Important Points Made by the Individual Speakers
- Incidental findings in whole genome sequences that are actionable can be a source of value rather than a liability.
- A range of options for providing different levels of sequence information exist, and all can provide benefits to patients.
- A lack of epidemiological information is often more of a factor than economic uncertainties in cost-effectiveness analyses of genomic screening.
- In receiving the results of a genetic or genomic test, patients tend not to learn about the potentially harmful effects that a test result can have.
- The expertise of physicians who specialize in particular areas and are highly qualified will continue to be an essential part of genomic screening systems.
- Much more genomic data on different racial and ethnic groups are needed.
- The costs of interpretation and delivery of information to patients need to be decreased in order to ensure equitable access to genomic technologies.
In the second scenario discussed at the workshop, the woman who underwent preimplantation screening in the first scenario has developed a health problem 5 years later:
The individual is seen at 40 years of age with progressive left lower extremity swelling and pain. Evaluation reveals an unprovoked deep vein thrombosis in her left lower extremity. She will be treated as an outpatient with low-molecular-weight heparin and warfarin. Targeted testing includes CYP2C9 and VKORC gene analysis.
Deep vein thrombosis is a common condition, said Frederick Chen of the University of Washington, when he introduced the case. About a half-million patients per year in the United States get a blood clot in the leg. About one-quarter of them end up with a clot that travels to the lung, causing a pulmonary embolism (Beckman et al., 2010). Clinicians are well trained and accustomed to dealing with this condition, which he suggested may change the bar for a new technology or medication to be used in treating a thromboembolism.
Various risk factors are associated with deep vein thrombosis in 40-year-old women, said Michael Murray, clinical chief in the Genetics Division of the Department of Medicine at Brigham and Women’s Hospital. These factors include smoking, pregnancy, immobility, extended travel, surgery, hypertension, obesity, and cancer. Genetic factors may also be involved, including the prothrombin mutation and Factor V Leiden.
Warfarin is thought to impede the synthesis of clotting factors, specifically through inhibition of the vitamin K epoxide reductase complex C1 subunit (VKORC1). Therapeutic use effectively lowers the amount of active vitamin K-dependent clotting factor by approximately 30 to 50 percent. The effects of anticoagulation therapy generally occur within 24 hours, with peak effect taking up to 96 hours.
As a result, clinicians give doses of warfarin and then have to wait 2 to 3 days to determine the effect. The goal is to achieve an international normalized ratio (INR) between 2 and 3 to minimize the risk of either clot complications or bleed complications. “It is a very unwieldy tool,” said Murray.
The U.S. Food and Drug Administration (FDA) label for warfarin explains its pharmacogenomics (Bristol-Myers Squibb, 2010):
A meta-analysis of 9 qualified studies including 2775 patients (99% Caucasian) was performed to examine the clinical outcomes associated with CYP2C9 gene variants in warfarin-treated patients [Sanderson et al.,
2005]. In this meta-analysis, 3 studies assessed bleeding risks and 8 studies assessed daily dose requirements. The analysis suggested an increased bleeding risk for patients carrying either the CYP2C9*2 or CYP2C9*3 alleles. Patients carrying at least one copy of the CYP2C9*2 allele required a mean daily warfarin dose that was 17% less than the mean daily dose for patients homozygous for the CYP2C9*1 allele. For patients carrying at least one copy of the CYP2C9*3 allele, the mean daily warfarin dose was 37% less than the mean daily dose for patients homozygous for the CYP2C9*1 allele.
In an observational study, the risk of achieving INR >3 during the first 3 weeks of warfarin therapy was determined in 219 Swedish patients retrospectively grouped by CYP2C9 genotype. The relative risk of overanticoagulation as measured by INR >3 during the first 2 weeks of therapy was approximately doubled for those patients classified as *2 or *3 compared to patients who were homozygous for the *1 allele [Lindh et al., 2005].
