endpoints, further multiplying the chances for erroneous conclusions (Simon, 2005).

An obvious solution to the problem would be to undertake prospective studies specifically designed to identify and validate predictive or prognostic biomarkers. However, that approach could be prohibitively expensive. A more viable alternative would be to combine prospective therapeutic clinical trials with biomarker studies (Dalton and Friend, 2006), defining appropriate criteria and analyses from the start. NCI should actively encourage and facilitate interaction between biomarker developers and groups involved in clinical research, including therapeutic, screening, prevention, and cohort studies, to enable the prospective collection of high-quality patient samples that are intended to test specific biomarker hypotheses. Open access to all interested biomarker developers (both industry as well as academia) should be a defining feature. True openness will be the best way to ensure these repositories and patient samples are leveraged to the greatest extent possible, and there is good precedent for this from the genome-wide genetic

association data now being placed into the public domain with all qualified researchers able to access the raw data and samples for discovery purposes.

NIH should also initiate and sustain funding of biorepositories that are created in conjunction with large cohort studies and clinical trials, and use of these prospectively collected samples should be encouraged for validating biomarkers. When clinical trials end or funding for cohort studies is not renewed, the ability to maintain the biorepositories created in conjunction with the study is often lost (Goodman et al., 2006). The samples collected in these prospective studies may be very valuable for biomarker research

and development, and NIH should consider continued funding for maintenance of biorepositories even if the original study itself is not continued.

In the long term, it may be more feasible to support such biorepositories through public–private consortia (see below). But regardless of how a repository is supported, funding must be sufficient to cover all essential components and activities, including involvement of pathologists to assess sample quality and confirm diagnosis, optimized sample collection and preparation, consistent capture and annotation of clinical patient records, medical informatics and database management, and general administrative

and maintenance costs. In addition, extramural experts should be broadly represented on any oversight committees.

Given the challenge and expense of developing biomarkers, as described above, it may be difficult or impossible for any single company or organization to successfully undertake the work alone. Once validated, biomarkers could prove useful for many stakeholders, including researchers, drug developers, and clinicians, but individual stakeholders may lack the necessary information and resources to effectively develop and validate markers. For example, validating a surrogate endpoint requires costly and lengthy clinical studies. As such, companies inevitably find it cheaper and faster to directly measure the primary endpoint of interest for a particular drug than to first validate the surrogate marker (Fleming, 2005). But once a surrogate marker has been fully validated for a pharmacologic class of treatment regimens, any drug developer could take advantage of the marker to streamline the development of drugs in that class.

Thus, the sharing of precompetitive data and cooperation in developing and validating biomarkers as common goods is paramount to progress in the field. By leveraging the strengths of different partners, consortia offer many advantages over individual efforts (Kettler et al., 2003; Schwartz and Vilquin, 2003; Nishtar, 2004; Chin-Dusting et al., 2005; Croft, 2005). Partnerships can lead to greater efficiency and effectiveness by pooling skills, technologies, and other resources. By sharing costs and risks while also reducing legal and IP barriers, consortia are more likely to take on challenges that, individually, the partners would be unlikely to tackle. Public– private partnerships (PPPs) in particular can more effectively leverage public funding and resources, increase the breadth and depth of representation in science and scientific agendas, and effect a more rapid translation from basic discoveries to public health applications (Mittleman, 2006). Industry, government, and nonprofit organizations all have a potential role to play in such partnerships, and could each make important and unique contributions to the endeavor.

Although private companies normally are inclined to protect their data to maintain a competitive edge, there are numerous precedents of successful PPPs developing tools and generating pre-competitive data to move a field forward, both in biomedical research and in other industries. For example, SEMATECH (Semiconductor Manufacturing Technology), established in

1987, helped U.S. semiconductor suppliers develop new production tools and establish industry-wide consensus on product specifications (Box 2-4). In biomedicine, PPPs constitute a common approach to tropical and neglected diseases (Kettler et al., 2003; Nishtar, 2004; Croft, 2005), and collaborative efforts have already been successful in biomarker development as well. For instance, the HIV Surrogate Markers Collaborative Group, established through multiple partnerships among academia, industry, and government, confirmed the usefulness of HIV RNA as a surrogate marker for testing new anti-HIV drugs. As a result, the Food and Drug Administration (FDA) began to approve drugs based on evidence of lower levels of plasma HIV RNA in response to drug therapy (Behrman, 1999; Mildran, 2006). Also, the International Life Science Institute, and its Health and Environmental Sciences Institute, which bring industry, government agencies, and academics together to share data and information on nutrition, food safety, toxicology, risk assessment, and the environment, has aided the development of biomarker candidates for toxicological assays (both genomic and proteomic).

The SNP Consortium (TSC) is a well-known example of an international public/private collaboration in biomedical research (TSC, 2006; Box 2-5) that demonstrates the willingness of multi-national drug companies to share information to achieve a common goal and shows the feasibility of this collaborative approach for fostering precompetitive work that could benefit the entire field. Single nucleotide polymorphisms (SNPs) are common small DNA variations that occur throughout the human genome. The primary objective of TSC was to create a high-quality, publicly available map of human SNPs, with the hope that it would aid genomic research and the development of genetic-based diagnostics and therapeutics. TSC exceeded its primary goal by identifying and mapping about 10-fold more SNPs than originally planned, while also completing the work in less time and with a smaller budget than had been scheduled. According to Arthur Holden, chief executive officer of TSC, several factors were critical to that accomplishment, including a clear, focused, and unifying objective, a carefully crafted work plan, strong teamwork with experienced management, and preeminent external advisers and investigators (Holden, 2006).

Given the power of partnerships to address unmet needs in biomedicine, a number of new consortia have recently been formed to develop and validate biomarkers (Feigal, 2006; Holden, 2006; IOM, 2006a; Mittleman, 2006). For example, the Critical Path (C-Path) Institute is a publicly funded nonprofit consortium consisting of pharmaceutical industry partners with

the goal of identifying and validating preclinical and clinical biomarkers for predicting human drug toxicities (Box 2-6). The companies that participate in this consortium share data and methods in order to test and cross-validate one another’s markers and methods.

