Ethical, Legal, and Social Issues (ELSI) of Responsible Data Sharing Involving Children in Genomics: A Systematic Literature Review of Reasons

Abstract

Background

Progress in precision medicine relies on the access to, use of, and exchange of genomic and associated clinical data, including from children. The ethical, legal, and social issues (ELSI) of such data access, use, and exchange may be accentuated in the pediatric context due in part to the highly sensitive nature of genomic data, children’s consent-related vulnerabilities, and uncertain risks of reidentification. Systematic analyses of the ELSI and scientific reasons for why and how genomic data may be shared responsibly are, however, limited. Methods: We conducted a modified systematic review of reasons according to Sofaer and Strech to examine the ELSI and scientific reasons for “responsible” sharing of children’s genomic and associated clinical data. Empirical articles, commentaries, and data-sharing policies indexed in Medline, Scopus, Web of Science, and BIOSIS were included in the analysis if they discussed ELSI and were published between 2003 and 2017 in English. Results: One hundred and fifty-one records met our inclusion criteria. We identified 11 unique reasons and 8 subreasons for why children’s genomic data should or should not be shared. Enhancing the prospect of direct and indirect benefits and maximizing the utility of children’s data were top reasons why data should be shared. Inadequate data privacy protection was the leading reason why it should not. We furthermore identified 8 reasons and 30 subreasons that support conditional data sharing, in which recontact for the continued use of children’s data once they reach the age of majority was the most frequently endorsed condition. Conclusions: The complete list of ELSI reasons and responsible conditions provides an evidentiary basis upon which institutions can develop data-sharing policies. Institutions should encourage the sharing of children’s data to advance genomic research, while heeding special reconsent and data protection mechanisms that may help mitigate uncertain longitudinal risks for children and families.

Keywords:

• Genomics
• data sharing
• pediatrics
• systematic review of reasons
• policy

Introduction

Translating discoveries in precision medicine to achieve improved individual- and population-level health outcomes necessitates broad sharing of genomic and associated clinical data. Such sharing is especially pertinent during childhood, when many Mendelian and other heritable genetic conditions first present and may be clinically actionable. Indeed, advancements in understanding genomic etiologies of childhood diseases as a result of concerted data sharing are innumerable (Downing et al. 2012; Mefford, Batshaw, and Hoffman 2012; Beaulieu et al. 2014), and whole genome sequencing used as a first-in-line test has been shown to increase diagnostic yield (Schofield et al. 2010; Xue et al. 2015), sometimes by nearly 27% (Lindstrand et al. 2019). The clinical successes of sequencing have furthermore motivated Congressional action (H.R.4144 116th Congress 2019). For children suspected of suffering from rare, undiagnosed genetic conditions, in particular, sharing their linked genotypic and phenotypic data in accessible variant databases may afford them the only opportunity for accurate diagnosis (Beaulieu et al. 2014; Boycott et al. 2014; ACMG Board of Directors 2017; Might and Wilsey 2014). Genomic data are, however, inherently identifying, and their broad exchange can pose informational risks not only to the pediatric patients, but also to their biological relatives (McGuire, Oliver, et al. 2011; Kaye 2012).

These general ethical, legal, and social issues (ELSI) give rise to unique considerations when data sharing involves children. Children’s inability to legally consent underpins their status as a vulnerable population in data-intensive research. While parents are authorized to make decisions in the best interests of their child, it is generally accepted that children assume decision-making rights once they reach the age of majority. Practical challenges can limit data-sharing preferences of future adults (Knoppers et al. 2016)—for example, when data are contributed anonymously and then aggregated for inclusion in larger, combined datasets. This aggregation makes retrieval and removal therefore unrealistic. Recent reports of patient reidentification from datasets originally thought to be anonymized further heighten data security concerns (Gymrek et al. 2013; Erlich et al. 2018), and the accuracy with which patients can be reidentified from such datasets forecloses on what scholars term the child’s “right to an open future” (Davis 2009; Davis 1997). Parents face ethical tensions, as enhancing immediate clinical benefits, such as diagnosis, may involve putting their child’s data at uncertain risk with little longitudinal control over how the data will be exchanged in the future. It is primarily this specific circumstance and those related to sharing genomic data from banked biospecimens involving children that this systematic review of the literature considers in depth.

