Introduction

Modern genomic analysis (whole genome or exome sequencing and array-based assays) is helping uncover the genetic architecture of psychiatric disorders.1, 2, 3, 4, 5, 6, 7, 8 Nevertheless, genomic testing raises complex ethical, scientific and procedural challenges for psychiatry researchers, including how to manage the increasing amount of clinically relevant information these technologies can generate.9, 10 For example, consider whether researchers should offer to return findings to individual participants in the following scenarios: (1) a genome-wide association study (GWAS) that will generate schizophrenia genetic risks scores for all case and control subjects; (2) a whole genome sequencing study of women with anorexia nervosa that will yield data on proven breast and ovarian cancer risk loci (for example, BRCA1 and BRCA2; OMIM #113705; 600185) which are not known to be related to eating disorders, but could generate clinically relevant findings for which there are medical interventions that decrease the risk of poor health outcomes (for example, bilateral mastectomy, oophorectomy, chemoprevention)11; (3) a genomic study of individuals with treatment-resistant psychosis and some degree of cognitive impairment that will yield data on genetic variation in HTT12 (OMIM #143100) and PSEN113 (OMIM #607822)—rare causes of Huntington’s disease and Alzheimer’s disease for which no treatments exist but of clear relevance to the study as well as research participant’s clinical status and prognosis.

Psychiatric research has seen a marked increase in the number of array-based GWAS2, 14 and to a lesser but growing extent whole genome and exome sequencing (WGS/WES). New genomic testing tools and decreasing costs15 will lead psychiatry researchers to generate a rapidly increasing number of clinically relevant findings (Table 1). For example, for less than $40 per sample, Illumina’s Global Screening Array contains approximately 50 000 probes for variants claimed to be clinically relevant in addition to its capacity to identify large copy number variants.16 In the next three years, at least 3 million samples are expected be run on the Global Screening Array.16 Many of these samples will likely come from psychiatric research studies.14 Numerous other psychiatric genomic studies will employ other arrays and WGS/WES, which could find even more clinically relevant findings, particularly ultra-rare, damaging or disruptive exon variants,17 that are not usually practical to genotype with array-based assays.

Table 1 Key terms for the return of results debate in genomics research
Full size table

In this Perspective we examine ethical, scientific and practical considerations about what findings should be offered to participants in psychiatric genomics research. Finally, we offer a framework for making determinations about the return of results (RoR) to participants.

Clinically relevant findings in genomics research

Researchers who use brain imaging technologies often identify incidental findings (‘incidentalomas’).18, 19, 20, 21 These are unrelated to the reason the brain scan was requested (for example, to measure the sizes of brain regions) but are detected nonetheless. The clinical relevance varies. Rarely, there might be an unsuspected finding that provides a general medical explanation for a psychiatric presentation—new onset major depressive disorder with brain metastases from a primary lung cancer or findings highly suggestive of multiple sclerosis. One of the more common incidental findings (although still <1%) is the detection of an asymptomatic primary brain neoplasm (for example, meningioma). The clinical significance of other incidental findings may be uncertain (for example, an old brain infarct or periventricular hyperintensities).

Genomics research presents a somewhat similar22, 23 situation when sequencing or array-based assays are used. In the genomics arena this issue is often referred to as ‘the incidentalome’.24 The original purpose is to generate generalizable knowledge about a particular disorder based on the study of large groups of cases and controls. Whatever the original intent, these data can contain information unrelated to the purpose for conducting the study, but of clinical relevance to individual research participants. Genomic researchers struggle with how to manage these incidental findings as well as clinically relevant findings generated when examining the primary and secondary target genes (Table 1).20, 25, 26 However, recent global debate about these issues have outlined some guiding principles.