Another new consortium is the Oncology Biomarker Qualification Initiative (OBQI), created to join the efforts of the FDA, NCI, and the Centers for Medicare and Medicaid Services (CMS) with the goal of improving cancer therapeutics and patient outcomes through biomarker development and evaluation. The OBQI aims to facilitate the codevelopment of diagnostic-therapeutic combinations and to reduce the time and cost of drug development by shortening clinical trials through enriched patient populations more likely to respond to therapies. The first project of OBQI will entail a PPP to qualify fluorodeoxyglucose positron emission tomography (FDG-PET) scanning as a marker for drug response in non-Hodgkin’s lymphoma.

The Pharmaceutical Biomedical Research Consortium (PBRC) was also recently formed, with the mission of “advancing the field of medicine, through the development and implementation of high quality pre-competitive biomedical consortia developed with its pharmaceutical members, in conjunction with appropriate other partners” (Holden, 2006; Box 2-7). The PBRC intends to undertake a number of independent projects, several of which will focus on biomarkers, including surrogate markers and predictive markers of serious adverse events.

NIH has also recently spearheaded several PPPs to develop biomarkers for common diseases, such as osteoarthritis and Alzheimer’s disease, with more projects planned for the future (Mittleman, 2006; NIH, 2006a). A primary goal of NIH is to establish policy regarding use of samples and information from study collections that have already been created. NIH also plans to develop common Institutional Review Boards for multi-centered studies so that a single determination can be provided by a single, common IRB, rather than multiple institutional IRBs providing multiple determinations.

In October 2006, the Foundation for the National Institutes of Health (FNIH), NIH, the FDA, and the Pharmaceutical Research and Manufacturers of America (PhRMA) announced the launch of another public–private biomedical research partnership, the Biomarkers Consortium, to search for and validate new biomarkers. FNIH has already secured $3 million in donations from major pharmaceutical companies for the consortium, and more funders are anticipated to join the effort. Like the OBQI, the first project

of this new consortium will be to evaluate the use of FDG-PET to measure response to treatment, but in this case, the group will simultaneously study lung cancer in addition to non-Hodgkin’s lymphoma. Other projects under consideration will focus on mental health and diabetes (FNIH, 2006).

All of these new collaborative efforts are commendable, but with the exception of the projects on FDG-PET imaging, none is focused on developing cancer biomarkers. Thus, the committee recommends that industry and other funders of biomedical research establish international public–private consortia to generate and share methods and precompetitive

data on the discovery, validation, and qualification of cancer biomarkers. A more cooperative and comprehensive approach that attempts to leverage and integrate all available data could have an enormous impact on the field. Such efforts could lead to better biomarkers for the entire spectrum of cancer health care, from early detection and disease classification to drug development and treatment planning and monitoring (Bast et al., 2005; Dalton and Friend, 2006). Organizers should examine and learn from past and ongoing biomedical consortia, especially the SNP Consortium, which provides a valuable model. Many complex issues must be addressed

in forming consortia, including governance, data sharing and access, intellectual property management, human subjects protections, and antitrust laws. Lessons from past experience could help to streamline the process and increase the probability of broad participation and success.

Most drugs are effective in only a fraction of the patients who receive them (Spear et al., 2001). This is especially true for cancer drugs, with an average drug response rate of less than 25 percent. The variability in response is due largely to the heterogeneity of specific molecular changes in tumors, which cannot be identified by current diagnostic methods. Differences in drug metabolism due to genetic variability (polymorphism) in patients can also contribute to the inconsistency in drug response.

The current approach to cancer treatment is still largely empirical and centered on population-based statistics (Dalton and Friend, 2006). Treatments are assigned according to diagnostic categories that are derived from cancer type and stage, rather than specific molecular changes. Biomarkers that would enable physicians to choose the treatment most likely to benefit a given patient could greatly improve treatment outcome, both in terms of improved effectiveness and in avoiding potentially debilitating but ineffective treatment. These improvements would also enhance the cost-effectiveness of treatment (see Chapter 4 for more information on cost-effectiveness), by increasing the probability that expensive treatments will be effective and by reducing the costs associated with managing toxic side effects.

However, once a drug is approved by the FDA, drug companies have relatively little incentive to develop biomarkers to guide treatment decisions, as this would likely restrict the population of patients treated with

the drug. For example, several new drugs that inhibit the epidermal growth factor receptor (EGFR) were recently approved by the FDA (FDA News, 2003, 2004a, 2004b, 2006). However, in each case, only a small minority of patients with the type of cancer for which the drugs are indicated actually responds to treatment, and to date no biomarkers have been shown to effectively identify patients who will and will not respond to each drug. Furthermore, several additional EFGR inhibitors are in clinical trials, so making the appropriate treatment choice could become even more chal-

lenging if and when these drugs gain FDA approval and enter the market (Grunwald and Hidalgo, 2003; Baker, 2004). Some efforts are under way to identify biomarkers to guide EGFR treatment decisions, but these studies are largely being done in academic settings. Furthermore, these efforts are not coordinated or unified through data sharing or by a common strategy.

Federal agencies and other funders should therefore support demonstration projects to discover and develop biomarkers that can predict the safety and effectiveness of FDA-approved oncology drugs in individual patients,

with the goal of selecting appropriate target populations for those drugs. A high-impact finding for a specific disease or drug-targeted pathway would not only improve treatment outcomes for patients, but also could define an optimal approach for the biomarker field and catalyze the diagnostic and pharmaceutical industries to undertake such studies for many cancers and therapeutics by establishing a viable route to market with profitable returns (i.e., viable business models) (IOM, 2006a). Such a finding would also establish precedents for health care providers to use biomarkers to guide therapy decisions and for payors to provide coverage for such tests.