Certain single nucleotide polymorphisms in the VKORC1 gene (especially the 1639G>A allele) have been associated with lower dose requirements for warfarin. In 201 Caucasian patients treated with stable warfarin doses, genetic variations in the VKORC1 gene were associated with lower warfarin doses. In this study, about 30% of the variance in warfarin dose could be attributed to variations in the VKORC1 gene alone; about 40% of the variance in warfarin dose could be attributed to variations in VKORC1 and CYP2C9 genes combined [Wadelius et al., 2005]. About 55% of the variability in warfarin dose could be explained by the combination of VKORC1 and CYP2C9 genotypes, age, height, body weight, interacting drugs, and indication for warfarin therapy in Caucasian patients [Wadelius et al., 2005]. Similar observations have been reported in Asian patients [Takahashi et al., 2006; Veenstra et al., 2005].
Murray estimated that 1 percent of the U.S. population takes warfarin, with a dose range between 1 milligram and 20 milligrams per day. Providers and patients try to stay between that INR value of 2 and 3 for 3 to 6 months and often longer. “You can imagine what a struggle that is,” Murray said.
Though the FDA label provides guidance for doses depending on the CYP2C9 and VKORC1 genotypes, the decision memo from the Centers for Medicare & Medicaid Services (CMS) for pharmacogenomic testing for warfarin response states that testing of the variants will not be covered unless the patient is in a trial developing the evidence base for the use of the test. This is an example where medical science and medical practice do not correspond, said Murray. This is one of the drivers for why genotyping to predict warfarin dosing is not being done.
Murray briefly described the Medco-Mayo Warfarin Effectiveness Study (Epstein et al., 2010), which found that warfarin genotyping reduces
hospitalization rates. An unusual feature of this study was that it did not provide genotype information to physicians before making the first dose decision. Rather, it provided that information about a month later, with 70 percent of physicians being told that a dose might be too high or low and that increased monitoring was warranted. Murray warned that the reduced hospitalization might be an instance of the “Hawthorne effect”—in which changes in patients’ behaviors and health outcomes are related to the special treatment they received—and the results have not yet been replicated. The National Institutes of Health (NIH) was conducting a major study called the Clarification of Optimal Anticoagulation through Genetics trial at the time of the workshop to examine genotype-guided dosing.
Warfarin is a cheap drug, but a large infrastructure has been built up around warfarin care involving both clinics and home-monitoring structures. “Patients stay on target with their therapy better when they are monitored by experts,” said Murray, which is “part of the cost of warfarin care.”
Regarding the patient in the scenario, Murray said that he, too, would encourage her to stop smoking. The prothrombin gene mutation would give insight into predisposition but not have any specific management implications at the time of care. The variations that Murray assumed the patient had in CYP2C9 and VKORC1 would lead to lower initial dosing of warfarin, with the expectation of a lower daily dose over time. Nevertheless, she would still need to be monitored, and her therapy would still need to be adjusted. Because the deep vein thromboembolism was unprovoked, she would receive at least 6 months of warfarin and probably more. Murray also noted that at the time of the workshop, genotypic information would not be routinely available to help providers in making their decisions.
In terms of receiving additional information from whole genome sequencing, Murray observed that a fair number of people with unprovoked deep vein thrombosis have an underlying cancer. If whole genome sequencing were to reveal that she has a syndromic cancer risk, that information might be valuable in this scenario. Also, some of the other cytochrome genes are relevant in the metabolism of warfarin, and variants in those genes might play a role in deciding on a dose. Murray noted, however, that he would not have use for further incidental findings in managing this patient’s acute case and suggested that conversations about ancillary information would be deferred to a later point in time.
Wylie Burke, professor and chair of the Department of Bioethics and Humanities at the University of Washington in Seattle, pointed out that costs will be associated with incidental findings from genomic screening. That information needs to be retained in a person’s medical record for future reference and action, even if no action is taken at the time the finding is made.
