Many ELSI concerns related to genetic risk prediction for multifactorial disease were raised in the ninety-two articles included in this study, which spanned from 1977 to 2021 (see Fig. 2). ELSI-related themes identified via thematic analysis are described in detail below.
Potential benefits
Potential benefits of genetic risk prediction for multifactorial disease discussed in the dataset are summarized briefly here, with representative quotes provided in Table 1. However, a comprehensive analysis of the clinical utility of PGS is beyond the scope of this paper. Indeed, benefit/risk analysis of PGS will depend on specific context: how the score was developed, its predictive power in a given population for a specific disease, and what decisions it is intended to inform (Kotze et al. 2015; Lewis and Vassos 2020; Yanes et al. 2020). Moreover, the predictive power of PGS will probably improve over time, due to increasing size of GWAS datasets and methodological improvements (Karavani et al. 2019; Lambert et al. 2019; Martin et al. 2019b; Ronald 2020). To improve predictive power, models may combine PGS with risk from rare variants typically not included (Briggs and Slade 2019; Choi et al. 2020; Fabbri and Serretti 2020; Fullerton and Nurnberger 2019; Lambert et al. 2019; Yanes et al. 2020) or other factors like health conditions or family history (Lambert et al. 2019; Polygenic Risk Score Task Force of the International Common Disease 2021; Ronald 2020).
Early determination of genetic risk may allow improved screening, earlier diagnosis, and/or increased opportunity to prevent or delay multifactorial disease via lifestyle changes, prophylactic treatments, or avoidance of risk-increasing environmental factors (Briggs and Slade 2019; Chaudhari et al. 2020; Driver et al. 2020; Ganna et al. 2013; Hall et al. 2004; Lambert et al. 2019; Palk et al. 2019; Pashayan et al. 2013). Other possible benefits of PGS are to improve care by clarifying diagnosis or guiding treatments (Duncan et al. 2019; Lambert et al. 2019; Lewis and Vassos 2020). Interestingly, PGS may provide insights into genetic overlap as well as differentiation between phenotypes and diseases (DiBlasi et al. 2021; Fullerton and Nurnberger 2019; Martin et al. 2019a; Ronald 2020; Yanes et al. 2020). PGS may also be useful in research (Manrique de Lara et al. 2019; Martin et al. 2019a).
PGS shares similarities with other clinical biomarkers (Saya et al. 2021), however, generally speaking, an individual’s genetic information does not change over time. Available even before birth, genetic-based risk can be assessed very early in human life (Torkamani et al. 2018). Ultimately, clinical or implementation research must empirically determine whether specific PGS or models incorporating PGS improve outcomes (Choi et al. 2020; Chowdhury et al. 2013; Duncan et al. 2019; O'Donnell 2020), including whether risk results motivate behavior change—which is not clear (Driver et al. 2020; Yanes et al. 2020).
Table 1
Illustrative quotes relating to potential benefits of polygenic risk prediction in the context of multifactorial disease. Citations within quotes omitted.
