Applying High-Resolution Variant Classification to Cardiac Arrhythmogenic Gene Testing in a Demographically Diverse Cohort of Sudden Unexplained DeathsCLINICAL PERSPECTIVE
Background—Genetic variant interpretation contributes to testing yield differences reported for sudden unexplained death. Adapting a high-resolution variant interpretation framework, which considers disease prevalence, reduced penetrance, genetic heterogeneity, and allelic contribution to determine the maximum tolerated allele count in gnomAD, we report an evaluation of cardiac channelopathy and cardiomyopathy genes in a large, demographically diverse sudden unexplained death cohort that underwent thorough investigation in the United States’ largest medical examiner’s office.
Methods and Results—The cohort has 296 decedents: 147 Blacks, 64 Hispanics, 49 Whites, 22 Asians, and 14 mixed ethnicities; 142 infants (1 to 11 months), 39 children (1 to 17 years), 74 young adults (18 to 34 years), and 41 adults (35 to 55 years). Eighty-nine cardiac disease genes were evaluated. Using a high-resolution variant interpretation workflow, we classified 17 variants as pathogenic or likely pathogenic (2 of which were incidental findings and excluded in testing yield analysis), 46 novel variants of uncertain significance, and 130 variants of uncertain significance. Nine pathogenic or likely pathogenic variants in ClinVar were reclassified to likely benign and excluded in testing yield analysis. The yields of positive cases by ethnicity and age were 21.4% in mixed ethnicities, 10.2% Whites, 4.5% Asians, 3.1% Hispanics, and 2% Blacks; 7.7% children, 7.3% in adults, 5.4% young adults, and 2.8% infants. The percentages of uncertain cases with variants of uncertain significance by ethnicity were 45.5% in Asians, 45.3% Hispanics, 44.20% Blacks, 36.7% Whites, and 14.3% in mixed ethnicities.
Conclusions—High-resolution variant interpretation provides diagnostic accuracy and healthcare efficiency. Under-represented populations warrant greater inclusion in future studies.
Sudden unexplained death (SUD) is a natural death of a previously healthy individual whose cause remains undetermined after scene investigation, complete autopsy, and medical record review. SUD affects children and adults, devastates families, challenges medical examiners and is a focus of research for cardiologists, neurologists, clinical geneticists, and basic scientists.1,2 As sequencing technology advanced rapidly from single gene testing to wider gene panels, genetic study has emerged as a leading tool in postmortem diagnostics in SUD. Testing yields from larger SUD cohort studies varied from 10% to 30%.3–7 Besides demographic and geographic differences of cohorts, we also see differences in forensic investigation among medical examiner offices, and the subjectivity of variant interpretation contributing to the variation in testing yield between groups. Conflicting variant classification is widespread in the literature and in ClinVar, which has become a critical issue in molecular diagnostics.8,9 A recent study showed there was only 34% concordance for variant classification across 9 molecular diagnostic laboratories8 using the guidelines for the assessment of variants in genes associated with Mendelian diseases published in 2015 by American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology (AMP).10 One of the recommendations from this study, on reviewing the causes of the inconsistency, is to develop disease-specific allele frequency thresholds to enable lowering of the stand-alone benign criteria from a minor allele frequency (MAF) of ≥5% to values specific to each disorder.8
See Editorial by Hamilton et al
A statistical framework of high-resolution variant interpretation—termed such as it utilizes details of genetic principles—which accounts for specific disease inheritance mode, prevalence, allelic heterogeneity, and reduced penetrance in the determination of the maximum tolerated allele count (AC) in ExAC for a likely disease-contributing variant, was developed.9 This framework removed two-thirds of variants from consideration, compared with the lenient MAF of 0.1% for autosomal dominant diseases,11 without discarding true pathogenic variants that would have been missed if filtering for singleton variants exclusively,9 which is a common practice in population genetics.
We adapted this statistical framework9 in classifying the variants identified through panel testing of cardiac channelopathy and cardiomyopathy genes in a large demographically diverse cohort where all cases underwent comprehensive, standardized investigation in the United States’ largest office of chief medical examiner. Applying maximum tolerated reference AC in gnomAD, a larger population database than ExAC, containing 277 264 sequenced chromosomes from various ethnic groups (eg, Europeans, Latinos, Asians, Ashkenazi Jewish, etc), we reclassified those variants with high AC yet are currently called mostly as pathogenic in ClinVar. Using the 2015 ACMG-AMP guidelines, we classified variants into pathogenic or likely pathogenic variants (P/LP), novel or rare variants of uncertain significance (VUS), and benign or likely benign variants (B/LB) in the cohort. We present the demographic variations of testing yields in this cohort, and remove incidental findings from yield determination that do not seem to directly impact the cause of death determination. We emphasize the importance of accurate variant interpretation as it directly affects diagnostic reliability as well as efficacy of family care by ensuring targeted testing and cost-effectiveness in broader health care. Our study also highlights the importance of conducting additional research on novel and rare VUS uncovered in under-represented populations.
