Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 1
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment INTRODUCTION Research has long sought to identify biomarkers that could detect cancer at an early stage, or predict the optimal cancer therapy for specific patients. Fueling interest in this research are recent technological advances in genomics, proteomics, and metabolomics that can enable researchers to capture the molecular fingerprints of specific cancers and fine-tune their classification according to the molecular defects they harbor. The discovery and development of new markers of cancer could potentially improve cancer screening, diagnosis, and treatment. Given the potential impact cancer biomarkers could have on the cost effectiveness of cancer detection and treatment, they could profoundly alter the economic burden of cancer as well. Despite the promise of cancer biomarkers, few biomarker-based cancer tests have entered the market, and the translation of research findings on cancer biomarkers into clinically useful tests seems to be lagging. This is perhaps not surprising given the technical, financial, regulatory, and social challenges linked to the discovery, development, validation, and incorporation of biomarker tests into clinical practice. To explore those challenges and ways to overcome them, the National Cancer Policy Forum held the conference “Developing Biomarker-Based Tools for Cancer Screening, Diagnosis and Treatment: The State of the Science, Evaluation, Implementation, and Economics” in Washington, D.C., from March 20 to 22, 2006.
OCR for page 2
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics At this conference, experts gave presentations in one of six sessions: Brief overview of technologies, including genomics, proteomics, metabolomics, and functional imaging Overcoming the technical obstacles, with presentations on informatics and data standards, and biomarker validation and qualification Coordinating the development of biomarkers and targeted therapies, with a clinical investigator and representatives from industry and the National Cancer Institute offering their perspectives Biomarker development and regulatory oversight, including current regulations governing biomarker tests as well as new clinical trial designs needed to incorporate biomarker tests that predict patient responders Adoption of biomarker-based technologies, with discussion on what motivates private insurers and Medicare to cover biomarker-based tests and what various organizations consider when recommending such tests be adopted into clinical practice Economic impact of biomarker technologies, with an exploration of cost-effectiveness analyses of biomarker tests and a payor perspective on the evaluation of such tests In addition, seven small group discussions explored the policy implications surrounding biomarker development and adoption into clinical practice: Clinical development strategies for biomarker utilization Strategies for implementing standardized biorepositories Strategies for determining analytic validity and clinical utility of biomarkers Strategies to develop biomarkers for early detection Mechanisms for developing an evidence base Evaluation of evidence in decision making Incorporating biomarker evidence into clinical practice This document is a summary of the conference proceedings, which will be used by an Institute of Medicine (IOM) committee to develop consensus-based recommendations for moving the field of cancer biomarkers forward. The views expressed in this summary are those of the speakers and discussants, as attributed to them, and are not the consensus
OCR for page 3
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics views of the participants of the workshop or of the members of National Cancer Policy Forum. OVERVIEW OF TECHNOLOGIES USED TO DISCOVER CANCER BIOMARKERS Technology is constantly evolving and recent technological advances have made it easier to discover many potential cancer biomarkers through high-throughput screens. Advances in imaging technology also are furthering the discovery and use of biomarkers. The goal of the first session of the conference was to provide a brief overview of the technologies currently being used to identify and develop cancer biomarkers (Figure 1). Genomics, Proteomics, and Metabolomics Todd Golub, MD, of the Dana-Farber Cancer Institute, began by discussing several of the genomics-based techniques commonly used to discover biomarkers for cancer detection or for patient stratification for therapy. Some of these techniques detect changes at the DNA level (are DNA-based), whereas others detect changes at the RNA level and are considered RNA-based. Dr. Golub explored which type of genomics test—DNA based or RNA based—would be likely to serve as a better biomarker if cost were not an issue. DNA-based tests are advantageous because DNA is more stable than RNA, and because most changes related to cancer occur at the DNA level, he said. But he noted that perhaps one could make a stronger argument for RNA-based tests because not only can they detect oncogenic RNA missteps, but molecular signatures at the RNA level also help reveal upstream DNA-level abnormalities that could contribute to a cancer. These abnormalities include base substitutions, and amplifications or deletions that alter the copy number or heterozygosity of specific genetic sequences. Dr. Golub noted that studying epigenetic changes in DNA, such as methylation, and genome rearrangements, such as chromosome translocations, can also lead to discovery of important cancer biomarkers, although he did not have time to address these topics in his presentation. Although early genetic analyses of cancers focused on detecting changes in the copy number of genes, Dr. Golub stressed that it is also important to screen for loss of heterozygosity (LOH). LOH can occur without a change
OCR for page 4
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics FIGURE 1 The spectrum of potential biomarkers suggests that no single technology can cover the enteir biomarker space. TG = triglycerides; Aβ = β-amyloid; HbA1c; PSA = prostate-specific antigen; CRP = C-reactive protien; CEG = comparative genomic hybridization; Snp = single nucleotide polymorphism; eCRF = electronic case report from; eMR = electronic medical record. SOURCE: Adapted from Schulman presentation (March 20, 2006).
