The current 'pipeline problem'
The translation of basic research to clinical practice is a priority for academia, government and industry, both in the United States and in other countries. The current focus on translation stems from concern that the level of investment in research is not being reflected in improved clinical outcomes, and there is particular concern that the benefits from the 'genetic revolution' have been slow to arrive.
In the United States, which is the primary focus of this article, an emphasis on translation is evidenced by the National Institutes of Health (NIH) Roadmap initiative (see Further information) and the US FDA's Critical Path Initiative (CPI) for drugs and diagnostics1. The CPI provides an analysis of the 'pipeline problem', which is broadly defined as a decrease in the rate of introduction of new drugs and diagnostics to the market in recent years, despite major scientific achievements and investment that might have predicted otherwise. The initiative is in part based on the premise that changes in regulation have the potential to stimulate the translation of science to the clinic and market.
Issues regarding biomarkers and diagnostics have emerged as crucial components of the CPI Opportunities Report2. It could be argued that the pathway from development of biomarkers and diagnostics to market is inherently different than that for drugs, and so changes implemented by the FDA to stimulate innovation in drug development might not similarly translate to innovation in development of biomarkers and diagnostics. As novel gene-based diagnostics proliferate they will be increasingly important to drug development, approval and clinical practice. Clearly, a translation problem in diagnostics will directly influence the translation of drugs to clinical practice, even for drugs that are developed and approved. An important step now is to synthesize available information to assess the relevance of regulatory and policy changes to facilitate the development of biomarkers and diagnostics. It is important to understand whether and how the 'pipeline problem' in diagnostics and drugs can differ, especially because it is anticipated that the co-development of drugs and diagnostics could soon be the prevailing model.
With this in mind, in this article, we characterize the 'pipeline problem' as it relates to diagnostics and biomarker development in the United States based on a review of the relevant published and web literature, interviews and meeting presentations (Box 1). Our review is based on interview and meeting data in addition to the peer-reviewed literature, with the aim of providing a more current and in-depth assessment than a literature review alone, and so also reflects our own interpretations, as well the opinions of our interviewees.
The factors driving medical device innovation may differ considerably from those that produce new drugs and biological products.
Is there a diagnostics pipeline problem?
A crucial concept in the development and adoption of diagnostics is the 'product pipeline': the development process from initial design or discovery to the finished diagnostic including implementation and post-marketing surveillance. One key measure of the pipeline problem for new drugs noted in the CPI is the increasing level of expenditures (public and private) with a simultaneous decrease in the number of submissions of New Drug Applications (NDAs) for new molecular entities (NMEs) and/or original Biologic License Applications (BLAs). The Advanced Medical Technology Association (AdvaMed), the primary trade association representing medical devices, subsequently examined data on new medical device submissions and concluded that during the same 10-year period there was actually an increase in the number of device-related submissions. In comments to the FDA, AdvaMed states, "The significance of this finding is that we believe the factors driving medical device innovation may differ considerably from those that produce new drugs and biological products."3
Yet despite the fact that diagnostic submissions might not be decreasing, the overarching belief of thought leaders is that a pipeline problem does exist for diagnostics development. However, it is believed that the pipeline problem for diagnostics is more at the 'front end' — in earlier stages of discovery and development. Data on approval submissions are therefore insufficient for understanding the pipeline problem.
What, then, are the key issues relevant to the diagnostic pipeline and how are these different from those for drugs? Below we discuss some of the issues raised in the literature and in our interviews.
What is the problem?
The pipeline problem for diagnostics is multifaceted. The extent to which all diagnostics (in-house or FDA approved) identify clinically relevant biomarkers — biomarkers truly indicative of improved clinical outcomes — is a key issue. For example, Tsongalis has noted that industry is rapidly providing reagents under the analyte-specific reagent (ASR) rule (Box 2), but raised the crucial question of which of these are most useful in patient care4. More generally, there is concern that diagnostics and biomarkers have not been widely proven to change clinical outcomes. Importantly, no regulatory authority for diagnostics considers clinical utility for test approval (discussed in more detail below). According to Janet Woodcock of the FDA, biomarker development will need data pooling, synthesis and analysis; identification of what is known, not known and what can be known through studies; and finally using what is identified, learned and developed about biomarkers to facilitate translation to clinical practice5.
