This workshop session focused on issues for consideration by the National Health and Nutrition Examination Survey (NHANES) as it goes forward in determining guidelines for returning results. Presenters in this session included Sharon Kardia, University of Michigan; Marc Williams, Geisinger Health System Genomic Medicine Institute; Muin Khoury, Centers for Disease Control and Prevention; and Laura Beskow, Duke University. Adam Berger, Institute of Medicine, chaired this session.
THE SCIENTIFIC VALUE OF INCORPORATING
GENOMIC DATA COLLECTION INTO NHANES
Sharon Kardia talked about changes in both epidemiology and genomics during the past 15 years. Most modern epidemiology does not have within its design a replication as a litmus test for whether or not significant results have been found. Genetics needed to move to this higher level, and with that came the need for very large sample sizes and for individual-level data to be shared across researchers, she explained. Understanding of gene mutations has been aided by pooling data from large cohort analyses and by the emergence of large research consortia that have pooled hundreds of thousands of participants across nations.
As Kardia described, momentum accelerated in the mid-2000s with the resources that the National Human Genome Research Institute (NHGRI) put together from the HapMap project and the creation of dbGaP (data-
base of Genotypes and Phenotypes). As technology continues to change, it is shifting the way in which science conceptualizes its measurements. The point of doing whole-genome sequencing on epidemiological studies has not been reached, but she predicted it will come. Researchers will encounter many issues related to the handling of incidental findings. “Where we are today is back in the GWAS space waiting for new data to come online and to find the rare mutations or any mutation that could be considered functional in the epidemiological space,” Kardia said.
What is particularly promising about NHANES, in Kardia’s view, is the availability of data in the tough areas in genomic study: environmental exposures, infectious diseases, mental health, nutrition, and risky behaviors. Research has mostly stayed away from these areas because of a lack of data. There are new territories at the intersection of infectious, chronic, and environmental health outcomes that have yet to be explored because researchers cannot study them jointly. Enhanced collection and use of genetic data in NHANES, combined with the extant large sample size and high quality measures, could enable key scientific contributions, she said.
Kardia touched on several options for genomic data in NHANES. The first option is to not measure genomes at all. She said that this would waste a national treasure, could significantly delay gene-environment interaction studies, and could also delay understanding of rare functional mutations. The second option is to measure genomes and not report the data. Today, there are not enough replicated findings for most variants to report on the probability of a disease given a particular genotype, and NHANES could help develop this knowledge base. Another option is to measure genomes and only report “Bin One” actionable variants. (See Figure 2-1 for a description of binning.) She characterized this as a “potentially frustrating option” where participants might be able to know their results, but researchers could not study or report data because of confidentiality policies. She asked why, at this point in time, one would measure something that would then have no opportunity to be incorporated into genomics research.
DETERMINING WHAT DATA ARE RETURNABLE
Marc Williams reported on work by the Geisinger Health System, a large integrated health care delivery system. Geisinger has a biorepository called MyCode that is a tool for patient engagement and ongoing participation research activities. Current work includes studying how to return whole sequencing results in a clinical setting, developing standardized institutional approaches to the return of results, and research on many of
the questions related to return of results that have been articulated in this workshop, he said.
When considering the types of genetic results that might be returned to study participants, Williams suggested the usefulness of contrasting the clinical perspective with the research perspective. From the clinical perspective, actionable results should be returned; these occur in genes with known clinical effect. There are different types of such genes, he noted. Some might be called deterministic, such as a known deleterious BRCA1 mutation that confers a very high risk for an individual of developing breast or ovarian cancer. There are predisposing mutations, such as a mutation in the HFE gene that makes one more likely to develop hemochromatosis, but where the likelihood is very much lower than with a deterministic variant. He further noted some variants could convey carrier status, such as variants in the CFTR gene for cystic fibrosis. And there are pharmacogenomic variants such as those in CYP2C19 that can affect metabolism of antiplatelet drugs like clopidogrel.
Williams pointed out that nonactionable variants occur in genes that are associated with clinical conditions, but for which there is no treatment or change in care available, such as ApoE4 and Huntington’s disease. In general, the genetics community thinks that information about nonactionable variants that are found incidentally rather than as part of a diagnostic testing protocol should not generally be returned, Williams asserted, because doing so would create concern in patients, would provide no benefit, and could increase health care costs. For variants of uncertain significance that occur in genes that are associated with a clinical condition but the effect of the variant is still unknown, he suggested the approach should be “we do not know what this means at the present time, but we will stay abreast of new information and be in touch with you if our knowledge changes.”
From the research perspective, Williams said, issues involving return of results will depend heavily on the type of research study. If it is an anonymized study, he asked, do highly actionable variants warrant breaking anonymization for return? Should participants be consented for the return of results? Can and should highly actionable results be returned even if there is no explicit consent for the return? If the purpose of the study is to examine the question of return of results, he continued, what if a participant with a highly actionable result is randomized to the non-return group?
