How to deal with uncertainty in prenatal genomics: A systematic review of guidelines and policies

Abstract Exome sequencing (ES) enhanced the diagnostic yield of genetic testing, but has also increased the possibility of uncertain findings. Prenatal ES is increasingly being offered after a fetal abnormality is detected through ultrasound. It is important to know how to handle uncertainty in this particularly stressful period. This systematic review aimed to provide a comprehensive overview of guidelines available for addressing uncertainty related to prenatal chromosomal microarray (CMA) and ES. Ten uncertainty types associated with prenatal ES and CMA were identified and defined by an international multidisciplinary team. Medline (all) and Embase were systematically searched. Laboratory scientists, clinical geneticists, psychologists, and a fetal medicine specialist screened the papers and performed the data extraction. Nineteen papers were included. Recommendations generally emphasized the importance of trio analysis, clinical information, data sharing, validation and re‐analysis, protocols, multidisciplinary teams, genetic counselling, whether to limit the possible scope of results, and when to report particular findings. This systematic review helps provide a vocabulary for uncertainties, and a compass to navigate uncertainties. Prenatal CMA and ES guidelines provide a strong starting point for determining how to handle uncertainty. Gaps in guidelines and recommendations were identified and discussed to provide direction for future research and policy making.

previously undetectable genetic anomalies. 1,2 Currently, some countries are introducing ES in prenatal genetics in cases of fetal malformations, 3,4 generating large amounts of information on the genome of the unborn child compared to karyotyping, CMA or targeted genetic testing panels. This raises the concern for an increased chance of an uncertain finding, such as genes or variants of uncertain significance (GUS/VUS). 5 Filters based on the presenting phenotype and for genes with valid phenotypic associations minimize this uncertainty, but may decrease the diagnostic yield. Alternatively, a more open analysis increases the chance of a diagnosis, but also increases the chance of finding an uncertain result. 6 This can be a challenge for all stakeholders involved. Laboratory specialists are challenged to interpret results. Clinicians are challenged to return results that have an element of uncertainty in a way that is understandable to parents experiencing an extremely stressful diagnostic process.
Parents are challenged to apply meaning to this information, and decide about the course of their pregnancy based on results that may not provide the certainty that they had hoped for. 7,8 Current implementation practices have therefore focused on developing strategies to deal with these uncertainties, both in laboratory and clinical settings. 9,10 One such strategy is the development of recommendations and guidelines. This systematic review provides an overview of guidelines and recommendations for practice that are available to support professionals dealing with uncertainty in routine clinical prenatal diagnostics.

| Proposed definitions of uncertainty
The authors conducted multiple discussions on the possible types and definitions of uncertainties that may be encountered during the process of providing ES diagnostics in clinical settings. Ten types of uncertainty associated with prenatal ES, from laboratory and clinical perspectives, are proposed (see Table 1 and S2). In this review we used these definitions to classify recommendations addressing uncertainty and developed a framework for analysing the papers included in this systematic review.

| Systematic review of guidelines and recommendations
We conducted a systematic review following PRISMA criteria 11 to identify guidelines and recommendations addressing uncertainties associated with prenatal diagnostic (genome wide) testing. As prenatal ES is a newly introduced technology we anticipated that there may not be many guidelines available yet and we therefore also included guidelines for prenatal CMA, which might also be relevant to prenatal ES. Diagnostic yield Likelihood to provide a diagnosis.

2) Uncertainties related to incomplete knowledge
Gene-disease correlations Phenotype associated with a variant is unknown (prenatally and postnatally), including its variability in expression and the natural history.
How a genetic anomaly presents prenatally New phenotypes associated with genes that have limited natural history information in the prenatal period. Or postnatal phenotype associated with pathogenic variant (e.g. mental disability) is not or only partially recognized prenatally.
Pathogenicity and variants of unknown significance (VUS) Insufficient evidence to classify variants as (likely) benign or (likely) pathogenic.

3) Uncertainties unrelated to the primary clinical question
Secondary findings Pathogenic variant(s) not related to indication of testing, but intentionally searched for as an additional analysis next to the standard test.
Incidental findings Pathogenic variant(s) not related to indication of testing and are identified inadvertently (unexpected result).