Materials and Methods
The authors declare that all supporting data are available within the article and its online supplementary files.
In the New York City Office of Chief Medical Examiner, forensic investigations in sudden death include scene investigation and family interview (by certified physician assistants), complete autopsy, cardiac pathology and neuropathology examinations, toxicological tests, microbiological tests (in infants), metabolic screen tests (in infants), and medical record reviews. Cases defined as SUD had negative or unremarkable results from the studies described above. Two hundred ninety-six SUD cases investigated in New York City Office of Chief Medical Examiner from 2001 to 2014 met these criteria. Data collected for each case include age, ethnicity, sex, body mass index, death circumstance (sleeping or strenuous activities), a family history of sudden death or cardiac arrhythmia, a personal history of seizure, fainting, heart murmur, or cardiac arrhythmia. However, data on family or personal history is often difficult to obtain, resulting in incomplete data entry for those categories. Race/ethnicity as defined by the US Census indicates the interpretation of what families of a decedent considered himself or herself to be: Black, Hispanic (Mexican, Puerto Rican, Cuban, Dominican, etc), White, Asian, and mixed ethnicity. This study is not regulated by 45 Code of Federal Regulations Part 46 because only cadaver specimens were used. Office of Chief Medical Examiner approved this study for diagnosis of the underlying cause of SUD.
Deep Next-Generation Sequencing of Targeted Genes and Variant Confirmation by Sanger Sequencing
Genome DNA was extracted from postmortem tissue samples or bloodstain cards from the 296 SUD cases using the Qiagen DNA kit and the M48 Biorobot (Qiagen, Germany). HaloPlex custom kit (Agilent Technologies) was used for target gene enrichment and Illumina Miseq for deep sequencing according to the manufacturers’ protocols. Paired-end sequencing (150 bp) produced a mean sequencing coverage of 1000×, >98% of the target base coverage ≥50×, and <0.1% have zero coverage. Variants in 89 cardiac disease genes responsible for cardiac channelopathy and cardiomyopathy were annotated and classified (Table I in the Data Supplement). Sanger sequencing confirmed all variants classified as P/LP, as well as novel VUS.
Sequencing Data Analysis
Agilent SureCall software (version 2.0) was used to perform adaptor trimming and alignment to the human-genome reference (GRCh37/hg19). Strand-ngs (version 2.5) was used for sequencing data quality filtering, reads local realignment, base quality recalibration, and variant calling with comparison to dbSNP138. The confidence of variant calling was set at the minimal coverage of 50 reads and the heterozygote variants’ percentage between 30% and 70%. Nonsynonymous variants—missense and equivalent (including in-frame indels, start lost, stop lost, and mature mRNA altering), and loss-of-function variants (nonsense, essential splice site, and frame shift) were annotated with Ensembl GRCh37.
Variant Interpretation Workflow
Once variants were annotated, we classified them as P/LP, VUS, or B/LB using the interpretation framework described by ACMG and the AMP in 2015,10 preceded by the evaluation of the AC of a variant in gnomAD (Figure 1).
We used disease-specific maximum tolerated AC in gnomAD (see section below) as a cutoff for P/LP versus VUS: the AC for a high penetrance P/LP variant is expected to be lower than the maximum tolerated AC. We then applied 2015 ACMG-AMP guidelines to categorize those low-AC variants into P/LP or VUS based on whether there are evidence collectively supporting the pathogenic role of a variant.
We used the prevalence of a disease as a cutoff for VUS versus B/LB: the AC for a variant with possible deleterious effect is not expected to exceed the prevalence of a disease tested in this study as these conditions are genetically heterogeneous and no single variant will be the cause of all cases. When the variant AC is less than the prevalence of the disease, 2015 ACMG-AMP guidelines were applied to separate those low penetrance VUS with possible clinical relevance from likely benign variants.