Williams posed a further question: If clinical relevance is the motivation for returning of research results, how does this process differ from returning test results in clinical care? Drawing on soon-to-be-published work in the American Journal of Medical Genetics, he offered a side-by-side comparison (see Table 6-1). In clinical care, the goal is to optimize the
TABLE 6-1 Clinical Care Versus Research
Clinical Care | Research |
|
|
SOURCE: Adapted from Williams (2014) presented at the Workshop on Guidelines for Returning Individual Results from Genome Research Using Population-Based Banked Specimens, February 10-11, National Research Council, Washington, DC. |
health of individuals, while in research the goal is the production of generalizable knowledge. In clinical care, the goal is to provide care in the best interest of the patient. Researchers want to protect participants from harm, preserve the integrity of the study, and avoid the therapeutic misconception. In clinical care, the patient has the right to access all clinical information. In research, there is no consensus or legal requirement that participants have access to research information. In clinical care, treatment takes place in the context of a provider-patient relationship. In research, the provider-patient relationship is not created through participation in a research study.
In his presentation, Muin Khoury argued that the primary issue involving the return of results in NHANES is the nature of the survey. It is a government statistical survey, not clinical care and not typical research. He suggested that the public health utility of NHANES goes well beyond gene discovery and genotype-phenotype correlations. NHANES is a unique, highly representative, population-based cross-sectional survey. In many ways it can serve as the ultimate control group for most genomic research done in the United States. The participants generally are not sick people. If one wants to find sequences or gene variants for rare conditions, one does not want to start with NHANES, he said.
Khoury addressed the potential evidentiary basis for return of results and described ongoing work at the CDC that seeks to develop an evidence-based approach for clinical and public health practice. The Evaluation of Genomic Applications in Practice and Prevention Initiative (EGAPP) Working Group is an independent, nonfederal, multidisciplinary panel, and one of its contributions has been to focus on the clinical scenarios for
which genetic testing is done and to differentiate between clinical validity and clinical utility.
Khoury noted there has been a lot of work on establishing clinical validity in terms of genotype-phenotype correlations and determining whether or not they rise to the level of actionability as far as health care. The concept of the binning of genome sequence results, described earlier by Kathryn Porter (Chapter 2), is now undergoing evolution. A prominent characteristic of Bin One right now is that it is rather empty, Khoury observed. Even if one takes the 56-gene list from the American College of Medical Genetics and Genomics (ACMG) and assumes most of the genes on the list are in Bin One, this probably reflects about one-half percent of the general population. In the context of NHANES, Khoury pointed out, that number is a small proportion of survey participants.
Khoury said that the EGAPP Working Group has developed and piloted an approach to evaluating the clinical relevance of genetic variation that is systematic, transparent, applicable, and credible (Goddard et al., 2013). The approach is to ask whether there is a practice guideline or systematic review for a genetic condition, whether the practice guideline or systematic review indicates that the result is actionable in one or more of several ways (e.g., patient management, surveillance/screening, family management), and whether the result is actionable in an undiagnosed adult with the genetic condition. This approach uses the criteria of actionability, penetrance, and significance to potentially sort conditions into and outside of Bin One (stage 1), followed by a stage 2 that involves further evidence review and expert consultation regarding potential Bin One items and then a re-sorting of these items into Bin One versus not-Bin One. Eventually, he said, the process ends up with tiers of evidence for the return of results (see Box 6-1). The 10 years of EGAPP work has now come to bear on a new project spearheaded by NHGRI and the National Cancer Institute that seeks to develop a more rigorous approach to the binning of genomic variants.
From the audience, Jeffrey Botkin raised a policy question about how to define clinical utility. Actionability is a key word, he said, but there is a huge difference between what is theoretically actionable and what might be recommended that people do with information versus what people actually do and how that impacts morbidity and mortality. For much of what EGAPP considered, there are not many data on the longer-term outcomes of actually providing that information. This raises the question, he said, “should we be satisfied with theoretical actionability, or should we set a higher standard and say we would like to have evidence that conveying this information back to people positively impacts their health?”
BOX 6-1 Tiers of Evidence Proposed by the Evaluation of Genomic Applications in Practice and Prevention Initiative (EGAPP) Working Group in the Evaluation of Clinical Actionability of Genetic Variants
FIRST TIER | Evidence from a systematic review, or a meta-analysis, or a clinical practice guideline clearly based on a systematic review |
SECOND TIER | Evidence from clinical practice guidelines or broad-based expert consensus with some level of evidence review, but using unclear methods or using sources that were not systematically identified |
THIRD TIER | Evidence from another source with nonsystematic review of evidence (e.g., GeneTest Reviews, OrphaNet, and Clinical Utility Gene Cards, opinion of a single or few (<5) experts) with additional primary literature cited |
FOURTH TIER | Evidence from another source with nonsystematic review of evidence (e.g., GeneTest Reviews, OrphaNet, and Clinical Utility Gene Cards, opinion of a single or few (<5) experts) with no citations to primary data sources |
A systematic review attempts to collate all empirical evidence that fits pre-specified eligibility criteria in order to answer a specific research question. The key characteristics of a systematic review as explicated by the Cochrane Collaboration (2011) are
- a clearly stated set of objectives with pre-defined eligibility criteria for studies;
- an explicit, reproducible methodology;
- a systematic search that attempts to identify all studies that would meet the eligibility criteria;
- an assessment of the validity of findings in the included studies, for example through the assessment of risk of bias; and
- a systematic presentation, and synthesis, of the characteristics and findings of the included studies.