OCR for page 5
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics in gene copy number, he noted, if both alleles for a specific gene have been mutated or epigenetically altered. This copy-neutral LOH may account for as much as half of all LOH in the genome. Two main types of arrays are used to detect changes in copy number or LOH linked to cancer. Single nucleotide polymorphism (SNP) arrays have between 50k and 500k SNPs across the genome and can detect both copy number changes and other forms of LOH. Comparative genomic hybridization arrays can detect changes in copy number of DNA content, but are unable to detect LOH in which the copy number remains the same. For this reason, Dr. Golub prefers SNP arrays for detecting cancer biomarkers. Higher density SNP arrays can give sharper resolution by reducing the signal-to-noise ratio than lower density SNP arrays, he pointed out. But the optimal amount of density that is the most cost-efficient means for detecting cancer biomarkers remains to be determined. Standard DNA sequence analysis of tumor samples as a means of detecting cancer biomarkers has numerous drawbacks, which Dr. Golub pointed out. Not only is it difficult and costly to do, but it is frequently inaccurate, causing false negatives because of normal tissue contamination of the tumor samples used. Most tumor samples contain a mixture of normal cells, such as inflammatory cells, as well as tumor cells. Because the Sanger sequencing results are an average of both the normal and tumor cells in a sample, normal genome contamination can obscure mutations in tumor cells that might serve as cancer biomarkers. However, newer techniques, such as single-molecule sequencing, may substantially lower the cost of sequencing, and should avoid the problems of normal cell contamination that plague standard sequencing efforts. “I think this is exactly the type of technology, even if cost neutral, that would dramatically accelerate our ability to detect important mutations in cancer,” Dr. Golub said. To exemplify this, Dr. Golub reported on results from his colleagues at Dana-Farber who used single-molecule sequencing to detect a mutation that was linked to resistance of the drug Iressa in a lung cancer patient. The lung fluid sample the researchers analyzed only had 3 percent tumor cells, and a standard Sanger sequencing analysis missed the mutation. Once a genetic signature with likely clinical relevance has been discovered, custom-made arrays that only have the gene sequences of interest need to be made for preclinical or clinical testing. Dr. Golub described a few genetic signature amplification and detection platforms useful for such testing, including a Luminex bead-based method. For this method, the genetic
OCR for page 6
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics material is amplified using polymerase chain reaction. The genetic signature is then read not on microarrays, but on miniscule color-coded beads that are detected by lasers in a flow cytometer. This is an inexpensive way to detect genetic signatures, costing about 50 cents for every 100 transcript signatures. One can also use the standard mRNA expression profiling platforms that are commercially available. These are all sufficiently accurate and precise to be used in a clinical setting to detect genetic signatures, according to Dr. Golub. Cost and throughput will be significant drivers of this technology, he added. The next presentation was on proteomics and metabolomics technologies, given by Howard Schulman, PhD, of PPD Biomarker Discovery Sciences. One of those techniques, which Dr. Schulman described as the traditional proteomic workhorse, uses two-dimensional polyacrylamide gels for the separation stage. This is a slow process that is less amenable to high-throughput. Surface-enhanced laser desorption/ionization is a high-throughput technology that can more quickly separate the proteins in a sample, but identifying the protein peaks is a challenge. That identification process can be bypassed by using software to differentially identify patterns of protein peaks to find a molecular fingerprint that can distinguish cancerous from noncancerous tissue. This fingerprint is based on the amounts of all the various proteins detected, without knowledge of what those individual proteins are, Dr. Schulman noted. However, it can be problematic to translate mass spectroscopy fingerprints into a clinical diagnostic test without identifying or further characterizing those proteins. One- and multidimensional liquid chromatography are also used to separate peptides in a sample (after protein digestion) that a mass spectrometer can differentially quantify and then identify (Figure 2). But the amplitude for each of the peptides can vary depending on the composition of the mixture, which makes it hard to compare one person’s sample with another’s, and one batch run versus another. This has proven problematic for researchers trying to develop cancer biomarkers based on differential quantification, otherwise known as molecular fingerprinting. To improve such differential accuracy, researchers developed a method called isotope-coded affinity tags several years ago. This technique labels a portion of a sample with a mass tag and runs both labeled and unlabeled samples through the mass spectrometer at the same time. The labeled sample serves as a sort of baseline control for the unlabeled sample. This helps normalize or eliminate a lot of the peak amplitude variability due to differences in mixture composition. But this is a more costly method because
OCR for page 7
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics FIGURE 2 One-dimensional and multidimensional liquid chromatography LC-LC/MS. LC = liquid chromatography; MS = mass spectroscopy; MW = molecular weight; HPLC = high-performance liquid chromatography; ESI = electrospray ionization. SOURCE: Schulman presentation (March 20, 2006). of the need for the reagents, and it has some bias introduced by the type of tag used, according to Dr. Schulman. The field is rapidly adopting a label-free approach in which chromatographic separation techniques and mass spectrometry are coupled with software-based solutions for normalizing the variation in amplitude signal due to differences in mixture composition to yield accurate differential expression data. Dr. Schulman concluded his talk by noting that the current state of proteomics is comparable to the early days of microarrays, which could detect about one-sixth the number of genetic sequences that can now be detected. But proteomics is still highly effective even without the ability to profile every protein, he said. He noted that one can profile more than a thousand proteins by using multidimensional chromatography. But the tradeoffs with more fractionation are lower throughput (due to slower processing) and higher costs. The advantage of proteomic and metabolomic profiling is that you can sample readily accessible tissues, such as plasma and urine, that are ideal for monitoring biomarkers in clinical trials and testing diagnostics.
