Important Points Made by the Individual Speakers
- Genomics cannot afford a bubble of unwarranted enthusiasm.
- Health economics research needs to consider both the decisions of individuals and the behaviors of populations.
- A major hurdle to genomic-based medicine is the lack of a national, dynamically updated, interpretative database of evidence for the clinical utility of genetic variants.
- Patients need time with their physicians to discuss options, but incentives are currently not aligned well to encourage this time.
- If genomic technologies add to the cost of health care without commensurate benefits, they will not be widely adopted.
- Data on the comparative effectiveness of genomic testing will take at least a decade to accumulate, which will delay the implementation of testing.
On the second day of the workshop, speakers reflected on the previous day’s discussions. Although none of the presenters purported to speak for an entire sector of stakeholders, they made points that relate to many of the economic issues underlying the development and use of genomic tests.
James Evans, who provided the introductory presentation summarized in Chapter 2, emphasized that genomics cannot afford a bubble of unwarranted enthusiasm. When people engage in wishful thinking without seeking out evidence, they can overvalue any commodity, from real estate to tulips to genomic tests. Genomic-based medicine has great potential and is likely to produce great advances in reducing patient suffering and improving care, but evidence is needed to prove these points rather than assuming that these advances will occur.
He also emphasized the importance of focusing on big problems where the maximum returns in terms of outcomes and cost will occur. Accumulating evidence will point to where the greatest outcomes are likely to occur, and this evidence should be heeded.
Free medical tests do not exist, said Evans, and the inappropriate use of a test adds to the burden of health care costs. Everyone has an interest in seeing that tests are used wisely and appropriately, because everyone pays for those tests through insurance. Evans argued against the concept that personal utility should be a factor in deciding whether a genetic test should be used because “it can mean anything to anyone. Before I help pay for your personal utility quest, I want to see some evidence that it improves outcomes.” Coverage decisions should be driven by commonly shared values, said Evans, rather than outliers. Determining and acting on these values is critical in many areas of genomic testing, whether informed consent, privacy, or cost. These decisions also need to maximize equity, though absolute equity is impossible. Existing medical tests are not used in what Evans called “shotgun or nontailored” ways, and genomic tests should not be treated differently.
Patients do not always choose to access all the information that is available to them. Many men and women who are informed about the benefits of particular screenings will still decide not to have the test. More information is not always better or desirable, said Evans, and extra information that is not of obvious value can be counterproductive.
All these points argue for the use of targeted genomic testing, Evans said, where particular information is gathered when it makes sense to do so. Targeted testing will also focus educational efforts so that people and providers can be educated in categorical ways rather than test by test.
Finally, Evans noted the tremendous obstacle of the cost of targeted drugs for small populations. Even if genomic testing can reveal the characteristics of a tumor, the cost of a drug may make treatment for that genetic lesion prohibitive.
Armstrong focused her comments on how resources are allocated within the health care system. For example, novel tests and treatments would have their biggest impact on common diseases, but they could also have a substantial effect on rare diseases if many different diseases could be treated. Decisions need to be made about how to allocate resources under such circumstances, and genomics itself can inform these decisions.
Much of the discussion at the workshop focused on decisions at the individual level, Armstrong observed. But that level represents only one end of a spectrum. Decisions range from the micro level of individual patients to the macro level of the health care system or payer. Health economics research should look across this spectrum and consider how a new technology can improve decision making and outcomes within large populations as well as in individual health decisions.
Armstrong also discussed several challenges at the intersection of genomic testing and economics. First, health economics research faces a workforce challenge in that many researchers have been working in the field for a long time. The next generation of researchers needs to be engaged to work on the major issues confronting the field.
Also, the field continues to struggle with data in that many data are lost or not available when needed. A system is needed to gather, store, and disseminate the data that will be needed to resolve major questions.
Finally, health economics research needs to be focused on helping to answer constructively whether a technology should be used, and resources need to be allocated for doing so, said Armstrong. For example, the recent focus on patient-centered outcomes provides an important forum for taking economic studies forward. The true impact of genomics may lie in preventing the use of unnecessary tests rather than in linking any particular single nucleotide polymorphism to a disease or in determining how a whole genome sequence is used, said Armstrong.