Emerging ethical consensus about the ror

Until recently, many institutional review boards limited the use of genomic data to research purposes only, and explicitly did not allow researchers to return individual findings.27, 28 As noted on Table 2, there are strong arguments for and against the RoR in genomics research. Some of the strongest arguments against the RoR are that the principal goal of research is to generate generalizable knowledge not to provide individual benefit, and there are important practical constraints such as the cost and resources necessary for the RoR. There are also technical and interpretative difficulties for determining the pathogenicity of variants, particularly novel ones identified in the course of research. Even the penetrance of variants that are known to be pathogenic may not be well estimated in the general population because most studies have examined the penetrance of these variants in clinical populations.29 Providing erroneous information regarding the pathogenicity and penetrance of a variant could lead to unnecessary treatments and harms for participants, and generate mistrust towards scientific research.

Table 2 Arguments for and against the return of genomic results to individual research participants
Full size table

Yet, for more than 15 years there has been growing support for offering some genomic findings to participants. On balance, arguments in favor of offering to return results (Table 2) led US advisory bodies and funding institutes to publish recommendations25, 56, 57, 61 in favor of offering to return findings that are analytically valid, clinically valid, medically important and medically actionable (Table 1). The US National Human Genome Research Institute, the US All of Us Research Program, the UK Genomes England Project, the Parliaments of Spain, Finland and Estonia, various Japanese government ministries, the Indian Council of Medical Research, and the H3 Africa Consortium, among others, also support participants’ access to findings.58, 59, 60, 62, 63, 64, 65, 66, 67, 68 Some regulatory bodies or advisory groups in places such as Singapore, Denmark and Taiwan do not support offering participant access to research data or findings.69, 70, 71 Overall, there is a clear movement towards favoring RoR across the globe.67

In 2014, genetic researchers and bioethicists from two US National Human Genome Research Institute-funded consortia—Clinical Sequencing Exploratory Research (CSER) Consortium, and Electronic Medical Records and Genomics (eMERGE) Network—published a consensus statement examining whether genomic researchers should adopt a policy of analyzing and offering findings about a specific set of genes to all participants. The CSER/eMERGE working group did not endorse the analysis and return of pathogenic variants in a specific set of genes, but concluded that analytically and clinically valid, medically important, and medically actionable findings should be the minimum findings offered by genetic researchers (‘minimum criteria’; Figure 1).27 The working group also concluded that because ‘resources for research should be primarily directed at scientific discovery’ researchers do not have a duty to hunt for these type of findings if they are not within the scope of the study, and that ‘participants have a right to decline the receipt of genomic results’.27 Notably, while many guidelines have been published, there are no regulations or laws that specifically address the RoR in the US and most other countries. Thus, determinations about which findings, if any, are offered are generally made by researchers and research ethics committees.

Figure 1
figure 1

Return of results framework. We propose that psychiatric genomics researchers offer to return findings generated in the course of research that meet the minimum criteria (Type 1): clinically valid, medically important, medically actionable, and identified within the scope of the research study. In our view, if resources allow, researchers should also offer to return clinically valuable findings generated in the course of research even if not medically actionable such as (Type 2) genomic findings that can help confirm or reject a diagnosis. In the example above, 22q11.2 deletion could help confirm a schizophrenia diagnosis, while NPC1 pathogenic variants may help reject such diagnosis. Similarly, we believe researchers should offer clinically valuable (Type 3) findings that suggest moderate to high genomic risks for a severe condition even if not medically actionable. Finally, we propose that researchers should offer (Type 4) likely clinically valuable findings, such as VUS potentially associated with a participant’s known symptoms. These findings will only be identified in a small subset of participants. If the results will be returned, results should ideally be corroborated by a certified clinical laboratory or some other reliable method and returned by a clinician (for example, genetic counselor) who can explain the results, implications and alternatives. ATP7B, gene associated with Wilson disease; BRCA1 and BRCA2, genes associated with hereditary breast and ovarian cancer; CNV, copy number variant; HTT, gene associated with Huntington disease; NPC1, gene associated with Niemman-Pick Disease Type C; PGx, pharmacogenetics; PSEN1, gene associated with early-onset Alzheimer disease; SCZ, schizophrenia; VUS, variants of uncertain significance.