OCR for page 8
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics He also noted that the lowest abundance proteins, such as cytokines or other signaling molecules, will likely require antibody-based protein chips to complement liquid chromatography separation techniques. Sensitivity to such proteins could also be increased by using samples likely to have higher concentrations of biomarkers of interest. For example, cerebral spinal fluid could be tapped to find biomarkers for lymphoma metastases in the central nervous system, or prostatic fluid could be used to detect biomarkers for prostate cancer. Affinity capture of protein subcategories, such as phosphorylated proteins, could also selectively profile lower abundance proteins of interest. Drs. Schulman and Golub stressed the need to experimentally validate the biological basis and importance of detected genetic or proteomic differences in a disease process. For example, researchers in Dr. Golub’s laboratory used high-density expression arrays to detect an RNA signature in bone marrow samples that correlated with response to a drug for myelodysplastic syndrome. They found a group of genes that were only highly expressed in patients who responded to the drug. Many of these genes previously had been identified as markers for late red blood cell differentiation, leading to the hypothesis that such differentiation may be predictive of drug response. To test this idea, they induced normal immature blood cells to differentiate into red blood cells. They found that all of the genes, whose boosted expression was linked to drug response in their biomarker discovery study, also had heightened expression during the red blood cell differentiation that occurred in their experiments. This validated their hypothesis and put the concept of genetic signature for drug response on firmer footing. “The most valuable and robust biomarkers will be those that have some component of experimental validation accompanying them,” Dr. Golub said. He added that “the challenge looking forward is going to be to move from simply cataloging mutations or genome abnormalities to coalescing those abnormalities into more of a molecular taxonomy that brings biological understanding to this catalog. The more we can integrate these anonymous molecular signatures with biological knowledge, the more they’re likely to stick.” Dr. Golub also pointed out the need to develop biomarker diagnostics that can easily be used on the paraffin-embedded or formalin-fixed tissues that are routinely collected in the clinic. “We need to make the technology work for those routinely collected samples rather than retrain the medical community to collect samples in a different way,” Dr. Golub said.
OCR for page 9
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics Drs. Golub and Schulman noted that a lack of good-quality samples can be a stumbling block for biomarker discovery. Rarely are enough samples collected in a clinical trial, and those samples that are collected are usually fixed in formalin, which can affect their ability to be analyzed in a mass spectrometer. Dr. Schulman suggested that pharmaceutical and biotechnology companies have experimental medicine groups that are best positioned to collect the samples required to discover biomarkers. But the biggest impediment for biomarker development, which Drs. Golub and Schulman both cited, was a lack of a critical mass of research in the discovery phase. “The bottleneck is not so much on the regulatory side or the validation side, but that not enough of the discovery effort has been made,” said Dr. Schulman. As to whether such efforts at biomarker discovery should take a hypothesis-driven or open-ended approach, Drs. Golub and Schulman agreed that both approaches were necessary. Open-ended discovery aims at uncovering a molecular understanding of a particular type of cancer that may eventually lead to useful biomarkers. A hypothesis-driven approach, in contrast, is more streamlined at finding molecular changes likely to predict a response to therapy or some other useful clinical endpoint. There is a role for both these approaches, Dr. Schulman said. But he added that pharmaceutical companies are unfortunantely more likely to conduct a hypothesis-driven search for biomarkers that predict drug response than to support a more open-ended search. Dr. Golub noted that the danger of conducting only hypothesis-driven research on biomarkers is that it does not address the challenge of “how do we get beyond discovering what we already know, in terms of biological knowledge?” Molecular Imaging Next, Michael Phelps, PhD, of the University of California, Los Angeles, discussed molecular imaging biomarkers for drug discovery, development, and patient care. He described how positron emission tomography (PET) can be used as a molecular camera to image in vivo processes at the molecular level. But PET is more than an imaging device, as it also can be used analytically to perform a variety of quantitative biochemical and biological assays. There are currently about 600 PET probes for metabolism, receptors, enzymes, DNA replication, gene expression, antibodies, hormones, drugs, and other compounds in nanomole amounts. Typical antibody probes get
OCR for page 10
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics broken down too quickly to be effective for PET imaging, but there are modified antibody probes that are small molecule versions of the original antibodies and retain the active end. Most PET probes were developed from probes used in in vitro assays so as to translate that assay into an in vivo measurement. Ninety percent of PET probes were developed from drugs, Dr. Phelps reported. Over the past few years, PET scanners have merged with computed tomography (CT) scanners to combine the anatomical definition of the CT with the biological assay capability of the PET scan. Researchers have also created microPET/CT machines to image biological processes and drug responses in mice. Because PET probes are administered in nanomole amounts, measures can be performed on biological processes without disturbing the processes or causing pharmacologic mass effects, Dr. Phelps noted. Not only can PET scans be safely done, but studies show they are more accurate than magnetic resonance imaging or CT scans for the diagnosis and staging of cancer, for assessing therapeutic response, and for detecting cancer recurrence. To