Whole genome or exome sequencing is initially entering clinical practice informally via academic medical centers and biotechnology laboratories operating under guidelines from the Clinical Laboratory Improvement Amendments (CLIA), observed Thomas White, retired chief scientific officer at Celera Corporation and regent’s lecturer at the University of California, Berkeley. Most of these tests are not being reimbursed by insurance companies, but when they are, compensation is made because the tests are “under the radar” and not consuming enough resources to warrant a thorough review for reimbursement.
Some researchers have argued that all genomic tests should be approved by the FDA, but that scenario is extremely unlikely, White said. For example, the path to FDA approval for a whole genome sequencing instrument and reagent system is currently unclear. Such a system would have complex intended uses, accuracy problems, no gold standard for comparison, rapid technical obsolescence, and a potential requirement for lengthy and costly prospective treatment-by-genotype clinical outcome studies. In addition, it is difficult for manufacturers that have to produce a system with validated reagents and software to compete with laboratories that develop tests under CLIA that use research-use-only instruments and that can update software and reagents as needed. “There is an uneven playing field that will prevent any of these tests from reaching FDA-approved instrument systems for decades,” said White.
A major hurdle to genomic-based medicine is the lack of a national, dynamically updated, interpretative database of evidence for the clinical utility of genetic variants, White said. Not every CLIA lab will be able to maintain and update a database of all its clinically useful variants and convey those updates to patients and physicians. Without evidence of clinical utility and cost-effectiveness, private and public payers may default to nonreimbursement. The government could have an extremely useful role in standardizing this information, said White, who suggested that the 100K Foodborne Pathogen Genome Project1 may be a model to emulate.
The basis for reimbursement of whole genome sequencing remains uncertain, he observed. Except perhaps for cancer indications, single clinical indications may not be cost-effective, said White. Furthermore, because of issues with accuracy, targeted confirmatory testing of some actionable variants identified by whole genome sequencing may be necessary and will increase overall costs. Given these uncertainties, White asked, how can the cost-effectiveness of whole genome sequencing be assessed for application over a lifetime?
Professional organizations need to develop guidelines for reporting genetic variants to patients and providers, according to White. Also, providers and patients are going to want evidence that a test will make a difference in treatment.
Stakeholders may need to accept different evidentiary standards for the clinical utility of tests for different medical conditions, White said. Also, the government will need to continue to support prospective randomized clinical outcome studies, with studies prioritized by disease area. Studies supported by the Patient-Centered Outcomes Research Institute could contribute to this goal.
Cost-effectiveness studies are not terribly expensive, said White, and
information from previous randomized clinical outcome studies can be used to assess cost-effectiveness. For example, data from the Framingham heart study have been used to look at genetic markers that increase the risk of heart disease and the costs of preventing heart disease in various ways (Shiffman et al., 2012).
Patients tend to differ from patient advocates, said Mary Lou Smith, cofounder of the Research Advocacy Network. Patients are more focused on themselves, their treatment, their families, and their future, whereas advocates look at a broader picture with the needs of a composite patient in mind. Smith is a two-time breast cancer survivor and spoke as a patient, not as an advocate.
Patients want to know what the best treatments are for them. They want to know whether they will benefit from a treatment, suffer the effects of toxicity, or both. If a genomic test will provide that information, the question becomes whether a physician will order the test, whether the physician will understand the test results, and whether the physician will use the test correctly, said Smith.
Patients need time with their physicians to discuss options, but physicians do not have incentives to make this time. They are pressured by their administrators to be productive, and chemotherapy is profitable for hospitals, which can distort incentives throughout the system, said Smith.
Research is needed on how information from genomic tests is used by patients, said Smith. For example, studies conducted by the Research Advocacy Network indicate that benefit matters more than toxicity to patients when making treatment decisions. The greater the benefit, the more likely a patient is to choose treatment. The higher the toxicity, the less likely a patient will choose a treatment, although the nature, duration, and severity of the side effects are also factors. “It’s not a simple process,” she said.
One important application of genomic testing will be to identify patients who will not benefit from particular treatments. Such testing, Smith said, could produce both tremendous cost savings and great improvements in quality of life.