Euan Ashley, director of the Center for Inherited Cardiovascular Disease at the Stanford University School of Medicine, described his evaluation of a colleague’s genome that had been sequenced as part of the Personal Genome Project (Ball et al., 2012). The colleague, Steven Quake, had asked Ashley about a variant in the myosin binding protein C gene, a gene that can be involved in sudden death in young people from hypertrophic cardiomyopathy (Maron et al., 2012). Ashley asked Quake whether anyone in his family had ever died suddenly at a young age, and Quake mentioned a cousin’s son who had. “With that, he became my patient, and somewhat inadvertently I became one of the first physicians to have [access to his patient’s] whole genome sequence.”
Ashley gathered several other colleagues from Stanford, and together they performed a thorough interpretation of Quake’s genome (Ashley et al., 2010). Shortly thereafter, Ashley was contacted by John West, chief executive officer of Personalis, Inc., who had recently had his own genome sequenced along with the genomes of several family members. (West describes his experiences in the next section of this chapter.) West had a family history of venous thromboembolism and had experienced a pulmonary embolism himself. Even after being put on warfarin, he had an unprovoked second pulmonary embolism. West was particularly interested in seeing whether he had a genetic condition that had been passed on to his children.
Ashley and his colleagues first determined whether West was a heterozygote for the Factor V Leiden mutation. This mutation is actually contained within the haploid human genome reference sequence because one of the anonymous contributors to the original sequencing project had the mutation. To cope with this and other deleterious variants that are contained within the reference genome, Ashley and his colleagues developed a synthetic reference sequence that contains the major population-specific allele at every position. They also used a newly developed technique to reduce sequencing error rates by up to 90 percent (Roach et al., 2010). These advances allowed them to build a robust platform from which to engage in genome interpretation.
Using these techniques, the Stanford group was able to deliver to West a list of potential risk alleles for a number of conditions that were shared among West, his wife, and their son and daughter (Dewey et al., 2011). The Factor V Leiden mutation had been passed from father to daughter, and other risk alleles had passed variously from the father and mother to the son and daughter. The group also put together assessments of onset, severity, actionability, lifetime risk, and variant pathogenicity for each risk allele,
rating each factor between 1 and 7. “These are very arbitrary numbers, but we had to start somewhere,” said Ashley.
“The genome has arrived,” Ashley concluded. The task ahead is to learn what to do with it.
Finally, John West provided his perspective on the genetic odyssey he and his family had taken. He had an unprovoked pulmonary embolism at age 43; the median age for such an event is 60 (Silverstein et al., 1998). After checking into the emergency room, he was told that his condition was life-threatening. “People were concerned about me even sitting up in bed, that this would dislodge the clot and could cause more serious complications,” he recalled. In the next 4 days, the hospital spent $22,000 on his tests and care. He was started on standard doses of heparin and warfarin without prior genetic testing. It was only after West was released from the hospital that he received results confirming he had a heterozygous mutation for Factor V Leiden.
His warfarin dosing turned out to be unstable even with very careful dietary restrictions and monitoring of INR. Six months later, West had a second pulmonary embolism. This incident led to suspicion that an occult cancer was causing the thrombophilia, though even after many additional tests, none was ever found. The warfarin dose gradually stabilized, however, and has been constant for more than 8 years.
On discussing his condition with his family, he found that his mother had been hospitalized for clotting in her legs when she was 40. The cause was unknown at the time, and the episode did not lead to any later screening in her children.
West had been involved with automated DNA sequencing since 1982, so he was an enthusiast for its use. When the company 23andMe started offering genotyping in 2007, he and his family had their genotyping done. The Factor V mutation was found in his mother and his daughter but not in his wife. But that test did not look at any other loci on the Factor V gene and did not assess structural variation.
In 2009 he, his wife, and their children had their whole genomes sequenced. The family then analyzed the sequences themselves, working with university groups. They found 13 other mutations in the Factor V gene. West’s daughter was found to have inherited four nonsynonymous variants from his wife along with his Factor V Leiden variant. Further analysis of the four variants in his wife revealed that they were probably benign. No structural variants were found in the gene.