Evidence collected for each variant and the strength of each type of evidence is coded using 2015 ACMG-AMP guidelines.10 The variant effect predictions were performed by dbNSFPv2.9.1 (eg, SIFT, PolyPhen 2, LRT, MutationTaster, Fathmm, Gerp++ and Phylop, CADD). In silico predictions were used as one form of evidence, for instance, when 8 programs consistently support a deleterious effect on a gene/gene product (PP3).
Statistical Framework in Determining Maximum Tolerated AC in gnomAD
We applied the statistical framework published by Whiffin et al9 to compute the maximum tolerated AC in gnomAD12 (Table 1). The cardiac arrhythmogenic diseases tested, such as long QT syndrome (LQTS), Brugada syndrome (BrS), catecholaminergic polymorphic ventricular tachycardia, arrhythmogenic right ventricular cardiomyopathy, hypertrophic cardiomyopathy, and dilated cardiomyopathy, are highly heterogeneous, with the majority of pathogenic variants manifesting in an autosomal dominant manner with reduced penetrance.
The prevalence of diseases used in Table 1 is the highest reported. The maximum allelic contribution is based on the commonest high penetrant causative variants reported for each disease,13 except for catecholaminergic polymorphic ventricular tachycardia, which is a rarer disease (1 in 10 000) for which there is no commonest pathogenic variant. The maximum allelic contribution for catecholaminergic polymorphic ventricular tachycardia is set at 10%, consistent with what was used by Whiffin et al.9 The penetrance is set to 0.5. The maximum credible population AF was calculated as: maximum credible population AF=prevalence×maximum allelic contribution×1/penetrance.
To calculate the maximum tolerated AC in gnomAD, which contains sequenced 277 264 chromosomes of 138 632 individuals (123 136 whole-exome sequencing and 15 496 whole-genome sequencing), a 1-tailed 99.9% confidence interval of a Poisson distribution was applied (Table 1). The maximum tolerated reference AC is calculated using the web tool (http://cardiodb.org/allelefrequencyapp).
We chose not to discard variants using the precomputed filtering allele frequency in ExAC published by Whiffin et al,9 as it risks losing low penetrance VUS. For genes that have been associated with multiple conditions, such as SCN5A in LQTS and BrS, variant AC is compared with the maximum tolerated AC in gnomAD calculated using the higher prevalence of disease.
Testing Yield in the Context of Cohort
By applying the variant interpretation framework described above, we downgraded several pathogenic variants reported in ClinVar14 to B/LB or VUS. We also removed incidental findings, where a P/LP variant is not considered to be relevant to the cause of death. The testing yield (percentage of cases that had P/LP) is calculated by removing the incidental findings as well as reclassified benign or VUS variants submitted previously to ClinVar as P/LP.
Characteristics of the Cases
The overall demographics of the 296 decedents are in (Table 2): 147 Blacks (49.7%), 64 Hispanics (21.6%), 49 Whites (16.6%), 22 Asians (7.4%), and 14 mixed ethnicities (4.7%); 142 infants (birth to 11 months), 39 children (1 year to 17 years), 74 young adults (18 to 34 years), and 41 adults (35 to 55 years); 125 females and 171 males.
Reclassification of Previously Reported Pathogenic Variants
In addition to the cardiomyopathy gene variants that were reclassified as benign or VUS by Whiffin et al,9 we reclassified 9 channelopathy gene variants submitted previously to ClinVar as P/LP to LB (Table 3), using the statistical framework and variant interpretation workflow described in Methods. An example is the NP_001120965.1:p.Glu1449Gly variant, also referred to as p.Glu1458Gly or p.Glu1425Gly, in ANK2. Evidence supporting the pathogenicity of this variant included cosegregation of the variant in a large kindred with autosomal dominant LQTS and functional studies in mouse model.15 ANK2 is a LQT gene, the maximum AC in gnomAD for a pathogenic LQT variant is 8 (Table 1). This variant was found in 96 of 126 426 European (non-Finnish) alleles and 143 of 276 786 alleles tested in gnomAD. In addition, the highest subpopulation MAF is 0.11% in Latinos, which exceeds the prevalence of the LQTS of 1/2000, making it highly likely benign. In addition, this variant is consistently predicted as benign by 8 programs; a Danish population-based study revealed that this variant is not associated with prolonged QT interval, syncope propensity, or overall mortality.16 In summary, the evidence strongly supports the benignity of this variant.