Smith also commented on informed consent. “I don’t think anyone in this room actually believes those 27-page documents inform patients,” she said. Even easy-to-read consent forms do not necessarily increase retention or understanding. “We need to start from square one and say, What does a patient really need to make an informed decision? And what is an informed decision?” Smith noted that she is a lawyer but still did not read the consent form given to her when she entered into a clinical trial. “I knew the
people, I trusted the people, I knew the research they were doing, I wanted it to go forward.”
She expressed concern about the delayed effects of treatment. If a woman develops a heart condition 10 years after receiving radiation treatments for cancer, there is no way to explore the connection between these two events because no system exists to make those linkages. And without knowing about such linkages, patients cannot be fully informed of the risks.
Smith emphasized the importance of engaging patients as partners if researchers want samples of tissue and data over time. This would provide a much fuller picture of the benefits and side effects of treatment.
Finally, evidence-based medicine requires new trial designs and end points for Phase I, Phase II, and Phase III trials, said Smith. “Right now, we don’t have them.”
Public health thinks about the individual within the context of a community, a nation, and the global population, observed Calonge. Public health is more than publicly funded health, though that is part of it. Viewed broadly, public health provides the following essential services:
- Monitoring health status to identify and solve community health problems.
- Diagnosing and investigating health problems and health hazards in the community.
- Informing, educating, and empowering people about health issues.
- Mobilizing community partnerships and action to identify and solve health problems.
- Developing policies and plans that support individual and community health efforts.
- Enforcing laws and regulations that protect health and ensure safety.
- Linking people to needed personal health services and assuring the provision of health care when otherwise unavailable.
- Assuring a competent public and personal health care workforce.
- Evaluating effectiveness, accessibility, and quality of personal and population-based health services.
- Conducting research for new insights and innovative solutions to health problems.
Calonge focused on more of a macroeconomic evaluation of genomics and public health. He noted that in Colorado, premiums for state Medicaid medical services will exceed school finance’s share of the general fund in
about the year 2017. By 2019, Medicaid and K-12 education will consume the entire general fund, leaving no money for any other state services. “We have to get control of costs,” he said.
If genomic medicine only adds cost, it will worsen rather than lessen this problem. Furthermore, genomic testing may not be covered by many of the insurance plans offered, said Calonge, even under health care reform.
Calonge also pointed out, as did James Evans (Chapter 2), that well more than half of all deaths result from causes that do not require genetic testing. Many are simply from old age. In Colorado, for example, of 30,000 deaths annually, 6,500 are from cardiovascular disease; 1,500 are from stroke; and 6,500 are from cancer, including 1,500 from lung cancer, 500 from colon cancer, 500 from breast cancer, and 40 from cervical cancer. Calonge estimated that more than half of these deaths could be prevented through efforts to control tobacco use, obesity, high cholesterol, and hypertension and by screening for colon, breast, and cervical cancer. Genomic testing could produce incremental improvements in health, but these improvements are much less beneficial than those available by applying what is already known.
Calonge concluded by pointing out that most genetics research is focused on discovery and the process of converting discovery to applications. Very little is devoted to the development of guidelines, the conversion of guidelines into practice, or converting practice into health impact in communities. Genomic medicine will fail if it rarely gets beyond discovery and applications to public health impact, Calonge said.
Hospital administrators need to make decisions on many legitimate requests, said Herbert Pardes, professor of psychiatry and former chief executive officer of New York Presbyterian Hospital. Cost is a critical factor in making these decisions, but it is not the only one. For example, teaching hospitals provide a large amount of care without reimbursement, whether through caring for indigent patients or by paying for promising innovations so that they can be developed to the point of reimbursement.
Hospitals have been deciding how much to invest in new genomic technologies for many years, but the situation is now beginning to change dramatically, according to Pardes. As the cost of sequencing drops and as electronic medical records are integrated into care, many advances are possible. Nevertheless, major questions remain. Who will pay for these advances? How much will they pay for them? How much of an impact will genomics have on improving clinical care? How much information will patients want? Will concerns about privacy be adequately addressed? Pardes also pointed to a practical problem: hospitals have been reluctant to
perform genetic testing for inpatients because of the current reimbursement structure. Still, patients and providers want the information that testing can provide.
