6
Future Considerations
Over the course of the workshop, several panel presentations concluded with general discussions of the major issues affecting the development and use of biomarkers. These discussions focused on three main issues: creating incentives for organizations to collaborate, moving forward even when a thorough understanding of biological mechanisms is lacking, and dealing with different levels of risk in biomarker development.
CREATING INCENTIVES FOR COLLABORATION
Janet Woodcock noted that approximately half a million reports of drug-induced injuries are submitted to the Adverse Event Reporting System annually. These injuries represent both a major public health problem and substantial health care costs. At the same time, observed Daniel Bloomfield, the expectations for safety and the amount of research needed to get a drug approved have increased, even though the typical commercial life of a drug has not changed. Given the reduced returns from drug development, fewer companies are pursuing difficult projects with the potential to reduce the toll of drug-induced injuries.
Woodcock emphasized that investing millions of dollars in basic research and hoping that the resulting knowledge will automatically become available for use in human populations may be insufficient. Instead, special initiatives often are necessary to translate new knowledge into results that can have an impact on health care. Woodcock cited two such projects that have been supported by the Foundation for the National Institutes of Health (FNIH). One is the Alzheimer’s Disease Neuroimaging Initiative,
which also has been supported by NIH and industry and in which the FDA participates. Another is the Osteoarthritis Initiative, a prospective investigation of a large number of potential biomarkers. In addition, Woodcock noted that the FNIH helped establish the Biomarker Consortium, a major public–private biomedical research partnership with participation from a broad and diverse group of stakeholders, including government, industry, academia, and patient advocacy and other nonprofit private-sector organizations. The goal of the Biomarker Consortium is to collaborate in rapidly identifying, developing, and qualifying potential high-impact biomarkers.
Many workshop participants stressed that such collaboration among industry, the FDA, and academic researchers could yield much more rapid progress in the development of biomarkers. The question then becomes whether incentives could be established to promote such collaboration.
One important set of incentives, according to Frank Sistare, would be clear agreement on the data that could be generated in regulated phases of drug development that would not need to be submitted to regulatory authorities. When drug development is in its earliest stages, companies need freedom to operate without worrying about having to submit all such data to regulators, who may then decide that the development process should be slowed so that certain concerns can be probed more thoroughly. The FDA has offered guidance on these decisions, and there is an ongoing dialogue with the agency to clarify the issues involved. But the current lack of clarity continues to inhibit industry from generating data that could be extremely useful for fear that the data could be used to slow drug development.
Government and industry need to be creative in implementing incentives and removing disincentives, Sistare continued. For example, could a company be offered a reduction in user fees for the submission of data related to the discovery or development of safety biomarkers deemed critically important by regulatory authorities, or could it gain a period of added exclusivity for a product? Although both of these steps would require legislation, they represent the kind of out-of-the-box thinking that is needed.
James Stevens of Eli Lilly suggested that incentives might include staggered goals for what can be done in 1 year, 3 years, and 5 years. Some research projects take relatively long to complete, and potential partners in collaboration may be unwilling to participate unless they know when particular goals should be achieved.
Other workshop participants questioned the practicality of establishing new financial incentives to foster partnerships. Given the many financial demands on the federal government, said Alastair Wood, incentives that require additional funding probably will not succeed. Unless collaborations have realistic objectives and expectations, the potential to make progress through cooperation may be forfeited. Wood also questioned why incentives are necessary if a partnership results in drugs being developed more
quickly and with less investment of resources. If a given partnership makes sense, why are incentives needed to foster it?
According to Robert Califf of the Duke University Medical Center, multi-institutional partnerships and collaborations with industry will be necessary for substantial progress to occur. This requires both “big science,” characterized by extensive cutting-edge technologies, and “big populations,” where associations can be detected and refined. An undertaking of this magnitude, Califf observed, is too big for individual companies, even large multinationals, no matter how global they are. The same is true for individual academic centers, even those with broad, interdisciplinary skills and knowledge. Califf described his own experience with the recently launched David Murdock Research Institute as an example of the type of partnerships that will be required in the future. Created by a major philanthropic gift, the Institute is a collaboration among Duke University, the University of North Carolina (UNC) system, and Dole Foods. The Institute also has links to universities and industries throughout the United States, and partnerships with organizations in India and Singapore. Substantial funding has enabled the Institute to combine large-scale biobanking and state-of-the-art technology with support for manufacturing and commercialization. Califf characterizes the Institute’s approach as a “factory approach to biomarkers development.”