The regulatory structure for laboratory tests (Box 2) results in many tests being offered outside of the purview of the FDA. Nonetheless, AdvaMed reports that the overwhelming majority of its members (84%) believe that FDA regulatory requirements are one of the top factors affecting their ability to develop new technologies6. As genomic-based diagnostics are increasingly co-developed with drugs or not approved as lower-risk devices or ASRs (as in the case of the AmpliChip7), the FDA is likely to have an important role in generating solutions to translating this research to clinical practice. Regulatory strategies for diagnostics will need to be considered within a model that adequately reflects the nature of the development of diagnostics (highly iterative and dynamic, as discussed below), while at the same time increasingly including drug development.
In addition to the regulation and development model, two issues stood out as major reasons for the diagnostic pipeline problem: money, and lack of samples for validating and testing biomarkers. With respect to the first issue, key concerns are the level of investment in diagnostics and the amount of reimbursement for their use (discussed in more detail below). The inability to 'capture value' creates disincentives for both industry and the venture capitalist community to invest in diagnostics. In particular, one of our interviewees stated that some diagnostics companies, particularly smaller firms, do not understand how to demonstrate value and what it takes to get a product to market.
With respect to the lack of samples, the availability of samples is considered instrumental to the development (or lack thereof) of diagnostics because an abundance of samples is crucial for determining test performance (analytical validity, clinical validity and clinical utility). Regulation, ethical and legal concerns, and practicality and feasibility issues are believed to hinder obtaining and using samples.
However, a key point is that as the FDA (and others) forge ahead with improving the translation of diagnostics (and drugs/devices), innovation must be appropriately balanced with evaluation. Drugs and diagnostics might fail to make it to market because they add little marginal value, and thus these failures should not be viewed as part of the pipeline 'problem'. Any discussion of pipeline issues must therefore also consider the value of what is in the pipeline. One proposed solution is to focus more effort on understanding how good a diagnostic would need to be to add value, which can be done via economic modelling and 'what if' analyses (Box 3).
It is important to recognise that there is an inherent tension between improving the pipeline and increasing the safety of new drugs. The FDA is under pressure to do both, but our interviews suggested that there is inadequate understanding by the public and policy-makers about the trade-offs involved. As one illustration, some thought leaders felt that there would soon be a public health crisis about 'home brew' tests (Box 2) because there would be a quality lapse that, in conjunction with recent safety-related events, would prompt a 'crisis'. However, home-brew tests are poorly understood by the public and therefore any 'crisis' might prove to be a false or inappropriate one.
Are diagnostics and drugs really different?
Development models. In responding to public comments on the FDA CPI report, AdvaMed argues that a key difference between drug and diagnostic development is that fundamental technologies and the medical devices that incorporate them are typically designed rather than discovered3. Consequently, diagnostic development might be characterized by continuous modifications throughout product life cycles. Typically, diagnostic development involves an iterative process (for example, performance characteristics are continuously fed back into the design) so that diagnostics can ultimately be shown safe and effective for their intended use, which are the FDA criteria for approval8. However, it is not uncommon to make modifications to existing devices for new uses3.
Figure 1 shows a comparison of the FDA CPI paper's general medical product development model and one proposed by representatives of the diagnostic industry at an industry/FDA roundtable to more specifically focus on diagnostics1, 9. The comparison of the two models highlights both differences and similarities, which will be particularly important to understand as the market moves towards more co-developed products. For example, the diagnostic development model notes that there are several points of entry to the market for diagnostics. Diagnostics can be directly accessed by the patient from a retail supplier, such as an online diagnostic test provider, a clinician, a hospital or academic medical centre or a laboratory (reference or clinical laboratory)10.
Figure 1 | Medical products and diagnostics development models.
Comparison of the FDA Critical Path Initiative paper's general medical product development model and one proposed by representatives of the diagnostic industry to more specifically focus on diagnostics. a | FDA medical product development model (adapted from Ref. 1). b | FDA model adapted for diagnostics9.
Regulation of diagnostics. Diagnostics can be viewed simultaneously as less regulated and more highly regulated than drugs. As described in Box 2, ASRs typically do not require full FDA pre-market review. However, some believe that diagnostics are inherently held to a stricter standard than drugs. One interviewee felt that, because a drug can be approved even if it only works in 10% of the population and/or could be compared with placebo for approval, drugs undergo less rigorous evaluation than diagnostics, which are often evaluated for equivalence to similar tests.
