Establishing a curated genomic-variant database is essential for bringing genomic medicine to fruition, a number of speakers said. However, many existing databases contain frequent errors, conventions are often different across databases, and funding for databases can be difficult to acquire. Madhuri Hegde of Emory University said that one approach would be to encourage individual laboratories to take responsibility for curating the data on particular genes. In that case, laboratories could develop and apply expertise to curation rather than simply dumping data into databases. However, such laboratories will need to have lines of communication with other laboratories and groups developing databases in order to convey information about important variants. They will also need funding, said Hegde, to carry out the responsibilities of curating and communicating the data.
Mark Boguski of Harvard Medical School countered that individual grant-supported databases do not constitute a sustainable business model and will not be able to achieve the kind of long-term quality and oversight needed for clinical applications. Henry Greely of Stanford University agreed and asked if, instead, a non-profit organization could examine genetic information, publicly disseminate a proposed interpretation, and then welcome comments from experts. It would be, he suggested, “almost like a continuous consensus conference that is always looking at the genome but in a transparent way, so that experts from around the world can see what they recommended, why, and try to change their minds.” Such an organization
could be funded both through grants, perhaps from a private funder, and through a small fee for the use of curated data at a national or an international level, Greely said.
Heidi Rehm of Partners HealthCare pointed out that the way to reduce errors in databases and in interpretations is to generate much more data so that results are more statistically significant. Furthermore, different kinds of models may exist simultaneously. There may be a role for databases with limited access for a small group of contributors who are willing to curate data at a much higher level, for example, or different levels of curation could be performed on an individual, an expert panel, or a professional guidelines level. “There are a lot of ideas we have to work through to figure out how to do this,” Rehm said.
Boguski recommended that the field move away from literature-based curation efforts because of the large number of errors and bias in publications. Instead, he suggested carrying out a large-scale sequencing effort that would re-compute all gene–disease associations from scratch once sequencing costs are low enough to do so.
Several of the panelists discussed whether the National Center for Biotechnology Information (NCBI) is an appropriate organization to maintain a genomic database. Rehm observed that NCBI has not had much experience with clinically curated environments, but she said that it “has done an amazing job at providing resources in real time to the community, and we should leverage that.” Boguski countered by noting that NCBI is not a health care delivery organization but rather a library that stores archival data that people use infrequently. “Working in a place where you see patients being wheeled past you on gurneys everyday is different than sitting at a data center somewhere and making decisions” about what is clinically relevant data, he said. Michael Christman of the Coriell Institute for Medical Research said that the data likely will end up residing in multiple places. What is important, he said, is that data have the equivalent of an audit trail so that it is possible to keep track of who sees data when and for what purpose. People may be reluctant to have a for-profit entity house their data, he said, although as Hegde pointed out, many clinical laboratories would be willing to pay a fee to get access to it. The problem is not who houses the database but rather who is going to curate it and whether the database is clinically validated.
Greely observed that an international organization that is not funded or run by any one country may have advantages, especially since countries like China are making significant investments in genomic research. Rehm added, however, that international projects often struggle with differing laws concerning what data may be put in a public environment. Bruce Korf of the University of Alabama at Birmingham asked whether a collaboration of professional societies could play a role, since that is where many of the
consensus discussions take place. But Federico Monzon of the Methodist Hospital observed that the size of the task for professional societies is “daunting,” especially since most of the work done for professional societies is performed by volunteers.
“We are talking about millions and millions of variants,” Rehm said. “I spend hours every day, every week, curating individual patient reports with individual variants to try and read through data and figure out what they mean. It is a difficult model to scale.” The only way to proceed, she said, is to start with the most common variants and gradually work toward the less common variants.
Hegde said that the development of databases should be carried out via a unified, rather than a fragmented, effort.
Funding will be a huge challenge, Rehm said. Many laboratories do not have the resources to place their data in the public domain even though they have agreed to do so. Grant support may be necessary to move data into the public domain so that experts can be engaged to curate it.