Califf further observed that existing public–private partnerships have been inhibited by uncertainty about how to manage conflicts of interest when public entities and for-profit corporations work together. A lack of clarity about the terms of engagement can stifle creative solutions.
Interests and incentives will vary even from one federal agency to another. For example, NIH has taken on important responsibilities, such as the Drug Induced Liver Injury Network (DILIN) and the Biomarker Consortium, that differ from the responsibilities of the FDA. Yet interagency collaborations have already begun to emerge, as exemplified by FDA and NIH interactions with respect to the DILIN initiative and the Biomarker Consortium. Successful partnerships hinge on finding common ground among agencies and between the federal government and industry. If important tasks are being overlooked within the federal government, it may be necessary to develop a new infrastructure within a federal agency to carry out those tasks. For example, a new, independent, cross-agency institute may be needed to foster biomarker development, suggested Richard Paules of the National Institute of Environmental Health Sciences.
John Bloom pointed out that partnerships could help establish standards for submission databases, review databases, and electronic medical records. Greater standardization throughout the biomarkers field also would encourage more sophisticated approaches to informatics. Bloom expressed the opinion that biomarker development faces no insurmount-
able barriers that cannot be overcome through a coordinated effort. But opportunities need to be seen as worthy of the attention and resources of institutions.
Wood also noted that many stages of biomarker development lend themselves to a noncompetitive structure. The more information that is shared among companies, the more productive research will be. Many companies see secrecy as essential to gaining an advantage, but secrecy also works in reverse. For example, other companies may have information about problems with another drug in the same class as the drug under development. A drug that proves to have problems early in the development process often is not extensively discussed outside the company that is developing it. Sharing such information could reduce the costs of research without compromising competitive positions.
Concluding the discussion, William Mattes of the Critical Path Institute suggested that any incentives put in place need to be carefully considered and structured so they do not create the appearance of favoring individual stakeholders. Incentives will be successful if they account for the varying interests of different groups. For example, academic researchers are rewarded for publishing their work and are unlikely to share information extensively before publication. Similarly, a company has incentives to work on its own compounds rather than in partnership with other companies on projects that are not directly product related.
MOVING FORWARD WITHOUT UNDERSTANDING MECHANISMS
As Califf pointed out, it is possible to make predictions with biomarkers that are probabilistically quite accurate without knowing much if anything about the mechanisms behind those biomarkers or the biological processes they reflect. This is already the case with cancer treatment, with physicians and patients being able to purchase multiple prognostic tests, each based on somewhat different arrays of biomarkers. While such options are available, however, it is always preferable to understand the mechanism involved because of the possibility of developing new targets for treatment or redesigning molecules to avoid toxicity by not engaging the mechanism.
Ravi Iyengar of Mount Sinai School of Medicine, whose workshop presentation addressed the role of systems biology in biomarker development (see Box 6-1), put the issue in a different context. Often a general mechanism is apparent for 90 percent of the cases of a disease or adverse drug reaction, and most of the other cases can be accounted for by using more tests and statistical associations. But 1 percent of cases may remain mysterious unless a biological mechanism is understood extremely well. If a signature for these outliers exists, Stevens asked, will clinicians be com-
BOX 6-1 Systems Biology and Biomarker Development In his presentation, Ravi Iyengar described the challenges facing systems biology, as well as the potential of this new perspective on biological processes to aid in the development of biomarkers. There are several definitions of systems biology. In the context of biomarker discovery, Iyengar described systems biology as the use of computational approaches to drive understanding. Network and statistical models that are implemented computationally are used to probe how the parts of a biological system function together. An understanding can be gained of how and why a complex biological function occurs as it does, although detailed mechanistic understanding of a molecular interaction may require different kinds of studies. Biological systems exist at different levels—from the organ level, to tissues and cells, to intracellular networks, to the molecular level. Many of the actual physiological measures in medicine are made at the level of clinical analysis and indicators. Systems biology models can often relate events at a lower level to clinical outcomes. A great challenge for systems biology, said Iyengar, is to integrate understanding of these different levels vertically. As an example of a correlation without detailed understanding, Iyengar cited an FDA-approved breast cancer diagnostic that is based on 70 genes, while an alternative diagnostic is based on 76 genes. Yet the two sets have only