SOURCE: Data from Green et al. (2008) and Goddard et al. (2013) presented in Muin Khoury’s presentation at the Workshop on Guidelines for Returning Individual Results from Genome Research Using Population-Based Banked Specimens, February 10-11, 2014, National Research Council, Washington, DC.
SURVEY PARTICIPANT ATTITUDES AND PREFERENCES
The role of participant preferences is an important factor to consider when trying to establish an appropriate NHANES policy for the return of genetic information, said Laura Beskow. In her presentation, she reviewed some of the literature regarding participant perspectives about return of results, highlighting a large online survey of U.S. adults about a proposed national genetic cohort study. One of the findings of the study was that 9 in 10 people agreed they would want to know all of their individual research results (Kaufman et al., 2008). People wanted research results about health risks even when there is nothing they could do about them. Other surveys have found similar results, she reported.1
In terms of the reasons people give for wanting this information, Beskow presented a number of common themes that emerge when looking across different studies (see, e.g., Murphy et al., 2008; Daack-Hirsch et al., 2013):
- anticipation that this information will be valuable, now and in the future;
- perceptions that respondents could then obtain treatment or have a course of action for preventing the risk to their health;
- understanding of genetic information as having benefit for family members or other relatives;
- reciprocity, the idea that if one helps with research, then one ought to expect and receive something in return;
- interest in assisting in other research or seeking out opportunities to participate in future research;
- life planning, that is, changes that might be made based on this information; and
- a general right to the information. Some people feel this is information about them and they have a right to it.
However, other studies shed light on factors that may influence what people say when asked if they want access to genetic information, Beskow explained. One factor is the very effect of asking, and the concept of involuntary curiosity. Curiosity often is triggered in people by asking them something that brings to their attention a gap in the information that they have. Similarly, for someone to learn that someone else might possess information about him or her can trigger curiosity. Reports of participant preferences are often conditioned by how questions are asked, and
__________________
1It should be noted that there is a literature on the limits of and problems with the use of surveys and polling in the formulation of ethically sound policies; see, e.g., Hausman (2004).
whether people are asked about what they would prefer to have happen as opposed to what they would find acceptable, Beskow noted.
She pointed out that when using hypothetical scenarios to elicit respondent preferences, how people respond may not accurately reflect what they actually do when the time arrives. This has been demonstrated in terms of the gap between what people have said about their interest in genetic susceptibility testing versus uptake when the test is available in clinical practice.
The context in which questions are presented is also important, Beskow stated. There is a difference between having researchers or practitioners develop a full-fledged scenario to present to people versus relying on simple verbiage around a question. Ideally, in a consent form, people are given the context of what they are being asked to do. That information influences the way that people then position themselves and perceive themselves in relation to the research. Beskow pointed out that people will acquire different expectations depending, for example, on the type of study it is, the relationship that is being offered to them with the research, and so forth.
During the discussion period at the end of the session, Gail Jarvik mentioned work done at the University of Washington with regard to participant preferences that involved semi-structured interviews with biorepository subjects. When people were asked if they wanted aggregate results, the answer was no, but when asked if they wanted their individual-level results, they generally said yes. However, Jarvik reported, when asked if they wanted their individual results if there was a significant cost to producing the results, most respondents said no, if it costs money that would otherwise be spent on research.
Further discussion centered on the concept of actionability. Robert Hauser raised the case of people who have important genetically determined characteristics, for example, people who are carriers for Huntington’s disease or have a very high probability of Alzheimer’s disease, but might not be informed of these characteristics because they are not medically actionable. Williams noted that most clinicians would not want to deal with this information because it is not medically actionable. But what if, Hauser asked, the information is actionable in other ways? If one is a carrier for Huntington’s disease and had this knowledge soon enough, one may decide not to get married or not to have children. Those are big decisions that may not be medically actionable, but they are certainly actionable. Williams commented further that there is traditional medical actionability, but there is also reproductive decision making, life planning, and other things that some characterize as personal utility. There may be
things that are very important to people that in fact exceed the importance that they place on medical actionability. According to Williams, it is incumbent on clinicians to discuss this with their patients, to understand their preferences and fears, and to contextualize results in that sense. This extends to research settings as well, he suggested.