Despite the need for practical guidance on how to responsibly share children’s genomic data, few analyses have sought to landscape ELSI reasons for responsible sharing as expressed in both the empirical bioethics and gray literature. Absent such analysis, institutions may inappropriately restrict or aimlessly permit access to sensitive genomic data. Both outcomes have negative consequences on advancing knowledge in pediatric disease, and ultimately misguide evidence-based policy development for responsible data sharing in the data-intensive sciences such as genomics.

Review methods/search

We conducted a systematic review of reasons according to Sofaer and Strech (2011) to fill the aforementioned knowledge gap and to guide institutional best practices for sharing genomic data involving children. Systematic reviews of reasons “take into account the specific conceptual and practical challenges of empirical bioethics” (Strech and Sofaer 2012), while preserving the systematicity associated with traditional reviews. This method of review enables reviewers to systematically search for and contextualize evidence contained in argument-based literatures that can be used to prioritize agenda setting and motivate policy action on complex issues that span medical, legal, and philosophical disciplines. Systematic reviews of reasons make arguments the primary outcomes of interest. The Sofaer and Strech model proposes four steps:

1. Formulate the review question and eligibility criteria.

2. Identify all of the literature that meets the eligibility criteria.

3. Extract and synthesize data.

4. Derive and present results: the answer to the review question.

Our synthesis of the data in step 3 involved first identifying passages where a reason is expressed (called a reason mention). We subsequently categorized reason mentions into two “types” (either a reason, or a subreason)¹. While Sofaer and Strech do not prescribe any one qualitative approach to categorizing reason types, all approaches aim to “compare text passages that mention reasons across papers and to match reason mentions from one paper with reason mentions from another, ensuring that a reason type captures similar reason mentions from different papers” (Strech and Sofaer 2012, 123).

Our review departs from the Sofaer and Strech (2011) model in that we ask two questions of the literature to more comprehensively capture the ELSI of “responsible” genomic data sharing involving children. We supplemented a traditional, reason-based review question—Which reasons have been given to support the view that children’s genomic and associated clinical data should/not be shared?—with a secondary review question that asked, What conditions are needed to support responsible sharing of children’s genomic and associated clinical data? Early data analyses informed this decision, as they revealed greater contention around the conditions that support responsible data sharing, rather than whether data should be shared ipso facto. Put simply, our results suggest broad support for sharing genomic data but is contingent on certain safeguards. The second review question details those conditions as reflected in the literature, and documents their supporting rationales. It is important to note a conceptual difference between conditions that support “responsible” and compliant data sharing. The latter conditions are both jurisdiction specific and institution specific. Compliant data sharing appeals to regulations governing human subjects research and data protection, among others, and for this reason is outside the authority of this article to define.

With support from a reference librarian, the review team searched the empirical and gray literatures indexed in MedLine, BIOSIS, Scopus, and Web of Science for records published in English between 2002 and 2017. A detailed search strategy for each database is provided in the supplementary materials. We included empirical studies; commentaries; editorials or position statements; conference proceedings; professional organization reports; and international or professional guidelines if they

1. Discussed ethical, legal, social implications (ELSI) of exchanging whole genome/exome data involving children.

2. The sharing of genomic or associated clinical data involved patients aged 0–18 years old (neonates to the age of majority in most jurisdictions).

3. The sharing of genomic or associated clinical data was derived in the context of either pediatric clinical research or pediatric care.

Records were excluded if they discussed pediatric data sharing only in relation to the return of incidental findings/individual research results or in newborn screening. Previously published systematic reviews (Mackley et al. 2017; Bertier et al. 2017) and empirical studies (McGowan et al. 2018; Burke et al. 2013; Holm 2017; Bishop, Strong, and Dimmock 2017; Petersen et al. 2017; Driessnack and Gallo 2011) address these issues at length. We also excluded technical papers where interoperability or security measures were detailed without explicit discussion of ELSI motivating these data protective measures.²

Data analysis

Lead author VR conducted the analysis of subreason/reason mentions on all retained records and policy documents using the NVivo 11 software. VR first scanned the document for all mentions of “data shar*,” “pediatric,” “ethic*,” “legal,” or “social.” Using qualitative content analysis (Sandelowski 2000), relevant reasons were coded to produce a preliminary coding scheme. All members of the review team consulted on the final codebook that was subsequently used to analyze subreason/reason mentions in the entire dataset.