detect cancer, technicians usually use a PET probe that images the heightened glucose metabolism of cancer cells. To predict or determine response to therapy, a number of different types of probes are used, depending on the type of cancer and type of treatment. The PET assay can enable stratification of patients according to whether they have the therapeutic target. For example, a probe that detects DNA replication may be used to predict whether a cancer will respond to a chemotherapy that blocks such replication. A probe for an estrogen receptor may be used to determine if breast cancer metastases are likely to respond to hormonal therapy. PET is especially useful for revealing whether a tumor is responding to therapy. It can detect within a day, for example, whether patients’ tumors are not responding to Gleevec, thereby quickly determining if patients should receive a different drug, Dr. Phelps pointed out. PET imaging also has advantages over standard techniques for assessing the pharmacokinetics and pharmacodynamics of drugs, he added. For example, standard pharmacokinetic assessments are based on measurements of how quickly a drug is cleared from the blood. In contrast, by using labeled drugs as probes, PET can precisely measure the concentration of the drug, not just in the blood, but in all tissues over time, he noted. Dr. Phelps described a recent innovation in PET technology that uses “click chemistry” to create PET probes. This technique involves combining two small molecules with low to moderate affinity to the target, but high
OCR for page 11
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics affinity to each other. They collectively latch onto the target as they bind to each other. The end result is that they bind to the target with an extremely high affinity that is the product of the affinities of the two molecules. These probes dramatically increase the sensitivity and physical resolution of PET imaging. Because the probes are such small molecules, they can access surface receptors, cells, and even the cell nucleus. Dr. Phelps concluded his talk by noting there are “PET pharmacies” scattered all over the world that use automated chemistry to make and ship labeled molecular PET probes. There are also “labs on a chip” that enable researchers to custom build their own PET probes using click chemistry and other techniques. In response to comments by Drs. Golub and Schulman regarding where the bottleneck is in biomarker development, Dr. Phelps noted that as one gets closer to introducing a biomarker into a clinical setting, Food and Drug Admininstration (FDA) premarket regulation can become very limiting. He suggested that regulatory bodies work with researchers to change the criteria by which drugs and molecular diagnostics are evaluated. MEETING THE TECHNICAL CHALLENGES OF BIOMARKER VALIDATION AND QUALIFICATION Appropriate analysis and interpretation of biomarker data presents enormous challenges, especially with the advent of genomic and proteomic technologies that can generate a tremendous amount of data on individual samples. Three speakers at the conference addressed the technical challenges involved with validating the accuracy and clinical relevance of cancer biomarkers. John Quackenbush, PhD, of Harvard University spoke about experimental design considerations and data reporting standards to aid the validation of biomarkers. David Ransohoff, MD, of the University of North Carolina also discussed experimental design, and the shortcomings of recent cancer biomarker studies that should be avoided in future studies. John Wagner, MD, PhD, of Merck and Co., Inc., gave a pharmaceutical company’s perspective on what is required to validate a cancer biomarker and establish its relevance to useful clinical endpoints. Dr. Quackenbush began this session by noting that with microarray technologies, researchers tend to do more hypothesis-generating experiments than hypothesis-driven experiments. But despite a lack of an experimental hypothesis, one still needs to think critically about experimental design and how data are collected, managed, and analyzed, he said. All of these steps
OCR for page 74
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics reveal whether tumors are progressing and how a treatment is affecting the targeted tumor. The development of surrogate endpoint markers, adverse reaction biomarkers, and pathway biomarkers would be precompetitive activities that should not require exclusivity. Therefore, all interested parties would benefit by pooling their activities and sharing the development costs, Dr. Sawyer noted. This is in contrast to diagnostics that will be used only when paired to the use of specific drugs, such as the HercepTest, which is used to predict response to Herceptin. The group suggested that both the diagnostic company and the drug company for a paired diagnostic and treatment share the development costs for these types of biomarkers. After group participants suggested that public-private partnerships could be established to facilitate development of candidate biomarkers, they explored further which groups should be involved in these partnerships. As previous discussions have noted, academia does some discovery work on biomarkers. But academia typically is not involved in the development of robust diagnostic assays because of a lack of expertise in the industrialization aspects and because of a lack of academic rewards and funding sources for this type of research. Start-up diagnostic companies also are not likely to develop biomarker assays because of the low profit margins of diagnostic tests, which make them unattractive to investors. “There was some discussion that if we wait and hope that this happens through free enterprise, we could be waiting awhile,” Dr. Sawyers noted. Consequently, group participants suggested a national effort to drive biomarker development, with NCI as the most likely agency to further this effort and support academic researchers doing this type of work. A public-private partnership that furthers biomarker development could be modeled after the SNP Consortium. This nonprofit foundation was organized to provide public genomic data, and it was supported by pharmaceutical and technical companies and the Wellcome Trust medical research charity. One discussant indicated that a main impetus for forming the foundation was to prevent academic institutions and industry from claiming intellectual property rights on each SNP they discovered in the human genome. Avoiding intellectual property claims could be an impetus for starting a biomarker consortium as well. The group noted that such claims on each possible biomarker could be a huge impediment to having diagnostic companies develop assays for the biomarkers. Several people in the group felt strongly that biomarker information should be in the public domain, with some stating “the real value of the intellectual property comes