A precedent for using a pharmacogenomic biomarker test to predict the major toxicity of a cancer drug already exists. Irinotecan, used primarily to treat advanced colorectal cancer, is a prodrug that is converted by a cellular enzyme to an active form. This active form of the drug is further modified by a second enzyme known as UGT1A1, allowing it to be eliminated from the body more efficiently (reviewed by Nguyen et al., 2006; Maitland et al., 2006). People with certain polymorphisms in the UGT1A1 gene have reduced ability to clear the active drug from the body and are therefore much more likely to experience dangerous adverse effects from the drug, including severe myelosuppression and diarrhea. As a result of these findings, in 2005 the FDA approved revisions to the safety labeling for irinotecan to recommend reduced dosing in patients who are homozygous for a specific UGT1A1 allele. A month later, the FDA also approved a molecular diagnostic test to identify patients with variations in the UGT1A1 allele that may be at increased risk of adverse side effects from irinotecan.

Another potentially informative case of how pharmacogenomics might be used to predict the effectiveness of a drug is a body of work undertaken to understand the variability in response to the antiestrogen tamoxifen in breast cancer patients (Box 2-8). Estrogen receptor status has been used for many years to identify patients who are likely to respond to tamoxifen, but more recent studies indicate that variations in the CYP2D6 gene might be used to identify nonresponders within that subgroup (Goetz et al., 2005). This could be useful information, as other treatment options, such as aromatase inhibitors, are now available to treat women with ER+ tumors. The investigators were fortunate to have access to a biomarker test already approved by the FDA,⁴ so other projects may face additional hurdles in

developing appropriate drug-targeting biomarkers. Nonetheless, this effort provides a working example of what might be possible if a concerted effort is undertaken to identify and validate treatment stratification biomarkers for drugs in use. Questions about how best to conduct such studies will need to be addressed early on. In particular, the studies must be well designed and adequately powered, but since patients are already taking the drugs, it should not be difficult to accrue participants for a study (IOM, 2006a).

It is now widely accepted that genetic mutations and epigenetic changes are primary driving forces in the initiation and progression of tumors. Cancers have been increasingly linked to changes that affect how proteins function within signaling pathways that control cell growth and death, motility, metabolism, and genomic integrity (Coleman and Tsongalis, 2006; Esteller, 2006; Varmus, 2006). That is, cancers arise and progress when cell-signaling pathways are altered. Furthermore, much of the heterogeneity among cancer patients can be traced to differences in the specific pathways that have been modified in each tumor. Decades of research have gradually led to the identification and delineation of the pathways that control these vital cell functions, and recent advances in developing molecularly targeted therapies, like imatinib and trastuzumab, are derived from that increased understanding of signaling pathways. In addition, acquired resistance to these cancer drugs is attributed to secondary mutations in critical signaling pathways (Baselga, 2006).

Thus, biomarkers that can detect alterations in specific signaling pathways would be extremely useful for the detection, diagnosis, and treatment of cancers. Classification of tumors by molecular changes rather than by organ site and morphologic appearance could radically change the approach to cancer care. Screening tests could be devised to detect altered pathways common to many cancers, rather than developing many organ-specific tests for cancers. Once cancer is diagnosed, identifying the altered pathways would aid in making individualized treatment decisions. Biomarkers tied to specific drugs, in contrast, can provide only a yes/no answer for that particular drug; they cannot suggest an optimal alternate treatment. In addition, it is widely believed that targeting multiple pathways will be necessary to effectively treat most cancers (Baselga, 2006). Pathway biomarkers would allow for a “systems” approach to diagnosis, treatment, and surveillance (van der Greef and McBurney, 2005), recognizing that pathways operate in the

context of interconnected networks. A recent study published in the journal Nature described a novel approach to define such pathway signature markers to aid prognosis and predict drug sensitivity (Bild et al., 2006).

Pathway biomarkers could also help identify new drug targets and streamline the drug development process (Stoughton and Friend, 2005; Dalton and Friend, 2006). It can be argued that a lack of financial incentives limits industry investment and efforts to develop narrowly targeted drugs for specific subsets of cancers. However, targeting pathways that are common to many different types of cancer could expand the potential use of and market for new molecularly targeted drugs. For example, pathway biomarkers used to develop drugs for common cancers could potentially

also be useful in rarer forms of cancer that are more difficult to study and would offer smaller returns to developers.

This emphasis on pathways could also invigorate the field of biomarker development itself. Biomarkers that are exclusively focused on a particular drug must be developed in conjunction with each new drug, at a high cost and with considerable risk. If an experimental drug does not achieve FDA approval, work on the associated biomarker would be for naught (IOM, 2006a). And even if FDA approval is obtained, the biomarker could still become obsolete if newer, more effective drugs become available and therapy guidelines change. In contrast, pathway markers are more likely to be applicable to the development of any new drug that targets an essential

pathway. Broad applicability would be optimized by emphasizing the development of objectively quantifiable biomarkers, rather than qualitative or semi-quantitative assays such as immunohistochemistry. This broader applicability will increase the potential market and also reduce the risk associated with the development process, thus improving the odds of profitability for the company. One caveat is that the requirements for sensitivity, specificity, precision, and accuracy of a biomarker may vary among different diseases, so markers would not necessarily be directly transferable. But the secondary development process for another disease is likely to be shorter and less expensive than starting from scratch.

