To open the workshop, keynote speaker Michael Murray, the director of clinical genomics at Geisinger Health System, described his organization’s MyCode initiative as an example of a genomic screening program, and he shared some of the lessons learned. This was followed by the first panel session, which focused on evidence considerations for integrating genomics-based programs into health care systems. Panelists shared examples of the types of clinical data and other evidence that are currently being collected by genomics-based programs at health care systems, and they considered opportunities for advancing knowledge about clinical utility. Katrina Goddard, a senior investigator at the Kaiser Permanente Center for Health Research, described some of the challenges faced in integrating genomic programs into the care delivery system at Kaiser. Bruce Korf, the Wayne H. and Sara Crews Finley Chair in Medical Genetics, a professor in and the chair of the Department of Genetics, and the director of the Heflin Center for Genomic Sciences at the University of Alabama at Birmingham (UAB) School of Medicine, shared lessons learned in implementing the Alabama Genomic Health Initiative, which is offering genomic analysis to 10,000 individuals in Alabama, returning clinically actionable results and compiling a research database and biobank. Debra Leonard, the chair of pathology and laboratory medicine at the University of Vermont (UVM) Medical Center, discussed the Genomic Medicine Program at the UVM Health Network, which intends to provide genome sequencing for all 1
million people who use the network as a way to improve patient outcomes and make care more cost effective.
The Geisinger Health System serves the rural northeast and north central parts of Pennsylvania and has recently expanded into southern New Jersey. The Geisinger MyCode Community Heath Initiative1 started in 2007 as a biobank initiative, Murray said. More than 170,000 Geisinger patients have signed up for MyCode, and more than 90,000 have undergone whole exome sequencing thus far (Carey et al., 2016). The only inclusion criteria for joining MyCode is that one must be a Geisinger patient, Murray said, and enrollees are not recruited for any health or disease parameters. Murray noted that the rural region of Pennsylvania that Geisinger serves is overwhelmingly European Americans (more than 90 percent Caucasian), which limits the ethnic diversity of the study population, though the rural location does allow Geisinger to address some aspects of socioeconomic diversity. The expansion to the Atlantic City area in New Jersey is expected to bring additional ethnic and racial diversity to the study, he said.
In 2014 Geisinger entered a collaboration with Regeneron Pharmaceuticals to launch the DiscovEHR study,2 which combines longitudinal electronic health records (EHRs) with DNA sequencing information to map genetic variation. The primary objective of the collaboration is discovery research, Murray said, and he referred workshop participants to a recent publication on gene variants and the risk of coronary artery disease as an example (Dewey et al., 2016). An important secondary objective of DiscovEHR for Geisinger is the clinical return of results to patients and their providers. The collaboration with Regeneron has enabled the whole exome sequencing and identification of secondary findings that the entire project builds on, Murray said. Support for the clinical confirmation and the return of results comes from multiple sources, including Geisinger internal funding, donor and foundation funding, and other grants. The model is not yet generalizable and sustainable, but Geisinger is working to create the evidence that would bring support. While there is currently no insurance payment for this work, payers have been supportive of covering cascade testing (identifying and screening family members of those at risk for certain genetic conditions), Murray said, as genetic testing in those
individuals is no different than genetic testing of anyone else who might warrant it based on risk.
Geisinger Return of Results Program
Exome data from the DiscovEHR study are assessed with a goal to identify secondary results of interest and importance to patients and providers. When a returnable variant is identified in the research data, it is clinically confirmed and then entered into the patient’s EHR. Then, there are three essential steps that happen once a result is entered into an EHR, Murray said. The first step is communication and counseling. The patient’s health care provider is informed of the genetic testing result 5 days before the patient is, so the provider can be prepared to advise the patient as needed. The patient is then directly notified and invited to participate further and meet with genetic counselors. A small number of patients have declined to continue participation when informed of their result, Murray said. The next step is for the patient to undergo condition-specific evaluation and management. The third step involves cascade testing of at-risk relatives, which Murray said multiplies the beneficial effect of the program since first degree relatives are at a 50 percent risk of having the same variant. The consent rate for MyCode with return of clinical results is 85 to 90 percent, and this high rate has been attributed to longstanding relationships with patients and trust of the system. It is made clear in the consent process that it is not possible to predict the impact of any results on such things as a patient’s disability and life insurance. To illustrate the process, Murray shared the case of a Geisinger patient and her family, which was recently featured in Science (see Box 2-1 and Trivedi, 2017). At this time, results are only returned to adults, Murray said. Planning is under way to expand the program to include children soon.