Systematic review results

Publication characteristics

In total, 151 records met our inclusion criteria; 118 records were retained from the database search, the bibliographies, and in-text references, which authors VR and GB hand searched for any normative guidance documents, such as policies, guidelines, or conventions, cited in the record. This snowball search resulted in an additional 33 records. Figure 1 provides a PRISMA summary of the search results, and a data extraction table of the following characteristics can be found in Table 1:

Figure 1. PRISMA chart.

Table 1. Publication characteristics

	No. (%) of publications N= 151
Publication Type
Peer reviewed article	125 (82)
Guideline/policy	9 (5.9)
Book section	3 (1.9)
Report	4 (2.6)
Journal published letters	4 (2.6)
Press release	4 (2.6)
Conference proceedings	2 (1.3)
Content Type
Commentary or opinion piece	70 (46)
Empirical study	43 (28)
Data sharing SOP	27 (18)
Data sharing guideline/policy	4 (3.3)
International convention	3 (3.3)
Professional guideline	1 (0.6)
Regulation	3 (1.3)
Journal field for 139 empirical studies, commentaries and reports
Genetics/genomics	42 (30)
Medicine	40 (27)
Bioethics	17 (12)
General science	12 (8)
Public Health	10 (7.2)
Social science	6 (4.3)
Health policy/law	6 (4.3)
Informatics	4 (3)
Nursing	2 (1.4)
Region of publication or policy/guideline/law
North America	93
Europe	43
Africa	4
Other (international)	4
Asia	3
New Zealand/Australia	3
Middle East	1

• Publication type.

• Content type.

• Journal field.

• Context in which pediatric data sharing was primarily discussed.

Sixty-five (43%) records discussed data sharing in an ELSI commentary/opinion piece, while 27 (18%) discussed ELSI in the context of organized data-sharing programs or pediatric research consortia. All 43 records reporting empirical data were critically appraised using the Mixed Methods Appraisal Tool (MMAT). Of these, 39 (94%) earned a MMAT score of 0.75 or higher (see Supplementary Material). Four records presented results from two empirical studies (Burstein et al. 2014; McGuire, Oliver, et al. 2011; Hens et al. 2009a; Hens et al. 2009b). Fourteen percent of all records for which a publication year was available were published in 2016; nearly half (49%) were published in between 2013 and 2017, inclusively.

Subreason/reason identity and incidence

We identified 604 total subreason/reason mentions, which we grouped under 19 main reason types and 38 subreason types. Enhance the prospect of direct and indirect benefits to individual as well as other pediatric patients was the reason most mentioned (27 mentions) for why children’s data should be shared. Consent was the condition most mentioned (29 mentions) supporting responsible sharing. The right to recontact/reconsent to continued data use at the age of majority garnered the highest number of subreason mentions (35) related to conditions that support responsible sharing, followed by limiting parental authority for broad consent (29) and managing data access via user restrictions (29).

Findings

Which reasons have been given to support the view that children’s genomic and associated clinical data should/not be shared?

Of 151 records that met our inclusion criteria, we identified one reason mention for why children’s data should categorically not be shared: “If whole genome sequencing to identify preventable diseases or SNPs or haplotypes associated with drug responsiveness does occur, the child’s DNA should be discarded afterwards and genomic information erased or stringently protected. If protective measures are not feasible, a decision will have to be made whether to genotype them at all” (Robertson 2003, 7).Two subreasons underpin this position: a lack of a compelling need to retain results derived from genotyping a patient who at present lacks competency to consent, and the potential for data misuse. The author reasons that whole genome sequencing for those patients rendered incompetent to consent at the time of genotyping should be reserved for rare circumstances when immediate clinical benefit justifies sharing, or when genome-wide comparisons are the primary aim of the research. Narrower genetic tests with time-limited information storage could suffice for meeting the direct clinical needs for the child, the author reasons, and therefore such options should be exhausted before considering broad sharing.

Time and regulatory history are key factors when interpreting the preceding subreasons/reasons. The year of publication (2003) marked the completion of the Human Genome Project, and the applications of whole genome sequencing in the clinic elicited both awe and trepidation. This record also predated important privacy legislations—most importantly the Genetic Information Nondiscrimination Act (GINA)—and human subjects research protections in the Revised Common Rule. Both GINA and the Revised Common Rule later included specific protections for genetic/genomic data, including for children. We posit that a stronger regulatory climate might have instilled greater trust in the data protections afforded under the law, and in turn a more permissive stance toward data sharing with appropriate oversight.