NIH and NCI have both recently stressed the need to support and facilitate translational research to ensure that critical basic science discoveries move “from bench to bedside” and that unmet medical needs in turn drive further bench research. The NIH Roadmap noted that “growing barriers between clinical and basic research, along with the ever increasing complexities involved in conducting clinical research, are making it more difficult to translate new knowledge to the clinic” (NIH, 2006b). An annual report from the President’s Cancer Panel concluded that “the translational research infrastructure is inadequate to enable the work that needs to be done; resources must be committed to develop the tools and workforce required. Increased funding for translation-oriented research—particularly collaborative, team efforts—is urgently needed across the translation continuum. Targeted Federal funding for translation-oriented research is drastically out of balance relative to financial commitments to basic science. Ways must be found to increase human tissue and clinical research resources without slowing the discovery engine. Supplemental funding may offer a temporary solution, but will be inadequate in the long term” (President’s Cancer Panel, 2005). Similarly, the NCI’s Translational Research Working Group (TRWG), recently appointed to evaluate the status of NCI’s investment in translational research and to envision the future, concluded that translational research is not well coordinated across NCI and that the resulting fragmented efforts are often duplicative and could lead to missed opportunities (Goldberg, 2006; NCI, 2006d). Specifically, a draft report from the TRWG concluded that

• the absence of clearly designated funding and adequate incentives for researchers threatens the perceived importance of translational research in NCI;

• the absence of a structured, consistent review and prioritization process tailored to the characteristics and goals of translational research makes it difficult to direct resources to critical needs and opportunities;

• translational research core activities are often duplicative and inconsistently standardized, with capacity poorly matched to need;

• the multidisciplinary nature of translational research and the need to integrate sequential steps in complex development pathways warrants dedicated project management resources; and

• insufficient collaboration and communication between basic and clinical scientists, and the paucity of effective training opportunities limits the supply of experienced translational researchers.

Support for translational research activities will be critical for developing and validating putative biomarkers. Initial discoveries of potential biomarkers are often published in high-impact journals, but subsequent work to confirm and validate those findings often does not merit publication in those same journals. Furthermore, such validation work often takes many years to complete and can require an interdisciplinary team approach to science that is not the norm in academia (reviewed by IOM, 2003; Gray, 2006; Kaiser, 2006a). The academic culture traditionally has not been supportive of faculty that engage in team science or translational research; promotion and reward structures are designed to recognize individual initiative and accomplishment. Thus, it will be important to consider how academic organizational structures, metrics for academic promotion, and the cultures of biomedical research can better support team building and multidisciplinary science. Key factors for success will include providing sufficient time, resources, and rewards for faculty who undertake translational research (Gray, 2006). Training programs that specifically deal with the many complexities of this work are also needed to help new translational investigators get started and become established (Kaiser, 2006a).

Although not a traditional NIH funding focus, several recent initiatives have been undertaken to foster translational research. For example, NIH’s new Clinical and Translational Science Award Program encourages institutions to develop new approaches to clinical and translation research, including new organizational models and training programs, and to develop novel clinical research methodologies (DHHS, 2006a; Gray, 2006; Kaiser,

2006b). Through this program, 12 institutions recently received 5-year awards totaling $108 million for the first year. The program is intended to eventually replace the 50-year old NIH program of General Clinical Research Centers, which currently consists of approximately 60 facilities with beds for patients participating in clinical studies. The NCI TRWG draft report also recommends a new organizational structure to coordinate NCI’s translational research, with designated leadership and budget and oversight by an external advisory committee. Some private funders, such as the Howard Hughes Medical Institute, the Burroughs Wellcome Fund, and the Doris Duke Charitable Foundation, have put a recent emphasis on translational research as well (Kaiser, 2006a).

Nonetheless, there is concern that funding and other support will not be maintained well enough to sustain a nascent, growing field of translational specialists (Kaiser, 2006a). But continued funding from federal and private sponsors of this work is essential for progress in reaching the goal of personalized medicine. Indeed, NCI has noted that the development of new diagnostic tests, cancer treatments, and other interventions that benefit people with cancer and people at risk for cancer will rely on strong translational research collaborations between basic and clinical scientists to generate novel approaches (NCI, 2006d).

The discovery and development of biomarkers entails a complex, multistage process, with many challenges that must be overcome to make meaningful progress in the field. Despite a few spectacular successes, the number of biomarkers used in drug development or clinical practice is very small, and most putative biomarkers never advance beyond the discovery stage. Moreover, the limitations of current technology render many discovery efforts inefficient and inadequate. Changes are needed to streamline the process and make the most of limited resources available for biomarker research and development.

First, a more organized, comprehensive approach to biomarker discovery is needed. Such an approach would more effectively foster technological innovation and could lead to more efficient, systematic searches for potential biomarkers. Second, international public–private consortia are needed to generate and share methods and precompetitive data on the validation and qualification of cancer biomarkers. Given the accomplishments of previous endeavors like The SNP Consortium, such collaborations are likely

to reduce the cost and risk of biomarker development and enable the field to move forward more efficiently.

Funders of biomedical research should also place more emphasis on developing pathway biomarkers and on developing biomarkers for drugs already in use. A focus on quantifiable biomarkers of signaling pathways rather than individual cancers or drugs could increase the applicability of biomarkers and thus increase the potential for return on investments by sponsoring companies.

Demonstration projects to develop biomarker tests that could determine which patients are most likely to benefit from drugs that are already in the clinic would not only improve treatment outcomes for patients, but also could catalyze industry and academia to undertake such studies by establishing a viable route to market and by delineating a viable business strategy.

Ensuring the availability of high-quality and well-annotated patient samples that have been collected in prospective studies will be crucial to progress in discovering and developing biomarkers. Thus, funders of biomedical research funders should initiate and sustain funding for biorepositories of such patient samples collected in conjunction with large cohort studies and clinical trials, and use of these samples should be encouraged for validating biomarkers. NCI in particular should actively encourage and facilitate interaction between biomarker developers and clinical trials groups to enable this prospective collection of patient samples.

Collectively, these strategies could lead to better biomarkers for the entire spectrum of cancer health care, from early detection and disease classification to drug development and treatment planning and monitoring, and they could bring personalized medicine closer to being a reality.

Aebersold R, Anderson L, Caprioli R, Druker B, Hartwell L, Smith R. 2005. Perspective: A program to improve protein biomarker discovery for cancer. Journal of Proteome Research 4(4):1104-1109.

AJHP (American Journal of Health-System Pharmacy) News. 2006. Genetics examined in tamoxifen’s effectiveness. [Online]. Available: http://www.ashp.org/news/ShowArticle.cfm?id=17501 [accessed November 2006].

Altman DG, Riley RD. 2005. Primer: An evidence-based approach to prognostic markers. Nature Clinical Practice Oncology 2(9):466-472.

