As further background for workshop discussion, the next session sought to review the range of information provided by genomic technology and its implications for the concept of clinical utility. The presenters were Gail Jarvik, University of Washington, and Benjamin Berkman, National Human Genome Research Institute. In the session, chaired by Ellen Clayton, Vanderbilt University, two recent ethical frameworks that address the issue of return of results were also considered.
Gail Jarvik talked about the current state of genetic technology and clinical utility. Genomics in medicine is still new, she began. Although being able to sequence a genome for $1,000 has not yet been reached, as some had predicted, the costs of genome sequencing have dropped substantially due to computing informatics. NextGeneration sequencing1 is much more rapid and less expensive than prior technologies, she explained, and sequencing is essential to identifying rare genetic variants. Rare variants tend to be more penetrant and more predictive of human health.
Jarvik said several important concepts around this new sequencing
1NextGeneration sequencing refers to a variety of methods for DNA sequencing originally developed in the mid-to-late 1990s.
technology should be understood. The first is coverage. When one refers to a whole-genome analysis, this does not mean that the entire genome is being fully analyzed. One cannot reliably determine trinucleotide repeat diseases, such as Huntington’s disease. In addition, there are genome regions that are of high homology to another region, and one does not know if the variant being observed is in the gene of concern or in another part of the genome that looks similar. She cited the pharmacogenetic gene CYP2D6 as a good example; although the desire to know people’s sequences for CYP2D6 is strong, the information cannot be determined using current technology because there is a highly homologous region. Many other regions of the genome are not read well for various technical reasons.
She identified a second concept, called capture. One can do a whole-genome analysis, but usually an exome or a panel of genes is analyzed, which is much less costly, in order to try to capture areas of the genome of greatest interest. She noted another advantage of capture, which is that it avoids most bacterial and viral DNA that is not important to a given analysis.
Read depth is another key concept. A whole-genome procedure may have a read depth of 20, meaning that everything has been looked at or read 20 times, and a sequence has been generated from those results. The greater the depth—one might have a read depth of 1,000 sequences for a particular region—the more certain are the results. Accuracy for common gene variants is extremely high with sequencing. Accuracy for rare variants, she said, is slightly less so, and is improved by higher read depth. Higher read depth also allows better evaluation for mosaicism.2
Copy number variation is also important, she said.3 SNP arrays are often used to determine copy number variation, as are other technologies that look for stretches of homozygosity4 across genotypes. There are numerous methods for detecting copy number variation in sequence data, but they are not perfect. They are better for smaller areas, and not as good for large runs. Jarvik explained that, clinically and in research, SNP arrays tend to be favored when looking for larger copy number variations.
The more challenging step in sequencing, she noted, is interpretation.
2Mosaicism indicates the presence of two or more cell populations with different genotypes in one individual who has developed from a single fertilized egg.
3Copy number variation refers to alterations of the DNA of a genome that result in a cell having an abnormal or, for certain genes, a normal variation in the number of copies of one or more DNA sections.
4Homozygosity is the state of having two identical forms of a particular gene, one inherited from each parent.
If something in the ClinGen5 database has a variant, or if a pathogenic variant has been clearly established, this helps greatly in interpreting results. However, there are many false positives in the literature and in the human genome mutation database, which is why the National Human Genome Research Institute (NHGRI) is investing a great deal of money in ClinGen for better annotation. Variant interpretation of items that one has never seen before can be quite challenging, Jarvik noted.
NextGeneration sequencing is primarily used in clinical panels rather than for whole exome or genome tests. As Jarvik explained, this is partly a cost issue; payers are more likely to cover panels. Panels have an advantage when one encounters unexpected results, because one does not need to order a first test, then possible subsequent tests, which was the common procedure prior to NextGeneration technology. Panels also reveal surprising results that may be thought of as incidental findings, she noted. A key question is how many incidental findings might likely be found in a genome. She pointed out that this has become important clinically, and referred to the American College of Medical Genetics and Genomics (ACMG) list of 56 genes that have been proposed for analysis and return to patients having genomic tests (Green et al., 2013).6 This list may grow in the future; the Clinical Sequencing Research Project (CSER)7 has proposed broader lists, including 118 genes that are medically actionable (Berg et al., 2013).