PowerPoint slide

Full size image

Which findings should be offered in psychiatric genomics research?

RoR Type 1: minimum criteria-medically actionable

Genomic analysis in psychiatric research can generate findings that meet the minimum criteria for RoR (Figure 1; Type 1). Medically actionable findings are expected in approximately 1% of the population.72 As with genomic research for other medical conditions,27 RoR of findings that meet these minimum criteria should ideally be offered in psychiatric research given potential clinical benefits, the universally acknowledged ethical principle of respect for participants,42, 43, 44 and several other reasons (Table 2). For example, in a study of the genomics of highly treatment-resistant psychosis, we are using WGS and a SNP array to search for variants that may provide an alternative diagnosis, help explain patients’ symptoms, and offer ideas as to why antipsychotics are ineffective in these patients. We look carefully at exonic variation in ATP7B which encodes a copper transporter that is an autosomal recessive cause of Wilson’s disease (OMIM #277900; clinically valid and medically important).73 Wilson’s disease is rare but can cause a clinical portrait initially confusable with schizophrenia (therefore, any such finding is within the scope of this study). Early detection is crucial as relatively benign therapies (for example, chelation and diet) can be highly beneficial, even curative, for psychotic symptoms (medically actionable). Thus, a pathogenic variant suggestive of risk for Wilson’s disease would meet the minimum criteria and we offer RoR for these types of findings.

Some psychiatry researchers may also examine secondary targets and identify findings that meet the minimum criteria for non-brain disorders (for example, pathogenic variants in BRCA1 or BRCA2). If that is the case, the emerging consensus is RoR should also be offered to participants. For general guidance about specific genes that may generate medically actionable findings, researchers can refer to the list of genes the American College of Medical Genetics and Genomics has deemed appropriate to analyze and offer whenever clinical genomic sequencing is performed (https://www.ncbi.nlm.nih.gov/clinvar/docs/acmg/)29, 74, 75, 76, 77, 78, 79, 80 and related literature in the research context.9, 10, 81, 82 Many of these genes are for cancer and cardiomyopathies (for example, APC and adenomatous polyposis coli or MYH7 and familial hypertrophic cardiomyopathy). Several are highly relevant for clinical psychiatry and these include multiple genes for long-QT syndrome (for example, KCNQ1 or SCN5A; psychiatric medicines worsen long-QT in these individuals) and for single-gene disorders with prominent psychiatric manifestations (for example, TSC1, TSC2, and tuberous sclerosis along with Wilson’s disease (ATP7B, discussed above).

Researchers may also consider returning genetic risk scores for disorders such as schizophrenia. However, the probability that an individual with a high number of markers will develop schizophrenia is far from being deterministic83; therefore, this finding would not be considered medically important and thus would not meet the minimum criteria. It would also not meet the criteria for any of the ‘beyond the minimum’ types of findings described in the next section.

‘Beyond the minimum’ findings

Psychiatry researchers can generate findings that do not meet the minimum criteria, but there may be compelling ethical arguments to offer RoR for some of them. At present, there is a lack of clear guidance in genomics research about which ‘beyond the minimum’ findings should be offered. The CSER/eMERGE working group recognized that: ‘Researchers might be ethically and scientifically justified in returning all genomic information (the “ceiling”) in some format and any level of information in between the “floor” of actionable results identified during the course of research and the “ceiling” of all genomic information’.27 Nevertheless, the report does not offer much direction about how to make these determinations, and there is little in the history of RoR in psychiatric genomics to guide researchers.