It is important to evaluate the allele frequency in all subpopulations as well as the total populations in gnomAD for comprehensive variant interpretation. For instance, a variant in KCND3 (Table 3), NP_004971.2:p.Leu450Phe, was found in a Black male infant in our cohort. Evidence supporting the pathogenicity of this variant included identification of this variant presented in a single patient with BrS among 86 unrelated patients and functional characterization of the effect of the variant in the HEK293 cells.17 This variant was not found among the 23 746 African alleles tested in gnomAD, but it has an MAF of 0.11% in Other, which exceeds the prevalence of the BrS of 1/1000, making it highly likely benign. In addition, this variant is not predicted consistently as benign or deleterious by 8 prediction programs. Therefore, we classified this variant as LB because of the high MAF.
Those reclassified benign variants in Table 3 were excluded from the analysis of testing yield for the positive cases presented below.
P/LP and VUS Identified in This Study
In our cohort, we identified 17 pathogenic and likely pathogenic variants that were studied previously (referenced by PubMed identifier number in Table 4): all variants are below the maximum AC in gnomAD and not enriched in any subpopulation group. We classified 2 variants as incidental findings that were excluded from the analysis of testing yield for the positive cases: the first is a variant in MYBPC3 found in case 2 (Table 4). This variant is typically age related and has not been reported to manifest as hypertrophic cardiomyopathy or sudden death in people younger than 20 years.18 We found this variant in a 1-month-old infant who had negative autopsy including a normal heart. It is unlikely that this variant is the cause of the death in the infant through the mechanisms that cause death in adults. The second variant is in GLA found in case 16, a 42-year-old Hispanic male who lacked any pathological evidence of lipid storage in Fabry disease, and the hypertrophic cardiomyopathy finding in the deceased is likely caused by a separate mechanism that was not uncovered by this study. However, we did not consider the p.Y221X variant in Plakophilin 2 (PKP2) detected in a 3-month-old infant (Case 5) without arrhythmogenic right ventricular cardiomyopathy (mixed ethnicity) as an incidental finding. Our unpublished data showed that sodium current amplitude (INa) in HL1-PKP2-KD cells transfected with this PKP2 variant failed to rescue the INa deficit observed in PKP2-KD cells, which is in contrast with results from PKP2-WT-transfected cells, supporting the pathogenic role of this variant through a similar mechanism affecting BrS. In addition, we did not consider the variant in MYBPC3 found in case 10, a 23-year-old Hispanic male as an incidental finding as the decedent has early sign of cardiac hypertrophy and fatal cardiac arrhythmia may occur in concealed stage of cardiomyopathy.
We also identified 46 novel (Table II in the Data Supplement) and 130 VUS present at low AC in gnomAD (Table III in the Data Supplement). One novel variant, p. L388M in SCN10A, worth discussion here is related to a reported variant, p.L388P in SCN10A, which has been functionally characterized as a BrS variant.19 However, because the role of SCN10A in BrS warrants further evaluation, we are considering this variant a VUS pending further family study.
Two genes, SCN5A and RYR2, were enriched with P/LP, novel and rare VUS. We found 7 P/LP, 2 novel, and 8 rare VUS in SCN5A and 3 P/LP, 5 novel, and 11 rare VUS in RYR2. Schematic representations of the amino acid positions affected by P/LP variants and VUS with the correlations of those variants in the functional domains of the respectively encoded proteins are shown in Figure I in the Data Supplement. Disease-causing variants in SCN5A and RYR2 have been reportedly associated with sudden deaths during sleep as well as those triggered by exercise or emotional distress.20 The ExAC data set also showed a high degree of intolerance to variation in RYR2 and SCN5A (Figure II in the Data Supplement), with the Z scores 5.21 and 2.53 for missense variants, respectively, and probability of loss of function intolerance=1 for both genes (pLI≥0.9 is for extremely loss of function intolerant set of genes).
Demographic Distribution of Positive, Uncertain, and Negative Cases
Among the 296 cases, 14 cases had at least 1 P/LP variant (positive cases, Table IVa in the Data Supplement), 124 cases had at least 1 novel or rare VUS (uncertain cases, Table IVb in the Data Supplement), and 158 cases had negative results (negative cases, Table IVc in the Data Supplement). The testing yields (percentage of positive cases) from high to low by ethnicity (Figure 2A) and age (Figure 2B) were 21.4% in mixed ethnicity, 10.2% Whites, 4.5% Asians, 3.1% Hispanics, and 2% Blacks; 7.7% children, 7.3% in adults, 5.4% young adults, and 2.8% infants. Furthermore, the percentage of uncertain cases varied by ethnicity (Figure 2A) and age (Figure 2B) were 45.5% in Asians, 45.3% Hispanics, 44.2% Blacks, 36.7% Whites, and 14.3% in mixed ethnicity; 48.6% young adults, 43.9% adults, 39.4% infants, 35.9% children. The percentages of negative cases by ethnicity and age were presented in Figure 2A and Figure 2B, respectively. Additionally, as it is difficult to distinguish missing family history versus a true negative family history in this retrospective postmortem study, we did not attempt to correlate the history with testing yield here, which instead should be one of the focuses in a prospective study.