4) Uncertainties related to the technology
Technical validity of a result False positives, false negatives, insufficient depth of read.
Possible incomplete result For example, One autosomal recessive variant compatible with the fetal phenotype, but no second variant is identified.

5) Uncertainties related to the condition
Incomplete penetrance Chance that a pathogenic variant presents with symptoms. Not everyone with the same genetic predisposition will be affected (reduced or incomplete penetrance).

| Search criteria and study selection
The search was conducted with professional assistance of the We excluded non-English, retired, and exclusively postnatal papers, as well as papers exclusively on specific syndromes, prenatal screening, predictive testing, preimplantation diagnosis, or genetic testing procedures (also see Figure 1). Disagreements were discussed until consensus was reached on final inclusion.

| Data analysis
To assess the quality of identified publications, a modified version of the Appraisal of Guidelines for Research and Evaluation checklist was used. 15 Each of the seven assessors scored papers independently. The F I G U R E 1 PRISMA flow diagram quality scores were not used as a criterion for inclusion. Qualitative data were then analysed using a framework analysis approach. 16 Each team integrated guidelines according to uncertainty type (Table 1). All seven assessors subsequently convened to discuss the analysis and come to agreement on the coding. Disagreements were resolved by consensus.

| RESULTS
Nineteen papers were eligible for quality assessment and data extraction (see Figure 1).  Table 3).

| Diagnostic yield
The analysis resolution defines the diagnostic yield of CMA, which should be incorporated in the laboratory report and in the informed consent. 26 Resolution recommendations varied between 200 kb resolution, 24,29 and 400 kb to minimize VUS and maximize yield. 27 19 Overall evidence for variant or gene implication should be assessed and integrated, including primarily statistical support (genetic analyses) and, if possible, informatic (e.g. conservation and predicted effect on function) and experimental evidence (e.g. functional studies). 18 To confidently implicate a new gene in disease, these genes should be replicated in independent families or population cohorts.
Null models (for e.g. de novo variants) should be used to compare against when detecting pathogenic variants, while also considering potential confounders (e.g. sample or gene size). 18 Detailed information on the phenotype is necessary to interpret the genotype, making clinical information important for the laboratory analysis as well as in the decision to report (to the clinician), 20 and should be available to the MDT before sequencing. 10,22 Providing all clinical information is not always feasible when time is limited. 10 If there is compelling information implicating a variant in a proband's phenotype it could be included in the test results. 3 Sharing phenotypic data paired with variants in databases further improves gene and variant interpretation. 10  Collecting (phenotypic) data allows for improved correlations between genetic data and a potential disease, although this is complicated when crucial clinical characterization of the fetus is not possible or non-specific. 29 Laboratories should set up systems where this clinical information can be submitted. 4 If the genetic data are the only other information available, post-test counseling was considered especially important. 20

| Pathogenicity and variants of unknown significance
Careful classification of pathogenicity of variants Careful classification of variants is crucial for correct reporting.
Sequence variants should be reported using the classification system T A B L E 2 Characteristics of the selected papers   Trio Data-S = Data storage or data sharing (and re-evaluation) is recommended.

Recommendations pertaining to lab technicians and clinicians
Limit-S = The scope should be limited (e.g. by filtering, lowering resolution).
Limit-N = The scope should not be limited.

MDT
= Discuss within multidisciplinary team how to handle the uncertainty.
Clin-Info = Supply clinical information/use multiple lines of evidence.
Protocol = A protocol, policy or set of agreed standards should be in place.

Trio
= Also sequence the probands' parents.