Results are currently returned for 76 genes which are associated with 27 conditions. The initial phases of this project have focused on the Centers for Disease Control and Prevention (CDC) public health Tier 1 conditions of hereditary breast and ovarian cancer (HBOC), familial hypercholesterolemia (FH), and Lynch syndrome.3 Within the cohort of about 50,000 individuals, 1 in 76 individuals (1.32 percent) was found to have a significant gene change that is associated with one of these three conditions, a frequency that Murray noted was higher than newborn screening, which produces a positive result to 1 in 800 individuals. Although the published literature suggests that the prevalence of these three Tier 1 conditions would be lower, Murray believes that this is a very conservative estimate of the
number of people with genetic variants associated with risk for these conditions. Extrapolating to the total population of the state of Pennsylvania (12.8 million people), screening would be predicted to identify more than 150,000 individuals with positive genomic screens for these three conditions. The return of results process for the MyCode program is summarized monthly on the Geisinger website.4
The health care system is already set up to routinely screen for and prevent key elements of the CDC Tier 1 conditions (i.e., breast cancer, coronary artery disease, and colon cancer) without using genomic screening, Murray noted. Genomic screening could identify a subset of individuals who are at high risk for these conditions and who might benefit from intensive screening and management. Approximately 80 percent of those who were identified as having a BRCA1 or BRCA2 mutation as part of the MyCode screen had not been tested as part of their prior routine health care, Murray said. About half of those individuals met the criteria for genetic testing but had not been tested, and about half did not meet the criteria for testing. During the course of their routine care, about 20 percent had been offered and had received genetic testing for BRCA1 and BRCA2. In other words, out of 50,000 individuals with an average of 14 years’ worth of EHR data at Geisinger and hundreds of laboratory and other diagnostic tests on record, only one in five who screened positive for a BRCA variant had been previously identified through clinical testing.
Geisinger has been working to build the infrastructure to support the return of results, which includes, for example, a clinical genomics team, oversight committees, telemedicine, condition-specific multidisciplinary clinics, a family history tool, patient-centered genomics reports, condition-specific educational modules for clinicians, EHR tools, a provider liaison, and a cascade testing facilitator. Leadership within the Geisinger system, including the current chief executive officer, the former chief executive officer, and the chief scientific officer, is an important factor driving the genomic screening program, Murray said.
Lessons Learned from Genomic Screening at Geisinger
Based on the Geisinger MyCode experience, Murray said, there are several lessons learned that will help in planning for integrating genomic screening into health care delivery systems. First, genomic screening makes some invisible risks visible, as the family health history example of the 57-year-old woman who tested positive for a BRCA2 variant shows. Second, traditional pretest genetic counseling for everyone undergoing screen-
ing will be difficult, given the nature of screening (i.e., the number of individuals expected to have a positive result) and the time needed for returning results to each participant. Third, unless or until it becomes a frequent event for a practitioner, primary care providers will, in most cases, defer patient management of screening results to specialists due to both practical time and expertise constraints. Finally, genomic screening will provide opportunities to correct clinical misattributions.
Genomic Screening Makes Invisible Risks Visible
Looking at FH as an example, Murray said that more than 40 percent of those in the 50,000-person MyCode cohort who screened positive for gene variants associated with FH had a low-density lipoprotein (LDL) cholesterol level below the typical cutoff of 190 that is used for diagnosis. This indicates that there is a large group of individuals who have the genetic change who would not be picked up within the health care system without this genetic finding, Murray said. Similar results have been published by others (Khera et al., 2016). Large studies have demonstrated that approximately 15 percent of people who present in emergency rooms or to health care providers with a myocardial infarction or acute coronary syndrome do not have identifiable risk factors for those conditions, and it seems clear that genetic risk such as FH will explain some of these cases, Murray said (Canto et al., 2011).
In response to a workshop participant’s question about how receptive providers and patients are to receiving genetic screening information for FH, Murray said that there can be confusion about the relevance of the diagnosis of FH. People with very high cholesterol, for example, are already taking medication, and those with an LDL below the cutoff may not fully understand the disease and the need to have lower cholesterol goals than usual. In general, Murray said, the finding of the genetic variant provides an opportunity to intensify the management of these patients, and he pointed out that because Geisinger has a single EHR system, it is possible to track patients over time to determine if they are complying with recommendations for routine screening or other disease management.
Murray also described a second example in which a woman with no personal or family history received a positive BRCA result. Prophylactic oophorectomy in this patient revealed stage 1 fallopian tube cancer, which was removed. Treatment at stage 1 offers an excellent prognosis for this gene-associated cancer, which Murray said would not likely have been identified until the woman presented with symptoms at stage 3 or 4 if it had not been for the genomic testing.
Traditional Pretest Genetic Counseling Is Difficult
Before a genetic test is performed as part of clinical care, patients routinely meet with a genetic counselor and talk in great detail about what the test is, what it can show, and what the results do and do not mean. For MyCode enrollees, about 3 to 4 percent are expected to have a positive screening result returned to them, Murray said. For the other 96 or 97 percent who do not receive a positive result back, pretest counseling would not necessarily benefit them in the near term. As such, Geisinger is looking at other approaches to provide information to people without using 45 minutes of a counselor’s time for each person enrolled in MyCode.