We identified 10 reasons and 7 subreasons why pediatric genomic and associated clinical data should be shared (Table 2), with a total of 149 mentions. Enhancing the prospect of benefit to individual and future pediatric patients was the reason most frequently cited for why pediatric data should be shared (27 reason mentions), followed by the need to maximize the scientific and clinical utility of children’s data (13 reason mentions).

Table 2. Reasons and subreasons given for why children’s genomic and associated clinical data should (n = 3) and should not (n = 149) be shared [number of mentions].

1. Genomic data privacy protection [1]  i. Potential for data misuse [1]  ii. Lack of compelling need to maintain genotypic information [1]

2. Enhance the prospect of direct and indirect benefits to individual, as well as other pediatric patients [27]  i. Direct clinical benefits for the patient [7]  ii. Indirect benefits to other, as well as future children [12]   • Indirect benefits justified on the basis of solidarity [1]  iii. Direct and indirect benefits to patient(s) [8] 3. Maximize the clinical and scientific utility of children’s data [19]  i. To improve statistical power and sample size [14] 4. Encourage research collaboration [13] 5. Advance pediatric medicine generally [12]  i. To improve diagnostic yields [4]  ii. To develop new drug therapies [5] 6. Fulfill a duty to share [8] 7. Support the learning health care system [1] 8. Prevent data waste [5] 9. To mobilize public health responses [3] 10. Enable verification and reproducibility of genomic results [6] 11. Minimize patient burdens and risks by not duplicating data collection [4]

Table 3. Reasons and subreasons given for the conditions that support “responsible” sharing of children’s genomic and associated clinical data³ [number of mentions] (Sofaer and Strech 2011).

1. Consent [29]  i. Right to recontact/reconsent for continued data use at the age of majority [35]  ii. Broad consent [11]   • Scope of parental authority to provide broad consent [29]   • Joint parental consent [2]   • Posthumous consent [1]  iii. Children’s assent and shared decision making [27]   • Familial implications of data sharing [5]  iv. Consent waivers [8]   • Ethics committee to determine recontact responsibility [7]   • Opt-in/-out for recontact [12] 2. Responsible data protection [18]  i. Data deidentification [25]  ii. Compliance with applicable laws [14]   • Technical strategies for data deidentification to comply with ethical–legal guidelines [10]   • Data linkage and implications for reidentifiability [19]  iii. Special protections for genomic data [9] 3. Balance benefits with informational risks [17]  i. Risks related to loss of privacy [12]   • Via third-party access [15]   • Via genetic discrimination [3]   • Uncertain risks via secondary/future use [4]  ii. Evaluation based on current available evidence [5]   • Evidencing the minimal risk standard [12]  iii. Physicians and investigators responsible for balancing benefits and risks [1] 4. Responsible data management [9]  i. Via interoperability and ease of use [13]   • Need for a shared lexicon for data interoperability [4]  ii. Role of institutions in data management [5]   • Need for harmonized norms [6]  iii. Via mechanisms of mutual accountability for data sharers and users [3]  iv. Clarification of data ownership [7] 5. Responsible data access [10]  i. Via data use(r) restrictions [29] 6. Sustaining patient willingness and trust [10]  i. Methods to enhance willingness and trust to share [7] 7. Continuing education and stakeholder outreach [13] 8. Incentivization and appropriate attribution for data sharers [6]

The interests of the child were often raised conjointly with the types and likelihoods of direct benefits anticipated, particularly when data are collected in clinical contexts. We subdivided the mention, Data sharing benefits, to distinguish among three discrete types of benefits. These include direct (7 subreason mentions), indirect (12 subreason mentions), and both indirect and direct benefits (8 subreason mentions). We also distinguished between intended beneficiaries of those benefits, including individual patients and other/future children. Twelve reason mentions endorsed data sharing on the basis that it may generate indirect benefits to other and future children, including one subreason that sharing manifests the solidarity principle.

Of 12 reason mentions that supported data sharing on the basis that it advances pediatric medicine, 4 subreasons explicitly discussed improving diagnostic yields and furthering pediatric-specific drug development (5 subreason mentions). While data protection and uncertain informational risks were previously cited as reasons against sharing, four reason mentions concluded that informational risks are reduced when sharing data that are “special or that would be difficult or expensive to duplicate” (Birmingham and Doyle 2009, 46). Achieving necessary sample sizes considering the relative rarity of some pediatric conditions substantiated 13 reason mentions given for why scientific collaboration vis-à-vis data sharing was both a scientific and ethical need in practice (Mccormack et al. 2013; Anonymous 2016).