Data on the comparative effectiveness of genomic testing will take at least a decade to accumulate, Pardes observed, which will delay advances in patient care with this technology. Furthermore, few sources of funds are available to pay for this research, whether from the public or private sectors.
Genomic testing in oncology is an area in which hospitals are deeply involved, and it is rapidly becoming commonplace. Hospitals know that if they do not offer state-of-the-art care, patients and physicians will go elsewhere. Furthermore, building a successful medical genomics program requires gaining experience and developing expertise. Hospitals need to be able to generate, interpret, store, and reinterpret genomic data over time and communicate this information to providers and patients, said Pardes. Genomic information also needs to be integrated into the electronic medical record with appropriate on-demand education and decision support.
Faced with this welter of new information, hospitals have adapted by launching pilot programs. For example, Columbia Presbyterian and Weill Cornell have focused on colorectal cancer, which represents about three-quarters of the gastrointestinal malignancies seen at the hospitals, Pardes said. They are using sequencing technology and bioinformatics to analyze about 50 genes identified as having the greatest potential impact on clinical care. The hospitals are investing about a half-million dollars in the first year in people and equipment to do the sequencing and interpretation in-house. In this way, they will improve inpatient care and learn how best to set up a genomic testing infrastructure and do so in a way that is financially feasible.
Another initiative involves 11 institutions in New York that have come together to create a new genome center. A strong and extensive genomic capacity will be available to all of the scientists within that initiative to understand diseases and improve health, said Pardes. At the same time, the center is designed to contribute to the overall economy of the region.
“Momentum is building with regard to the value of precision medicine,” Pardes said. “Ultimately, our intent is to navigate this system by integrating the twin goals of improving health care and reducing cost. Some may think that’s impossible; I do not agree with that. I think we can do both.”
To realize this potential, Pardes encouraged the academic community to communicate much more extensively with policy makers and politicians. The academic community needs to support research while they also emphasize the need to cut costs without hindering improvements in quality and innovation.
Pardes also encouraged the medical community to involve patients in
decisions. Patient involvement can lead to more conservative approaches to procedures. It also helps avoid imposing decisions on patients by outside groups. “When you bring patients into the discussion, good things happen,” he said.
Health care decisions are not generally based only on cost-effectiveness ratios, Veenstra remarked. A physician offering a test to a patient is trying to produce the greatest benefit for the patient. The health care system offers genetic counseling because of the value it provides for patients. These considerations apply as well in trying to reduce costs in health care. “Health economics isn’t about choosing the cheapest thing,” said Veenstra. “It’s about cutting the things that have the lowest value.”
Patient-centered research is focused on what patients need and want to know. They need to know the potential harms and benefits of a test, said Veenstra. They need to understand that a test may produce false positives and false negatives. And they need to have a role in making decisions.
Genomics is probably the most complex area in health care in which to determine value, said Veenstra. The economics of genetic testing is a “morass” and will probably remain so for at least the next two decades.
Grosse observed that cost-effectiveness is one factor in making decisions but not the only one. For example, provider and patient preferences and reimbursement systems also can drive practice. People receive some medical tests more often than cost-effectiveness considerations would dictate, but they receive those tests because patient and provider preferences override other considerations. Even where a test has very little demonstrated clinical utility, a provider may prescribe it, and a patient may agree.
Ramsey pointed out that cost-effectiveness is not a perfect tool for informing decision making, but it is a better tool than the system currently has to determine resource allocation. Most other developed countries have integrated cost-effectiveness analysis into their decision-making processes, and the United States is likely to move in that direction under the Affordable Care Act. For example, accountable care organizations are going to be making decisions about the value of interventions, and they will be promoting high-value interventions and discouraging low-value interventions. As the U.S. health care system seeks to reduce costs while maintaining or improving the quality of care, cost-effectiveness is a tool that can help do that, said Ramsey.
Integrating value into decision making will require getting more stakeholders involved. Stakeholder input is needed to determine what questions to ask and what end points to evaluate, and such input is needed not only at the beginning of a study but throughout the process. “That’s the way to
move cost-effectiveness from its current state to a state where there’s some more acceptance,” Ramsey said.