Primary Care Providers Are Likely to Defer Management
In the initial phase of the MyCode program, 270 notifications to providers were made. Of those, Murray said, 187 patient screening results went to internal primary care providers, 76 went to external primary care providers, and the remainder went to other providers. The 270 notifications went to 184 unique providers. This means that many primary care providers hear about the MyCode project for the first time when they receive these results. In almost every case, the providers ask if the Geisinger Genomic Medicine team can manage the patient, Murray said. There is educational support, and many providers have tried to learn more as a way to encourage their patients to follow up and to guide their patients to receive the right care. However, most primary care providers are not seeing a lot of MyCode patients, and they can refer them to a specialist as needed. “Where we really need to engage with [primary care providers] is in supporting the program and in taking the next step, such as preventing cancers or heart attacks,” Murray said. He likened this scenario to how most providers in the United States do not see patients with positive skin tests for tuberculosis with any frequency and refer any such patients to an infectious disease or pulmonary specialist. For the foreseeable future, this is probably what will happen when providers see the occasional positive genetic screen, he said.
Correcting Misattribution Is Possible and It Matters
As an example of correcting misattribution, Murray described the first 14 MyCode patients who were identified as having hypertrophic cardiomyopathy gene variants. Seven had no diagnosis in their EHR and no history of having been evaluated for the condition. Seven had been evaluated, and two had been told prior to their genetic screening result that they had hypertensive heart disease causing their structural heart disease. The screening results showed that one of those people actually had non-obstructive
cardiomyopathy resulting from a genetic change, not hypertensive heart disease. This finding has value not only to the patient, but also potentially to his family members. For the second person, both genetics and hypertension were found to be contributing. Misattribution is common within diagnostic clinical care, Murray said, and genetic screening identifies genetic changes, and provides an opportunity to rectify that misattribution.
One workshop participant asked how the leadership at Geisinger was thinking about penetrance of a disease in the long term. Murray said that they take a very stringent interpretation of genetic variants, which leads to high confidence that the pathogenic variants identified are a risk for driving clinical disease. Even so, some people with a pathogenic variant will not develop disease, and genetic counselors coach patients and providers on this. As data are collected over decades to come, the percentage of non-penetrance will become clearer, Murray said, and there are systems in place to reassess variants over time should the evidence level change. While some non-penetrance might be attributable to luck, he said, some will be because of biology, and there will be interesting research opportunities to determine if there are protective variants in families.
Exploring Challenges Moving Forward
In closing, Murray highlighted three challenges as the field of genomic screening advances:
- Avoiding false reassurance. The MyCode project has had to keep reinforcing the message that “no result means no result,” Murray said. Not hearing back from the program does not mean that a patient is in the clear for a particular disease; it simply means that there was not a positive result signifying a significant risk among the genes that were screened. MyCode is a research program, and Geisinger strongly recommends that people who are at risk based on personal or family history have appropriate clinical testing done and be evaluated by the usual methods. Appropriately communicating the limitations of genetic screening will be important as the applications expand into public health or other similar settings, Murray said.
- Understanding non-penetrance. Currently, it is not possible to distinguish those individuals without disease who will eventually develop the disease from those without disease who will never develop it. Some individuals will have a risk variant and yet will go through life without ever developing the disease. It is important to make that message clear and to continue to work to better understand non-penetrance.
- Making cascade testing work. To be able to test first-degree family members where they live, Geisinger has worked with many clinical sites across the country, beyond the Geisinger system. The benefits of screening will be multiplied if systems for effective cascade testing can be implemented, but there is currently no roadmap for such testing.
Goddard described some of the challenges that Kaiser Permanente experienced with its genomic programs. Before offering several examples of these challenges, she highlighted three key areas of research that are being evaluated to inform the implementation of genomic programs within their delivery systems and listed examples of measurable outcomes of each program within the time frame of a typical research study. The first area is the potential benefits of the intervention, including patient management, information-seeking behavior of the patient, other changes to health behaviors, the potential for positive psychosocial impact, and change in health outcomes, which is typically not measurable within the time frame of a study. The second area being evaluated is the potential harms of the intervention, including negative psychosocial impact, misunderstanding, stigma or discrimination from receiving genomic results, health disparities, and costs associated with genomic tests. The third area is implementation choices for the delivery system, including impact on resources, workflow and logistical barriers, patient motivations and preferences, and the extent to which an intervention can be piloted before implementation.
Lynch Syndrome Screening
In 2011 Kaiser Permanente Northwest began a study on integrating universal Lynch syndrome screening into care for colorectal cancer patients.5 Seventy-three patients were identified within the health system to evaluate whether Lynch syndrome screening resulted in a change in care management. The population prevalence of Lynch syndrome (1 in 440) suggests that around 1,100 people should have been identified in the region. The first barrier to be overcome, then, is the fact that people with these conditions are not being identified, Goddard said.
The study found that when patients do receive a result suggesting a diagnosis of Lynch syndrome, the patients and their providers would like a
5 For more information on the study, see https://www.clinicaltrials.gov/ct2/show/study/NCT01582841?term=goddard&rank=5 (accessed January 10, 2018).
lot more support in care management. Nearly all of the 73 patients had a different primary care provider. Viewed another way, nearly all of the primary care providers likely have only one patient in their practice with this condition. Educating all of these primary care providers on the management needs for these patients is a large challenge, Goddard said, and she agreed with Murray about needing to rethink how patients with positive genetic screening results have their follow-up care managed.