Subtheme | Illustrative Quote |
Early disease/risk prediction | “Early disease detection, prevention and intervention are fundamental goals for advancing human health. Meanwhile, genetic risk estimation is, for all intents and purposes, the earliest measurable contributor to common heritable disease risk.” Torkamani et al. (2018) |
| “In conclusion, it is becoming clear that PRS have value in disease risk prediction and there are multiples areas where this may have clinical utility.” Lambert et al. (2019) |
Potential clinical utility | “Broadly speaking, the clinical utility of PRS can be divided into four categories: (i) informing population screening programs, (ii) refining risk for individuals undergoing genetic testing for monogenic risk genes, (iii) guiding therapeutic interventions and (iv) facilitating diagnosis and predicting health outcomes” Yanes et al. (2020) |
Value in research | “Currently, such scores are quite useful in research, and they are telling us much about the relationships between different disorders and other indices of brain function.” Fullerton and Nurnberger (2019) |
| “A potential future use of PRS is in clinical trials, potentially enabling more effective drug treatment targeting in high-risk patients” Martin et al. (2019a) |
Potential mis- or over-interpretation and the need for clear and accurate communication
A significant challenge for PGS is the potential for mis- or over-interpretation, which may lead to inappropriate actions, such as overdiagnosis or overtreatment (Lewis and Green 2021). General understanding of genetics may not translate into an ability to interpret PGS, the simplicity of which, provided as numbers or percentiles, belies its complexity (Driver et al. 2020). Studies have demonstrated that individuals struggle with probability literacy and risk numeracy, and that recipients of health information understand absolute risks better than relative risks (Peay 2020). Also, inappropriate expectations “for distinctly separated phenotypic subpopulations” may be anticipated from a monogenic paradigm (Ozdemir et al. 2009). PGS do not perfectly differentiate individuals with and without disease: the mean score for cases will be higher than for controls (Pashayan et al. 2013), but cases and controls can have low and high scores, respectively (Fullerton and Nurnberger 2019; Lambert et al. 2019). Individuals with extremely high PGS are at higher genetic risk for a disease, but if the absolute risk for the disease is very low, this increased risk may not be clinically meaningful (Rosenberg et al. 2019). PGS may actually change how we perceive multifactorial disease—instead of a binary state (someone has a disease or doesn’t), it could be viewed as more of a continuum (Lewis and Green 2021). Another article astutely points out that genetic testing for complex diseases raises “fundamental questions of what should be considered a ‘genetic’ disorder in the first place—
and what the psychological, social, ethical, and legal implications are of labeling a common condition like heart disease as ‘genetic’"(Andrews and Zuiker 2003).
Another issue involves gene-environment interactions. Although PGS intends to measure genetic contributions to a particular phenotype, some variant-associated effects may be conferred by gene-environment correlations (Ronald 2020). Although controlling for and separating genetic and environmental effects in PGS is a “long-studied issue… it is still not a ‘solved’ problem in genetic studies… [and] care is needed to ensure this powerful tool is applied appropriately”(Blanc and Berg 2020). Appreciation for the limitations of polygenic scores is also required when interpreting phenotype differences across populations: “genetic contributions to traits, as estimated by polygenic scores, combine with environmental contributions so that differences among populations in trait distributions need not reflect corresponding differences in genetic propensity”(Rosenberg et al. 2019).
Overinterpretation is also possible. Some caution that even though PGS “demonstrate the importance of genetic variation in the etiology of the disorders,” this does not validate “the value of the risk score proposal in disease prediction (i.e., screening)”(Wald and Old 2019). Although moderate relative risks (e.g., three- to six-fold) “can have considerable significance in determining causes of disease… estimates of the relative risk between a disease marker and a disease have to be extremely high for the risk factor to merit consideration as a worthwhile screening test”(Wald and Old 2019). In these authors’ estimation, PGS will not meet this requirement, and they warn: “It is important that the potential applications of genomic medicine are not compromised by raising unrealistic expectations in medical screening”(Wald and Old 2019).
Clear and accurate communication of the benefits and limitations of PGS will be critical to its ethical implementation (Ronald 2020). To mitigate the potential for mis- or overinterpretation of PGS, many articles stress the need for training healthcare professionals who will encounter and interpret PGS and counsel patients (Andrews and Zuiker 2003; Briggs and Slade 2019; Chowdhury et al. 2013; Kotze et al. 2015; Lea 2003; Lewis and Green 2021; McLaren et al. 2016; Motulsky 2002; Palk et al. 2019; Pashayan et al. 2013; Pashayan and Pharoah 2012; Rashkin et al. 2019; Torkamani et al. 2018; Vassy et al. 2018). Also acknowledged is that responsibility for accurate communication is shared (Ronald 2020; Rosenberg et al. 2019).
