Information from five ongoing studies in the United States was presented in the next workshop session to understand how other population-based studies approach the issues of genomic data collection and related research findings. One prominent difference between the National Health and Nutrition Examination Survey (NHANES) and the other studies is that the latter follow people over time and thus have a longitudinal relationship with their participants, whereas NHANES is a cross-sectional survey of the U.S. population. Session presenters included Carolyn Tucker Halpern, University of North Carolina at Chapel Hill; David Weir, University of Michigan; Martha McClintock, University of Chicago; Robert Hauser, National Research Council; John Moye, National Institute of Child Health and Human Development; and Tina Hambuch, Illumina, with the session chaired by Mildred Cho, Stanford University.
Carolyn Halpern provided an overview of the National Longitudinal Study of Adolescent Health (Add Health).1 It began in 1994 with the goal of assessing the health of adolescents in the United States and understanding the different processes that contribute to health in both social and environmental contexts. The study has a core sample of about 12,000
adolescents who were teenagers in grades 7-12 in 1994-1995, with additional oversamples of race-ethnic groups and of individuals with cognitive and physical disabilities. There is a saturation sample of 16 schools where all students in the schools were invited to complete an in-home survey. An embedded genetic sample has more than 3,000 pairs of mono-zygotic twins, dizygotic twins, full siblings, half siblings, and unrelated pairs who live in the same household.
At the first wave in 1994, more than 20,000 adolescents in grades 7-12 were interviewed in their homes, and more than 17,000 of their parents were also interviewed. There have been four study waves to date, and biological data collection has increased with time. In the most recent study wave in 2007-2008, Add Health added the use of dried blood spots to capture indicators of metabolic function, immune function, and inflammation. Cardiovascular indicators included blood pressure and pulse. The study requested buccal cell DNA from the entire wave IV sample and also did a complete prescription medication inventory.
Participation in providing biospecimens has been good: 96 percent of the sample consented to provide blood spots, and 78 percent consented to archiving those bloodspots for future testing. Halpern mentioned that genome-wide genotyping of the archived specimens is under way. Add Health has returned some measurement information to participants over time, but to date it has not returned any bloodspot or genetic result data, largely because these data are viewed as experimental rather than clinical-type data.
According to Halpern, the main genetic-related issues that face this study involve questions about research versus clinical techniques, the lack of clinically actionable markers that have been used to date, the lack of guidance from researchers about balancing harm and good for different kinds of results delivered in different ways, the lack of study resources to answer respondents’ questions and provide tailored counseling if needed, and deciding what information is important to whom, given that the strength of associations may vary substantially across sub-populations.
David Weir explained that the longitudinal Health and Retirement Study (HRS),2 which started in 1992, shares some similarities with NHANES. Both studies aim for nationally representative samples and have multidisciplinary content. Both are characterized by rapid public data release, and both are fully committed to creating a national data resource for researchers.
In 2006, the HRS started collecting biomarkers. These biomarkers included dried blood spots as Add Health uses, visible performance measures, and genetic data on approximately 20,000 people. The biomarker data were stored in a repository, with no funding and no definite plans for what would be done with these data. HRS considered candidate gene models, Weir said, but decided that the GWAS (genome-wide association study) model might be most appropriate. Broad coverage of the genome would permit exploratory work on a wide range of health and behavioral phenotypes, especially the longitudinally defined phenotypes that HRS sees as its comparative advantage.
It was preferable for HRS to manage GWAS centrally and then permit access under a common set of rules. In so doing, HRS managed the samples and could preserve the repository for future use in sequencing or some type of more detailed analysis. The Johns Hopkins Center for Inherited Disease Research does the genotyping, and the University of Washington has done quality control as well as imputation to 1,000 genomes, giving the HRS about 22 million SNPs. Different types of HRS data have different release policies, Weir explained. GWAS data are considered to be in the most-restricted category, and there are rather onerous requirements to access the data. The data are now in dbGaP,3 and to date there are 76 authorized requests associated with the study.