Medical records were reviewed to determine what care was recommended for each of the 73 patients and how many actually received that care at the appropriate intervals (e.g., colonoscopy, endoscopy, genetic counseling, urinalysis, ultrasound, and other tests and procedures). Patient adherence (the extent to which patients received the recommended care at the appropriate intervals) was not 100 percent, and Goddard suggested that there is an opportunity to improve adherence to these recommended treatments.
Lynch syndrome screening among newly diagnosed colorectal cancer patients has now fully transitioned from a research program to a component of the care delivery system at Kaiser Permanente Northwest, Goddard said. The data from the research program provided the opportunity to show that genetic screening could be done systematically, she said. In addition, because the study randomized participants to either the usual care arm or the systematic screening approach, it was found that the majority of the people in the usual care arm (i.e., patient self-referral or provider referral to a medial geneticist) never went to geneticists and people with Lynch syndrome were not being identified.
The NextGen Study: Examining Expanded Preconception Carrier Screening
One of the projects within the Clinical Sequencing Exploratory Research Consortium is the NextGen study6 considering the impact of genome sequencing for expanded preconception carrier screening of women and their partners who are planning pregnancy in the near future. In particular, the study looked at whether there was a misunderstanding of “negative” preconception carrier screening results. In this context, Goddard explained, a negative result for a couple planning a pregnancy occurs when either the female partner was not found to be a carrier for any of the conditions, or the female partner is a carrier of at least one autosomal recessive health condition, but her partner is not a carrier for the same condition or was not tested.
This example helps to illustrate some of the potential harms that delivery systems are concerned about, Goddard said. One harm of particular concern was whether expanded carrier screening would lead to an increased use of health care services after genome sequencing (compared to those who did not receive genome sequencing or expanded carrier screening). The study assessed face-to-face, telephone, and e-mail encounters across primary care and mental health care. There were concerns about whether receiving carrier results might have some negative psychosocial consequences, Goddard said. The study found no difference between the two groups (genome sequencing versus usual care) in terms of overall use of health care services (Kraft et al., 2018). This could be reassuring to delivery systems to see that implementing an expanded carrier screening program would not lead to a significant uptake in unnecessary care services, she said. Another question that may arise is whether receiving carrier screening results will drive an increased use of mental health services, and in this particular study there was no evidence that people were seeking additional mental health services as a result of receiving their carrier results, said Goddard.
Another concern was whether women who had received negative carrier results would inappropriately decline recommended care during a subsequent pregnancy (e.g., ultrasound, amniocentesis, non-invasive prenatal testing, quad screen, other genetic testing). Again, Goddard reported, there was no difference between the two groups (genome sequencing versus usual care). Concerning refusals of pregnancy-related services, in one case the participant had declined a test because of a misunderstanding of her genetic test result, Goddard said. After a discussion with her provider, she decided to get the test that she had initially declined. This was only clear because the provider had documented it in the EHR, Goddard noted. In most cases it is challenging to understand the reasons for refusals as there is no documentation of them in the EHR.
Exploring Challenges and Opportunities Moving Forward
In closing, Goddard described several of the challenges she has encountered within her research studies and spoke about where more work may be needed to advance implementation. One of the challenges is that prospective studies have limited follow-up time during which to evaluate health outcomes, meaning that surrogates must be used. For patients, it is unclear what care can be attributed to the genetic test result as well as the reasons why care is refused. Another challenge is the lack of shared understanding of what is actionable genetic information, which Goddard said is one area where there is considerable opportunity to share information across programs and reach consensus. The Actionability Working Group of the Clini-
cal Genome Resource, or ClinGen, a National Institutes of Health (NIH) resource, is focused on defining the clinical actionability of genetic variants. The working group is attempting to provide resources to the community to help build a consensus around what is actionable for each genetic disorder under consideration or, in other words, what are the well-established, clinically prescribed interventions that can prevent disease or delay the onset of the disease, lower clinical burden, or improve clinical outcomes.7
The Alabama Genomic Health Initiative is a collaboration between UAB and the HudsonAlpha Institute for Biotechnology in Huntsville, Alabama.8 The initiative had a $2 million allocation from the State of Alabama for 2017, with an additional $2 million for 2018, Korf said. The goal of the program is to offer genomic analysis to 10,000 individuals in Alabama in order to both return clinically actionable results and compile a research database and biobank.
Participants are being recruited into one of two cohorts. A population cohort of adults not selected for any particular phenotype is being given genotyping using the Illumina Global Screening Array, which Korf noted was chosen because of the low cost and adequate coverage of genes of interest. An affected cohort of individuals expressing phenotypes suggestive of a rare disease will receive whole-genome sequencing in an effort to achieve a diagnosis. Participants in the population cohort will be tested for variants of at least the 59 genes on the American College of Medical Genetics and Genomics list, although this may be expanded as the program develops. The detection rate of pathogenic variants is estimated to be about 50 percent, Korf said. Individuals who are positive for a pathogenic variant will receive free genetic counseling and will be connected to supportive longitudinal care based on their diagnosis. Participants in both of these cohorts will also have the opportunity to participate in the initiative’s biobank effort, which will store DNA and other participant information for research use.