There is a widespread belief that co-developed drugs and diagnostics will soon be the prevailing business and development model (for example, similar to HER2/neu tests and the anti-HER2 monoclonal antibody trastuzumab (Herceptin; Genentech), which is used to treat HER2-positive breast cancers), and it is felt that the FDA is being proactive in adapting to these changes. However, the FDA has traditionally independently regulated the two areas11, and so there is some concern about the capacity of the FDA to jointly review co-developed products adequately and the ability of the FDA centres to work together.
Reimbursement. Reimbursement can be considered the ultimate incentive for industry to bring products to the market, yet it is a poorly understood topic. Indeed, reimbursement rates have been a highly charged issue for industry, government and insurers (see, for example, Ref. 12). It is believed that inadequate reimbursement is a key (if not the key) reason for not developing diagnostic products. At least two government-commissioned reports have recommended a re-evaluation of reimbursement rates for diagnostics13 and for genetic tests more specifically14. Although there are ongoing discussions to amend rates, the high potential costs of novel technologies has been of particular concern within the paradigm of cost containment15.
Some believe that industry will be driven to use pharmacogenomics and develop biomarkers and companion diagnostics by payers (and to a lesser degree, regulators) who will demand some sort of selection of patients where drug prices were skyrocketing. However, others feel that payers can only respond reactively rather than proactively. Diagnostics have historically been perceived as being of lesser value than drugs and thus there is more reluctance to pay high prices for diagnostics. The concern among insurers is that using expensive tests for large populations will wipe out any potential savings from targeting drugs. Diagnostics might also be undervalued because the entire sequence of events is not considered — for example, using a test can reduce the need for expensive procedures. This is an area where modelling and economic analyses can be very insightful (Box 3).
Inadequate reimbursement rates might not only limit availability but could also create an incentive not to seek FDA approval. One of our interviewees noted that if the test is FDA approved then it is more likely to get Medicare reimbursement approval and subsequently private insurer approval, resulting in potentially wider utilization but lower rates of reimbursement. Conversely, if the test is developed for in-house use, it is more likely that prices are negotiated directly with stakeholders who value the test information and are who are therefore willing to pay more for the test itself — but then the test will be less widely used. There is a similar incentive not to get FDA approval for 'esoteric' tests. As a result, companies developing new diagnostics (potentially based on pharmacogenetics and other novel therapeutic strategies) leading to alternative therapeutic plans might initially pursue non-Medicare payers for reimbursement because these plans were more flexible and faster decision makers in the United States than the Centers for Medicare and Medicaid Services (CMS). However, such a strategy leaves coverage decisions fragmented and locally based, rather than national and standardized.
What are the roles of key stakeholders?
There are many stakeholders in the translation of diagnostics to clinical practice including patients, clinicians, industry (diagnostics, pharmaceutical, biotechnology and venture capital), laboratories, government, payers and academics. In this commentary, we focus on the role of industry, the FDA and payers.
Industry: the ascent of the diagnostic industry. The increasing focus on genetically based interventions is causing the role of diagnostics — and the diagnostic industry — to become more crucial to pharmaceutical development. Larger pharmaceutical firms have therefore engaged with diagnostics firms directly and developed internal diagnostics foci16. Pharmaceutical firms are increasingly embracing investment in genomics-based diagnostics as they use these diagnostics throughout the clinical development phases17, 18.
It might therefore be important for industry to embrace new models for diagnostics and drug development. Three key industry players — diagnostics, pharmaceuticals and biotechnology — are likely to shift towards more collaborative business models that focus on co-developed products. Research has shown that pharmaceutical products developed collaboratively have a higher probability of success19 and at least one large pharmaceutical firm (Novartis) has advocated sharing biomarker data as one way to speed their translation to clinical practice20. The FDA's focus on voluntary genomic submissions (VGDS) is further evidence of a belief that sharing data is an opportunity to improve translation21. The FDA reports that a promising number of companies to date have submitted data under the VGDS guidance22, but industry also remains sceptical of submissions.
What would a co-developed product model look like? One such model is presented in Fig. 2 (Ref. 23).
Figure 2 | Prototype of possible approach to developing and regulating combined test–drug products.