Table 2
Illustrative quotes related to risks and challenges associated with clinical translation of polygenic risk prediction in the context of multifactorial disease. Citations within quotes omitted.
Theme | Illustrative Quote |
Potential for mis-or overinterpretation | “There are four important considerations of the information content of a polygenic score, and how it can be interpreted: 1) The known information, which shows where an individual lies compared to others on the risk scale 2) The unknown information from incomplete genetics or unmodelled environment 3) The potential for incorrect information, for example, where the individual differs from characteristics of the research study used to estimate the effect size of each genetic variant by genetic ancestry, age, environmental load, or disease definition, or where there is a technical bias in data collection 4) The intended use of the PRS, for example, more complete information would be required for justifying a pharmacological intervention than for using the PRS to motivate behaviour change” Lewis and Vassos (2020) |
| “Other barriers to PRS utility include physician and public education regarding the interpretation of polygenic risk, especially in the understanding of various and dynamic risk metrics.” Torkamani et al. (2018) |
Need for accurate communication | “This conceptual transfer from monogenic disorders to polygenic disease is quite inappropriate, because polygenic disease involves the co-inheritance of several genetic determinants that usually have to interact with environmental factors before the disease becomes manifest. The genetic determinants for a phenotype can be variable and they may interact with different ways; some of the genetic factors can even be protective for the occurrence of the disease.” Galton and Ferns (1999) |
| “we argue that particular attention should be paid to the difficulties associated with the communication and interpretation of results. This would be due, in part, to the fact that, given the etiological complexity of psychiatric disorders, a PRS in the top percentile would be an indicator of risk, not a definitive prognosis. For this reason, nuance and skill would be required in articulating and ensuring correct understanding (both of counsellors and patients) of ‘complex’ risk. While the difficulties associated with feedback of complex genetic risk are not necessarily unique to PRS, they nevertheless warrant consideration given its recency.” Palk et al. (2019) |
Potential for stigma and/or discrimination | “Further investigation of the effect of PGS to raise or lower other foundational genetic counseling outcome measures like decisional conflict, adjustment of health related behaviors, feelings of stigma/depression, and retention rates are necessary.” Rashkin et al. (2019) |
| “As with all areas of science, a strong backlash against PGS would threaten progress. One potential way to prevent this is to avoid overhyping and overinterpreting the utility of PGS while also acknowledging their considerable and unique strengths. As we saw above, a PGS is not a person’s destiny. A second threat is the miscommunication of PGS research to the general public. Engagement with a wide range of user groups, patient and public involvement activities and promoting and funding the important area of genetic counselling are all key.” Ronald (2020) |
Potential for conflicts of interest/premature commercialization | “A potential worrisome development is the increasing employment of genetic counselors by commercial testing companies since it will be difficult for such professionals to be objective about the pro’s and cons of testing.” Motulsky (2002) |
| “Despite limited sensitivity and clinical utility of PRS, direct-to-consumer (DTC) advertising for psychiatric genetic testing that utilizes PRS is increasingly common.” Docherty et al. (2021) |
| “There has also been a rapid rise of direct-to-consumer assays and for-profit companies (23andMe, Color, MyHeritage, etc.) providing PGS/PRS results to customers outside of the traditional patient-provider framework.” Wand et al. (2021) |
Need for standards and possible regulation | “As GWAS sample sizes increase and PRS become more powerful, they are set to play a role in research and personalized medicine. However, despite the growing application and importance of PRS, there are limited guidelines for performing PRS analyses, which can lead to inconsistency between studies and misinterpretation of results.” Choi, et al. 2020 (2020) |
| “Although we have provided explicit recommendations on how to acknowledge study design limitations and their effects on the interpretation and generalizability of a PRS, future research should attempt to establish best practices to guide the field.” Wand at al. (2021) |
Potential for stigma and discrimination
Even if PGS are interpreted appropriately, the power of genetic testing for polygenic disease “to identify individuals at increased susceptibility… creates a set of complex ethical, legal and social issues”(Manrique de Lara et al. 2019) that are intertwined with risk prediction as well as perceptions and misperceptions about genetics. Although it is important to consider how PGS results might impact family members (Lewis and Green 2021), this may not be as critical for polygenic prediction as with genetic testing for single-gene disorders (Briggs and Slade 2019; Rashkin et al. 2019), as complex trait theory suggests that most cases will arise in individuals without family history of the disease (Lambert et al. 2019).