Patients need to demand value, as patient advocacy groups are increasingly doing. These advocacy groups “see families become bankrupt or [living] in severe financial straits being talked into very expensive therapies that actually offer very modest benefit,” said Ramsey. Provider groups are starting to understand the costs of their recommendations, though guideline developers have not yet accepted this reality as much as they should. As an example, Ramsey cited treatments for gastrointestinal cancers that provide very similar outcomes yet differ in cost by an order of magnitude.
One option would simply be not to cover, prescribe, or pay for new cancer therapies that offer few benefits, Ramsey observed. Other countries are doing this, but the fragmented delivery system in the United States makes it difficult for all the stakeholders to speak with one voice. Thus, multiple groups will have to understand the benefits of such an approach and advocate for it to bend the curve of rising health care costs.
In response to a question about whether government agencies should seek to place a value on a drug as it progresses through the approval and coverage processes, Ramsey said he did not think that the FDA would get into the business of approving value, “and I’m not even sure that’s the right role for them.” In addition, Congress has forbidden CMS from making decisions based on QALYs.
Finally, in his reflections on the previous day, Billings concentrated on the role of industry and its view on how value is developed and delivered. Industry does not think about value only for itself, observed Billings. It seeks to provide value to clinicians, payers, and patients when developing new tests and treatments. Nevertheless, clinicians actually represent a very diverse group. They cannot be thought of as a single entity, and no single set of guidelines will apply to every health care provider. Leaving value to be expressed solely in guidelines would be inadequate, he said. Payers are also very heterogeneous in the United States, with some willing to cover molecular diagnostic tests and others not. Industry seeks to deliver value to patients with effective treatments and provides benefit to patients in terms of access to drugs and treatments that might otherwise be unaffordable in return for participating in research. The knowledge gained from this research can then provide further value to patients’ families and the community by being a source of information that industry can use to meet patient needs better.
Industry needs some certainty about what value means to these different elements of the health care system, Billings said. Only in that way can it innovate and conduct research to meet that need. At the same time, industry needs the system to be flexible, because its methods are changing rapidly.
A large reference lab may offer between 4,000 and 5,000 kinds of tests, and few of those tests will have gone through rigorous, evidence-based, randomized clinical trials, said Billings. In recent decades, more rigorous, evidence-based standards have been developed, particularly for therapeutic interventions. But requiring that randomized clinical trials be conducted for every test is a large burden to put on innovators in very rapidly changing fields, Billings observed, especially when a test targets small patient groups.
A new model is needed for gathering evidence, said Billings. He suggested that electronic medical records systems provide a potentially valuable resource for gathering evidence about value. By combining data from electronic medical records, researchers can study even rare patient groups to determine which variations in a genome are likely to be important for a particular question.
Ramsey said that the burden of collecting evidence that decision makers want for new tests is too high and not feasible for companies. Instead, he suggested coverage with evidence development, which has been piloted at CMS and is now being considered at many health plans. It “would apply very well to this domain,” he said. Under coverage with evidence development, insurance companies agree to cover the test with the provision that additional testing is conducted to collect evidence in a structured way. Final approval then depends on whether the evidence demonstrates that the test improves outcomes compared with the standard of care. This model can be risky for test development in that the test may be rejected, which is one reason why manufacturers have been hesitant to pursue this route, said Ramsey. Nevertheless, the time may be right today for manufacturers to place greater reliance on this sort of model.
Ramsey pointed out that coverage with evidence development requires the collection of evidence, but, he said, “I would argue that we have the data.” Insurance claims are available that contain information on costs and outcomes, and electronic medical records in the near future will cover most patients. “What we don’t have is a mechanism for stitching that information together in real time or in the time that we need to get the data,” he said. Building such a mechanism is feasible, though doing so would require examining and resolving a number of issues, including privacy protections.
One participant, however, pointed to the difficulties of collecting evidence under a coverage system in which patients of like circumstances expect similar treatments. “If you want the best evidence development, it needs to be done not through coverage per se but through some kind of an authoritative body making priorities about research,” the participant said.
Veenstra agreed with the need to set priorities for research and observed that because gold standard evidence cannot be produced for every test or variant, multiple stakeholders need to come together to identify where this
type of evidence is needed. In addition, innovative approaches are needed for gathering evidence, which will require the development of infrastructure and resources to collect that evidence.