Jarvik highlighted other problems with reporting incidental findings. The current stock of knowledge is often insufficient for positive identification of findings, and there may be major differences in interpretation. CSER screened the human genome mutation database and concluded that 90 percent of what the ACMG said was pathogenic was not necessarily pathogenic (Dorschner et al., 2013). A larger problem she identified involves gene variants of uncertain significance; the vast majority of these
5ClinGen is a National Institute of Health-funded resource dedicated to harnessing both research data and the data from the hundreds of thousands of clinical genetics tests being performed each year, as well as supporting expert curation to determine which variants are most relevant to patient care. See http://www.nih.gov/news/health/sep2013/nhgri-25.htm [June 2014].
6There has been vigorous debate about the development of the ACMG list and other such lists, focusing in part on the processes by which recommendations are generated (see, e.g., Evans and Rothschild, 2012; McGuire et al., 2013; Wolf, Annas, and Elias, 2013). In 2014, the ACMG revised its recommendations with regard to patient opt-out of analyses of medically actionable genes. See https://www.acmg.net/docs/Release_ACMGUpdatesRecommendations_final.pdf [June 2014].
7The Clinical Sequencing Exploratory Program, initiated by NHGRI in 2010, is intended to support the development of methods needed to integrate sequencing into the clinic, and also the ethical, legal, and psychosocial research required to responsibly apply personal genomic sequence data to medical care. See https://www.genome.gov/27546194 [June 2014].
have turned out to be benign as more knowledge is accrued, yet they may be over-interpreted by clinicians.
Jarvik explained that a combination of two consortia—CSER and eMERGE—is engaged in ongoing work aimed at finding a consensus about return-of-research results, motivated by discussions of whether the ACMG 56-gene list and policies around it should apply to research. This work is moving toward agreement that researchers should offer information on actionable variants discovered, whether purposively or incidentally, in the course of their analyses. If a research finding is pathogenic and actionable, Jarvik said in closing her presentation, it should be returned.8
The workshop delved further into the return-of-results discussion with Benjamin Berkman referencing the active debate in the bioethics literature about whether there is an obligation for researchers to return incidental findings. He said while there seems to be an evolving majority view that there is some obligation, the contours of that obligation remain unclear. According to Berkman, this lack of clarity is at least partially due to the fact that there is no consensus about the principle(s) on which such an obligation might rest.
Many reasons have been cited in the literature for why there might be an obligation to disclose genetic incidental findings (GIFs) to research participants, some of which have been mentioned previously during this workshop, Berkman pointed out. These reasons include beneficence, a broad principle that researchers should have the welfare of research participants as a goal, and a more narrow duty to rescue or warn participants if they are in significant imminent danger. Berkman described different formulations of respect for personal autonomy and for participants’ right to know about their own information. There is the notion of reciprocity, the idea that investigators owe something to individuals participating in a research project. There are role-based reasons that may be similar to the doctor/patient relationship, and reasons relating to professional responsibility. Justice and fairness have been invoked in various ways, and there are other, more instrumental reasons involving legal liability, public trust in research, and the professional reputation of institutions.
In contrast, Berkman described several arguments against an obligation to return genetic incidental findings. There have been challenges to the notion that some of these broad principles—beneficence, respect for persons, reciprocity, and justice—are violated by lack of disclosure.
8For further information on genomic information and clinical practice, see Institute of Medicine (2012b).
He said it has been argued that the purpose of research is not to benefit the individual research participant but rather to produce generalizable knowledge, and that this is a fundamental difference between clinical and research enterprises. There are risks of conflating research and clinical care (the therapeutic or diagnostic misconception),9 according to this argument, as well as the very important problem of resource limitations.
Berkman described the relevant sections from two major recent efforts that address the return-of-results issue, one issued by the Presidential Commission for the Study of Bioethical Issues (2013) and one published by the American College of Medical Genetics and Genomics (Green et al., 2013). The Presidential Commission distinguishes primary findings, that is, things one is looking for, from anticipatable incidental findings, those which occur when one is looking for A and finds B, but one is reasonably sure that one will find B. An example is misattributed parentage. Then there are unanticipatable incidental findings, where one is looking for A and perhaps discovers B but there was no way priori that one could have anticipated finding C.