![Figure 2 : Prototype of possible approach to developing and regulating combined test|[ndash]|drug products. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.com](/nrd/journal/v5/n6/thumbs/nrd2033-f2.jpg)
One possible model for co-developed products is presented in the FDA Co-Developed Products document23. This figure shows a potential regulatory scenario for a single test that would be used in conjunction with a single drug in the clinical management of a patient. The figure highlights key events for both diagnostics and drug regulation but it is the overall coordination of the regulatory process so that the products launch together that is unique in the co-developed products model.
FDA: the importance of a proactive regulatory approach. It has been suggested that early and frequent industry and FDA meetings can speed translation to clinical practice particularly if accompanied by clear expectations on deliverables and outcomes, as well as flexibility24. The importance of clarity is evidenced by the ongoing discussions on acceptable biomarkers for drug and diagnostics development and microarray experiments25 (see also Minimum Information About a Microarray Experiment, Further information). Industry is encouraged by FDA's continuing efforts in producing guidance documents such as the Pharmacogenomic Data Submissions Guidance21 and Co-developed Products Concept Paper23.
A role for the FDA could also be to cultivate strong internal organization. Key FDA centres will need to work collaboratively with each other because novel products will increasingly require their collective expertise. However, some industry spokespersons remain sceptical that the FDA's organizational structure and resources are sufficient for improving translation, and wonder whether the CPI has the potential to increase regulatory burden26. There has been a perception that certain FDA centres are less rigorous than others in their evaluations, and because combination product sponsors recommend which centre should evaluate the application there is an incentive to recommend the 'easier' centre27. Some further argue that the current regulatory system, which focuses on definitive safety and effectiveness, is not well suited for probabilistic diagnostics or those aimed more specifically at subpopulations26. There is a more recent concern that the resources devoted to new safety initiatives in the wake of the COX2 debacle will further strain an already inadequately funded agency28. It will be especially important for the FDA to develop and foster adequate expertise (and funding) for future regulatory decisions about novel diagnostics.
The FDA might also have a role in working with industry to develop new evaluation methods and paradigms for data collection. There are opportunities for the FDA with respect to the collection and use of biological samples and the evaluation of data for regulatory purposes. The FDA could reconsider the use of stored samples for retrospective analysis and for uses not initially proposed, although this will require consideration of ethical and legal issues. It has been argued that genetic markers are not significantly different for regulatory purposes than other diagnostic tests29, but the issue of whether genetic samples are unique is unresolved30. AdvaMed believes that the FDA could develop guidelines for waiving informed consent and Institutional Review Board (IRB) review for studies that use left-over or banked samples that are unlinked or unidentified3.
The FDA could improve the pipeline problem for diagnostics and biomarkers through any action that would improve access to samples. Data such as the VGDS data from which industry and the FDA could identify promising drugs, diagnostics and biomarkers together also offer potential. A concern expressed was that the FDA was overly cautious on retrospective data analysis for regulatory purposes, yet as technology for data collection and data mining improve, retrospective analysis and models will increase in importance.
The FDA might also indirectly encourage investment in diagnostics through labelling changes. It is believed that including a diagnostic in the label increases the value of diagnostics, which creates an incentive for future diagnostics development. The FDA could require that test information appear in labelling, but for a test to be included in the label the test must be available (Lesko, L., personal communication). The FDA, the Centers for Medicare and Medicaid Services (CMS) and industry might have to work together to develop a publicly available comprehensive list of diagnostic tests. Nonetheless, whether labelling with biomarker and diagnostic test information actually influences clinical practice is uncertain. The potential influence a label has on encouraging the use of a diagnostic might more specifically relate to how the label describes the test, such as whether it is recommended or required. For example, the label for 6-mercaptopurine was recently amended to include information on recommended testing for thiopurine S-methyltransferase (TPMT) but it is believed that clinical practice has not significantly changed following the labelling revisions31. Other labels with pharmacogenetics-related information (for example, mentioning poor metabolizers) provide limited utility to the prescribing clinician or patient32.
Finally, another role for the FDA is to encourage strong external relationships. The FDA has an important role to play that goes beyond their actual mandate because of their place in the pipeline. Labelling issues illustrate a broader concern about whether the current structure of the FDA and the Clinical Laboratory Improvement Amendments (CLIA) is adequate for the future33, 34, 35. Because tests are not typically subject to full review or tracking by either the FDA or the CMS, there is little information available on what tests are being performed by which laboratories, for what purposes and with what results4. This lack of information has been noted as a barrier to the development and approval of combination products. For example, the approval of pharmacogenetic drugs will often require the simultaneous review and approval of the appropriate test36. One issue is that the FDA has decided not to engage in more regulation of home-brew tests because of the large resources required and the lack of a mandate37. Similarly, although the FDA does not consider the cost or economic value of products, their decisions ultimately affect those outcomes and payers increasingly want such information.