Initial funding was received in October 2016, and much of the period from October 2016 until May 2017 was spent getting institutional review board approval for the project and assembling working groups. Recruitment began on a pilot basis in May 2017 (with an initial 100 participants) and open enrollment began in July. More than 1,000 individuals were recruited from July to September 2017, and recruitment is continuing at the pace
of about 100 people per week. Currently, participants have been enrolled from 45 of the 67 counties in Alabama. The age span of participants is 18 to 89 years, with enrollment numbers peaking in the 50 to 70 age group. Thus far, participants are about 73 percent female and 27 percent male, and 78 percent are white. The initial racial distribution during enrollment may have been a result of recruitment being done outside a major outpatient clinic at UAB, Korf said. Recruitment is now being expanded to other areas of the state, including Huntsville, Montgomery, Selma, and Tuscaloosa, to better mirror the state population.
During the consent process, individuals are asked whether they would like the results of their analysis to be shared with a primary care provider and whether they are willing to have their samples entered into the biobank. Nearly all participants (93 percent) consent to having their samples stored in a biobank; however, about half do not want their primary care provider to receive their results. Korf said that this might be the result of concerns about insurability. The program is now at the stage where it is beginning to return results. Every person that has an actionable or pathogenic finding will have genetic counseling provided, and counselors will be working to connect those people to providers who can help them to manage any risk that has been identified, Korf said.
For 2018, in addition to expanding recruitment to other regions of the state, the initiative is working to build trust in the community through the engagement of participants and providers and through public education about genomic medicine, Korf said. The initiative will also continue to develop the biobank and genomic database to support precision medicine research for years to come. Korf acknowledged the large number of people involved in the initiative, including an oversight committee, principal investigators, and working groups on bioethics, data and bioinformatics, education, genomics, and participant and provider engagement.
Lessons Learned by the Alabama Genomic Health Initiative
In the relatively short time that the project has been under way, Korf said that his team has learned valuable lessons, which he shared with the workshop participants, including comments about how his team is thinking about these issues as they move forward.
Adding Value for Participants
It is estimated that 1 to 3 percent of participants in the population cohort might receive an actionable or pathogenic result, Korf said. Because the global screening array only picks up about half of the potential pathogenic variants, fewer than 1 to 3 percent of participants would likely receive
results, he said. Although the number of participants who will directly benefit is modest, the return of results and the potential impact of those results for an individual’s family members does seem to be the key motivator for participation. The program has considered other approaches to add value for participants if needed, such as returning ancestry information, providing carrier status, or providing information about pharmacogenetic variants that might predict how a person responds to a medication. Korf noted that returning carrier status or pharmacogenetic information to the participants might require much greater genetic counseling capacity to meet the needs of those who receive this information. Furthermore, Korf said, it is likely that a participant will have long forgotten receiving a finding of a pharmacogenetic variant if the participant needs that particular drug in the future, and he raised a concern about potential liability for providers if that result is buried in a health record or lost due to the movement of patients among health care systems.
There are various systems issues to be solved, including the need to increase the diversity of the population sampling, Korf reiterated. Another concern is that some people participate as a way of getting the genetic testing they need. The education and consent process is designed to make clear that people are participating in research, not clinical testing. Similar to points made by Murray and Goddard, Korf highlighted the education program’s efforts to communicate that a negative result does not exclude the possibility that a participant has a pathogenic variant, even among the genes being analyzed (as the pickup rate is not 100 percent). If there is a clinical indication for counseling, it should happen in a clinical setting. The education program has recognized that this is an area that needs extra attention, he noted.
The Alabama Genomic Health Initiative is not just an opportunity to generate evidence, Korf concluded. It is also an opportunity to raise awareness and educate providers and citizens across the state to more readily embrace genomics as it matures in the future.
The UVM Health Network serves about 1 million people in Vermont and northern New York. The network consists of six hospitals and
the Medical Group, which is the network-wide physician organization, Leonard said. There is an affiliation agreement among the hospital network and UVM, the UVM College of Medicine, and the College of Nursing and Health Sciences. The network has also established two accountable care organizations (ACOs),9 OneCare Vermont and AdirondacksACO, and OneCare Vermont has joined with two federally qualified health centers (FQHCs) to form the Vermont Care Organization. This is important, Leonard explained, because at the end of 2016 the state of Vermont signed an all-payer ACO model agreement10 with the Centers for Medicare & Medicaid Services. The agreement calls for 70 percent of all eligible residents, including 90 percent of Vermont’s Medicare beneficiaries, to be in an ACO or other value-based payment model by the end of 2022. Vermont is moving very rapidly from a fee-for-service model (where doing more results in more payments) to a global payment model focused on keeping people healthier, thereby reducing costs and accruing shared savings.