three genes in common, which raises the question of how the sets are related. Research in Iyengar’s laboratory has shown that both sets of genes are linked to overlapping sets of upstream transcription factors and signaling. In turn, transcription factor activity profiling and network analyses can help identify relationships between mutated disease genes and prognostic gene expression signatures. This is one way to connect events at different levels, enabling oncologists to use molecular markers in treatment decisions. Iyengar’s laboratory also has been looking at congenital and drug-induced arrhythmias. Using genes identified as being related to long-QT (LQT) syndrome, he and his colleagues built a disease gene network to see how the genes are related. From a very large network of 15,000 nodes and 70,000 interactions, they identified an LQT gene “neighborhood” of about 1,400 nodes. They found that unique networks can be constructed around genes involved in disease states, and the properties of these networks can help explain some of the characteristics of different states. Iyengar said that these networks also can explain drug side effects because there is a relationship between the genomics and systems pharmacology of LQT syndrome. Networks of biomarkers are likely to perform better than single biomarkers for complex diseases because networks across genes integrate multiple sources of information. In this way, systems biology approaches can provide insight into the pathogenesis of adverse events and suggest alternative targets for treatment. It may even be possible to predict clinical outcomes 2–5 years into the future on the basis of information from cellular or molecular networks. SOURCE: Iyengar, 2008. |
fortable using it to make clinical decisions without knowing the mechanism behind a response?
Several workshop participants responded that biomarkers can provide valuable information even when biological mechanisms are largely unknown. At a fundamental level, Califf observed, many biological mechanisms remain at least partly unknown. Woodcock stated that medicine is conducted among many uncertainties, and reliable information that can distinguish who is and is not at risk is an advance beyond not having such information. Also, Woodcock pointed out that the discovery of predictive biomarkers can lead to research on their reliability and on their association with outcomes.
Bloom emphasized the importance of not interpreting the term “biomarker” too narrowly. A biomarker is a piece of information that can be used correctly or incorrectly in making a decision or seeking additional information. The term “biomarker” can even be misleading if it is interpreted as denoting a single measurement without a broader biological context.
DEALING WITH DIFFERENT LEVELS OF RISK
Bloomfield described a hypothetical scenario involving a drug that is effective at treating depression but causes a mean blood pressure rise of 2 millimeters (mm) of mercury in a test treatment population. Should such a drug be approved? The ultimate question in such cases, he said, is the level of risk that patients, physicians, and society are willing to accept.
Woodcock emphasized the complexity of this issue. The older anti-psychotics, for example, posed major risks, but at one point they were the only available treatments, so they were widely used. Regulators know that a 2 mm rise in blood pressure will translate to a mortality difference if a drug that causes it is used long enough. In the past, calculations of risks and benefits were left largely to physicians and patients; today, other groups play a role in these calculations as well. This is one example of how biomarkers could be pivotal. If it were possible to identify subgroups who would experience the 2 mm rise in blood pressure or would have a good response to the antidepressant, the risk/benefit calculation would be easier to make.
Califf suggested that an effective drug for depression would save lives, and therefore should be available on the market. At the same time, however, an outcome study should be done to determine the true effect of the drug on the balance of risk and benefit. The more biomarkers that can be identified to gauge the effects of a drug, the stronger the signal will be as long as the research reflects an awareness of the complex methodology that must be applied to understand the joint effects of multiple markers. Iyengar
pointed out that most predictions take the form of probabilities, which do not tell a patient or physician exactly what to do, and proper decisions will be more likely if all parties involved understand the role of probabilities in decision making.
Insel proposed a promising way to involve the public in the biomedical enterprise and inform them about its results. He suggested that every patient should become a partner in a research program addressing the condition affecting that patient. This has already happened in some areas, such as cystic fibrosis and particular kinds of childhood cancer. It could occur as well for much broader groups, such as everyone with cardiovascular disease.
Califf responded by saying that one of the most encouraging aspects of establishing the David Murdock Research Institute is that the organizers have been overwhelmed by calls from people in the surrounding region who want to be enrolled in epidemiological studies. Involving these volunteers in research will take careful planning, but they represent a largely untapped resource that could speed the pace of scientific progress.
REFERENCE
Iyengar, R. 2008. Systems biology of biomarker sets. Speaker presentation at the Institute of Medicine Workshop on Assessing and Accelerating Development of Biomarkers for Drug Safety, October 24, Washington, DC.