Anderson NL, Anderson NG. 2002. The human plasma proteome: History, character, and diagnostic prospects. Molecular and Cellular Proteomics 1(11):845-867.

Azzazy HM, Mansour MM, Kazmierczak SC. 2006. Nanodiagnostics: A new frontier for clinical laboratory medicine. Clinical Chemistry 52(7):1238-1246.

Baker M. 2004. EGFR inhibitors square off at ASCO. Nature Biotechnology 22(6):641.

Baker SG, Kramer BS, Prorok PC. 2004. Development tracks for cancer prevention markers. Disease Markers 20(2):97-102.

Baselga J. 2006. Targeting tyrosine kinases in cancer: The second wave. Science 312(5777):1175-1178.

Bast RC Jr, Lilja H, Urban N, Rimm DL, Fritsche H, Gray J, Veltri R, Klee G, Allen A, Kim N, Gutman S, Rubin MA, Hruszkewycz A. 2005. Translational crossroads for biomarkers. Clinical Cancer Research 11(17):6103-6108.

Behrman RE. 1999. FDA approval of antiretroviral agents: An evolving paradigm. Drug Information Journal (33):337-341.

Bernard S, Neville KA, Nguyen AT, Flockhart DA. 2006. Interethnic differences in genetic polymorphisms of CYP2D6 in the U.S. population: Clinical implications. The Oncologist 11(2):126-135.

Bild AH, Yao G, Chang JT, Wang Q, Potti A, Chasse D, Joshi MB, Harpole D, Lancaster JM, Berchuck A, Olson JA Jr, Marks JR, Dressman HK, West M, Nevins JR. 2006. Oncogenic pathway signatures in human cancers as a guide to targeted therapies. Nature 439(7074):353-357.

Bledsoe M. 2004. HIPAA models for repositories. ISBER (International Society for Biological and Environmental Repositories) Newsletter 4(1):3-8.

Chin-Dusting J, Mizrahi J, Jennings G, Fitzgerald D. 2005. Outlook: Finding improved medicines: The role of academic-industrial collaboration. Nature Reviews Drug Discovery 4(11):891-897.

Citron ML. 2004. Dose density in adjuvant chemotherapy for breast cancer. Cancer Investigation 22(4):555-568.

Coleman WB, Tsongalis GJ. 2006. Molecular mechanisms of human carcinogenesis. EXS (96):321-349.

Cottingham K. 2006. Speeding up biomarker discovery. Journal of Proteome Research 5(5):1047-1048.

Cristoni S, Bernardi LR. 2004. Bioinformatics in mass spectrometry data analysis for proteomics studies. Expert Rev Proteomics 1(4):469-483.

Croft SL. 2005. Public–private partnership: From there to here. Transactions of the Royal Society of Tropical Medicine and Hygiene 99 Suppl 1:S9-S14.

Dalton WS, Friend SH. 2006. Cancer biomarkers—an invitation to the table. Science 312(5777):1165-1168.

DARPA (Defense Advanced Research Projects Agency). 2006. DARPA Webpage. [Online]. Available: http://www.darpa.mil/ [accessed July 2006].

De Bortoli M, Biglia N. 2006. Gene expression profiling with DNA microarrays: Revolutionary tools to help diagnosis, prognosis, treatment guidance, and drug discovery. In: Gasparini G, Hayes DF, Biomarkers in Breast Cancer: Molecular Diagnostics for Predicting and Monitoring Therapeutic Effect. Totowa, NJ: Humana Press Inc. Pp. 47-61.

de Hoog CL, Mann M. 2004. Proteomics. Annual Review of Genomics and Human Genetics 5:267-293.

Desta Z, Ward BA, Soukhova NV, Flockhart DA. 2004. Comprehensive evaluation of tamoxifen sequential biotransformation by the human cytochrome P450 system in vitro: Prominent roles for CYP3A and CYP2D6. Journal of Pharmacology and Experimental Therapeutics 310(3):1062-1075.

DHHS (Department of Health and Human Services). 2005. December. Advanced Proteomic Platforms and Computational Sciences for the NCI Clinical Proteomic Technologies Initiative. [Online]. Available: http://grants.nih.gov/grants/guide/rfa-files/rfa-ca-07-005.html [accessed July 11, 2006].

——. 2006a. Institutional Clinical and Translational Science Award. [Online]. Available: http://grants.nih.gov/grants/guide/rfa-files/RFA-RM-06-002.html [accessed July 2006].

——. 2006b February. Clinical Proteomic Technology Assessment for Cancer. [Online]. Available: http://grants.nih.gov/grants/guide/rfa-files/RFA-CA-07-012.html [accessed July 11, 2006].

Diamandis EP. 2004. Mass spectrometry as a diagnostic and a cancer biomarker discovery tool: Opportunities and potential limitations. Molecular and Celullar Proteomics 3(4):367-378.

EDRN website. 2006. [Online]. Available: http://edrn.nci.nih.gov [accessed October 2006].

Eiseman E, Haga, S. 1999. Handbook of Human Tissue Sources. Washington, DC: RAND.

Eiseman E, Bloom G, Brower J, Clancy N, Olmsted S. 2003. Case Studies of Existing Human Tissue Repositories. Arlington, VA: RAND Corporation.

Englbrecht CC, Facius A. 2005. Bioinformatics challenges in proteomics. Combinatorial Chemistry and High Throughput Screening 8(8):705-715.

Esteller M. 2006. Epigenetics provides a new generation of oncogenes and tumour-suppressor genes. British Journal of Cancer 94(2):179-183.

Fan TW, Lane AN, Higashi RM. 2004. The promise of metabolomics in cancer molecular therapeutics. Current Opinion in Molecular Therapeutics 6(6):584-592.

FDA (Food and Drug Administration). October 18, 2006. Summary Minutes of the Advisory Committee Pharmaceutical Science Clinical Pharmacology Subcommittee October 18-19, 2006. [Online]. Available: http://www.fda.gov/OHRMS/DOCKETS/AC/06/minutes/2006-4248m1.pdf [accessed November 2006].