Building and curating a clinical variant database will be a long, multi-step process, Debra Leonard of Weill Cornell Medical Center said. A clinical-grade genome sequence and phenotype repository is needed first, and the curation at that point will revolve around collecting the proper information about the data being deposited. A clinical variant database can then be derived from those data by grading and assessing the sets of sequence and phenotype information in order to build decision-support tools. Breaking down the process into these steps may provide a way forward in developing this resource since groups may be willing to collaborate to create and maintain the repository, she suggested.
Maintaining databases for cancer variants raises somewhat different issues than maintaining databases for germline variants. As Monzon observed, sequencing cancer genomes also uncovers germline sequence information, but in sequencing cancer genomes there tends to be a much more direct link between acquired mutations and the disease. “There are intrinsic differences in the data that you are generating about somatic mutations … that would warrant having a separate database that you could interrogate very specifically for interactions with drugs, the ability to treat [the cancer], or even to prevent the development of tumors,” Monzon said. Boguski observed that including cancer genomes in a master database could create difficulties. People with Stage IV melanoma who have a life expectancy of 6 months are not going to be worried about a hypothetical risk of Alzheimer’s disease in 20 years. “To comingle presymptomatic genetic test databases with things that are directed at solving an acute clinical problem, I think, is a mistake,” he said.
Rehm countered that variants can be important in both acute clinical problems and long-term problems, which would argue for putting all the data in one place. But the data should be annotated to specify whether a variant has somatic effects, germline effects, or both. If that were done, everything that people would want to know about genetic variation would be available in one place. Furthermore, Korf said, other diseases are likely to involve variants in somatic cells, which will increase linkages to germline variants.
Muin Khoury of the CDC asked how genomic medicine can progress from clinical validity to clinical utility. How will it be possible to document the value added and the social benefits of genomic information in large populations? Will there be a protracted period during which researchers and clinicians may not know enough to offer useful advice? (See Box 8-1.)
Nicholas Schork of the Scripps Translational Science Institute responded that there are probably just a handful of common diseases where enough is known to predict who is more likely to develop the disease, such as type 1 diabetes. But type 2 diabetes and other common diseases remain difficult to predict from genomic tests alone, partly because the increased risks demonstrated by tests today are not sizable enough to convince most people to change their behaviors. Still, Schork said, there may be various other advantages to having such genomic information. For some patients, having this information is empowering, even if very little can be done to change that risk. “There may be individuals out there who would like that information and capitalize on it in whatever way that they see fit.”
Christman argued that it will be a major development when sequencing costs are lowered to the point that the technology is widely used. Even if the clinical utility is limited at present, it will grow as extensive databases are developed from much larger cohorts.
Robert Nussbaum of the University of California, San Francisco Medical School argued for pilot projects “because I don’t expect to actually find something in every person.” Nevertheless, enough is known to move forward even if most people do not have actionable findings. Furthermore, genomic findings are complete and enduring, so a person’s genomic data can be useful whenever something changes in a person’s life.
Monzon pointed out that the experience with cancer, where conditions that once seemed to be the same can now be divided into molecular subtypes that call for different treatments, may be extended to many other diseases. In that case, genomic medicine should be used on a populationwide basis to spread its benefits as widely as possible.
• The genomics field should take a systems approacha to whole-genome sequenc ing, which will require important changes by government, health care providers, and patients. (Greely)
• There should be more collaboration between clinical entities and laboratories, a greater emphasis on the fact that some parts of the genome will remain refractory to analysis, and NIH funding to laboratories to establish databases that can be used to refine and deliver genomic medicine. (Hegde)
• Informatics capabilities should be leveraged to create clinical genotype– phenotype databases, education should be improved, and reimbursement should be set at levels that make it possible for the health care system to do analytical thinking about how best to serve patients. (Monzon)
• There should be greater interoperability of medical records systems so that information relevant to care follows people throughout life and so that genomic information can act as an incubator for innovation. (Korf)
• A public–private partnership should be established to test the full introduction of genomic information into health care to demonstrate its feasibility.(Nussbaum)
• A universal health care information technology system should be established that includes both genetic and clinical information, and barriers to data sharing should be reduced. (Rehm)
• There should be funding for education, novel research to explore gene– phenotype relationships, and improved sequencing technologies. (Schork)
• More emphasis should be placed on genetics and genomics in medical schools. (Christman)
• A study should be carried out to determine whether patients can be encouraged to use genomic information in productive ways. (Shelton)
• Intellectual property issues surrounding gene patents should be resolved, whether through licensing, litigation, or legislation. (Boguski)
a Systems approach is defined as an interdisciplinary method of study that involves consideration of all the components involved in a process and their interactions with each other.