UVM Heath Network Genomic Medicine Program
Genotype drives phenotype, and a person’s genome contains fundamental medical information that is not being used in medical care, Leonard said. The promise of genomic medicine, she continued, is to improve patient outcomes, improve population health, and improve the cost effectiveness of care, and this promise aligns with the current health care reform agenda in Vermont. Leadership at the health care system level is important for driving genomics programs forward, Leonard said, acknowledging the support of UVM Health Network chief executive officer, John Brumsted, in supporting genomic medicine for Vermont.
The vision of the Genomic Medicine Program in Vermont is to provide “genomes for all,” that is, for all of the million or so people in Vermont and northern New York whom the UVM Health Network serves. The program has a clinical genomic medicine component, a genomic translational research component, and a genomic education component, all of them built around a number of central resources including a biobank, genome database, and health care database. When she was recruited as the chair of pathology and laboratory medicine, Leonard said, she received a half million dollars to start a genomic medicine program. An additional $2.7 million was allocated to build a laboratory for genomic medicine, with
9 Accountable care organizations are groups of coordinated health providers that take responsibility for delivering high-quality care to patients. For more information, see https://www.cms.gov/Medicare/Medicare-Fee-for-Service-Payment/ACO (accessed January 16, 2018).
10 For a further explanation of the Vermont all-payer ACO model, see https://innovation.cms.gov/initiatives/vermont-all-payer-aco-model (accessed January 16, 2018).
operational funding coming through the pathology and laboratory medicine budget from the UVM Medical Center. The genomic medicine team includes a director, four faculty members, a technical director, three technical staff, a genetic counselor and a pre-authorization specialist. Bioinformatics are handled through a partnership with PierianDx.
Clinical Genomic Medicine
The Genomic Medicine Program began genomic testing in 2016 with a cancer gene panel, the GenePanel Solid Tumor test, a screening panel for 29 actionable gene variants useful for diagnosis, prognosis, or treatment of solid tumors such as those in the breast, colon, and lung. The panel was initially ordered only by oncologists, Leonard said, but it is now being ordered by anatomic pathologists on all unresectable colon cancers, lung cancers, and melanomas (i.e., genomics has been incorporated into cancer care delivery). A rapid GenePanel for acute myeloid leukemia (AML) is currently being developed and validated. This needs to be a rapid test, Leonard said, because AML patients are very sick when they come to the hospital, and timely diagnosis is needed. A 100-gene panel for blood cancers (leukemia, lymphoma, multiple myeloma) is also being developed. In addition to cancer gene panels, a pharmacogenomics gene panel of 50 to 80 genes will be developed and will include clinical decision support built into the EHR based on the Clinical Pharmacogenomic Implementation Consortium guidelines. The next phase will move to exome or genome sequencing for inherited disorders, starting with patients with specific diseases or symptoms, such as cardiovascular disorders, neurologic/neuromuscular disorders, and unidentified inherited disorders in children. Additional patient cohorts will be added by disease type until, eventually, testing will be provided for everyone as long as the value can be demonstrated, Leonard said.
Genomics-Based Translational Research
Research on the clinical value of genomic testing is being done in collaboration with PierianDx, which is coordinating between Genospace, a cloud-based data storage and analysis platform, and Precision Health Economics, an economics research group looking at the value of precision health. This work is being funded by the UVM Health Network and the UVM Health Network Medical Group for an initial 2-year period, after which external funding will be needed, Leonard said. The available data that can be used to assess the value of genomic tests include genomic data, treatment and health outcomes data, cost data from claims and billing information, and patient demographics. For the GenePanel Solid Tumor test, for example, a recent (2013–2015) historical control group of solid--
tumor patients will be compared to current solid-tumor cancer patients who have received genomic testing. Oncology care will proceed after the gene panel testing, and patients will be grouped according to the intervention received: (1) those that have received a targeted therapy, (2) those for whom no targeted therapy was indicated, and (3) those for whom targeted therapy was indicated but not given. Data from a 36-month period will be analyzed against the historical controls and among the three intervention groups for health outcomes (e.g., progression-free survival, overall survival, tumor response) and total cost of care.
Research is also under way on the implementation of genomic medicine and on functional genomics, Leonard said. The genomic medicine implementation research will identify issues for using genomics in clinical care, develop and implement strategies to address those issues, measure effectiveness and efficiency, and analyze and use the data and information. The functional genomic research will study biological impact variants of uncertain significance by building these variants into model systems to determine the functional effects and then provide that information to those involved in clinical care.
Several ongoing genomic education activities are taking place at UVM, including an undergraduate honors college course called Controversies in Modern Genomics. For medical students, grant funding from the National Cancer Institute is being used to develop a national curriculum in genomics, and residents and fellows are learning about genomics as part of the molecular pathology rotation. To engage health care providers, 73 leaders across the UVM College of Medicine, Health Network, and Medical Center had their genomes sequenced through an Illumina Understand Your Genome program. The UVM Genomic Medicine Program is also hosting multidisciplinary conferences for health care providers to educate them about genomics through case study discussions that include researchers as well as specialists from applicable disciplines. To encourage patient, family, and public engagement and education, the Genomic Medicine Program is using a range of venues for outreach such as blogs, the press, community talks, and focus groups.