FDA News. May 5, 2003. FDA Approves New Type of Drug for Lung Cancer. [Online]. Available: http://www.fda.gov/bbs/topics/NEWS/2003/NEW00901.html [accessed September 2006].

——. February 12, 2004a. FDA Approves Erbitux for Colorectal Cancer. [Online]. Available: http://www.fda.gov/bbs/topics/NEWS/2004/NEW01024.html [accessed September 2006].

——. November 19, 2004b. FDA Approves New Drug for the Most Common Type of Lung Cancer. [Online]. Available: http://www.fda.gov/bbs/topics/news/2004/NEW01139. html [accessed September 2006].

——. September 27, 2006. FDA Approves New Drug for Colorectal Cancer, Vectibix. [Online]. Available: http://www.fda.gov/bbs/topics/NEWS/2006/NEW01468.html [accessed October 19, 2006].

Feigal E. 2006. Partnerships to Accelerate Innovation. Presentation at the meeting of the National Cancer Policy Forum, Washington, DC.

Feng Z, Prentice R, Srivastava S. 2004. Research issues and strategies for genomic and proteomic biomarker discovery and validation: A statistical perspective. Pharmacogenomics 5(6):709-719.

Fleming TR. 2005. Surrogate endpoints and FDA’s accelerated approval process. Health Affairs 24(1):67-78.

FNIH (Foundation of the National Institutes of Health). 2006. Public–private partnership forms The Biomarkers Consortium to advance the science of personalized medicine: Lung cancer and lymphoma set for initial investigations, prospective projects in major depression and diabetes cited. Bethesda, MD: Foundation of the National Institutes of Health.

Friede A, Grossman R, Hunt R, Li R, Stern S (editors). September 2003. National Biospecimen Network Blueprint. Durham, NC: Constella Group, Inc.

Gjerde J, Hauglid M, Breilid H, Lundren S, Varhaug JE, Kinsanga ER, Mellgren G, Steen VM, Lien EA. 2005. Relationship between CYP2D6 and SULT1A1 genotypes and serum concentrations of tamoxifen and its metabolites during steady state treatment in breast cancer patients. Abstract. San Antonio Breast Cancer Symposium (June):5086.

Goetz MP, Rae JM, Suman VJ, Safgren SL, Ames MM, Visscher DW, Reynolds C, Couch FJ, Lingle WL, Flockhart DA, Desta Z, Perez EA, Ingle JN. 2005. Pharmacogenetics of tamoxifen biotransformation is associated with clinical outcomes of efficacy and hot flashes. Journal of Clinical Oncology 23(36):9312-9318.

Goldberg K. 2006. NCI’s translational research uncoordinated, advisors say. The Cancer Letter 32(24):4-5.

Goldberg P. 2003. A tissue bank to break the bank? NCI, Dialogue plan expensive resource. The Cancer Letter 29(32):1-8.

Goodman GE, Thornquist MD, Edelstein C, Omenn GS. 2006. Biorepositories: Let’s not lose what we have so carefully gathered! Cancer Epidemiology, Biomarkers & Prevention 15(4):599-601.

Gray M. 2006. Broad cultural shift needed to maximize potential of translational research. Research Policy Alert. [Online]. Available: http://www.aimbe.org/assets/library/66_012406.pdf [accessed August 2006].

Grunwald V, Hidalgo M. 2003. Developing inhibitors of the epidermal growth factor receptor for cancer treatment. Journal of the National Cancer Institute 95(12):851-867.

Hartwell L. 2005. How to build a cancer sensor system. The Scientist 18-19.

Hayes DF, Trock B, Harris AL. 1998. Assessing the clinical impact of prognostic factors: When is “statistically significant” clinically useful? Breast Cancer Research and Treatment 52(1-3):305-319.

Holden A. 2006. IOM Session on Biomarker Consortia. Presentation at the meeting of the National Cancer Policy Forum, Washington, DC.

Hudis CA. 2005. Clinical implications of antiangiogenic therapies. Oncology (Williston Park) 19(4 Suppl 3):26-31.

IOM (Institute of Medicine). 2003. Large-Scale Biomedical Science: Exploring Strategies for Future Research. Nass S, Stillman B, eds. Washington, DC: The National Academies Press.

——. 2006a. Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics. A Workshop. Patlak M, Nass S, rapporteurs. Washington, DC: The National Academy Press.

——. 2006b. Effects on Health Research of the HIPAA Privacy Rule: A Workshop . National Cancer Policy Forum Herdman R, Moses H, rapporteurs. Washington, DC: The National Academies Press.

Irwin D, Klenlow P. 1996, November. SEMATECH: Purpose and Performance. Presentation at the meeting of the Proceedings of the National Academy of Sciences, Washington, DC. National Academy Press.

Jacobs JM, Adkins JN, Qian WJ, Liu T, Shen Y, Camp DG 2nd, Smith RD. 2005. Utilizing human blood plasma for proteomic biomarker discovery. Journal of Proteome Research 4(4):1073-1085.

Jin Y, Desta Z, Stearns V, Ward B, Ho H, Lee KH, Skaar T, Storniolo AM, Li L, Araba A, Blanchard R, Nguyen A, Ullmer L, Hayden J, Lemler S, Weinshilboum RM, Rae JM, Hayes DF, Flockhart DA. 2005. CYP2D6 genotype, antidepressant use, and tamoxifen metabolism during adjuvant breast cancer treatment. Journal of the National Cancer Institute 97(1):30-39.

Johnson MD, Zuo H, Lee KH, Trebley JP, Rae JM, Weatherman RV, Desta Z, Flockhart DA, Skaar TC. 2004. Pharmacological characterization of 4-hydroxy-N-desmethyl tamoxifen, a novel active metabolite of tamoxifen. Breast Cancer Research and Treatment 85(2):151-159.

Kaiser J. 2006a. A cure for medicine’s ailments? Science 311:1852-1854.

Kaiser J. 2006b. NIH funds a dozen “homes” for translational research. Science 314(5797):237.