All this work was done on a direct-to-consumer basis, said West, “not
because we wanted to work outside the medical system but because you couldn’t do it inside the medical system.”
When West’s daughter turned 18, she was prescribed estrogen-based contraceptives to treat her acne. His daughter refused her dermatologist’s prescription because she knew her genotypic information and the significantly increased risk of developing deep vein thrombosis and pulmonary emboli (Vandenbroucke et al., 2001). “Had she not had that information—and we got a lot of grief about having children sequenced—she would have been prescribed something that would probably have increased her chances of blood clots by 10 to 20 times,” West said. Other precautions that West and his daughter take include using Tylenol rather than aspirin; avoiding foods high in vitamin K, such as spinach and miso; and avoiding injuries that could provoke internal bleeding. “This is not a big disaster in my life,” he said. West maintains careful compliance with his warfarin dosing and has monthly INR tests to confirm the coagulation results. He noted, “This is an example where there is a genetic test where there is, in fact, a great deal that is actionable.”
West addressed the question of the “incidentalome,” or information from the genome that was not being sought and is not necessarily wanted. He described this issue as misplaced. The biggest medical issue he has had in his life is that the genetic testing was done after the deep vein thrombosis occurred. “You need to do the testing ahead of time. I do not want to be in that hospital, being told that I shouldn’t move because the clot might get dislodged. I do not want to go through 8 months of testing to try to find a cancer that apparently wasn’t there because people hadn’t really looked at my genome.” In his case, the incidentalome would have included the Factor V Leiden variation and other variants that could have led to practical and inexpensive lifestyle changes. “If I can make these kinds of changes and avoid the deep vein thrombosis and pulmonary embolism, that is where I want to be. And I think that is probably where we should explore what are the economic balances of heading in that direction,” he added.
Information on the genome can be delivered at a wide range of price points, he observed. Whole genome sequencing is still expensive, but panels that cover many variants are much less expensive. Even though people have different financial situations and insurance coverage, a range of options could deliver substantial benefits. West said that he supports consumer choice and thinks that genome sequencing should occur within the context of medical practice, yet consumers who are interested in these options often have to go around the medical system “because the medical system has not dealt with this.”
West also issued several cautionary notes. One is that genome sequences are not as accurate as people might think. False positive and false negative rates on variant detection are high and do not appear to be declining.
The technology has been improving, but the remaining errors are largely systematic, so accuracy has not dramatically improved.
Second, as with the Factor V variant, the human reference sequence contains many disease-related alleles. Also, many public databases have no mechanism to remove old data, and the errors in those databases have not been corrected, said West. Finally, no system exists to combine risks when multiple variants are known to predispose to a disease.
All these problems are solvable, West said. But he reiterated that many challenges remain in the technology arena as well as in the policy arena.
Finally, West made the point that the era of genomic testing need not cost the health care system a lot more money. If genomic screening were an add-on insurance option, like dental insurance, he would opt for it, have his family tested, and respond appropriately to any findings in which there was reasonable confidence and something actionable that could be done. “We haven’t found that there is such a lack of actionable things that we have to delve into all the uncertain and vague results.”
As more people learn about what genomic testing can offer, they will demand that the nation take advantage of the benefits to be gained, West said. “If we can solve some of these diseases and get rid of the actual burden of disease, we will have a much bigger impact on the economics of health than we will by rearranging different parts of the insurance system.”
Veenstra began his presentation by noting that the development of genomics will have an effect on the U.S. economy as a whole. It will influence job growth, the formation of companies, reimbursement policies, and health care policies in general. But the focus of the workshop, he noted, was on the value that genomics could bring to the health care system, patients, and society. Only after this value is understood can other policies be generated.