Recommendations pertaining to patients and clinicians
Pre-T = Obtain informed consent or information on diagnostic possibilities should be stated in informed consent document, patients leaflet or given during pre-test counselling.
Post-T = Issue to discuss during post-test counselling after reporting. Information about limitations of the test or a disclaimer should be included in the report (to the clinician).
(Not-)R* = Advice to (not) report, with specific exceptions. interpretation, (6) significance of the finding(s). 32 It was recommended that clinical laboratories submit (de-identified) variant data to public databases, including data on potential pathogenicity, VUS, relevant clinical information, and frequency data. 3,10,17,19,22 Information in these databases should be updated and re-evaluated continuously. 17,18 Such databases and sharing within these databases will, over time, reduce the amount of VUS as more data are pooled together on unknown variants improving identification and classification of variants. 10,19 This data sharing should be an integral part of reporting and thus supported by an international committee. 32 Old data should not be routinely re-analysed. 10 Patients should be informed about the possibility of finding/ reporting VUS and other potential outcomes beforehand. 3,4,21,23,29,31 Pre-and post-test counselling were again considered important in guiding patients through uncertainty. 4,25,31 During pre-test counselling, a genetic healthcare professional should obtain clinical information and consent. 29 The types of variants that will be reported should be included in the consent document, including how this is different for the reporting policy of incidental findings (IF's) and secondary findings (SF's). 17 Parents should be able to opt-out of receiving variants in non-disease genes. 23  Generally, laboratories should only report known or (likely) pathogenic SF's. 17 Limiting the scope of testing may be possible for complex findings (e.g. pseudogenes), but this may be undesirable for SF's on the ACMG list. 33,34 In those cases only variants on the active copy of the gene should be reported, and validated with additional testing (i.e. Sanger sequencing) if needed. 17 Vears et al (2018) 17 In trio analysis, every sequenced person (other than the proband) should give separate consent. 17 The chance of finding SF's and whether SF's are included or excluded should be discussed during pre-test counselling and during the informed consent procedure. 4,23,29 Offering an opt-out of receiving SF's (from the ACMG list) is recommended. 23  should not be reported. 10,24,28 Prenatal ES guidelines stated to (1) not report variants without a known fetal or childhood phenotype, (2) not report when heterozygous for autosomal recessive disorders or Xlinked disorders, and (3) report highly penetrant pathogenic IF's that may cause moderate or severe early-onset disorders. 23 Armour et al (2018) suggested that if parents indicated they want to know all relevant results, laboratories should report pathogenic IF's. 24 Other papers recommended to only report known and (likely) pathogenic IF's, 3,17,23 or (likely) pathogenic, actionable IF's. 21,28 Early onset disorders are usually reported, but late onset disorders are not. 27 An opt-in for treatable late-onset disorders, but not for non-treatable late-onset IF's may be offered. 29 Deletions in genes found through CMA that are associated with recessive disorders not fitting the fetal phenotype should only be reported if the carrier frequency is higher than 1:50. 27 Vears et al (2018) proposed to report heterozygosity of recessive disorders, but to honor an adults' informed consent and/or choice to opt-out, even if the IF may be relevant to their health. 10 Pre-test counseling should address the possibility of IF's and which IF's are (not) reported, 3,4,17,19,[23][24][25][26]29,31 Pre-test counseling should also discuss the possibility to detect non-paternity and how this influences the interpretation of the genetic results. 29 Matthijs et al (2016) recommended supplementing this information with written leaflets or online information. 19 Patients should be able to opt-in or out of receiving IF's and their choice should be clearly indicated 3,10,17,23 to prevent disclosure of unwanted results. 29 Extensive post-test counseling should be offered when reporting pathogenic IF's. 24 Difficult IF cases need to be discussed by a MDT on a case-bycase basis. 3,4,[27][28][29] The psychological impact and potential insurance risks of receiving such findings should be taken in consideration. 17 If treatable pathogenic IF's are reported, it should be made clear that these findings are not associated with the indication of testing. 28

| Technical validity of a result
In case of a finding based on insufficient read-depth or ''noisy'' CNV's validation by a different technique was recommended. 26,31 To reduce the impact of unwanted false positives meticulous evaluation and subsequent re-evaluation of candidate variants in databases was deemed important. 18 If possible, estimated diagnostic specificity may be provided in the report to the clinician to indicate the risk of a false positive. 32 The limitations of enrichment methods and sequencing platforms should determine whether additional testing is required and, if so, what type of additional testing is required. 19 Disclaimers and limitations concerning coverage should be clearly described in the laboratory report. 17,31

| Possible incomplete result
This situation was not specifically discussed by any of the guidelines that were reviewed.

| Penetrance and expression
Penetrance and expression were rarely discussed in ES guidelines.