Payers: the lynchpin. The role of valuing and paying for diagnostics is essential to translation. On the one hand, expensive, novel, high-tech diagnostics have the potential to increase costs to insurers. On the other hand, diagnostics could streamline expensive therapies based on efficacy and avoiding adverse events, and therefore decrease costs to payers. It is not surprising that some of our interviewees expressed the opinion that payers will not only support the implementation of diagnostics but also require them — although others are more sceptical that payers will be able to drive the agenda.
A role for payers will be developing new criteria for reimbursement that adequately values novel diagnostics. An Institute Of Medicine (IOM) report13 in 2000 noted that the process for integrating new technologies into the payment system, including determinations of coverage, assignment of billing codes and development of appropriate prices, is slow, administratively inefficient and closed to stakeholder participation. The IOM also predicted exacerbation of the problem as laboratory medicine evolved, and concluded that there was insufficient means by which to determine the value of individual tests and services with respect to reimbursement rates. Final recommendations from the Secretary's Advisory Committee on Genetics, Health and Society (SACGHS) on coverage for genetic tests are expected in the near future38, although the draft recommendations clearly indicate that there is some belief that reimbursement rates for genetic-based diagnostics are insufficient14.
Payers are also crucial because they are in a position to encourage the analysis of the utility of new diagnostics and therapeutics. It is likely that payers will increasingly demand evidence that new products either improve care at a reasonable cost or decrease expenditures, and it is therefore likely that economic analyses of the value of such products will become increasingly important. Similarly, venture capital firms will only invest in companies that are developing products that have demonstrated value. Historically, venture capital firms have been more reluctant to invest in diagnostics companies, although this might change as the business model moves to co-developed products.
Conclusions
With the increasing emphasis on personalized medicine and targeted therapies, diagnostics and biomarkers will have an increasingly important role in drug development and healthcare delivery. It is therefore crucial to address the pipeline problem for diagnostics and biomarkers, and the unique aspects distinguishing the problems in diagnostic and drug development.
Several recommendations can be made to address the pipeline problem through both research and policy changes (Box 4). A key recommendation is that more information must be made available on the use and value of diagnostics and biomarkers. One approach is being undertaken by the EGAPP Project (Evaluation of Genomic Applications in Practice and Prevention) (see Further information). EGAPP is a 3-year model project sponsored by the Centers for Disease Control and Prevention (CDC) that began in 2005. The goal is to establish and evaluate a systematic, evidence-based process for assessing current and future genetic tests as they make the transition from research to practice. The 13 members of the Working Group have expertise in evidence-based review; health technology assessment; primary care, specialty care or nursing care; epidemiology; clinical genetics/genomics; economics and decision analysis; laboratory practice; and ethics, law and policy development. This Working Group is prioritizing and selecting topics, establishing methods and processes, overseeing expert and peer review of commissioned evidence reports, and developing conclusions or recommendations based on the evidence. Economic models and analyses can also be developed to help address the lack of demonstrated clinical utility and provide evidence for greater investment in diagnostics and for payers' reimbursement decisions39, 40.
An important role for industry will be to embrace new models for diagnostics and drug development. Three key industry players — diagnostics, pharmaceutical and genomics/biotech firms — must shift towards more collaborative business models that focus on co-developed products. The FDA also has the important role of facilitating a streamlined approval process that provides clarity; coordination across centres; cooperation and collaboration with other stakeholders outside the FDA; facilitating solutions to the sample problem; facilitating solutions for using data for approval, including standards development; and developing approaches for review of co-developed drugs/diagnostics. A role for payers will be developing new criteria for reimbursement that adequately value novel diagnostics.
Finally, it is clear that there must be collaboration across different sectors (such as the FDA, CMS, private insurers, industry and academia) in resolving the pipeline problem. The issues are complex and will require cooperative efforts, including efforts to educate the relevant stakeholders appropriately. Although not specifically discussed by our interviewees, the understanding and knowledge about the uses and benefits of new diagnostics and therapeutics is crucial to translation. There remains much scope for improving the knowledge of payers, providers, the FDA, CMS and patients through education and evaluation41.