The HRS has separate consents for the performance measures, the bloodspots, and the DNA. Consent rates in 2010 for dried blood and DNA exceeded 84 percent among white participants and ranged between 78 and 84 percent among Hispanic and African American respondents. For DNA, HRS adopted a broad general use release. The DNA specimen (saliva) is collected separately and the consent is separate, so, as Weir said, there is no confusion in the participant’s mind about what it might be used for. HRS does not promise any return of genetic results.
Weir discussed several issues that HRS grapples with, including that the clinical value of the information contained in HRS DNA samples is increasing with time. As science learns more, how far does the obligation to report extend, and how does one draw the line? A second issue he discussed involves the recognition that the genome is heritable and that return of results has implications for later generations. If one is using philosophical notions of care and obligation, does one not have an obligation to later generations? That potentially puts a study such as HRS into conflict with the privacy of its participants, he suggested. And, if a no-
3The database of Genotypes and Phenotypes (dbGaP) was developed by the National Center for Biotechnology Information to archive and distribute the results of studies that have investigated the interaction of genotype and phenotype. See http://www.ncbi.nlm.nih.gov/gap [June 2014].
return consent constitutes an agreement, then the criteria to violate that agreement by reporting should be higher than just some possible small benefit—the question is what is the standard. Where does one draw the line for a compelling need? Weir queried.
The National Social Life, Health, and Aging Project (NSHAP),4 according to Martha McClintock, seeks to describe and re-conceptualize the health of older adults by including psychological and nervous system measures of sensorimotor frailty, sensory function, cognition, emotional states, health behaviors, activity, and sleep. The study began in 2005-2006 and now includes a nationally representative sample of about 6,000 community-dwelling older adults. Overall consent rates are on par with other studies, as is biomeasure consent given that participants already are in the study. Consent to use DNA from participant-collected Oragene kits is 83 percent and reaches 94 percent for future biomeasure analyses.
McClintock explained that in-home biosamples and biomeasures are collected by nonmedical field interviewers. NSHAP collects bloodspots in order to study C-reactive protein, Epstein Barr virus, hemoglobin A1c, and other markers. The study involves older adults’ sexuality and asks about male impotence, as well as including direct measures of women’s comfort and a bio-assay of women’s estrogenization and balance with testosterone and progesterone. NSHAP also collects saliva to measure the circulating levels of these steroids. She noted NSHAP is interested in looking at genes for the receptors of these hormones and at variants for the glucocorticoid receptor, which has been shown in the past year to be associated with vulnerability or resilience to stressors that then lead to cardiovascular disease.
With regard to data reporting, McClintock said that there is limited return of nongenetic data, mainly well-established health indicators such as blood pressure and obesity and suggestions that participants consult a physician. NSHAP does not yet have policies for genetic data as DNA extraction is just now being done. She said conversations with people in the medical community, including IRB clinicians, suggest that it would be not only inappropriate but also irresponsible to report back genetic data from this kind of study. However, in conversations with social scientists, advice is more along the lines of “if you know the gene, you should report to participants.” She said that the policy for return of results in NSHAP is currently caught between these two cultures.
Robert Hauser explained the Wisconsin Longitudinal Study (WLS)5 follows a cohort of about 10,000 Wisconsin high school graduates in 1957, supplemented by data from random selections of siblings and graduates’ spouses. Almost all are white, which makes them rather like two-thirds of Americans in their cohort, but they clearly are not like everybody, he pointed out. For genetic purposes, they represent a good study population because of their homogeneity.
Hauser noted a tradition in British social anthropology of community studies in which one starts with a survey and then engages with people in the community and helps them improve their lives. He stressed, however, the WLS does not fit this model; the WLS is engaged in observation and has very rarely strayed beyond that.
The study has not collected a great deal in the way of biomarkers. He said items that might be thought of as biometric or biological include several measures of body mass index; facial characteristics such as attractiveness, smiles, and facial mass (from high school yearbooks); DNA collection with Oragene saliva kits; experience with Medicare part D; and home interviews in the latest study round (with respondents aged 71-72) that include anthropometric and performance measures. The only biospecimen collected is saliva.