Several articles note the connection between the potential for misunderstanding PGS and possible downstream negative consequences (Andrews and Zuiker 2003; Fabbri and Serretti 2020; Galton and Ferns 1999; Ikeda et al. 2021; Palk et al. 2019; Saya et al. 2021; Torkamani et al. 2018). While some advocate that PGS should be treated like other non-genetic laboratory tests and biomarkers (Andrews and Zuiker 2003; Saya et al. 2021), others worry that genetic information in the context of multifactorial disease may elicit stigma or discrimination (Chowdhury et al. 2013; Ikeda et al. 2021; Kious et al. 2021). Some advocate appropriate counseling (Briggs and Slade 2019; Ikeda et al. 2021; Palk et al. 2019), while others call for stronger legal protections against genetic discrimination (Torkamani et al. 2018) and privacy regulations (Li et al. 2020).
There are also concerns about genetic determinism: “Without appropriate communication of the uncertainty around [PGS], large-scale deployment…could potentially reinforce and amplify false genetic-determinism attitudes”(Polygenic Risk Score Task Force of the International Common Disease 2021). Relatedly, people may discount environmental and lifestyle influences on diseases (Driver et al. 2020; Peay 2020; Warren 2018). Particularly in certain therapeutic areas, such as mental health, communication of PGS should not encourage problematic, stigmatizing assumptions or “reductive interpretations” which could lead to discrimination (Palk et al. 2019). Biogenetic explanations and the deemphasis of social determinants may be associated with lower social acceptance for individuals with mental health disorders (Palk et al. 2019). Another issue relates to the “widespread genetic overlap across psychiatric and substance use disorders, indicating that many genetic variants will increase risk for multiple outcomes… These complexities must be considered when designing best practices for return of genotypic feedback”(Driver et al. 2020).
More research is needed to evaluate these issues (Driver et al. 2020; Rashkin et al. 2019), including “the impact of PRSs on knowledge, self-concept, symptom burden, and treatment adherence for affected individuals. For at-risk individuals, studies may evaluate knowledge and risk perception, the positive and negative psychological and social impact of learning the risk information, and any resulting behavior changes for participants”(Peay 2020).
Potential conflicts of interest/ premature commercialization and the need for standards and possible regulation
The potential benefits of identifying those at high genetic risk for common diseases “have fallen on fertile ground among politicians, health-care providers and the general public, particularly in light of the increasing costs of health care in developed societies"(Hall et al. 2004). The perceived value of PGS propels research in this area, but also leads to concerns about potential conflicts of interest and premature commercialization (Motulsky 2002). Interpretation of data, including PGS, rests with humans who may have particular values, biases, associations, or interests that impact their analysis (Ozdemir et al. 2009). Despite the uncertain clinical value of PGS at the present time (Curtis 2019; Parens et al. 2020; Rosenberg et al. 2019), PGS are increasingly becoming available through direct-to-consumer (DTC) testing companies, including in sensitive and controversial contexts, such as psychiatric conditions and preimplantation genetic testing of embryos after in vitro fertilization (Docherty et al. 2021; Lewis and Vassos 2020; Motulsky 2002; Rashkin et al. 2019; Treff et al. 2019).