Accurate variant interpretation directly affects the precision of clinical diagnosis and enables effective heath care in the genome medicine era. Setting a robust, high-resolution, disease-specific variant interpretation statistical framework and workflow is a crucial first step toward avoiding false-negative (because of retaining only novel or singleton variants) or false-positive results (because of overly lenient in MAF cutoff).
Adapting the statistical framework recently developed,9 we further developed disease-specific variant interpretation workflow using gnomAD as reference population. We chose gnomAD over ExAC for 2 reasons (1) the number of unrelated individuals sequenced in gnomAD is twice as high as that in ExAC (138 632 versus 60 706 individuals), so it provides better estimate of MAF; and (2) gnomAD includes specific ethnic groups, such as Ashkenazi Jewish, that are not specified in ExAC, and increases the size of under-represented populations in ExAC, such as Latinos, and Other which can provide power and generalizability to the results.
To facilitate variant interpretation, we used 2 cutoffs, 1 for P/LP versus low penetrance VUS, and 1 for VUS versus B/LB (Figure 1). The precomputed filtering AF in ExAC used by Whiffin et al9 can be used to automate the candidate variant selection process; we think similar work is needed for gnomAD for large-scale whole-exome bioinformatics pipelines.
We reclassified 9 previously called P/LP variants in ClinVar to LB (Table 3). It is important for the original submitters to reevaluate their calling, as a false-positive presents challenges to other diagnostic laboratories in reporting the significance of the same variant, and even more concerning, a physician may treat families with benign variants whose status as P/LP remains outdated.
The recognition of incidental findings in large-scale genetic testing is important and interpreting their contribution to the cause of death should be made with great caution. Taking the pathogenic variant in MYBPC3 found in the infant (case 2; Table 4) as an example, although the testing result is useful to guide any at-risk adult family members for clinical screening of hypertrophic cardiomyopathy to prevent sudden death, this variant is unlikely to be the cause of the death in the infant. Therefore, we made a clear distinction between testing positive of a P/LP variant and solved cases, where the cause of death remained undetermined.
Alongside the small number of pathogenic variants, we identified in the cohort, we uncovered a large number of novel (Table III in the Data Supplement) and rare VUS (Table IV in the Data Supplement) that should not be discarded as genetic noise before clinical and function studies prove they are benign. This is particularly important in non-Whites, as our low testing yield in non-White minorities (Figure 2A) correlates with the dearth of reported studies in those populations in the literature—it is reflected by the discrepancy in uncertain cases between non-White and White ethnicities, as a majority of studies are based on populations of European descent.3–7 Our study highlights the critical need for more studies of ethnic minorities in future research for both case information and normative control data for these populations. At the same time, we emphasize that these VUS should not be considered diagnostic and lead to treatment decisions for asymptomatic family members who carry them, or for making prenatal decisions.
Compared with testing yield reported from other large cohort studies,3–7 our testing yield (Figure 2) is lower, for which there are several possible explanations: (1) as a result of our high-resolution variant interpretation workflow, we removed half of the variants from consideration compared with the lenient frequency of 0.1% for autosomal dominant diseases, without discarding true pathogenic variants. Using the interpretation workflow presented here, several previously reported P/LP variants3–7 would be reclassified either as VUS or benign. (2) We have removed incidental findings of P/LP that is unlikely to be the cause of death—a practice not done previously. (3) Our cohort has a more diverse representation of ethnicity and age (Table 2) than previous studies. The high testing yield in the mixed ancestry group warrants validation from additional studies by including larger number of participants with mixed ancestry background as our finding might be biased by small sample size. In addition, our study is also limited by family-reported ethnicity for the decedents that may not correlate closely with their genetic ancestry.