What conditions support “responsible” sharing of children’s genomic and associated clinical data?

We categorized 442 total reason mentions into 8 main reasons, and 25 subreasons for why genomic data sharing should be conditional. Table 2 organizes these subreasons/reasons by frequency of mentions. The most-mentioned subreason/reason identified in our analysis centered on conditions related to consent. We identified 95 total subreason/reason mentions related to data protection, 69 related to informational risk evaluation, 47 concerning genomic data management, and 39 for data access. The involvement of community stakeholder groups in establishing responsible data-sharing practices comprised 3 additional reasons, namely, the importance of community outreach and continued education (13), stakeholder willingness and trust (17), and incentivizing researchers to share via appropriate attribution (6).

Consent models

Debate on whether and for how long parents should authorize continued use of pediatric data at the age of majority (29 subreason mentions) was among the most mentioned practices supporting conditional data sharing; the most heterogeneous consent modality related to broad consent (11 subreason mentions). There was general consensus that broad consent referred to a one-time permission to reuse data for future research, pending requisite ethics oversight (Riggs et al. 2019; Sanderson et al. 2017). Importantly, eight reason mentions discussed various waivers of consent to facilitate greater access to larger datasets while still satisfying permissions outlined in original consent documents. Irrespective of consent model, 27 subreasons supported that children should be involved in decision-making processes in developmentally appropriate ways, and 35 strongly endorsed recontacting children at the age of majority when logistically possible.

Responsible data protection

Nine reason mentions claimed that genetic/genomic data categorically warrants special protections above those required for other health-related data due to its inherent identifiability. This position was irrespective of whether the data contributor was considered part of a vulnerable population. Generally, data protections were considered proportionate to the (i) sensitivities of data shared, (ii) intended purposes for sharing, and (iii) likelihood of unauthorized reidentification (Brothers et al. 2014; Fisher, McCarthy, and Harrington 2013; Hens, Lévesque, et al. 2011; Scholtens et al. 2015; SACHRP 2018; Brothers 2011). Records argued that data protection responsibilities rest with clinical professionals (World Medical Association 2013) or data stewards in research contexts (Manolio et al. 2007). It is also worth noting that privacy and data protection were often discussed interchangeably, despite their conceptual differences (Gostin 1995).

The most mentioned reason (25) involved data deidentification, under which 10 subreason mentions justified how technical deidentification strategies satisfy both regulatory requirements and overall data utility. Although linking multiple datasets together was perceived to enhance the likelihood of reidentifiability (19 subreason mentions), such linkage was overwhelmingly endorsed in combination with adequate data governance and oversight (Canadian Institutes of Health Research Natural Sciences and Engineering Research Council of Canada and Social Sciences and Humanities Research Council of Canada 2014; Lea 2013; Fisher, McCarthy, and Harrington 2013; Lloyd et al. 2015; Platt et al. 2014; Driessnack and Gallo 2011; Giesbertz, Bredenoord, and van Delden 2015; Heeney et al. 2011; Suresh et al. 2005; Mascalzoni, Paradiso, and Hansson 2014; Majumder et al. 2017; Ebner et al. 2016; Wjst 2010; Said et al. 2017; Manolio et al. 2007; Brett and Deary 2014). Our finding is consistent with ongoing initiatives to improve technical infrastructures (Wan et al. 2017; Decouchant et al. 2018; Humbert et al. 2017).