WLS uses three consent forms: one for use of Medicare records, one for saliva, and one for use of Social Security earnings. Hauser noted that response rates for Medicare data linkage and saliva collection are above 90 percent, both for the graduates and for the siblings, while consent for Social Security data is about 10 percentage points lower. The consent forms provide for unrestricted use for group analysis. The labs that WLS uses are not qualified to provide feedback for any medical purpose, Hauser said, and study participants are told that they will not receive any feedback. Analyses of genomic and Social Security data are only permitted within a secure data enclave.
A number of disclosure issues have arisen in terms of questions asked and not asked, Hauser stated. The WLS once thought about asking people about suicidal ideation but decided against doing so because of the issue of informing proper authorities. Respondents have been asked whether they were abused as children, including sexual abuse, but these respondents were consenting adults and there was no disclosure issue. Some respondents have been asked about their end-of-life plans, which Hauser said raises concerns not about disclosure but about whether respondent behavior is affected by the kinds of questions that are asked. In its latest
round, the WLS asked about elder abuse, an area where there is an obligation to report, especially because some of the interviewers were licensed in professions that required them to report such incidents. Another recently added study question involves whether people have had surgeries requiring a general anesthetic, asked in a way that would permit the WLS to identify through hospital records which anesthesia was used. Hauser said this is potentially useful because of the implications for later cognitive functioning and could conceivably lead to some kind of participant feedback, which has not happened yet.
The National Children’s Study (NCS),6 now in its Vanguard or pilot testing phase, is a new national longitudinal study of environmental influences on children’s health and development authorized by the Children’s Health Act of 2000. John Moye explained that the Main Study will assess exposures and outcomes by following children to age 21. The NCS is a data acquisition resource, not a conventional study, with the majority of samples and information (including biological information) intended to be analyzed in the future.
Moye contrasted several aspects of NCS and NHANES. The NHANES has a long history, a cross-sectional design, and sample sizes in the range of 5,000 per wave. The NCS, he noted, is a recent undertaking with a longitudinal design and a planned sample size of 100,000. NCS will enroll prenatally or at birth, collect a wide variety of sample types, and request consent to store material for genetic testing from all participants. NHANES varies entry age with survey cycle, collects and stores linked blood and urine samples, and requests consent for DNA storage and genetic testing only from participants aged 20 or older. NCS focuses on pregnant women and children, while NHANES surveys a broader demographic.
Moye explained that the NCS consent booklet titled “What You Should Know About Joining the National Children’s Study Vanguard Study” specifically addresses how respondents can find out about the results of the study, and it indicates that NCS will share what is learned from the study as a whole and will provide individual information when it becomes available. The booklet mentions that most samples will be tested in the future, that the “when or which tests” are not yet known, that an oversight committee will advise on which tests may be returned, and that participants may be asked if they wish to receive results when they are available. The consent signature page says that the participant under-
6See http://www.nationalchildrensstudy.gov/Pages/default.aspx [June 2014].
stands samples may be used for a variety of tests in the future (including genetic tests), that the participant will not routinely get results back from samples they contribute, and that permission is granted to use samples to obtain genetic information.
The NCS return strategy was developed in consultation with the study’s advisory and oversight groups, the NCS Federal Advisory Committee, and the NCS Independent Study Monitoring and Oversight Committee (ISMOC). To inform the return strategy, Moye said, results were categorized according to health impact. Some results such as height and weight are simply descriptive information. Some have no clinical significance, such as routine genomic copy number variation and the salivary amylase gene. Some are clinically significant and amenable to medical interventions, such as elevated blood pressure or blood lead or gene mutations associated with colon, breast, or ovarian cancer. Others, such as Alzheimer’s disease or spinocerebellar ataxia, are clinically significant but not medically actionable at present. Some are of unknown clinical significance; for example, environmental chemicals with no recognized safety thresholds or genetic mutations not yet associated with particular health or other outcomes. The consensus, according to Moye, was that clinically significant and medically actionable results warranted return. These are serious conditions that can be mitigated by timely intervention, or as he said, “something bad that we can do something about.” The logistics of return require the participant to be recalled, permission to receive the information elicited, and arrangements made for the evaluation to be repeated, and, if confirmed, for appropriate referral to be made.
The NCS experience to date is limited to environmental contaminants and clinical chemistry obtained as part of the Vanguard pilot, he said. No genetic test results have yet been returned. An additional consultation will be planned for future analyses, and the development of the return strategy remains an ongoing process.