Standards and guidelines are needed as they are critical for quality control with PGS (Choi et al. 2020). Since different research groups use different methods to develop PGS, results can diverge: “We believe this lack of consistency to be a prime concern for the PRS field, and additional resources, such as a centralized public database of published polygenic scores, are necessary to increase PRS comparability and evaluation and thus improve their potential for translation”(Lambert et al. 2019). Problems with PGS include overfitting caused by using the same dataset for generating and training (Yanes et al. 2020), or errors caused by misclassifying a phenotype or treatment of missing information (Li et al. 2020). Inaccuracies may have significant downstream repercussions, especially if commercialization of tests and/or health policy is based on biased or substandard studies (Ozdemir et al. 2009). However, guidelines for PGS will depend on the specific disease, as well as the availability of GWAS summary data for discovery and target populations (Yanes et al. 2020). In response to the need to standardize development and reporting, and enable evaluation, the PGS Catalog (www.pgscatalog.org), an open database of PGSs, was developed as a resource for the community (Yanes et al. 2020). The “Polygenic Risk Score Reporting Standards,” a framework defining minimal information needed to interpret and evaluate PGS, was put forward in 2021 (Wand et al. 2021).
Regulation of PGS also needs consideration: too much could delay availability and limit access, but too little could jeopardize safe and appropriate use (Knoppers et al. 2021). An article by Docherty et al. on PGS for suicide prediction includes a robust discussion of FDA regulation of DTC genetic tests (Docherty et al. 2021). Risks of DTC genetic tests include false positives, false negatives, and errors in interpretation (Docherty et al. 2021). The authors believe the current regulatory environment enables “oversimplification and exaggeration of research results for marketing purposes” and provision of genetic tests “without demonstration of clinical validity”(Docherty et al. 2021). “At a minimum, companies offering DTC genetic testing for polygenic risk should publish guidelines for interpreting their results that, in layperson terms, acknowledge a lack of clinical utility,” the authors write. “Again, however, the current failure of DTC companies to do so may be difficult to remedy without regulatory changes”(Docherty et al. 2021). Appropriate governance and regulation of PGS in the context of embryo screening is another concern (Karavani et al. 2019; Lázaro-Muñoz et al. 2021; Munday and Savulescu 2021).
Companies offering genetic testing “with varying clinical utility” may give “consumers an expanded sense of agency and autonomy around their genetic information”(Rashkin et al. 2019). Although some individuals are interested in genomic tests “to be empowered with personal risk information”(Saya et al. 2021), many people are more interested in genetic information if it can guide actions (Driver et al. 2020). Notwithstanding individuals’ desire for genetic information, it “is not automatically empowering” and “if the results are not carefully communicated, patients may be confused about their impact, and unsure of what steps to take next”(Kious et al. 2021). To support individuals’ decisions, it is important to set policies around what information should be provided (Pashayan et al. 2013). While DTC companies and even healthcare systems may argue that individuals have a right to learn about what is currently known about their (or their future child’s) genomic liability, some voice strong concern about the potential for significant harm (Docherty et al. 2021; Fabbri and Serretti 2020).
Concerns related to ethical research: equitable access to benefits, diversity in research, and responsible data sharing
Other identified themes included equitable access, diversity in research, and responsible data sharing (see Table 3 for additional representative quotes). Many articles voice concern that PGS may exacerbate health inequities: current algorithms have varying accuracy across different population groups due to the Eurocentric bias in genetic databases (Briggs and Slade 2019; Cavazos and Witte 2021; Chowdhury et al. 2013; Dikilitas et al. 2020; Fernandez-Rhodes et al. 2020; Lambert et al. 2019; Martin et al. 2019b; Palk et al. 2019; Slunecka et al. 2021; Warren 2018; Yanes et al. 2020; Zhou et al. 2021). Some commercial tests for polygenic risk are restricted by ancestry (Lewis and Green 2021). According to the PRS Task Force, “responsible use” of a PRS is achieved when “there are clear benefits that outweigh risks, and where effort is taken towards a goal of equitable benefit for all”(Polygenic Risk Score Task Force of the International Common Disease 2021). To ensure equitable access to PGS, many recommend increased diversity in genetic research (Chowdhury et al. 2013; Fernandez-Rhodes et al. 2020; James et al. 2021; Knoppers et al. 2021; Rubin and Glusman 2019). Other possibilities “include the creation of separate PRS algorithms or risk stratification methodologies for individuals from distinct genetic ancestry groups, as well as the creation of a singular algorithm trained on holistic training data that are representative of diversity in genetic ancestry (i.e., PRS scores with cross-population portability)”(Knoppers et al. 2021). Aside from ensuring that PGS realize “comparable performance across sub-populations and across human genetic diversity,” it is also important to make sure that there is equitable access to risk-stratified care and follow-up (Knoppers et al. 2021). There is also concern about how to ethically use race, ancestry, and ethnicity in PGS reporting (Fernandez-Rhodes et al. 2020; Lewis and Green 2021; Mudd-Martin et al. 2021).