Potential Considerations Moving Forward
Evidence Generation and Data Sharing
The Genomic Medicine Program in Vermont is generating genomic data through genome sequencing, collecting other types of data (e.g.,
pathology, radiology, treatments, responses, costs), and using those data to assess whether a treatment for a patient was based on the genomic results received. While the information is not currently being shared with other organizations or health care systems, Leonard said, the genomic data could be submitted to ClinVar and ClinGen. A genomic cancer database is not yet available, nor is a place to share information related to the total cost of care. Leonard suggested that the All of Us Research Program at NIH could build a genomic medicine database to gather all of the data being generated by programs such as the UVM Genomic Medicine Program as a way to see the landscape of genomic medicine implementation across these systems.
Impact on Clinical Care
Prior to implementation of the genomic medicine program, other departments (e.g., pediatrics, obstetrics and gynecology, oncology, pathology, and laboratory medicine) were finding ways to address staffing and support for clinical genetics activities, Leonard said. They are currently developing a strategic and business planning process to better provide clinical genetics services across the UVM Health Network and to support the use of the genomic information that will be generated through the program, she added. The “genomes for all” approach to genomic testing will start with the testing of identified disease cohorts. In the all-payer model, test access is not based on an ability to pay. As such, the program does not foresee access issues based on ability to pay for the testing, and participation will be based more on patient choice. This begs the question of whether patient choice should be an option if, in a population health management model of care delivery, genomics does improve health outcomes and reduce costs, Leonard said.
Measuring Outcomes and Addressing Implementation Challenges
The important outcomes to measure depend on the purpose of the genome test, Leonard said. For cancer tumor response, progression-free survival and overall survival can be measured. For pharmacogenomics, adverse drug reactions, drug choice, and dosing adjustments can be assessed. For genetic disorders, diagnosis, secondary findings, and treatment options could be measured. The cost of care and of harms can be assessed for all. Tracking harms is not something that has been done thus far, Leonard said, but it is something that will be addressed as the program moves forward. She went on to say that if it is found that the program does not have sufficient value, it will be discontinued, although the hope is that the program will have value that can be demonstrated. Regarding challenges in implementation, Leonard said, one challenge that has been found is that oncolo-
gists are often not treating patients with the targeted therapy indicated by genomic testing.
Data Aggregation Across Programs
One workshop participant emphasized the magnitude of the amounts of data from genomic testing and the need to create a system to aggregate data from different programs, which is particularly important when dealing with rare diseases and even rarer genetic variants. Individual programs, even those that gather data from entire populations, will be limited in their ability to learn about rare conditions, Leonard said. That is, in part, why the focus of many genomic screening programs is currently on the more common, more prevalent variants and conditions. There are hurdles to overcome when aggregating data, including how to ensure data protection (i.e., patient privacy) and keeping the database updated, she said. Participants in the Alabama Genomic Health Initiative are given the opportunity to consent to having their information being used for future research, Korf said. This can include longitudinal data collection about outcomes in terms of both disease penetrance and screening and management activities undertaken as a result of the identification of the pathogenic variant. Korf also expressed support for the concept of a system for aggregating data. There is often a lack of evidence on the effectiveness of interventions that are undertaken following a genetic test result, Goddard said. In addition, when information about a specific population is not available, researchers often extrapolate from similar situations. Goddard noted the importance of defining what evidence needs to be captured when collecting and aggregating data.
Integrating Genomic Results into the EHR
A barrier to integrating genomics-based programs into health care is integrating genomic results into the EHR, a workshop participant said. In Geisinger’s MyCode program, the genomic variant results of importance are entered into the problem list of the EHR, Murray said.11 In this way, the result is always within the view of providers who might only open the chart casually or in urgent situations. For example, if a patient with a variant in a gene associated with long QT syndrome calls his or her physician on the weekend complaining of heart palpitations and instead reaches the provider
11 The problem list in an EHR is the section where the most important health concerns for a patient are listed.
on call, that provider will see the genomic finding in the problem list and send the patient to the emergency department to be evaluated, rather than telling the patient to call back on Monday. The UVM Health Network is working with Epic to develop a standardized EHR for use across the health network, Leonard said. It is also providing input as part of Epic’s molecular genomics workgroup. There is no single EHR in use across the state in Alabama, Korf said, and the Alabama Genomic Health Initiative has no control over entry of information into the EHR. Much of the communication is on paper, and the focus is on making sure anyone who has a positive result for a pathogenic variant is connected to a management plan.