OCR for page 75
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics from developing the assays and not just linking an mRNA to a possible outcome,” Dr. Sawyers reported. This raised the problem of how to give diagnostic companies exclusive rights so that they are encouraged to fully develop and commercialize a biomarker. The group came up with several incentives for biomarker development. Defining the FDA approval path for a biomarker diagnostic more clearly, and linking the approval path for paired diagnostics and therapeutics so both companies share the risks and development costs would provide incentives for biomarker development. It was also suggested that there be patent extensions of innovative biomarker diagnostics to reward the ground-breaking work that one or two companies do that is then used by competing companies to develop similar products. Precedents exist for this enhanced exclusivity in the development of pediatric interventions, and have been proposed for the development of anti-infectives, Dr. Frank noted. Finally, group participants suggested working with payors to define the cost effectiveness of biomarker tests. “There was a sense that the cost effectiveness of a biomarker was not really appreciated,” he said. “If it were, then reimbursement paradigms could be built in that would incentivize companies to make them sooner.” Group discussants also suggested working with payors to establish alternatives to basing reimbursement decisions on evidence generated from large, long-term clinical trials. CMS and other insurers often require more evidence than does the FDA for a biomarker’s effectiveness prior to reimbursing its clinical use, Dr. Sawyers noted. Several group members suggested that evidence could be generated after the test enters the clinic via community-based postmarketing studies. These studies could be facilitated by using an electronic medical records infrastructure. Dr. Sawyers concluded his summary by discussing his group’s suggestion that there be a demonstration project to develop biomarkers for drugs already on the market. This project could show the value of using biomarkers to identify the group of patients most likely to respond to the drug, or to identify and exclude those likely to have severe adverse reactions to the drug. Such a proof-of-concept experiment could lay out a path for developing biomarkers and could provide lessons about the appropriate business model to follow and regulatory issues to consider. The reason to use approved drugs for the demonstration project is because patients already taking the drugs can be easily accrued into a study, Dr. Sawyers said. One discussant suggested demonstrating the usefulness of biomarkers that indicate the safety of a number of drugs in a class. Another discussant suggested
OCR for page 76
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics using the demonstration project to show the value of biomarkers in predicting responsiveness for two or three drugs widely used in oncology. If a demonstration project had high-impact findings, it could serve as a catalyst that would spur investment into diagnostic companies and lead more academic institutions and industry to pursue biomarker discovery and development, the group pointed out. Several discussants thought some “success stories” via such a demonstration project would overcome the inertia that is preventing extensive biomarker development. The science needed to do such work is already in place, they noted, and what is lacking is leadership and funding. As an example of a biomarker demonstration project, Dr. Sawyers mentioned the pilot project already under way that was previously described by Dr. Woodcock in her presentation. This is a nonprofit public-private partnership to qualify FDG-PET as a marker for drug response in non-Hodgkin’s lymphoma. Dr. Sawyers’ group also reiterated the need for annotated, quality-assured patient samples that are readily available to further efforts to develop biomarkers. EVALUATION OF EVIDENCE IN DECISION MAKING DISCUSSION Dr. Ramsey was the moderator who provided the summary of the discussion on evaluation of evidence in decision making. This discussion group noted that many biomarker-based tests in wide use today were never thoroughly evaluated for analytic validity, clinical validity, or clinical utility in relation to standards. Consequently, their value is often unknown. Group members suggested that this lack of standardized evaluation be eliminated for new tests because the developmental and clinical costs of these tests are quite expensive, and costs also can be incurred if tests are used inappropriately and/or cause undue harm to patients. Some group participants agreed there is a need for more uniform standards for biomarker evaluation. Dr. Ramsey said there is no consistency regarding standards among organizations and regulatory programs such as the FDA, CLIA, the College of American Pathologists (CAP), and the American Society for Clinical Oncology (ASCO). Each organization has its own set of standards for biomarker tests that are based on different criteria. There is even variability within these organizations, the group noted. In a discussion following Dr. Ramsey’s summary, Dr. Dai pointed out that scientific journals also have their own set of standards for biomarkers. For example, if researchers want to publish gene expression biomarkers, journals
OCR for page 77