Greely agreed that genomic medicine may not be cost-effective today but said that it may become cost-effective soon as costs go down and efficacy goes up. Furthermore, people are interested in the technology. The best approach for scientists and clinicians, he said, may be to figure out how the technology can be used in ways that are the most helpful and least harmful.
Rehm observed that many cancer patients are involved in a research study even as they are being treated. This is “one of the reasons that cancer has made much more progress,” she said. “It is because we are doing so
much research in real time.” If information can be gathered from clinical care in low-cost ways, it should be, she said. The focus needs to be on what is clinically useful, “especially from the standpoint of what payers will reimburse and where to distribute limited funds.”
Nussbaum said that a quality improvement research model may be the most appropriate approach for genomic medicine. “Educate, implement, assess, and then go back and do it again. It is an ongoing, iterative process.” The databases and informatics tools are necessary complements to this process and their development will require usage data.
“For rare Mendelian disorders, genomics is here and it can help,” Schork said. “For cancer, I think there is no question [that genomics is useful]. For pharmacogenomics, there is [also] no question. For prevention of common chronic conditions … there has to be more research in that area.”
One workshop participant noted that some patients may not want to know the results of genetic tests. Rehm added that patients with late-onset diseases for which there are no treatments have good reasons not to want to know the results of genetic tests. Patients should be able to decide what information they want or do not want, she said. However, as information becomes more actionable, the health care provider may have to take a more active role in disseminating that information. Greely agreed that physicians will be less comfortable letting their patients refuse to learn information that may save their lives, although he also observed that they often do not have much success getting patients to change behaviors that are harmful to health.
Increasing physician interest in genomic tests when their own work may not be closely linked with such tests is a challenge, some participants noted. Korf suggested that it is necessary to provide some motivation to learn about these technologies while the field is still forming. For example, a surgeon who will be performing an anterior cruciate ligament repair may be interested in how genetic variants may affect wound healing, recovery from traumatic events, or response to pain medications. The challenge is to test these factors and demonstrate their clinical utility. “I don’t think any kind of physician or health provider is looking for more work to do unless they can be shown that it is going to somehow make outcomes better or make their jobs easier,” Korf said. Boguski added that in some cases it may be necessary for patients to help educate health care providers about the utility of genetic tests.
While cardiologists and radiologists receive ample reimbursement for reading an electrocardiogram, interpreting an echocardiogram, or reading a CT scan, the same is not true for laboratory testing. The challenge in this is the laboratory coding system, Rehm said. “It is like trying to reimburse for how many times you picked up a scalpel [instead] of the actual surgical procedure.” Monzon added that some interpretative services are reimbursed, but reimbursement decisions depend on the service and on the payer. “There is no consistency in the system.” Rehm noted that this issue is not specific to laboratory reimbursement and that the fee-for-service, as opposed to an outcome-based, reimbursement system is “a universal problem that every specialty needs to address.”
Private-sector companies, such as sequencing technology manufacturers and pharmaceutical companies, are integral players in genomic medicine, Rehm said, but those companies must recognize the differences between operating in a clinical environment and a research environment. For example, changes in technologies can undermine extensive validation that has been done with specific platforms. Having regulations in place that would establish standards for raw reagents and software systems could be helpful in this regard, Rehm said.
Hegde added that private companies are very interested in the establishment of genomic databases and that perhaps it would be possible for them to create a common fund that would support this work. Schork said that many possible business models could work and that private-sector interpretation of genomic information is likely to grow.
Bruce Blumberg of Kaiser Permanente concluded the workshop by saying that “this is going to happen… . It is not a matter of whether it should. This [workshop] was an effort to begin to flesh out some of the issues that clearly need to be solved… . It was a very good start.”