Based on ethical and legal analysis of RoR policies, relevant bioethics literature, and our recent experience returning results in psychiatric genomics research, we propose four types of findings that we consider appropriate to offer to participants (Figure 1). In our view, in addition to Type 1 findings (actionable results that meet the minimum criteria), if resources allow, psychiatric genomics researchers should ideally also offer non-medically actionable findings, that are ‘clinically valuable’ (defined in Table 1). Such findings may include: Type 2-clinically valid findings that help corroborate or reject a psychiatric diagnosis; and Type 3-clinically valid findings that provide information about important health risks. We propose that researchers should consider offering Type 4-‘likely clinically valuable’ findings such as variants of uncertain significance (VUS) potentially related to a participant’s symptoms.

An informed consent process which clearly states that while individuals are participating in a research study, the researchers may generate and return clinically relevant information will be critical. Many potential participants may not be comfortable with having clinically relevant information managed in a research context as opposed to a clinical context where there are different regulatory protections and clinicians who have a fiduciary relationship with the individual.84 A description of the different types of findings that will be analyzed and offered will also be critical to allow participants (or their legally authorized representative) to decide which findings, if any, will have more utility for individual participants and are in their best overall interest to learn.85, 86, 87

In the US, new regulations will require studies to specify whether or not clinically relevant data may be returned and under what circumstances.88 Furthermore, researchers will be able to request ‘broad consent’ for future unspecified secondary research using identifiable biospecimens or information. This has important implications for researchers, biobanks and repositories regulated under the Common Rule. If these groups want their collected identifiable biospecimens or information to be used for future secondary research, they will need to provide ‘sufficient information to allow a reasonable person to expect that the broad consent would permit the types of [secondary] research conducted’.88 Psychiatry researchers conducting secondary research with these identifiable biospecimens or information should ideally also consider offering to return the clinically relevant results proposed here (Figure 1). To do this, investigators will need to contact their IRBs to evaluate whether RoR is permissible based on what participants were informed during the initial broad consent process.88 Therefore, researchers, biobanks and repositories will need to address the possibility of RoR in the original consent process to make it more feasible for future secondary research studies to offer RoR.

Why offer ‘beyond the minimum’ findings?

Some may argue against offering each of the clinically valuable and likely clinically valuable types of findings. Here, we describe these types of findings in more detail and explain why we propose it is appropriate to offer these. The clinically valuable findings (Types 2 and 3) are clinically valid but not medically actionable. For years, medical actionability has perhaps been the main argument for offering the RoR in research.9, 10, 27, 40, 56, 57, 61, 82, 89, 90, 91 One could argue that the research ethics principle of non-maleficence implies that if there is nothing the participant can do to reduce the risk of poor health outcomes associated with the genomic risk, researchers should not burden participants with this information. However, as described below, this genomic risk information may still be clinically valuable, and returning these findings—if the research participant or representative provided consent—would be consistent with the research ethics principles of beneficence, respect for persons and autonomy.

RoR Type 2: diagnosis

We propose that, if resources allow, genomic findings that help corroborate a psychiatric diagnosis be offered to research participants, even if not medically actionable. For example, a 22q.11.2 deletion92, 93 (OMIM #188400) in a participant diagnosed with treatment-resistant schizophrenia would not be medically actionable with regards to the schizophrenia symptoms because it is already known that the participant does not respond to available antipsychotics—although one could argue that given this variant’s pleiotropic effects it could be medically actionable for other purposes (that is, documented risks of impaired immunity, thrombocytopenia and hypocalcemia).93, 94, 95 Nevertheless, because of the association between 22q.11.2 deletion and risk for schizophrenia,92, 93 this finding can help substantiate the schizophrenia diagnosis. This is clinically valuable given the numerous disorders, diseases, injuries or agents that may cause symptoms that mimic schizophrenia. Similarly, pathogenic variants in NPC1 (OMIM #257220)—which increases the risk for Niemman-Pick disease type C and can present with psychosis96—in a participant diagnosed with schizophrenia would not be medically actionable, but it would suggest that the primary diagnosis is not idiopathic schizophrenia, but rather a single-gene disorder which is clinically valuable information.