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics may ask them to compare the biomarkers to what is already available. They may even require that researchers use a specific statistical modeling technique when making such comparisons. Group members thought the ASCO guidelines for tumor biomarkers for breast or colorectal cancer21 could serve as a potentially useful model in terms of how one might set standards for evaluating whether biomarkers are ready for clinical use. These guidelines established the appropriate levels of evidence needed for different types of clinical decisions made based on biomarker test results. For example, the highest level of evidence was required for a biomarker assay that would indicate the need to deny specific care, that is, one that indicates drug resistance. However, there was no group consensus on what standards should be required or recommended for cancer biomarkers. This lack of consensus stemmed, in part, from recognizing that there is no gold standard for many of the new kinds of assays used to detect cancer biomarkers, and the evolving nature of those technologies. This made many in the group reluctant to specify standards. In addition, the group recognized that broad, generalized standards alone are not sufficient; guidelines may also need to be use specific and even target specific. Because the technologies for genomics and proteomics assays are rapidly evolving, the group noted, standards have to be adaptable to the changes in technology that are continually occurring. There is also such a wide range of uses for biomarkers in the cancer arena that standards for one use, such as a surrogate endpoint in clinical trials, may not be applicable to another use, such as a predictor of patient responsiveness. In addition, the standards for a biomarker that predicts responsiveness to a drug may vary depending on the type of cancer on which it is tested, such as lung or breast. However, basic generalized criteria should be met for all clinical tests including biomarker-based tests, the group members recognized. They agreed with Dr. Ransohoff’s assertion in his presentation that the standards of clinical epidemiology still apply to biomarker-based tests. Working against a common desire to fully evaluate biomarker tests and ensure they meet certain standards is the desire of companies to bring such tests to market as quickly as possible to generate revenues to compensate for development costs. In addition, companies that are developing biomarker tests to be used in combination with specific drugs are often under time pressure to put the drug and the diagnostic on the market at the same time. 21 http://www.jco.org/cgi/content/full/19/6/1865.
OCR for page 78
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics Because diagnostic development often lags behind drug development for paired diagnostics and therapeutics, shortcuts may be taken in evaluations of the diagnostic, some discussants pointed out. Because of such financial and time pressures, companies usually seek the fastest and easiest entry into the market, such as CLIA certification for a home-brew laboratory test, rather than a more rigorous evaluation process by the FDA that might require them to conduct clinical studies. Consequently, few biomarker-based tests are designated Class III devices, which require clinical evidence of their effectiveness and safety. Competition with other companies also prods the makers of biomarker diagnostics to lower the standards bar in order for their products to go to market before those of their competitors. As Sharon Kim, MBA, of Precision Therapeutics observed, “The challenge has been not just to set your own quality standards for yourself, but you worry and wonder what your potential competitors might be held to because there is no standard, and so are you holding yourself to too stringent of a standard, knowing there may be someone else out there that may place a lower level-of-evidence bar or variability bar out there? While the FDA has the ability to come in and regulate, they have elected not to, and so it is more self-regulated. Even for CLIA-governed or CAP-governed labs, there is no specific cookbook or guidance you can go to.” Industry representatives in the discussion group pointed out that companies often evaluate their biomarker diagnostics in phases, with a more complete evaluation of their broader applications not occurring until after the tests enter the market. For example, a cancer detection test may at first only be evaluated for its accuracy and predictive value in high-risk populations because this evaluation can be done relatively quickly compared to one done in the general public. But to create a greater market for their products, companies may evaluate them for broader uses once they are already on the market for a more restrictive indication. In that way, companies can quickly bring their products to market and begin gaining revenue on them to help cover the costs of further evaluations. But once a test is on the market, there are few ways to control, beyond coverage decisions, how the test is clinically used. As was noted in Mr. Heller’s presentation at the conference, the high-variability and rapidly evolving approach to the FDA’s regulation of biomarker diagnostics has created uncertainty as to what evaluations industry needs to do of their tests and what standards to apply, Dr. Ramsey said. The group spent some time discussing whether health insurance payors should
OCR for page 79
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics set the standards for biomarker diagnostics. The group noted that if they did, it would add another layer of variability, uncertainty, and complexity that would be problematic for the developers of the tests, especially if there was no agreement among health insurers in this regard. The group also considered whether the FDA, CMS, and perhaps other stakeholders should work together to develop more uniform standards for the evaluation of biomarkers. But consensus was not reached on this issue, in part because of the tradeoffs involved. Having these agencies set uniform standards would be beneficial in that companies would know what to expect and what would be required of them regarding the evaluation and performance of their biomarker diagnostics. “As long as they are not overly burdensome, they would help us defend our experimental design if we could refer to something else that had been published and widely accepted. That way when the data were reviewed our study design wouldn’t be questioned, which could help speed things through [an FDA approval process],” said Lynne McBride of BD BioSciences. But Dr. Aronson, the session moderator, said, “there are decisions that come out of CMS and the FDA that are more political than rational and health plans do not follow them.” But she added that it would be valuable to gather together a community of stakeholders