Related to concerns about maximizing the accuracy of PGS and diversifying genetic research is an expressed need for responsible data sharing and safeguarding (Briggs and Slade 2019; Daniels et al. 2021; Pashayan et al. 2013). Ethical obligations for data collection and analysis are significant, given that relative to monogenic testing, PGS will “affect a larger number of people, are less focused, require the compilation of a larger set of private data about each individual, and raise greater concern about informed consent”(Andrews and Zuiker 2003). The potential for stigma and/or discrimination also necessitates special attention to “how and when genetic samples and data are acquired, stored, and used”(Chowdhury et al. 2013). However, barriers to data sharing limit the power of PGS (Knoppers et al. 2021). One organization, the Global Alliance for Genomics and Health, has developed policies to guide ethical sharing of genomic and clinical data (Mudd-Martin et al. 2021).
Table 3
Illustrative quotes related to concerns about equitable access, the need for diversity in genetic research and responsible data sharing. Citations within quotes omitted.
Theme | Illustrative Quote |
Concern about equitable access | “we consider the consistent observation that they [PRS] are currently of far greater predictive value in individuals of recent European descent than in others to be the major ethical and scientific challenge surrounding clinical translation and, at present, the most critical limitation to genetics in precision medicine.” Martin al. (2019b) |
| “Furthermore, in polygenic risk, where estimates are derived from risk estimates from European populations, transferability to non-European populations is limited; use of PRS risk stratification therefore has the potential to widen health inequalities in the absence of non-European risk estimates…. It should also be noted that personal utility, ethical considerations and risk/benefit considerations will also be affected by cultural and social background. These issues must be acknowledged and steps taken to remedy the resulting disparities.” Briggs and Slade (2019) |
Need for diversity in genetic research | “We discuss the important ethical, legal, and social implications of increasing ancestral diversity in genetic studies of cardiometabolic disease and the challenges that arise from the (1) lack of diversity in current reference populations and available analytic samples and the (2) unequal generation of health-associated genomic data and their prediction accuracies.” Fernandez-Rhodes et al. (2020) |
| “The lack of representative GWAS has been recognized as a key obstacle in the project of Precision Medicine. Initiatives like the All of Us study, the Clinical Sequencing Evidence-Generating Research Consortium (CSER), the Human Genome Reference Program (HGRP), the PRS Diversity Consortium, and others, have been designed to address both the underlying science and clinical translation, as well as longstanding debates about the role of race in medicine, genetics, and genomics.” James et al. (2021) |
Need for responsible data sharing | “The nature of research on common, complex disorders makes the potential breach of confidentiality both more likely and more risky.” Andrews and Zuiker (2003) |
| “In addition, many of the approaches used in research (e.g., anonymization, de-identification) are not applicable to genetic information because the genome is the ultimate identifier. Thus there is a requirement for additional strategies that preserve the privacy of genomic data while not compromising the accuracy of the results.” McLaren et al. (2016) |
| “To balance the advantage of advancing healthcare using large-scale EHR data and potential concerns of privacy violations, more up-to-date regulatory measures are needed to match the pace of technological development.” Li et al. (2020) |