What is especially striking about our yield is the discrepancy between the yield in Blacks as compared with Whites and the higher frequency of uncertain cases in Blacks versus Whites (Figure 2A). A large proportion of our cohort is Black (49.7%), a population that represents only 25% of New York City, suggesting Blacks are disproportionately affected by SUD. We have seen a major representation of Black victims in other natural death conditions, such as fatal pulmonary embolism that we reported previously.21 We cannot determine the reasons, beyond speculation, as to why the cohort has such high Black representation. However, disparities in access to, and quality of health care have been documented in the Black population.22–24 Further investigation into parsing out biological differences versus socioeconomic differences on health outcomes and mortality should be pursued.
Furthermore, our testing utility is rather low in infants. Additional studies could focus on (1) parental testing to determine whether those novel VUS identified in infants arose de novo, which then increases the possibility of cardiac arrhythmia as a plausible explanation of COD; and (2) tissue specific gene expression studies in heart and brain to uncover novel disease-causing genes, as the heterogeneous causes of sudden death in infants may in fact not be directly cardiac related but rather neurological in nature as we discussed previously.25
Precise variant interpretation reduces rates of false positives and false negatives which directly impact the accuracy of diagnosis for at-risk family members of decedents of SUD, and enhance the efficiency and cost-effectiveness in health care. The large number of novel and rare VUS uncovered in this study demand further evaluation for clinical relevance. Testing yield differences among demographic groups highlights the need for the inclusion of under-represented populations in future studies.
Drs David B. Goldstein and Matt Halvorsen from the Institute for Genome Medicine at Columbia University Medical Center have provided valuable comments and advice.
Sources of Funding
National Institute of Justice Grant No. 2011-DN-BX-K535.
The Data Supplement is available at http://circgenetics.ahajournals.org/lookup/suppl/doi:10.1161/CIRCGENETICS.117.001839/-/DC1.
- Received May 31, 2017.
- Accepted October 31, 2017.
- © 2017 American Heart Association, Inc.
- Mazzanti A,
- Maragna R,
- Priori SG
- Burns KM,
- Bienemann L,
- Camperlengo L,
- Cottengim C,
- Covington TM,
- Dykstra H,
- et al
- Sanchez O,
- Campuzano O,
- Fernández-Falgueras A,
- Sarquella-Brugada G,
- Cesar S,
- Mademont I,
- et al
- Lahrouchi N,
- Raju H,
- Lodder EM,
- Papatheodorou E,
- Ware JS,
- Papadakis M,
- et al
- Amendola LM,
- Jarvik GP,
- Leo MC,
- McLaughlin HM,
- Akkari Y,
- Amaral MD,
- et al
- Whiffin N,
- Minikel E,
- Walsh R,
- O’Donnell-Luria AH,
- Karczewski K,
- Ing AY,
- et al
- Richards S,
- Aziz N,
- Bale S,
- Bick D,
- Das S,
- Gastier-Foster J,
- et al
- Walsh R,
- Thomson KL,
- Ware JS,
- Funke BH,
- Woodley J,
- McGuire KJ,
- et al
- Page SP,
- Kounas S,
- Syrris P,
- Christiansen M,
- Frank-Hansen R,
- Andersen PS,
- et al
- Hu D,
- Barajas-Martínez H,
- Pfeiffer R,
- Dezi F,
- Pfeiffer J,
- Buch T,
- et al
- Tang Y,
- Sampson B,
- Pack S,
- Shah K,
- Yon Um S,
- Wang D,
- et al
- Sims M,
- Lipford KJ,
- Patel N,
- Ford CD,
- Min YI,
- Wyatt SB
- Sampson BA,
- Tang Y
The clinical implications of this study include (1) precision variant interpretation is the crucial first step in diagnostic accuracy. At the cohort level, it provides an accurate estimate of testing yield and utility; at the case level, it directly relates to diagnosis and cause of death determination, as well as clinical management of at-risk family members. Applying the high-resolution variant interpretation workflow in this study, we removed previously reported pathogenic variants in ClinVar from further consideration, which if remained in consideration as falsely disease-causing could lead to treatment decisions for persons carrying these variants that would not be warranted and potentially detrimental. (2) Our study calls for special attention to the novel and rare variants of uncertain significance uncovered in this study. Those variants should not be discarded as genetic noise before clinical and function studies prove they are benign. At the same time, we emphasize that these variants of uncertain significance should not be considered diagnostic or lead to treatment decisions for asymptomatic family members who carry them, or for making prenatal decisions without the further analysis outlined. This is particularly important as our low testing yield in non-White minorities correlates with the dearth of reported sudden unexplained death studies in those populations in the literature, where most studies were based on populations of European descent. Our study highlights the critical need for more studies of ethnic minorities in future research.