to help establish the evidence base needed for biomarkers used clinically. In a discussion following Dr. Ramsey’s summary, Dr. Waring stressed the need to engage the pathology community when setting standards for biomarker tests. “When we are talking about predictive tests that determine treatment decisions for patients with serious life-threatening diseases, I think that the pathologists and the pathology community are often the afterthought in this process. We need to engage them very early and make sure they understand the consequences of the decisions and that they maintain quality testing,” he said. He noted that CAP and ASCO would be meeting in a few weeks to try to develop common guidelines for HER2 testing. The group discussed further Dr. Waring’s presentation on the variability among laboratories on the accuracy of the IHC test for HER2. Part of that variability stemmed from the manual, visual, and subjective nature of the test, the group noted. But it is likely that such variability in accuracy will crop up again for other biomarker tests, Dr. Ramsey said. The group debated whether there should be additional measures of quality assurance in such tests. Suggested quality assurance measures included proficiency testing akin to what is now required for cytotechnologists who read Pap
OCR for page 80
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics smears, volume requirements akin to what is required for radiologists who read mammograms, and requirements for collecting, analyzing, and reporting data on test performance. There was no consensus on which measures, if any, should be pursued to improve the quality of biomarker testing. INCORPORATING BIOMARKER EVIDENCE INTO CLINICAL PRACTICE DISCUSSION Moderator Robert McDonough, MD, of Aetna U.S. Healthcare summarized his group’s discussion on incorporating biomarker evidence into clinical practice. He noted that there are many sources of information on biomarkers that reach clinicians, including journals, colleagues, product vendors, patients, popular media, practice guidelines, clinical trials abstracts, meetings, and continuing medical education. But when the group evaluated what prompts clinicians to adopt biomarker tests into their clinical practices, evidence-based information was not high on the list. “If you are looking at the screening for cancers, there is no correlation between the strength of the evidence and adoption,” said discussant Mark Fendrick, MD, of the University of Michigan. For example, an impressive 75 percent of the target population undergoes regular screening for prostate cancer, despite the fact the USPSTF gave it an unimpressive I rating. This is in contrast to the 50 percent of the target population who undergo regular colon cancer screening, which the USPSTF gave its highest rating because of its proven effectiveness. Academic practitioners appear to be more influenced by evidence, however, and may delay adopting a new test until there is evidence showing its effectiveness, several discussants agreed. This is in contrast to community practitioners, who may more readily adopt a new test or drug, even when there is little to no evidence of its clinical value. As a consequence, once a product enters the market, it may be impossible to gather the evidence on a test’s clinical value because of difficulties accruing patients to serve as controls for the trials needed to gather that evidence. Other factors beyond evidence appeared to be more important in influencing the incorporation of biomarker tests into clinical practice, the group noted. The most influential factor they identified was reimbursement for a test at a sufficient level. “If you look at the adoption of CT scans, PSA testing, or even COX-2 inhibitors, until they were paid for, they were not used,” said Dr. Fendrick. Because most diagnostics are relatively inexpensive, insurers are more likely to reimburse their costs without scrutinizing
OCR for page 81
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics the evidence base for the test, the group also noted. “If they didn’t pay for even low-cost biomarkers unless they were validated in a proper way, that would be an incentive to do those [validation] studies,” said discussant Dr. Carbone. The promotion that health insurers and employers do for various tests also influences their use, some discussants pointed out. For example, insurers often promote preventive health tests, such as those used to screen for various cancers, via informational mailings and their websites. “Some employers give discounts on health insurance to employees who undergo a self-assessment that indicates what types of screening and other health maintenance measures they should undertake,” Dr. Carbone said. “I think it is widely adopted when you give people a buck to do it.” Another highly influential factor was whether the test was adopted by what the group called “thought leaders.” A thought leader is someone who other members of a group look to as an authority. A thought leader may be misinformed, but he or she is still influential. In academic settings, thought leaders tend to be the lead investigators of clinical studies or the chairs of departments. In clinical practices the thought leader “is the clinician down the hall who seems to be knowledgeable about what is new in medical technology,” Dr. McDonough said. He said one discussant noted that physicians who practice in groups seem to adopt technology more rapidly than solo practitioners, possibly because of the presence of thought leaders in group practices. Another potential driver for the uptake of new biomarker tests is patient requests for the tests, the group noted. Studies reveal that if a patient asks for a drug by name, there is an 80 percent chance that a physician will prescribe it, Dr. Fendrick observed. Presumably patients have the same influence over the tests they request, he suggested. Through promotional efforts, product manufacturers also influence doctors and patients to use their biomarker tests, Dr. McDonough noted. “What I always thought was an important factor was the guy who knocks on your door—the vendor of the new device or new drug or new test,” he said during his group’s discussion. Dr. Waring also noted that for a test such as the FISH test for HER2, used to determine patient responsiveness to a specific treatment, the pharmaceutical company that provides that treatment may pay the costs of the test if it is not covered by an insurance provider. This is especially the case in Europe where national health plans may not offer the test as part of their services. “Roche until recently was paying for those tests to be performed in their own central laboratories,” he