The Presidential Commission developed several principles related to return of results and provided a list of relevant practical considerations that Berkman summarized. The Commission recommends that, during the informed consent process, studies/researchers need to tell participants that incidental findings might arise, describe the types of findings that might arise, and state whether or not such information will be disclosed. Studies should decide in advance how to honor participant preferences (including their right not to know). It is important to develop a plan to manage anticipatable and unanticipatable findings, subject to IRB approval, according to the Commission principles. If disclosure is very difficult or impossible (which may be the case with some biobank research), researchers must justify their plans for nondisclosure. The Commission stated that there is no duty to look for secondary findings.
With regard to the ACMG, Berkman said it is important to remember that its recommendations are explicitly limited to the clinical context. Nevertheless, researchers are starting to import the recommendations into their protocols, and there is considerable debate about whether or not this is appropriate. As noted earlier, the ACMG provides a minimum list of 56 findings to report from any clinical sequence. He said ACMG thinks there
9The therapeutic misconception, a term introduced by Appelbaum, Roth, and Lidz (1983), is said to occur when a research subject fails to appreciate the distinction between the aims of clinical research and of ordinary treatment, and therefore inaccurately attributes therapeutic intent to research procedures. The therapeutic misconception may present an ethical problem in clinical research insofar as the failure to distinguish the aims of research participation from those of receiving ordinary treatment may seriously undermine the informed consent of research subjects.
is a duty to look for incidental findings, and that these findings should be delivered to a clinician who then manages the information in the context of patient-specific circumstances. They do not favor the so-called right not to know or that patient preferences should be solicited. They are arguing that beneficence outweighs personal autonomy once the data have been collected.
Berkman then described recent work that seeks to develop the first extensive national study of institutional review board (IRB) professionals and their understanding, experience, and beliefs covering a range of domains. A survey of 796 IRB members and IRB professionals in the PRIM&R10 consortium asked how they are grappling with questions about incidental findings. The intent was to understand the ethical and practical principles that they appeal to, and the extent to which they recognize limitations on a potential obligation.
As Berkman reported, almost 80 percent of respondents thought that there is always or sometimes an obligation to return incidental findings, with about 15 percent saying rarely or never. Respondents were asked to endorse different reasons why there might be an obligation, and three reasons received strong support: duty to warn, respect for autonomy, and beneficence. Reciprocity received low support, a surprising result to Berkman, given what is often encountered in the literature. The notion that research participants are equivalent to patients, and that the two groups should be treated similarly, also received low support, which tends to weaken the ACMG recommendation in the research realm. As the study showed, IRB members understand that research subjects are different than clinical subjects, and feel that something developed in the clinical context is not necessarily appropriate in the research context.
Perhaps most relevant to the workshop discussion, according to Berkman, respondents were asked what reasons or arguments might they accept to mitigate, limit, or reduce an obligation to disclose incidental findings. Only two received strong support: inadequate clinical or analytic validity, and inadequately demonstrated clinical utility. Other arguments that people cite fairly frequently as reasons why there should not be an obligation to disclose, in particular lack of funding resources and the time and effort required, got remarkably weak support. This suggests that IRBs might not accept arguments that projects involving the National Health and Nutrition Examination Survey (NHANES) data should be
10Public Responsibility in Medicine and Research is a membership organization for biomedical researchers and professionals responsible for ensuring research protections. Its principal activities include education, certification programs, public policy initiatives, and professional development programs. See https://www.primr.org/ [June 2014].
exempt from incidental findings obligations due to the difficulty in contacting participants.
During the discussion period at the end of this session, workshop participants further explored the question of whether there is an obligation to look for incidental findings. One participant noted that there are two ways not to look for findings. One is simply not to make the extra effort to look. Another is to actively not look, in other words, filter results in a way that will not produce results that one does not want to have to respond to. In a sense, the participant commented, NHANES has been taking the latter approach by not accepting protocols that would produce results that would potentially provide an obligation to disclose results. Jeffrey Botkin said he felt that it is perfectly acceptable to filter results so that incidental findings are reduced.