He also observed that traditional 1- to 2-year analyses to determine cost-effectiveness are not adequate. More efficient approaches are needed that may be more qualitative than quantitative in nature.
In addition, evidence thresholds are needed to determine at what point patients, physicians, and payers are willing to act on the basis of the evidence in hand, said Veenstra.
A participant pointed out that molecular diagnostic testing is being conducted largely in academic medical centers and not in community hospitals, where most cancer patients get care, particularly minority patients. Enhancing the relationships between academic and community hospitals could allow greater access for patients to this type of testing.
Another participant noted that genomics did not create disparities and that it is not going to solve them. The focus should be on whether genomic technologies will exacerbate or ameliorate existing disparities. One issue is the lack of reference genome sequences for underrepresented populations.
Offit observed that genomics can be applied presymptomatically, diagnostically, prognostically, and therapeutically. All these applications are different, and the economic analyses of these applications are different as well.
In addition, a participant pointed out that consumers will increasingly choose the level of insurance coverage they want. The less expensive plans are less likely to cover genomic testing, which could exacerbate disparities.
Another individual pointed to the inevitable involvement of families in genetic testing. Today’s health care system is focused on individuals, yet genomic testing often involves individuals and their parents to determine whether a given genetic variant has an effect or not. “The health care system is very poorly organized to deal with families as opposed to individuals,” he said. Calonge agreed and noted that this issue brings up additional ones regarding how to perform this testing. Many family members have different insurance companies, which makes coverage of testing difficult even in situations where the medical information may be beneficial. “This is a huge gap that I do not think anyone has quite addressed,” Calonge said.
Feero observed that a number of research needs had emerged from the workshop (see Box 7-1). He particularly highlighted the need for better methods of measuring and determining the value of patient choice and pref-
erence. Practical considerations, such as the ability to afford insurance that includes genomics coverage, need to be accounted for in these measures.
He also noted that better infrastructure is needed to measure economic-related outcomes in addition to traditional measures. Electronic medical records systems need to anticipate the need for the eventual integration of genomic information so that downstream outcomes can be captured.
In the final session of the workshop, David Veenstra and Scott Ramsey revised and reorganized the workshop themes that W. Gregory Feero presented (see Chapter 1) into four categories of research needs. This box compiles these needs as Veenstra and Ramsey presented them. These themes should not be seen as recommendations of the workshop, but they are promising concepts that warrant further discussion and possible action.
- Need for development of the evidence base—collaboration, infrastructure with clinical trials groups.
- Need for innovative approaches to the prioritization of comparative-effectiveness research.
- Determining if and how genomic sequence information modifies health care provision and patient outcomes.
- Impact of increasing accuracy of sequencing on patient outcomes and costs.
- Evaluation of proper use of family history to guide medical decision making, integrated into health information technology infrastructure.
Health Economics Methods
- Need better (quicker) approaches and frameworks for performing health economic evaluations of genomic testing.
- Evaluation of evidence thresholds for data in hand versus data that must be obtained, and cost of further research.
- Divergence of economic assessment models in public health, clinical care, and academics.
- In the setting of a disruptive technology and a zero-sum game/shrinking pool of resources, what/who will be replaced, and how will genomic interventions be funded?
Health Economics Applications
- When is genomic sequencing cost-effective? For example, only when it is performed during newborn screening with data being used over the lifespan?
- Better education of genomic scientists regarding economic analysis/integration of economic analysis and ongoing studies.
- Methods/infrastructure (including informatics) in health systems to follow downstream consequences of providing sequence data.
- Is cost reduction demonstrable? Do accountable care organizations provide a possible mechanism for more efficient health care delivery of genomic technologies?
- Study of provider preferences for provision of genomic medicine—evaluation of barriers to implementation.
- Economic incentives for test and evidence development with value-based and specific pricing versus old system (current procedural terminology code stacking).
- Determination of relative contribution of environment/setting on cost-effectiveness.
- Developing outcomes data on informed consent/study of efficient methods for patient education regarding informed consent.
- Stakeholder engagement; methodology to increase participation in clinical trials.
- Development of improved methods for assessing value/personal utility/patient preference in economic analysis.
- Potential for genomic medicine to exacerbate disparities, including applicability of information to minority populations and socioeconomic status disadvantages. Focus on interventions.