| Local policy and protocols
Local decision-making was highlighted as an important factor in handling uncertainty. The decision to report problematic variants (VUS, IF's, SF's, etc.) for example, often depends on MDT and/or policies that are put in place either by a country or laboratory. Guidelines were therefore often more general and non-specific (e.g. report treatable (late-onset) IF's, but not non-treatable IF's). Rarely were recommendations given on what a policy should specifically look like or include. This acknowledges that local policies exist on how to handle uncertainties and recognizes that there are differences between laboratories and healthcare contexts. With global institutions using different protocols and policies that are often only known internally, it is difficult to achieve an overview of all the local policies that currently exist. Developing more universal guidelines incorporating policies that can apply within each institution is therefore challenging and may even be unnecessary.

| Providing clinical information
The importance of clinical information has been mentioned in the context of most uncertainty types (see Table 3). However, it should be kept in mind that it is not feasible to provide all the clinical information on the fetus due to the situation depending on imaging techniques, the fact that development and function of organs is incomplete, for instance of the brain, and time constraints. New features may become evident after birth, which may lead to re-interpretation of prenatal results. 35

| Use of databases and data sharing (including validation and re-analysis)
Clinical and empirical evidence is important to classify causative results.
Empirical data should be fed into and retrieved from shared databases, which serve as a platform for knowledge building and can help interpret variants that are of uncertain or unknown significance. 10,19 However, some issues on use of databases and data sharing were not explicitly discussed in the guidelines: (1) studies may be biased, e.g. data of a particular population, (2) curation of the databases is a very important factor and data should be validated and updated continuously at a rapid pace. This requires extensive effort and depending on available resources may not always be feasible, 36 (3) data sharing can be done nationally and/or internationally. Sharing data within an extended network (i.e. centralizing) will enable optimal knowledge building, but requires calibration of different systems. Yet, there are already some publicly available international databases that are used to share anonymized pathogenic variants and data originating from healthy individuals. 29

| Multi-disciplinary team
An MDT approach is the norm when it concerns handling particularly uncertain findings. The MDT serves to aid the lab specialist and referring clinical geneticist to interpret results and determine whether to report uncertain findings (e.g. where pathogenicity is unclear). The MDT should include representatives of all needed disciplines that can bring technical, empirical, and clinical insights together. This is mostly discussed locally, but the MDT can be useful on a larger scale as well. Extending the MDT network to include, for example, other laboratories or centres can help with optimizing data sharing and expands the consultation options. 29 The latter is also the case when using MDT on an international level, enabling collaboration of an expert group that can then serve as an additional resource. 28

| Pre-and post-test counselling
Generally, recommendations stated that uncertainties should be addressed during pre-or post-test counselling. Also, pre-and posttest counselling should enable parents to make informed decisions about which results they wish to receive. Only two papers offered more specific direction on which points should always be discussed in pre-and/or post-test counselling and in which format(s) (e.g. written and orally). [28][29][30] Several papers agreed counselling should be provided by a specialized genetic professional. 3,4,23,25,30,32 Offering psychosocial care during counselling was rarely recommended, while the prenatal setting causes significant psychological distress as parents are challenged to make (irrevocable) decisions about their pregnancy. 38

| Implications of guidelines for patients
Although guidelines were often rigorously developed, input of parents on guidelines was rarely sought out. This may be the case, because these guidelines are aimed at the healthcare professionals. Uncertainty that may have originated in the laboratory for example occurs mostly behind the scenes before the result reaches the parents. However, as guidelines reflect different views of healthcare professionals on how to handle uncertain results (e.g. reflected in the least amount of consensus on VUS, which is associated with the largest amount of uncertainty), they also reflect the views of patients. Especially when the importance of counselling is highlighted by most recommendation papers, it may be worthwhile to include the patient as part of the MDT when developing guidelines.

| Strengths and limitations of the systematic review
This systematic review was strengthened by the participation of an MDT of experts in reviewing the guidelines. Another strength is the proposal of 10 distinguishable uncertainty types. International MDT discussions were held until there was consensus on clear and mutually exclusive definitions, which were used to identify guidelines as well as provide vocabulary to internationally discuss prenatal ES uncertainties.
A limitation of the systematic review is the comparison of prenatal CMA and ES guidelines. CMA was included because a lack of guidelines was expected for prenatal ES. However, there are diagnostic differences between CMA and ES. ES has a higher resolution, which has the advantage of widening the diagnostic yield, but at the same time there is a greater chance to encounter VOUS, and IF's.