Benefits and risks

We found three distinct types of risk–benefit balances: (1) protecting children and advancing the science of pediatric disease (Bertier et al. 2017; Manhas et al. 2016; McGuire, Oliver, et al. 2011; Burstein et al. 2014; Nature Biotechnology 2015; Hens, Lévesque, et al. 2011; Platt et al. 2017; Fisher, McCarthy, and Harrington 2013; Hens, Wright, and Dierickx 2009; Dowty and Korff 2009; Manolio et al. 2007); (2) respecting individual rights versus serving public interests (Ahrens et al. 2006; Petrini et al. 2012); and (3) preserving data privacy versus activating a right to data access (Golding 2009; Nooner et al. 2012). The majority of records implied that balancing benefits with informational risks (17 reason mentions) is complicated, given the uncertainties of data privacy/and security at the time of data contribution (Dowty and Korff 2009; Bertier et al. 2017; Manhas et al. 2016; McGuire, Oliver, et al. 2011; Platt et al. 2014; Burstein et al. 2014; Fisher, McCarthy, and Harrington 2013; Hens, Wright, and Dierickx 2009; Hens, Lévesque, et al. 2011; Nature Biotechnology 2015; Nooner et al. 2012; Ahrens et al. 2006; Anderson and Merry 2009; Golding 2009; Lunshof and Gurwitz 2013; Manolio et al. 2007; Petrini et al. 2012; Hens, Lévesque, et al. 2011). Threats to loss of privacy were discussed in general terms as reasons on which to condition data sharing (12 mentions), whereas unauthorized data access by third parties was the most mentioned subreason for this loss (15). Appealing to the argument that benefit–risk evaluation should be based on current available evidence, 12 subreason mentions indicated that sharing genomic data was, on the whole, considered minimal risk, especially when data were deidentified (25 subreason mentions).

Responsible data management

The records we synthesized interpreted data management to include administrative tasks that support data formatting and standardization for exchange of data across information systems (9 reason mentions). We identified two management approaches to achieve this level of data management: via mechanisms of mutual accountability for both data sharers and users (3 subreason mentions), and by improving data interoperability and ease of use (13 subreason mentions). Four subreason mentions explicitly endorsed a shared lexicon to facilitate such interoperability (see, e.g., the Global Alliance for Genomics and Health (GA4GH)) 2016). Clarifying data ownership was an additional responsibility of data managers, according to 7 subreason mentions. Several records outlined conflicting viewpoints on whether patients or institutions own the shared data. Authors supporting the former argued that data derived from biospecimens are an extension of the patient’s physical body and patients retain rights to control the data produced. Others reasoned that entities that expend resources to maintain samples/data ultimately own them, as well as all commercial/financial benefits derived therefrom.

Responsible data access

Conditions for responsible access were contingent on the data type, where the data were being deposited, and by whom. Controlled access via user restrictions was the most mentioned subreason (29) supporting responsible access to identifiable or coded data. Many records discussed user authentication and verification (Dyke et al. 2016) as practical means by which to ensure bona fide users—for example, researchers, physicians, and so on can access data of a potentially sensitive nature. Seven of 10 reason mentions described the role of data access committees, data use agreements, and other confidentiality agreements to manage controlled access requests (Pediatric Imaging 2011; Ries, LeGrandeur, and Caulfield 2010; Birmingham and Doyle 2009; Wjst 2010; Fisher, Bromley, and Mansfield 2016; Jernigan et al. 2016; Manolio et al. 2007), while 3 subreason mentions named ethics review boards as primary gatekeepers of data access (Pinto et al. 2015; Mccabe and Mccabe 2011; Tindana et al. 2012).

Stakeholder involvement

Of 10 reason mentions centered on sustaining patient willingness and trust in the research and clinical enterprise, 7 subreason mentions proposed specific methods for improving patient engagement. Direct stakeholder involvement, either as part of the shared decision-making process, or as participants in future public perceptions research, was among the most frequently mentioned methods. These methods were separate from continuing education and stakeholder outreach (13 reason mentions), including educating patients about their rights to access their own medical records that might contain genomic data, and more informed ELSI guidance through public priority setting in research. Unsurprisingly, researchers were among the most frequently discussed stakeholder communities among which to foster responsible data sharing practices. Six reason mentions advocated for improved data attribution as an incentive structure for researchers to share raw genomic data, especially in the context of large national genomics initiatives (Holland et al. 2009; Majumder et al. 2017), pursuant to Article 27(General Assembly of the United Nations 1948) of the Universal Declaration of Human Rights. Information regarding data release and the associated structures governing such release (10 reason mentions) both played central roles in enhancing patient willingness and public trust.