Kelloff GJ, Lippman SM, Dannenberg AJ, Sigman CC, Pearce HL, Reid BJ, Szabo E, Jordan VC, Spitz MR, Mills GB, Papadimitrakopoulou VA, Lotan R, Aggarwal BB, Bresalier RS, Kim J, Arun B, Lu KH, Thomas ME, Rhodes HE, Brewer MA, Follen M, Shin DM, Parnes HL, Siegfried JM, Evans AA, Blot WJ, Chow WH, Blount PL, Maley CC, Wang KK, Lam S, Lee JJ, Dubinett SM, Engstrom PF, Meyskens FL Jr, O’Shaughnessy J, Hawk ET, Levin B, Nelson WG, Hong WK. 2006. Progress in chemoprevention drug development: The promise of molecular biomarkers for prevention of intraepithelial neoplasia and cancer—A plan to move forward. Clinical Cancer Research 12(12):3661-3697.

Kettler H, White K, Jordan S. 2003. Valuing Industry Contributions to Public–Private Partnerships for Health Product Development. Geneva, Switzerland: The Initiative on Public–Private Partnerships for Health, Global Forum for Health Research.

Kiviat NB, Critchlow CW. 2002. Novel approaches to identification of biomarkers for detection of early stage cancer. Disease Markers 18(2):73-81.

Knox SK, Ingle JN, Suman VJ, Rae JM, Flockhart DA, Zeruesenay D, Ames MM, Visscher DW, Perez EA, Goetz MP. 2006. Cytochrom p450 2D6 status predicts breast cancer relapse in women receiving adjuvant tamoxifen. Journal of Clinical Oncology 24(18S)(June 20 Supplement):504.

Kolch W, Mischak H, Pitt AR. 2005. The molecular make-up of a tumour: Proteomics in cancer research. Clinical Science (London, England) 108(5):369-383.

LaBaer J. 2005. So, you want to look for biomarkers (Introduction to the special biomarkers issue). Journal of Proteome Research 4(4):1053-1059.

Lee KH, Ward BA, Desta Z, Flockhart DA, Jones DR. 2003. Quantification of tamoxifen and three metabolites in plasma by high-performance liquid chromatography with fluorescence detection: application to a clinical trial. Journal of Chromatography B, Analytical technologies in the biomedical and life sciences 791(1-2):245-253.

Lim YC, Desta Z, Flockhart DA, Skaar TC. 2005. Endoxifen (4-hydroxy-N-desmethyl-tamoxifen) has anti-estrogenic effects in breast cancer cells with potency similar to 4-hydroxy-tamoxifen. Cancer Chemotherapy and Pharmacology 55(5):471-478.

Maitland ML, Vasisht K, Ratain MJ. 2006. TPMT, UGT1A1 and DPYD: Genotyping to ensure safer cancer therapy? Trends in Pharmacological Sciences 27(8):432-437.

McGowan D. 2006. SEMATECH Media Relations. Personal communication. May 31, 2006.

Mildran D. 2006. Former Co-Chair, HIV Surrogate Markers Collaborative Group. Personal communication. June 2006.

Mittleman B. 2006, June. Biomarkers and Partnerships. Presentation at the meeting of the National Cancer Policy Forum, Washington, DC.

MMRC (Multiple Myeloma Research Consortium). 2006. Welcome to the MMRC. [Online]. Available: http://www.themmrc.org [accessed June 2006].

MMRF (Multiple Myeloma Research Foundation). 2006. Multiple Myeloma Research Foundation. [Online]. Available: http://www.multiplemyeloma.org [accessed June 2006].

NCI (National Cancer Institute). 1999. Unconventional Innovations Program (UIP). [Online]. Available: http://otir.cancer.gov/programs/ati_uip.asp [accessed November, 2006].

——. September 2004. National Cancer Institute Announces Major Commitment to Nanotechnology for Cancer Research. [Online]. http://www.nci.nih.gov/newscenter/pressreleases/nanotechPressRelease/print?page=&keyword= [accessed November 2006].

——. March 2005. The Early Detection Research Network. DHHS. [Online] [accessed July 2006].

——. 2006a. OBBR First-Generation Guidelines for NCI-Supported Biorepositories. [Online]. Available: http://biospecimens.cancer.gov/biorepositories/guidelines_full_formatted.asp [accessed July 11, 2006].

——. 2006b. OBBR (Office of Biorepositories and Biospecimen Research) Mission and Goals. [Online]. Available: http://biospecimens.cancer.gov/biorepositories/obbr_mission.asp [accessed July 11, 2006].

——. 2006c. Prostate SPORE National Biospecimen Network Pilot. [Online]. Available: http://prostatenbnpilot.nci.nih.gov/ [accessed July 11, 2006].

——. 2006d. Translational Research Working Group. [Online]. Available: http://www.cancer.gov/trwg/ [accessed June 21,2006].

NCI Cancer Diagnosis Program. 2006. NCI-Supported Specimen Resources. [Online]. Available: http://www.cancerdiagnosis.nci.nih.gov/specimens/finding.html#resources [accessed July 2006].

Nguyen H, Tran A, Lipkin S, Fruehauf JP. 2006. Pharmacogenomics of colorectal cancer prevention and treatment. Cancer Investigation 24(6):630-639.

NIGMS (National Institute of General Medical Sciences). 2006. Protein Structure Initiative. [Online]. Available: http://www.nigms.nih.gov/Initiatives/PSI [accessed July 2006].

NIH (National Institutes of Health). 2006a. NIH Public-Private Partnerships. [Online]. Available: http://ppp.od.nih.gov/ [accessed June 2006].

——. 2006b. NIH Roadmap for Medical Research: Re-engineering the Clinical Research Enterprise. [Online]. Available: http://nihroadmap.nih.gov/clinicalresearch/overview-translational.asp [accessed July 2, 2006].

Nishtar S. 2004. Public–private “partnerships” in health—A global call to action. Health Research Policy and Systems 2(1):5.