To determine the incremental value of genomic testing, researchers need to compare the use of the test with a different course of action, said Veenstra. For example, in the case of warfarin treatment, is testing being compared with treatment without testing or with no treatment at all? If testing is performed, when do the provider and patient receive the test results? What are the consequences of those results, both for future treatments and for other decisions that a provider or patient might make?
Furthermore, economic analyses need to take all of the possible consequences into account. For example, warfarin treatment is designed to mediate between the risk of bleeding and the risk of clotting in a patient. Either of those outcomes could have severe consequences for a patient, including death and long-term disability. “When you are working through
an economic evaluation, you look at that decision, and you think about every single thing that could happen to that patient. And we want to try to include all of those aspects as best we can,” Veenstra said.
In the current scenario, a CEA would take into account the costs of the tests, continued monitoring, the drug, any adverse clinical outcomes, and other factors. An outcome of this analysis might then be the cost per clot avoided or cost per life year saved. A CUA then would add an assessment of the patient’s quality of life, such as whether a patient could have a long-term debilitating outcome. This latter technique, which often measures patient outcomes in terms of QALYs, needs to account for patient preferences—for example, by conducting surveys that ask people to rate their health state if they were debilitated by a stroke—though not necessarily for whether they get genetic testing. CUAs are preferred from a theoretical perspective, observed Veenstra. Still, real-world decisions are often more influenced by CEA studies.
Grosse pointed out that the weakest links in CEAs usually involve epidemiology and clinical effectiveness rather than economics. Does an intervention actually make a difference in terms of the health outcomes? “If we don’t have evidence of effectiveness, we do not have cost-effectiveness,” he noted.
Ashley made the interesting point that relatively little evidence exists to guide treatment decisions for most of the patients he sees, even though his specialty, cardiology, has better clinical trial evidence than most fields of medicine. But these clinical trials have very tight inclusion criteria for patients so that statistically significant results can be obtained when examining small effects, and most of Ashley’s patients would not qualify for the trials and are therefore not necessarily described by the trials’ results. “The reality of medicine as we practice it is that we don’t have evidence for most of the things that we do.”
Grosse acknowledged that diverse sources of evidence, not just the results of clinical trials, are necessary to make conclusions about cost-effectiveness. But these varied sources of evidence might still not be sufficient to translate into a recommendation for coverage or clinical practice.
Veenstra agreed that evidence is a key component of being able to make a health economic assessment and that for genomics a real-world standard is being applied for how much evidence is needed. He noted, however, that stakeholders are not aligned on that standard and that the varying thresholds impact real-world policy and decision making. The question becomes, said Ramsey, “how comfortable are you with the chance that you might be making a mistake,” whether it be in providing a test that turns out to have no benefit or withholding a test that turns out to have great benefit.
One participant pointed out that medicine is much more sensitive to errors of commission than errors of omission. In other words, medicine is
much more willing to risk harm by not providing something than by doing so. Patients are not happier to die of a clot than bleeding, he said. Some of this bias results from concerns about liability, but the harms can be equivalent. “We need to be honest about assessing the whole landscape,” he said.
Also, the chain of evidence between a decision and an outcome can be long and complex, the participant continued. Most anticoagulation outcomes are secondary outcomes, such as time in range or time to stable dosing as opposed to direct outcomes of bleeding and clotting. The question is how much confidence there is that those secondary outcomes are predictive of primary outcomes. Intermediate end points such as genetic testing may be able to help bolster that confidence by assessing risk.
The participant also noted that West’s family history already captured the risk of thrombophilia regardless of the status of his Factor V Leiden gene. Not using this family history represents an opportunity cost that needs to be recognized. “If we are spending our money on genomics, then that means we are ignoring other things. And in the realm of anticoagulation, if we focus on genomics, are we ignoring the opportunities to do other things like clinical decision support with guided dosing,” he asked. Similarly, is genotyping misplaced in urban settings where there are anticoagulation clinics?