Limitations

Several limitations should be noted when interpreting our results. First, our review questions demanded inclusion criteria that were broad enough to capture relevant ELSI reasons across an interdisciplinary array of literatures. For similar reasons, we did not distinguish between types of data-sharing environments in our analysis, such as research only or clinic only. Such distinctions certainly impact the generalizability of our findings to other pediatric patient populations, legal jurisdictions, and data use cases. Two reviewers resolved discrepancies in the qualitative coding prior to analysis, but it is possible that different reviewers would make different reason assignments. That is, “[reason] types could be made narrower or broader; some broad types cover diverse narrow types; and there may be good reasons to class a narrow type of reason under different broad types” (Sofaer and Strech 2011, 178). As a result, the sum totals for reasons reported in this article would also change. Lastly, the conditions we described that support responsible sharing suggest there are many negative consequences of sharing for children and their families if such conditions are not met. We do not consider these consequences, in and of themselves, reasons against data sharing. Rather higher frequency with which some reason/subreason mentions appear in the literature may suggest, for instance, how the data-sharing community weighs the significance of those conditions or reflect the most publicized reasons for conditional sharing.

Conclusion

This review is an evolving synthesis of ELSI reasons that ground “responsible” data sharing in theory and practice. It not only identifies specific practices that support responsible sharing but also contributes to the policy conversation by highlighting ELSI priorities around which to base empirical research for policymaking purposes. We thematically categorized 58 unique subreasons/reasons for why, and under what conditions, to support responsible data sharing into the following groups: (1) parent and family involvement, (2) stakeholder involvement, (3) benefits and risks and (4) data governance and release (Table 3). Our conclusions from this review later informed an ethical–legal framework of responsible data sharing for pediatric genomics published elsewhere (Rahimzadeh et al. 2018).

Maximizing clinical and scientific utility of children’s data was the most mentioned reason for why children’s data should be shared. The convergence of the evidence-based medicine movement and rapid advances in computational power since first sequencing the human genome in 2003 could have underscored why the aforementioned reason was most frequently discussed. Inadequate data protection was the foremost reason cited for why data should not be shared, as well as a primary theme supporting reasons for conditional sharing. Ensuring that consent reflects the potential for future secondary use via recontact at the age of majority was the most mentioned subreason for conditional data sharing.

Taken together, the complete list of reasons and conditions this review synthesizes makes two practical contributions to the field. First, it guides pediatric research institutions in developing data-sharing policies that align with contemporary ELSI priorities and scientific aims (see, e.g., Taylor et al. 2019). Second, our review supports current and future policy decision makers by identifying data-sharing practices that, based in part on the frequency of reasons identified, could be strengthened. Recontact at the age of majority and data access restrictions are two noteworthy examples in this regard. Future empirical research is needed involving individuals whose genomic data were shared as children to assess how the ELSI reasons identified herein translate across the life course.

Author contributions

Authors VR, GB, and BMK conceptualized the literature review study as part of a doctoral dissertation project. VR and GB screened articles for inclusion and developed the coding strategy for the content analysis of all retained articles. VR led in the drafting of the article and peer review revisions. All authors approved the final version of the article.

Ethical approval

This study was approved by the McGill University Institutional Review Board.

Acknowledgments

The authors sincerely thank the anonymous peer reviewers for their insightful comments and rigorous review of earlier versions of this article, as well Genevieve Gore, chief librarian for the McGill University Department of Family Medicine, for her guidance in developing the search strategy.

Conflicts of interest

No potential conflict of interest was reported by the authors.

Additional information

Funding

Author VR received funding from the Canadian Institutes of Health Research Vanier Canada Graduate Scholarship (CIHR #359258) and BMK was supported by the Canada Research Chair in Law and Medicine (950-231245).

Notes

1 Sofaer and Strech provide examples both broad and narrow reason “types.” We distinguish between these different types as “reasons” and “subreasons,” respectively, in the review. A reason (e.g., avoid exploitation) is thematically related to, but conceptually broader than, a subreason (e.g., avoid exploiting research participants).

2 While return of incidental or secondary findings is considered a form of data sharing, this review of reasons focused on responsible sharing outside the nucleus of care. Similarly, newborn screening programs may imply responsible sharing, yet are obligatory forms of data collection and management mandated by the state and under the auspices of public health.

3 The total number of reason mentions is not the sum total of its subreason mentions for many of the conditions analyzed here. This is because a main reason was counted if the record discussed the condition but did not specify how to enable this condition. For example, we identified 10 reason mentions for data access as a feature of responsible sharing; 29 subreason mentions explicitly suggested user restrictions to facilitate such data access. We assigned the latter (data user restriction) as a subreason under the main reason category responsible data access.