Norton L. 1997. Evolving concepts in the systemic drug therapy of breast cancer. Seminars in Oncology 24(4 Suppl 10):S10-3-S10-10.

Norvell J, Berg J. 2005. The protein structure initiative, five years later. The Scientist 19(20):30.

NSABP (National Surgical Adjuvant Breast and Bowel Project). 2006. Study of Tamoxifen and Raloxifene. [Online]. Available: http://www.nsabp.pitt.edu/star/index.asp [accessed July 6, 2006].

Omenn GW, States DJ, Adamski M, Blackwell TW, Menon R, Hermjakob H, Apweiler R, Haab BB, Simpson RJ, Eddes JS, Kapp EA, Moritz RL, Chan DW, Rai AJ, Admon A, Aebersold R, Eng J, Hancock WS, Hefta S, Meyer H, Paik Y, Yoo J, Ping P, Pounds J, Adkins J, Qian X, Wang R, Wasinger V, Wu CY, Zhao X, Zeng R, Archakov A, Tsugita A, Beer I, Pandey A, Pisano M, Andrews P, Tammen H, Speicher DW, Hanash SM. 2005. Overview of the HUPO Plasma Proteome Project: results from the pilot phase with 35 collaborating laboratories and multiple analytical groups, generating a core dataset of 3020 proteins and a publicly-available database. Proteomics 5:3226-3245.

Ong S, Mann M. 2005. Mass spectrometry-based proteomics turns quantitative. Nature Chemical Biology 1:252-262.

Pepe MS, Etzioni R, Feng Z, Potter JD, Thompson ML, Thornquist M, Winget M, Yasui Y. 2001. Phases of biomarker development for early detection of cancer. Journal of the National Cancer Institute 93(14):1054-1061.

President’s Cancer Panel. June 2005. Translating Research into Cancer Care: Delivering on the Promise. [Online]. Available: http://deainfo.nci.nih.gov/ADVISORY/pcp/pcp04-05rpt/ReportTrans.pdf [accessed July 10, 2006].

Quackenbush J. 2006. Microarray analysis and tumor classification. New England Journal of Medicine 354(23):2463-2472.

Ransohoff DF. 2004. Rules of evidence for cancer molecular-marker discovery and validation. Nature Reviews Cancer 4(4):309-314.

——. 2005. Bias as a threat to the validity of cancer molecular-marker research. Nature Reviews Cancer 5(2):142-149.

Ross JS, Gray K, Mosher R, Stec J. 2004. Molecular techniques in cancer diagnosis and management. In: Nakamura RM, Grody WW, Wu JT, Nagle RB, Cancer Diagnostics: Current and Future Trends. Totowa, NJ: Humana Press, Inc. Pp. 325-360.

Schwartz K, Vilquin JT. 2003. Building the translational highway: Toward new partnerships between academia and the private sector. Nature Medicine 9(5):493-495.

SEMATECH. 2006. SEMATECH webpage. [Online] Available: http://www.sematech.org/ [accessed June 2006].

Simon R. 2005. Roadmap for developing and validating therapeutically relevant genomic classifiers. Journal of Clinical Oncology 23(29):7332-7341.

——. 2006. Guidelines for the design of clinical studies for the development and validation of therapeutically relevant biomarkers and biomarker-based classification systems. In: Gasparini G, Hayes D, Biomarkers in Breast Cancer: Molecular Diagnostics for Predicting and Monitoring Therapeutic Effect. Totowa, NJ: Humana Press, Inc. Pp. 3-15.

Skaar, T. 2006. Associate Director, General Clinical Research Center Pharmacogenetics Core Laboratory, Indiana University. Personal communication.

Spear BB, Heath-Chiozzi M, Huff J. 2001. Clinical application of pharmacogenetics. Trends in Molecular Medicine 7(5):201-204.

Srivastava S, Wagner J. 2002. Surrogate endpoints in medicine. Disease Markers 18:39-40.

Stearns V, Johnson MD, Rae JM, Morocho A, Novielli A, Bhargava P, Hayes DF, Desta Z, Flockhart DA. 2003. Active tamoxifen metabolite plasma concentrations after coadministration of tamoxifen and the selective serotonin reuptake inhibitor paroxetine. Journal of the National Cancer Institute 95(23):1758-1764.

Stoughton RB, Friend SH. 2005. How molecular profiling could revolutionize drug discovery. Nature Reviews Drug Discovery 4(4):345-350.

Tinker AV, Boussioutas A, Bowtell DD. 2006. The challenges of gene expression microarrays for the study of human cancer. Cancer Cell 9(5):333-339.

TSC (The SNP Consortium). 2006. The SNP Consortium Ltd. [Online]. Available: http://snp.cshl.org [accessed June 2006].

Tyers M, Mann M. 2003. From genomics to proteomics. Nature 422(6928):193-197.

van der Greef J, McBurney RN. 2005. Innovation: Rescuing drug discovery: in vivo systems pathology and systems pharmacology. Nature Reviews Drug Discovery 4(12):961-967.

Varmus H. 2006. The new era in cancer research. Science 312(5777):1162-1165.

Wagner JA. 2002. Overview of biomarkers and surrogate endpoints in drug development. Disease Markers 18(2):41-46.

Wagner, JA. 2006, March 20. Biomarker Validation and Qualification: Fitness for Use. Presentation at the IOM workshop on Developing Biomarker-based Tools for Cancer Screening, Diagnosis, and Treatment. Washington, DC.

Weckwerth W, Morgenthal K. 2005. Metabolomics: from pattern recognition to biological interpretation. Drug Discovery Today 10(22):1551-1558.

Yezhelyev MV, Gao X, Xing Y, Al-Hajj A, Nie S, O’Regan RM. 2006. Emerging use of nanoparticles in diagnosis and treatment of breast cancer. Lancet Oncology 7(8):657-667.

Young S. 2006. MMRC Executive Director. Personal Communication. June 7, 2006.

Zolg JW, Langen H. 2004. How industry is approaching the search for new diagnostic markers and biomarkers. Molecular and Cellular Proteomics 3(4):345-354.