Additionally, genomic information that can help corroborate or reject a diagnosis would provide a more complete clinical picture of the participant, including the potential pathogenesis of the symptoms. In a statement on clinical genetic testing, the International Society of Psychiatric Genetics recognized that some genomic information can be valuable even if not medically actionable: ‘Although there are no effective therapies yet for Fragile X or HD [Huntington’s disease], confirming the diagnosis provides the clinician and the family with useful information about how the patient’s illness is likely to progress and can help anticipate the needs of patients and their caregivers’.97 Furthermore, if treatments that target the pathogenesis of their symptoms are developed, participants informed of these variants will be in a better position to seek and access novel treatments.

RoR Type 3: significant health risks

In our view, clinically valid genomic findings associated with important health risks should ideally also be offered even if not medically actionable such as PSEN1 pathogenic variants associated with early-onset Alzheimer.13, 98 Non-medically actionable findings could be emotionally burdensome since there are, at present, no clinical interventions available to help decrease the risk of poor health outcomes. However, as with the return of any finding, informed consent would be paramount to help participants decide if knowing this type of information is in their best overall interest.85 Allowing participants to decide whether they want these findings returned would be consistent with respect for persons and their autonomy. Furthermore, participants could benefit in numerous ways, for example: by not being unnecessarily surprised with the onset of symptoms if the disease is ever expressed; seeking genetic and mental health counseling to learn more about the disease and how to cope with the risk; joining support groups; being attentive to novel therapies or clinical trials; planning certain aspects of their lives such as finances, insurance, and housing arrangements; and informing relatives so they can decide if they want to get tested.

RoR Type 4: VUS potentially related to symptoms

We propose a fourth—and controversial—type of finding for which we believe RoR might be offered: a subset of VUS potentially related to a participant’s known symptoms. Some argue against the return of any VUS on the grounds that by definition these are variants for which not enough evidence has been gathered to determine their pathogenicity, and therefore not clinically valid.83 Participants and their clinicians36, 37, 38, 39 may misinterpret the finding and order unnecessary tests or medical interventions that may generate harms with little prospect of benefit.

Nevertheless, we propose offering a small subset of VUS that may help explain a participant’s symptoms because they meet the following criteria: (1) very rare; (2) nonsense or damaging missense variants (particularly if they occur in genes or exons known to be intolerant to variation; (3) occur in genomic loci known to be associated with a psychiatric disorder or related neurological disease; and (4) the participant has known symptoms that are consistent with that disorder or disease.

We believe that RoR for these VUS should be offered to participants for the following reasons. They have characteristics that suggest they may be associated with a participant’s known symptoms. Psychiatric genomics is developing its knowledge base, and there are still high numbers of variant–phenotype associations that may eventually be shown to be clinically valid, but for which the field has simply not collected enough data yet. A number of reports argue that researchers should not have a duty to return any findings beyond their funding period, given the lack of resources to do so, among other practical obstacles.27 However, with the current pace of data collection, it is likely that in the near future the pathogenicity or the role of many of these VUS will be identified. If participants have access to these findings and, for example, any of these VUS are later identified as pathogenic, they could help provide a more complete clinical picture of the participant and information about the pathogenesis of symptoms, which could potentially improve clinical management. If these VUS are not made available by psychiatry researchers, the vast majority of participants are not likely to have access to this genomic information through other means until the cost of genetic testing and analysis decreases significantly more.