Clinical Responsibility for Returning Reinterpreted Results
What is the responsibility downstream, one workshop participant asked, if there is a reinterpretation of variants or new evidence? For the Alabama Genomic Health program, Korf said, it was thought that attempting to keep participants informed on a regular basis was a daunting task. People frequently move, and there is often no way to re-contact them after their initial participation. The consent process is explicit in describing the screening as a one-point-in-time encounter, and it explains that, although knowledge is likely to change over time, the program cannot promise that it will be able to contact participants if something relevant to their profiles does change. There is an opportunity for participants to stay in touch with the program, he added, and if they want reinterpretation or would like to talk to someone about the significance of the results, that can be arranged, though the onus is on them as the participants. For the program at Geisinger, there are no systems or budget for re-contacting participants. Because all participants are in the Geisinger health care and EHR system, it could be possible to reach them in the future, but there is no promise made of a continuous review of results into the future. Systems are not yet in place for long-term re-contact, Goddard agreed, though it is also important to make sure that people understand that their results may be different if they are tested again in the future. Because the testing in Vermont is being done in a clinical setting, the plan is to build in reanalysis and send new results to health care providers, genetic counselors, and medical geneticists who would be responsible for reporting those results to participants, Leonard said.
A theme that arose multiple times during the discussion was the importance of institutional leadership support for genomics-based programs. There is an opportunity cost to investing in genomics (i.e., money that
is invested in genomics could have been invested elsewhere), and all of the program leaders have a vested interest in demonstrating the utility of genomic testing, said a workshop participant who went on to ask how the programs plan to measure the long-term added value of benefits versus harms so that their institutional leadership will continue to support the programs. The Alabama Genomic Health Initiative is not designed to be a test of the value of genomic medicine, Korf said. Rather, it is a test of the particular approach that is being tried in a state which, historically, has not had many large community health programs. Program leaders have been cautious and measured in what has been promised. If realistic expectations are set, then a set of questions can be answered, he said. Kaiser collaborates with FQHCs, Goddard said, which have extremely limited resources and approach genomic testing in terms of the opportunity costs (i.e., what else their providers could be doing for their patients instead of genomic screening). Patients and providers in those systems have emphasized the importance of equity, noting that if genomic solutions are available in other health care settings, they would like to also see them implemented within their systems. Leonard explained that she was given a 10-year window in which to demonstrate utility because, during discussions of setting up the laboratory, she suggested to leadership that in 10 years the Genomic Medicine Program would be sequencing the genome of every patient who comes through the health network. Her leadership agreed and was willing to integrate genome sequencing into the clinical laboratory in the health network so that the process would be ready when the evidence was available.
Panelists were asked to think about how to combine their data with data from other institutions with genomics-based programs in order to provide the economic evidence needed for other health care systems to initiate their own genomic programs. Perhaps, a workshop participant said, an economic model could be developed of the outcomes that would be expected in Lynch syndrome or HBOC over time in the absence versus in the presence of the data already generated by genomics-based programs. The Alabama Genomic Health Initiative would welcome an opportunity to be networked into a larger community and share data, Korf said. The array that the initiative is currently using has a lower sensitivity but an affordable cost per patient, which could potentially translate to large-scale screening if the initiative can demonstrate cost effectiveness. Different systems are doing genomic testing in different ways, Leonard said, and while some are testing in a clinical setting where there are EHR outcomes and cost data, others are not. Any coalescing of data should be among systems that have EHR data and cost data (i.e., billing and claims data) so that both outcomes and financial impacts can be measured, Leonard said.
Genomic Testing as an Essential Health Benefit
Including genomic testing as an essential covered health benefit12 would make it more widely available to more diverse populations, a workshop participant suggested. However, the participant continued, it is not clear at this time whether it might be possible to convince policy makers to include genomic testing as one of the essential health benefits that insurance plans must cover. Genomic testing is not yet at a cost point that it would be feasible to develop and implement such a policy, Leonard said. It might be more realistic to consider at some point in the future—after genomic testing has been implemented in certain settings and if there is research demonstrating the long-term usefulness of genomic testing across a population. Korf agreed and said the field is currently in evidence-generating mode. Data being collected now will be the basis for such future policy decisions. The Alabama Genomic Health Initiative was funded by the policy makers in the state, and this demonstrates a real interest on their part in the potential application of genomic testing, Korf said. The evidence-generating process that resulted in the list of essential health benefits was based on the work of such groups as the U.S. Preventative Services Task Force and on recommendations from the National Academies and CDC, Isham said. It is important to consider the type of evidence cascade that will be needed to bring policy makers to a consensus about genomics as an essential health benefit in the future.
Participant and Public Engagement
The genomics-based programs described in the session had infrastructure built in for participant and public engagement, a workshop participant observed. She asked about the vision for conducting research to inform how programs engage with and consent different populations, given the different funding models and different contexts of the programs. The Alabama Genomic Health Initiative was envisioned as an opportunity to generate evidence, Korf said, as well as an opportunity to raise awareness and educate the public and providers about the growing field of genomics. There is a significant budget for education and outreach in the program, and there are a variety of engagement activities planned and under way. (For further discussion about participant engagement, see Chapter 5.) There is probably much more opportunity for data collection than there are resources to make the most use of the data at this time, Korf said.
12 Essential health benefits are categories of services that health insurance plans must cover under the Patient Protection and Affordable Care Act. More information about the essential health benefits is available at https://www.healthcare.gov/glossary/essential-health-benefits (accessed February 14, 2018).
This page intentionally left blank.