Decision points can differ on the basis of the available evidence, Grosse said. For example, considering an option has a lower threshold of evidence than deciding to choose an option. Grosse also noted that clinicians will continue to make judgments about treatments for their individual patients regardless of standards that are set across a population, given the heterogeneity of preferences and treatment effects. “It is not one-size-fits-all.”
Ramsey said that the best approach would be to have a cost-effectiveness study for the average population and then to modify those results on the basis of the characteristics of each patient. A system can also be structured to encourage the more cost-effective approaches and discourage the cost-ineffective practices, with flexibility for individual decisions on the basis of patient characteristics.
Veenstra said that changing clinical management poses its own risks. For example, monitoring of phenotypes, as with INR measurements, works fairly well, and genomic-based treatment decisions need to be compared with that standard.
Finally, Ramsey mentioned that patients may not fully comprehend the impact of receiving a test result. They may test positive and be relieved that the condition was caught in time and they can be treated or they may test negative and feel they do not have to worry about developing a genetic condition. But patients tend not to learn about the untoward effects that a test can have, such as distress or anxiety to patients and their families, the development of a false sense of security regarding risk of disease, results
being uninformative for decision making, the potential need for additional confirmatory testing for positive tests, or actual harm from unnecessary procedures based on a false positive result. “It leaves those of us on the front lines in a real dilemma, particularly when there are advocates who want that test and there is evidence that does not support that advocacy,” he said.
In response to a question, several panelists discussed the related topics of access and equity with regard to genomic technologies. In the past, investigators have often experimented on themselves or their colleagues first, because those individuals can give well-informed consent. But that approach focuses on justice as protection from research harms, not on justice as access.
Ashley observed that the analyses of both Quake’s and West’s genomes fell into the category of research as opposed to health care. These individuals were chosen first because both understood the potential benefits and harms of whole genome sequencing. Equity will be important moving forward from this point, he said. Ashley also noted that he participates in a free clinic once a month so that cardiology patients who cannot come to his clinic at Stanford can see him in an alternate setting, and “I would be as willing to entertain the idea of whole genome sequencing in that setting as I would in the clinic at Stanford.”
West pointed out that when he approached Ashley’s group, he and his family had already had their genomes sequenced. He also stated that the largest costs in the future will not be the generation of a sequence but the interpretation of that sequence and communicating that information to patients. If those costs stay high, the health care system will ration the use of the technology in some way. To provide more equitable access, researchers will need to reduce these costs, whether through Web-based tools or other means.
West was asked whether he was concerned that some of the actions he took in response to his genetic results were not based on solid evidence. West responded that he and his family sought to find the best evidence they could. They relied on advice from physicians who specialize in these areas and are highly qualified. “I expect the physician to have the judgment to know what the recommendations are. And those may be based on their personal experience in some cases,” he said. Billings agreed, pointing out that this professional expertise is always going to be a necessary part of the system and will need to be taken into account in projecting costs.
Evans observed that West’s Factor V genetic variant is a risk factor for deep vein thrombosis, but it is a modest risk factor. For example, discourag-
ing women from taking birth control pills who are heterozygous for Factor V Leiden would create many more medical complications from unwanted pregnancies than it would prevent clotting problems. “We have to think very hard before we say that we are going to tell millions of people that they are at a high risk for clotting, when the reality is that it is a very modest risk factor,” he said. In addition, as Calonge pointed out, false positives will be generated by these genetic tests in a pre-event prevention setting, and harms could be associated with those mistaken test results. He also noted that, for most individuals, just knowing you are at increased risk for a disease is insufficient to change behavior and suggested that West’s family may not be representative of the average population in adopting changes.
Another topic of discussion was the focus to date of genomic data on some populations and not others. A number of panelists agreed that much more genomic data on different racial and ethnic groups are needed and that caution should be used when interpreting information for different groups. Ashley pointed out that databases of single nucleotide polymorphisms are being developed that are much more heterogeneous in terms of populations studied.