Practical challenges for the RoR in psychiatric genomics research

There are two key practical challenges for the RoR in psychiatric genomics research: conflicting RoR policies across countries and cost. Psychiatric genomics research is often conducted through multinational consortia.14 However, countries have different policies regarding the RoR: some prohibit the return, others provide general guidelines, and many do not have any guidelines.67, 99 These conflicting policies are problematic because, within an international consortium, some participants may benefit from the RoR and others not. Additionally, projects could decide not to offer the RoR to avoid conflicts with regulators in countries that restrict or prohibit the RoR. Therefore, it is important to begin a dialogue that can help harmonize guidelines regarding the RoR to facilitate research collaborations and maximize the benefits of the research endeavor by directly benefiting participants with clinically relevant information. Developing a consistent informed consent document would be an important step for the RoR. Another key step will be to develop a website or software with up-to-date information about clinically relevant variants for psychiatric research that helps standardize the variants offered and allows researchers to sift through data more efficiently in order to identify clinically relevant findings. This could also help minimize the amount of individual resources and time specific projects devote to the analysis portion of the RoR.

The most important challenge for the RoR in psychiatric genomics research is cost. The principal costs of RoR include corroborating research findings in clinical laboratories and having a clinician conduct the RoR. Countries such as the US require that researchers validate findings in a certified clinical laboratory100 before returning any results to participants, which significantly increases the cost of RoR. Amending regulations to allow less expensive ways of corroborating findings is another way to help decrease cost while still protecting participants from the return of erroneous findings. Given the complex nature of genomic information, it is important that the RoR is performed by a clinician (for example, a genetic counselor, clinical geneticist, or psychiatrist with genetics training) who can carefully explain the finding, its implications and suggest specific next steps. Psychiatrists and other clinicians generally report a lack of competence or preparation to manage genomic testing and findings and, thus, might overestimate risk and order unnecessary tests and procedures.37, 38, 39, 97, 101, 102 Therefore, ideally, the RoR clinician will also be available to communicate with the research participant’s physician. However, the use of clinicians to conduct RoR significantly raises costs.

Funding agencies in the US and other countries should provide funds for the RoR. This may be difficult because of limited research budgets. However, as we described above, funding agencies and advisory bodies in many countries have recognized the importance of offering to return clinically relevant findings. Thus, funding agencies should make every effort to act in accordance with those statements and the emerging consensus about the importance of offering the RoR of certain clinically relevant findings, by providing funds to allow researchers to offer RoR. One possibility could be to offer supplements for psychiatric genomics studies that are most likely to identify clinically relevant findings. Some studies are using a less expensive ‘outsourcing’ approach to RoR by offering participants their raw genomic data, which allows participants the possibility of getting it interpreted by a third party.86, 103 A drawback of this approach is that many participants may not have resources to get their data interpreted.

It is important to note that researchers will not identify clinically relevant findings in the vast majority of participants. Under the proposal presented here, researchers would only offer clinically relevant findings they will generate within the scope of their study. Current estimates are that only about 1% of participants are expected to have medically actionable findings.72 The number of individuals with clinically valuable or likely clinically valuable findings in psychiatric genomics is currently difficult to estimate, may vary between cases and controls, by disorder (for example, greater with intellectual disability and autism), and, within disorder, might vary with clinical severity and age of onset. There will also be a number of participants who decline the RoR completely and studies suggest that approximately 39% of participants may refuse some types of findings.104 In addition, RoR is one of the main motivations51, 52, 53 for participating in research, therefore, by offering findings researchers will likely save significant time and resources in recruitment. The primary goal of research is to generate generalizable knowledge but, if psychiatric genomics researchers have the resources to return results while achieving the scientific goals of their studies, offering these findings can help maximize the societal benefits of psychiatric genomics research.

Conclusion

With this Perspective, we hope to spark a discussion about which kinds of findings may be offered in psychiatric genomics research considering the particularities of this field and the potential risks and benefits to participants. We propose that, as in genomics research for other medical conditions, psychiatric researchers should ideally offer to return medically actionable findings identified in the course of research. If resources allow, researchers should consider offering clinically valuable and likely clinically valuable findings. There are obstacles that need to be addressed to facilitate the RoR. However, the RoR from psychiatric genomics research can help maximize the benefits of this research for society and promote the best interest of participants.