As part of the discussion of return of results, the workshop steering committee included experience in the private sector. Illumina is primarily known as a technology company, a maker of sequencers and chips that many people use for genomic work. Illumina also has a clinical laboratory that is Clinical Laboratory Improvement Amendments (CLIA) certified and CAP (College of American Pathologists) accredited.
Tina Hambuch, part of the group responsible for evaluating the information that emerges from the Illumina clinical lab, noted that today’s genetics-testing landscape is increasingly complex. Individual reference sequences typically involve about 3.1 billion positions, and researchers see
on average 3.3 million variants, 1.5 million of which are in gene regions. Approximately 11 percent of the genes in the genome may be clinically tested. Clinical tests tend to focus on genes that are deterministic; that is, monogenic diseases where there is high confidence that observed gene mutations are related to disease in a predictive way. There are other genes that are related to disease and have strong predictive value for the probability of developing disease, but are not absolute. Other genes have clear relationships to disease and contribute to a risk of developing disease, but those risks are each very small. Given this complexity of information, which is now being gathered at the population level, it is unlikely that everyone who is part of the analytic process has the same level of understanding of that information, she commented.
Hambuch described an Illumina-sponsored project that brings together a range of stakeholders (physicians, primary health care workers, policy makers, administrators) and engages them in the process of genetic testing, including an informed consent process that involves the stakeholders’ own doctors. Illumina does not counsel patients directly, but relies on a team of genetic counselors to support participants’ physicians and help them understand what a given test does or does not do and how it can be used. Genetic samples are sequenced to Illumina standards, results are annotated and interpreted, and reports are returned to the participant’s physician.
In this fairly healthy, high-functioning group of adults, Illumina has identified a set of 1,600 genes that correspond to about 1,200 monogenic, high-penetrance conditions. For every genome sequenced in this group of participants, there are on average 5,343 variants in the set of 1,600 genes. Researchers go through each of those variants and evaluate what they mean.
Results are returned to physicians who then presumably give that information back to their patients. Since this is more of a screen or predisposition, she said, it is questionable whether there even are secondary findings. However, Illumina follows the American College of Medical Genetics and Genomics (ACMG) guidelines for reporting on 56 genes. The clinical report only includes the pathogenic and likely pathogenic variants according to the ACMG standards. A section of the report highlights variants that are weakly suspicious for causing disease, not because they are thought to be pathogenic but because there might be reasons for the doctor and patient to make some follow-up decisions around additional testing or screening.
The report that is produced for the participants and their physicians also has an appendix that includes all annotated variant calls that pass quality scores, and the interpretations around those calls. Over time, Illumina tries to update any variants that were originally considered clini-
cally reportable and have changed status, or vice versa, and to update the physicians about those variants. As coverage of the genome improves, Illumina wants to offer people the opportunity to come back and have their genomes reinterpreted in light of technological improvements.
This process has turned what used to be series of clinical tests into a resource for additional research, Hambuch said. Some Illumina participants request their raw data as well as their complete genome-level variant calls. They may want to share their genomes with specific other individuals or donate genomes to help further research. Participants’ right to their data raises many questions, she noted. What exactly constitutes their data? Does this mean raw data?7 A clinical report? All the gene variant calls that were made in sequencing? She said there is not yet a common understanding of what people are entitled to in terms of levels of return of results. For example, if someone wishes to give results to a medical college one day and a church the next day, does this require separate consents for each use? Do the data recipients need to consent participants again? If a genome is donated for additional study, is a recipient institution required to obtain IRB input? There are no clear policies for these questions, she stated, but what is clear at least for Illumina is that its study participants would like their information back and would like to have more control over their information.
During the discussion at the end of this session, several participants raised the possibility that use of the term “return of results” may need to be more nuanced. There are three types of return that all probably require policy, a participant commented. One is investigator-initiated return or biobank-initiated return. The second is the offer of aggregate results, which may lead to further interaction with participants. A third involves participant-initiated requests for results. Various statutes and regulations come into play, particularly if a government authority is collecting information, a participant stated.
7See Lunshof, Church, and Prainsack (2014) for a discussion of issues concerning access to raw personal data.