OCR for page 82
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics said. “So these tests were becoming available not because of reimbursement issues—they were being made available by the pharmaceutical company for business reasons.” Other influences on the clinical adoption of a biomarker test hinge on features of the test itself, the discussion group said. Ease of interpretation is one such feature. If the test is easy to interpret and has a simple positive versus negative result, it will be adopted more readily than a test whose results require “some kind of complex algorithm to understand,” said Dr. McDonough. Clinicians are also more inclined to adopt tests that are reliably accurate and have timely results. “If you need to make a decision today, and the test is going to take 2 weeks, regardless of how easy or reliable that test is, it may not be very clinically useful,” said Dr. McDonough. Clinicians are also more likely to adopt tests if there is little to no risk in using them, and there are no alternative tests or test-linked treatments. Insurers are also more likely to reimburse for both the test and treatment, for those that are linked, if there are no treatment alternatives and the disease the drug targets is life threatening, the group noted. Inconvenience to the patient is another important test feature that influences its adoption in the clinic. Physicians are more likely to prescribe a simple blood test than an endoscopic procedure or a test that requires a stool sample, Dr. McDonough pointed out. Practitioners are also more likely to use a test that will influence their clinical decision making. “Is it a test that might give you some idea of the prognosis of lung cancer, but will not actually influence the type of therapy you might actually give to the patient? If the test does not seem to have any influence on the clinical management we would hope that would make it less likely that a clinician would use it,” Dr. McDonough said. Like other discussion groups, Dr. McDonough’s group recognized that low profit margins on diagnostic tests act as a disincentive to the development of biomarker tests and their evaluation in clinical trials. This led to the suggestion by Dr. McGivney that payors help subsidize some of this clinical research. “A payor who is asking for evidence should actually support, in part, the development of some of that evidence,” he said. Dr. McDonough said that some insurers, such as Aetna, do pay for routine costs of their patients in clinical trials. But Dr. McGivney countered that there is an increasing trend for payors not to cover such costs. Given that reimbursement levels highly influence the adoption of clinical tests, other discussants suggested that payors tailor their copay amounts for biomarker tests based on a test’s value or degree of evidence to support
OCR for page 83
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics any positive impact on patient outcomes. Zero copayment amounts could be allotted for those biomarker tests that are highly cost effective and likely to affect clinical management. High copayments could be required for tests whose cost effectiveness is questionable due to a lack of evidence on their benefits. But the group recognized that “it would not be easy to structure a benefit program to that fine a degree of assigning copays based on someone’s assessment of cost effectiveness,” Dr. McDonough said. There would be legal issues that might be difficult to overcome, such as varying state regulations that affect copayment levels. In addition, both legislators and the insurance clientele might look askance at plans that specify high copayments for treatment-linked tests for life-threatening illnesses. For payors to more adequately influence the adoption of biomarker tests, those tests need to have their own Current Procedural Terminology (CPT) codes, group members noted. These identifying codes are established by the American Medical Association and are used to report medical procedures and services to health insurers. Health insurers then specify reimbursement rates for each code. CPT codes are also used for developing guidelines for medical care review. “Many of these biomarkers do not have specific CPT codes,” said Dr. McDonough. “They are defined by process steps so that the insurer, even if they were willing to scrutinize biomarkers, often find it difficult to know what type of biomarkers are being used. What this means is that many of these biomarkers are being incorporated into clinical practice without much scrutiny.” This is especially true for home-brew tests, which are always defined by process steps. These tests, therefore, bypass scrutiny by both regulators and reimbursers, the group noted. Even when a test has been approved by the FDA, some discussants said, there is no guarantee that laboratories will use that test. Instead, they may offer their own home-brew version of the test, which may not be as acurate. Home-brew versions of the HercepTest, Dr. Waring said, help explain the variability in accuracy among laboratories. In a discussion following Dr. McDonough’s summary, Dr. Ramsey gave an overseas perspective of health care payors playing a role in gathering clinical data to evaluate new products. For example, the United Kingdom’s National Health Service pays for a new drug at an agreed upon price, with the requirement that data on the drug’s effectiveness be collected in a patient registry. If the drug does not show effectiveness at the expected level, the drug’s price is reduced so that the total reimbursement over time
OCR for page 84
Developing Biomarker-Based Tools for Cancer Screening, Diagnosis, and Treatment: The State of the Science, Evaluation, Implementation, and Economics reflects the actual quality of life gain observed. He thought such risk sharing in drug development was valuable, and noted that the group’s suggestion that payors cover the costs of clinical trials on biomarker tests would put all the burden of risk on insurance companies. He suspected they would balk at such a suggestion and reiterated that risk sharing has some value.
Representative terms from entire chapter: