Yield on Reinterpretation of Genetic Variants in Pediatric Cardiomyopathy

Abstract

Background

Variant interpretation can change over time as new knowledge emerges. Our aim was to determine the frequency and causes of variant reinterpretation on systematic reevaluation in pediatric patients with cardiomyopathy.

Methods

Overall, 227 unrelated pediatric patients with cardiomyopathy enrolled in the Heart Centre Biobank harbored a pathogenic/likely pathogenic (P/LP) variant or a variant of uncertain significance (VUS) on clinical genetic testing (2005–2022). Variant pathogenicity was reevaluated using the American College of Medical Genetics and Genomics guidelines. Additional extension cohorts (n=4547, cases) were analyzed to assess variant burden in cases versus controls (gnomAD 4.1.0).

Results

A total of 382 variants (110 P/LP, 272 VUS) in 227 patients were reevaluated. Forty‐nine variants in 49 patients (21.6%) changed classification. Twelve (10.9%) P/LP variants were downgraded to VUS in 14 patients. Leading criteria were high population allele frequency and variant not located in mutational hotspot or critical functional gene domain. Thirty‐seven (13.6%) VUS were upgraded to P/LP in 35 patients. Leading criteria were variant location in mutational hotspot for gene and deleteriousness on in silico prediction. Only 8 reclassified variants had been reported back by the clinical genetic testing laboratory at the time of the study. Ten of the 37 VUS upgraded to P/LP were significantly enriched in cardiomyopathy cases (n=4796) versus controls.

Conclusions

One in 5 patients with cardiomyopathy had a clinically relevant change in variant pathogenicity on systematic reevaluation that would require modifying family clinical screening and cascade genetic testing. These findings underscore the clinical importance of regular variant reinterpretation on follow‐up.

Nonstandard Abbreviations and Acronyms

ACMG

American College of Medical Genetics and Genomics

AMP

Association for Molecular Pathology

DCM

dilated cardiomyopathy

HCM

hypertrophic cardiomyopathy

P/LP

pathogenic/likely pathogenic

VUS

variant of uncertain significance

Clinical Perspective.

What Is New?

• Systematic reevaluation of genetic variants in pediatric cardiomyopathy revealed that 21.6% of patients had a clinically meaningful change in variant classification; this represents one of the largest studies of pediatric variant reinterpretation and identifies the most common causes of variant misclassification that should be taken into consideration by clinicians and genetic testing laboratories.

What Are the Clinical Implications?

• Variant reclassification affected care in 1 of 5 families with cardiomyopathy, requiring renewed clinical screening in genotype‐negative family members of cases that underwent variant downgrading and providing an opportunity for cascade genetic testing for family members of cases that underwent variant upgrading.

• This emphasizes the importance of regular variant reinterpretation to inform genetically‐guided risk assessment and follow‐up in families affected by cardiomyopathy.

Cardiomyopathy is primarily a genetic disorder and a common cause of childhood heart failure. ¹ Cardiomyopathy phenotypes include arrhythmogenic cardiomyopathy, dilated cardiomyopathy (DCM), hypertrophic cardiomyopathy (HCM), restrictive cardiomyopathy, and left ventricular noncompaction cardiomyopathy. Dozens of genes are implicated in causing the disease, with most genotype‐positive cases being autosomal dominant and caused by a rare DNA variant ¹ and with considerable overlap in the genes associated with different cardiomyopathy subtypes. ¹ , ²

The genotype of a patient informs cascade clinical and genetic screening of first‐degree relatives. Genotype‐positive patients (patients who harbor a pathogenic/likely pathogenic [P/LP] variant in a cardiomyopathy gene, especially those with P/LP variants in sarcomeric and other cardiomyopathy‐associated genes), tend to have a worse prognosis and their genotype may offer clinically meaningful insights for risk stratification and treatment decisions. ³ Importantly, when a patient is genotype positive, family members are offered cascade testing for the variant, and only genotype‐positive family members require ongoing echocardiographic screening for cardiomyopathy. ⁴ When a patient is genotype negative that is, does not harbor a P/LP variant, then all family members are recommended serial echocardiographic screening. ⁴ A previous study in HCM reported that, among 1361 probands with nonbenign variants, 917 (67.3%) harbored P/LP variants and 444 (32.6%) harbored only variants of uncertain significance (VUS), the latter reflecting incomplete or conflicting evidence for pathogenicity. ⁵ However, variant interpretation can change over time as new knowledge, additional functional validation data, and more accurate computational prediction tools emerge. A change in variant interpretation can change family screening recommendations. The 2024 American Heart Association and American College of Cardiology guidelines for HCM and the European Society of Cardiology guidelines for cardiomyopathies emphasize the importance of periodic variant reevaluation. ⁶ , ⁷ To improve the efficiency of variant reevaluation, clinicians are advised to document the date when the genetic test results were initially assessed and to consider any new medical and scientific information that has become available since that time. ⁸ Small pediatric cardiomyopathy cohort studies have reported variant reclassification rates ranging from 8.9% to 10.3%. ⁹ , ¹⁰ Here we performed a real‐world systematic reevaluation of variants in a larger cohort of unrelated pediatric cardiomyopathy cases to identify the frequency and criteria for clinically “actionable” variant reclassification on follow‐up.

METHODS

Data availability

Deidentified data analyzed in this study are available in the main and supplemental tables, and additional data are available from the corresponding author on reasonable request. We used the Strengthening the Reporting of Observational Studies in Epidemiology cohort reporting guidelines. ¹¹

Study Cohort

For this single‐center retrospective study, a total of 280 pediatric unrelated patients with primary cardiomyopathy were enrolled in the Heart Centre Biobank Registry at the Hospital for Sick Children (Toronto, Ontario, Canada). Of these, 249 underwent clinical genetic testing. A total of 227 patients with a P/LP variant or a VUS were included in the study cohort and genotype‐negative patients were excluded (Table 1 and Table S1). Patients with secondary cardiomyopathy caused by chromosomal malformations, neuromuscular disorders, mitochondrial or metabolic disorders, congenital heart defects, or reversible causes were excluded. Patient demographics, diagnosis, type of clinical testing (single gene testing, targeted panels, multiple gene panels including expanded cardiomyopathy panels, and whole exome sequencing), year of clinical genetic testing, and the original (and updated) variant classification by the genetic testing laboratory were captured. In addition, to perform case–control burden analysis for relevant reclassified variants, data was accessed from an extension cohort of 4547 cases derived from international studies and published literature (Table S2). The gnomAD (v4.1.0 genome data set) was used to assess allele frequencies (n=76 215) for variant reinterpretation, but for case–control burden analysis, we used gnomAD (v4.1.0, exome data set) (n=730 947) to avoid any bias related to using overlapping patients as controls. The study was approved by the Institutional Research Ethics Boards and written informed consent was obtained from all patients or their parents/legal guardians and the study protocol adhered to the Declaration of Helsinki.

Table 1.

Baseline Study Cohort Characteristics

Characteristics	No.=227 (%)
Male sex	133 (58.6%)
Ancestry
European	140 (61.7%)
Asian/South Asian	56 (24.7%)
African	22 (9.7%)
Mixed	5 (2.2%)
Unknown	4 (1.8%)
Cardiomyopathy phenotype
DCM	104 (45.8%)
HCM	76 (33.5%)
Left ventricular noncompaction	20 (8.8%)
Restrictive cardiomyopathy	10 (4.4%)
Arrhythmogenic cardiomyopathy	9 (4.0%)
Mixed type	8 (3.5%)
Type of test
DCM panel	25 (11.0%)
HCM panel	54 (23.8%)
Multiple panels	124 (54.6%)
Whole exome sequencing	4 (1.8%)
Others	3 (1.3%)
Unknown	17 (7.5%)
Variants identified
P/LP	110 (28.8%)
VUS	272 (71.2%)
Patient genotype
Harbored P/LP variant	71 (31.2%)
Harbored VUS	122 (53.7%)
Harbored P/LP variant and VUS	34 (15.0%)

DCM indicates dilated cardiomyopathy, HCM; hypertrophic cardiomyopathy, P/LP, pathogenic/likely pathogenic; and VUS; variant of uncertain significance.

Reevaluation of Variant Pathogenicity

Variants identified as P/LP or VUS on clinical genetic testing were reinterpreted for pathogenicity using the American College of Medical Genetics and Genomics/Association for Molecular Pathology (ACMG/AMP) guidelines for interpretation of variant pathogenicity. Using these criteria, a variant can be interpreted as P/LP, VUS, likely benign, or benign. ¹² , ¹³ For variants with multiple isoforms, the Matched Annotation from the National Center for Biotechnology Information and European Molecular Biology Laboratory‐European Bioinformatics Institute transcript (MANE Select and MANE Plus Clinical) was used to annotate and evaluate variant effects in standardized and clinically relevant gene transcripts. A clinically actionable variant reclassification was defined as either the upgrading of a VUS to P/LP or the downgrading of a P/LP variant to a nonpathogenic variant that is, VUS or benign/likely benign variant.

Several criteria were applied for variant reinterpretation including confirmation of patient phenotype, the current association between gene and cardiomyopathy phenotype on the most recent ClinGen version 11 curation, ¹⁴ updated ClinVar records (2024‐01‐29 Web Release), literature evidence, control population allele frequencies, and prediction scores derived from computational tools. When available, the results of clinical genetic tests from parents or siblings were used to ascertain de novo status and to perform segregation analysis for a familial variant. Reclassifications were verified by our Return of Research Results committee comprising pediatric cardiologists, clinical geneticists, genetic counselors, and bioinformaticians. ¹⁵

A detailed approach to variant reinterpretation is described next. Of note, all eligible variants reported on clinical testing were single nucleotide variants and small insertions‐deletions.

1. PVS1 (Pathogenic very strong) criterion was applied to loss‐of‐function variants in haploinsufficiency‐intolerant genes using gnomAD 4.1.0, defined as a probability of loss‐of‐function intolerant score ≥0.9. PVS1 criterion was not applied to MYH7 variants as per ClinGen guidelines because the contribution of loss‐of‐function variants in this gene to inherited cardiomyopathy remains incompletely understood. ¹⁶ TTN variants were considered pathogenic only if they were protein‐truncating variants (frameshift, nonsense, canonical splice site) predominantly in the A‐band and I‐band regions and isoforms. ¹⁷ , ¹⁸

2. The PS2 (Pathogenic strong) criterion was applied when the variant was confirmed through genetic testing to be de novo with parentage confirmed and no family history of cardiomyopathy.

3. The PS4 criterion of higher variant burden in cases compared with controls, that is, gnomAD (v4.1.0, exome data set; n=730 947) was conducted by jointly analyzing cases from our cohort and additional cases from extension cohorts (n=4796).

4. The PM1 (Pathogenic moderate) criterion was applied if the variant was located in a mutational hotspot or critical functional domain for the gene. This was determined using DiscoVari, the Cardiomyopathy Variant Curation Expert Panel from ClinGen, or literature evidence. ¹⁹ , ²⁰ For MYH7, the relevant transcripts analyzed were ENST00000355349 and NM_000257.4, covering codons 167 to 931. In the case of MYBPC3, the analysis focused on transcripts ENST00000545968 and NM_000256.3, specifically targeting codons 485 to 502 and 1248 to 1266. For TNNI3, the key transcripts used were ENST00000344887 and NM_000363.5, covering codons 141 to 209. For TNNT2, the transcripts ENST00000367318 (codons 79 to 179), and ENST00000656932.1/NM_001276345.2 (codons 89 to 189) were included in the analysis.

5. PM2 criterion was applied if population variant allele frequency in gnomAD 4.1.0 (genome data set) was <0.0001 for autosomal dominant and <0.01 for recessive inheritance, while accounting for Grpmax filtering allele frequencies (95% confidence) as ancestry‐specific allele frequencies. ²¹

6. The PM5 (Pathogenic moderate) criterion was applied when the variant incorporated a novel missense change at an amino acid residue where a different missense change is determined to be pathogenic in ClinVar or Leiden Open Variation Database v.3.0 supported by literature. ¹³ , ²²

7. The PP1 (Pathogenic supporting) criterion was applied when the variant segregated with disease among family members. According to the ClinGen guidelines, in cardiomyopathies with reduced penetrance, PP1 is applied depending on the number of affected individuals demonstrating cosegregation with the variant (not including the proband). Specifically, PP1 supporting is applied when cosegregation is observed in a single affected family member, PP1 moderate for 2 to 3 affected members, and PP1 strong for ≥4 affected members. ²³

8. The PP2 (Pathogenic supporting) criterion was defined as a missense variant in a gene with a low rate of benign missense variation, where missense variants are a known mechanism of disease. Based on ClinGen Cardiomyopathy Expert Panel Specifications to the ACMG/AMP Variant Interpretation Guidelines in the Criteria Specification Registry for cardiomyopathy, ²⁰ PP2 was applied only to variants in the gene TPM1.

9. The PP3 (Pathogenic supporting) criterion for missense variants, which requires multiple lines of computational evidence supporting a deleterious effect on the gene, was applied when all 3 pathogenicity scores (Combined Annotation Dependent Depletion v1.7, Rare Exome Variant Ensemble Learner, and AlphaMissense.v2023.hg38) reached predefined thresholds ²⁴ , ²⁵ , ²⁶ These thresholds were adopted from the ClinGen recommendations for Combined Annotation Dependent Depletion (>25.3) and Rare Exome Variant Ensemble Learner (>0.644). ²⁷ As the threshold for AlphaMissense was not specified in this recommendation, it was derived from the original publication as a score >0.564. ²⁵

Statistical Analysis

Frequency of variant reclassification was compared using Fisher exact test by year of clinical testing (before versus after 2015, ie, publication of ACMG guidelines), cardiomyopathy phenotype (HCM versus DCM), ancestry (European versus non‐European), sex (male versus female) and type of genetic test (single gene testing/single targeted panel versus multiple panels/exome sequencing). Patients with left ventricular noncompaction cardiomyopathy, arrhythmogenic cardiomyopathy, or restrictive cardiomyopathy were excluded from subgroup analysis due to limited sample size. The proportion of variants reclassified from VUS to P/LP and from P/LP to VUS was also compared using Fisher exact test. We applied the Benjamini–Hochberg method to correct for multiple testing and calculate adjusted P values. A total of 5 independent statistical tests were included in this adjustment, examining the association between variant reclassification and year of clinical testing, cardiomyopathy phenotype, ancestry, sex, and type of genetic test. An adjusted P value <0.05 was considered significant. Burden analysis was performed by comparing the frequency of each variant in cases versus controls, with allele count and allele number extracted for each variant from gnomAD v4.1.0 exome data set. Fisher exact test was used to assess the difference in allele frequencies between cases and the gnomAD v4.1.0 exome data set used as controls. Following the guidelines of the Cardiomyopathy Variant Curation Expert Panel from ClinGen, the strength of evidence was classified based on the lower bound of the 95% CI for the odds ratio (OR), with strong evidence requiring a lower bound of ≥20, moderate evidence requiring ≥10, and supporting evidence requiring ≥5, and anything <5 considered not met.

RESULTS

Study Cohort Characteristics

The baseline characteristics of the study cohort (n=227) are summarized in Table 1. Among the participants, 133 (59%) were male; 64% were of European descent, 25% were of Asian descent, 10% were of African descent, and 1% were of mixed descent or other ancestries. Of 227 participants, 206 underwent one or more gene panel testing, 4 underwent whole exome sequencing, and 17 underwent other types of genetic tests of unknown types. The 227 patients harbored a total of 382 variants (110 P/LP, 272 VUS) (Table S1). 105 participants (46.3%) harbored at least 1 P/LP variant, 122 (53.7%) harbored only VUS, and 34 (15%) harbored a combination of P/LP variants and VUS. Three variants that were found in 2 families each—MYBPC3 (c.442G>A, p.Gly148Arg), MYH7 (c.1759G>A, p.Asp587Asn), and LMNA (c.868G>A, p.Glu290Lys)—were reported differently for each family by the clinical laboratory: P/LP for one family and VUS in another family.

Reinterpretation of Variant Pathogenicity

Reclassification of P/LP Variants

After reinterpretation, 12 of 110 P/LP (10.9%) variants were downgraded to VUS in 14 of 227 patients (6.2%) (Table 2). Genes affected by the downgrade included MYH7 (5 variants), MYBPC3 (4 variants), and 1 variant each in LMNA, DSG2, and RRAGD. Notably, the MYBPC3 c.3628‐41_3628‐17del (Δ25bp) variant was identified in 6 South Asian cases. The criteria primarily contributing to downgrading of P/LP variants were an allele frequency higher than expected in gnomAD (n=7), the variant not being in a mutational hotspot or critical functional gene domain (n=5), the variant not being predicted as deleterious by in silico tools (n=4), the variant recently reported as VUS in ClinVar (n=3), and failure to meet MYH7‐specific ACMG/AMP criteria (n=1) (Figure 1A).

Table 2.

Pathogenic or Likely Pathogenic Variants Reclassified to VUS

Participant ID	Cardiomyopathy phenotype	Year of clinical test	Affected gene	CDS change	Amino acid change	Current ACMG criteria	Final variant classification
95	RCM	2016	DSG2	c.3039C>A	p.Tyr1013*	PVS1, BS1	VUS
216	DCM	2009	LMNA	c.1912G>A	p.Gly638Arg	PM1, PM2	VUS
73	DCM	2013	MYBPC3	c.1654G>T	p.Ala552Ser	PM2, PP1 (supporting)	VUS
26/76/101/116/145/175	DCM, 3 RCM/HCM, HCM 2	2010/2010/2013/2017/2018/2020	MYBPC3	c.3628‐41_3628‐17del	NA	PP1 (supporting), PP5, BS1	VUS*
59	HCM	2009	MYBPC3	c.3548T>G	p.Phe1183Cys	PM2, PP1 (supporting), PP3 (moderate)	VUS
211	HCM	2009	MYBPC3	c.2170C>T	p.Arg724Trp	PM1, PM2	VUS
62	DCM	2010	MYH7	c.1322C>G	p.Thr441Arg	PM1, PM2	VUS
67	DCM	2009	MYH7	c.5507C>T	p.Ser1836Leu	PM2	VUS
168	HCM	2009	MYH7	c.1003G>T	p.Ala335Ser	PM1, PM2	VUS
16	LVNC	Unknown	MYH7	c.732+1G>A	NA	PM2, PP5	VUS
209	LVNC/DCM	2021	MYH7	c.5395G>A	p.Glu1799Lys	PM2, PP3 (moderate)	VUS
12	DCM/RCM	2022	RRAGD	c.289_303del	p.Thr97_Cys101del	PM2, PM4	VUS

ACMG indicates American College of Medical Genetics and Genomics; BS, benign strong; CDS, coding sequence; DCM, dilated cardiomyopathy; HCM, hypertrophic cardiomyopathy; LVNC, left ventricular non‐compaction; PM, pathogenic moderate; PP, Pathogenic supporting; PVS, pathogenic very strong; RCM, restrictive cardiomyopathy; and VUS, variant of uncertain significance.

^*One variant was reclassified by the clinical laboratory.

Figure 1. Criteria for variant reclassification.

A, Downgrading P/LP variants to VUS, (B) Upgrading VUS to P/LP variants. ACMG indicates American College of Medical Genetics and Genomics; BS, benign strong; P/LP, pathogenic/likely pathogenic; PM, pathogenic moderate; PP, pathogenic supporting; PVS, Pathogenic very strong; and VUS, variants of uncertain significance.

Reclassification of VUS

Of 272 VUS (13.6%), 37 were upgraded to P/LP in 35 of 227 patients (15.4%) (Table 3). Genes affected by the upgrade included ACTC1 (1 variant), ACTN2 (2 variants), LMNA (1 variant), MYBPC3 (2 variants), MYH7 (13 variants), NEXN (3 variants), PLN (1 variant), TNNC1 (1 variant), TNNI3 (2 variants), TNNT2 (4 variants), TPM1 (3 variants), TTN (1 variant), and VCL (2 variants). The primary criteria contributing to upgrading of VUS to P/LP were variant location in a mutational hotspot for the gene (n=23), variant interpretation as deleterious by modern in silico prediction tools (n=19), a novel missense change occurring at an amino acid residue where a different pathogenic missense change had been observed before (n=11), a change in the gene association with cardiomyopathy according to ClinGen (n=8), published experimental findings suggesting that the variant affects gene function (n=7), cosegregation with disease in multiple affected family members (n=6), the variant being de novo (n=4), the variant being recently reported as pathogenic in ClinVar (n=2), and the finding of a different variant that results in the same amino acid change being recently reported as pathogenic in ClinVar (n=1) (Figure 1B). A total of 47 VUS were downgraded to benign/likely benign variants, but these were not further evaluated because they do not reflect a clinically actionable change. An overview of variant reclassification frequency is provided in Figure 2A.

Table 3.

VUS Reclassified to Pathogenic or Likely Pathogenic

Participant ID	Cardiomyopathy phenotype	Year of clinical test	Affected gene	CDS change	Amino acid change	Current ACMG criteria	Final variant classification	ACMG PS4 criteria (supporting, moderate, strong, not met)
207	LVNC/DCM	2014	ACTC1	c.809G>A	p.Gly270Asp	PS2, PM1, PM2, PP3 (moderate)	Pathogenic	NA
150	DCM	2018	ACTN2	c.784‐2A>G	NA	PVS1, PM2	Likely pathogenic*	NA
167	DCM	Unknown	ACTN2	c.36C>G	p.Tyr12*	PVS1, PM2	Likely pathogenic	NA
87	DCM	2012	LMNA	c.868G>A	p.Glu290Lys	PS2, PM2, PP3 (moderate)	Likely pathogenic†	Strong
213	HCM	2007	MYBPC3	c.442G>A	p.Gly148Arg	PM2, PP1 (strong), PP5	Likely pathogenic*, †	Not met
217	HCM	2010	MYBPC3	c.1828G>A	p.Asp610Asn	PS3, PM1, PM2, PM5, PP3 (supporting)	Pathogenic	Strong
31	DCM	2011	MYH7	c.2171T>A	p.Ile724Asn	PS1, PM1, PM2, PP3 (moderate), PP5	Pathogenic	NA
57	DCM	2016	MYH7	c.5158A>G	p.Asn1720Asp	PM1, PM2, PP3 (moderate)	Likely pathogenic	NA
115	DCM	2019	MYH7	c.3667G>A	p.Glu1223Lys	PM1, PM2, PM5, PP3 (supporting)	Likely pathogenic*	NA
126	DCM	2019	MYH7	c.4276G>A	p.Glu1426Lys	PS3, PM2, PP3 (moderate), PP5	Likely pathogenic*	NA
197	DCM	2022	MYH7	c.1423C>A	p.Gln475Lys	PM1, PM2, PM5, PP3 (supporting)	Likely pathogenic	NA
222	DCM	2009	MYH7	c.2683C>G	p.Gln895Glu	PM1, PM2, PP1 (strong)	Likely pathogenic*	NA
84	HCM	2017	MYH7	c.4136C>A	p.Ala1379Asp	PM1, PM2, PP1 (supporting), PP3 (supporting)	Likely pathogenic	Strong
151	HCM	Unknown	MYH7	c.530C>T	p.Thr177ile	PM1, PM2, PP3 (moderate)	Likely pathogenic	Moderate
205	HCM	2006	MYH7	c.1759G>A	p.Asp587Asn	PM1, PM2, PM5	Likely pathogenic*, †	Strong
221	HCM	2010	MYH7	c.1220G>T	p.Gly407Val	PM1, PM2, PM5, PP3 (supporting), PP5	Likely pathogenic	Strong
179	LVNC	2021	MYH7	c.1118C>A	p.Ala373Glu	PM1, PM2, PM5, PP3 (supporting)	Likely pathogenic	NA
202	LVNC	2015	MYH7	c.532G>A	p.Gly178Arg	PM1, PM2, PP1 (moderate), PP3 (moderate)	Likely pathogenic	NA
79	LVNC/DCM	2015	MYH7	c.1180G>A	p.Asp394Asn	PM1, PM2, PM5	Likely pathogenic	NA
126	DCM	2019	NEXN	c.1473 + 1G>T	NA	PVS1, PM2	Likely pathogenic*	NA
210	DCM	2015	NEXN	c.373C>T	p.Arg125*	PVS1, PM2	Likely pathogenic	NA
209	LVNC/DCM	2021	NEXN	c.201G>A	p.Trp67*	PVS1, PM2	Likely pathogenic	NA
54	HCM	2015	PLN	c.2T>C	p.Met1	PVS1, PM2	Likely pathogenic	NA
104	DCM	2018	TNNC1	c.376G>A	p.Glu126Lys	PS2, PM2	Likely pathogenic	Strong
5	RCM	2011	TNNC1	c.23C>T	p.Ala8Val	PS3, PM2, PP3 (supporting), PP5	Pathogenic	Supporting
117	DCM	2008	TNNI3	c.550G>A	p.Glu184Lys	PS2, PM1, PM2	Likely pathogenic	NA
76	Restrictive cardiomyopathy/HCM	2013	TNNI3	c.484C>T	p.Arg162Trp	PM1, PM2, PM5, PP5	Likely pathogenic	Moderate
14	DCM	Unknown	TNNT2	c.481C>T	p.Arg161Cys	PM1, PM2, PP3 (moderate)	Likely pathogenic	NA
89	DCM	2016	TNNT2	c.552G>T	p.Lys184Asn	PM1, PM2, PP1 (supporting), PP3 (supporting)	Likely pathogenic	NA
137	DCM	2020	TNNT2	c.316G>A	p.Glu106Lys	PS3, PM1, PM2, PP3 (moderate), PP5	Pathogenic	NA
215	LVNC/DCM	2014	TNNT2	c.862C>T	p.Arg288Cys	PS3, PM1, PM5, PP5	Likely pathogenic	NA
28	DCM	2009	TPM1	c.45G>T	p.Lys15Asn	PS3, PM1, PM2, PP2, PP3 (moderate)	Likely pathogenic	NA
151	HCM	Unknown	TPM1	c.413A>G	p.Glu138Gly	PM2, PM5, PP2, PP3 (strong)	Likely pathogenic	NA
161	HCM	2020	TPM1	c.560A>T	p.Glu187Val	PM1, PM2, PM5, PP2, PP3 (moderate)	Likely pathogenic	NA
106	DCM	2012	TTN	c.16353C>G	p.Tyr5451*	PVS1, PM2, PP1 (supporting)	Likely pathogenic	NA
101	DCM	2010	VCL	c.2949del	p.Gln971Profs*	PVS1, PM2	Likely pathogenic	NA
122	DCM	2018	VCL	c.313C>T	p.Arg105*	PVS1, PM2	Likely pathogenic	Strong

ACMG indicates American College of Medical Genetics and Genomics; CDS, coding sequence; DCM, dilated cardiomyopathy; HCM, hypertrophic cardiomyopathy; LVNC, left ventricular noncompaction; NA, not applicable; PM, pathogenic moderate; PS, pathogenic strong; PP, pathogenic supporting; PVS, pathogenic very strong; and VUS, variant of uncertain significance.

^*Seven variants were reclassified by the clinical laboratory.

^† Three variants were reported as pathogenic/likely pathogenic in one family and as VUS in another family by the clinical testing laboratories.

Figure 2. Variant‐ and patient‐level reclassification.

A, Variant‐level reclassification: Of 110 P/LP variants, 12 were downgraded to VUS (10.9%). Of 272 VUS, 37 were upgraded to P/LP (13.6%). In addition, 47 VUS were downgraded to B/LB. B, Patient‐level reclassification: Of 71 patients harboring only P/LP variants, 11 patients were downgraded to having only VUS. Of 122 patients harboring only VUS, 16 patients were upgraded to having only P/LP variants and 13 patients were upgraded to P/LP+VUS. Of 34 patients harboring both P/LP+VUS, 2 patients were downgraded to only VUS, 3 patients were upgraded to only P/LP variants and *4 patients had a change in the gene harboring a P/LP variant following reinterpretation of coexisting variants. B/LB indicates benign/likely benign; P/LP, pathogenic/likely pathogenic; and VUS, variant of uncertain significance.

Variant Burden Analysis

Among the 37 variants that were upgraded from VUS to P/LP (Table 3), 11 distinct variants (in 27 cases) were identified across several extension cardiomyopathy cohorts (n=4547) (Table 3). Burden analysis confirmed 10 out of the 37 reclassified variants also met ACMG PS4 criterion for higher burden in cases compared with controls (gnomAD v4.1.0, exome data set) (Table 3). For example, LMNA c.868G>A showed an OR of 505.1 (95% CI, 52.5–4857.1), and MYH7 c.1759G>A had an OR of 1011.2 (95% CI, 121.7–8400.7), both supporting strong evidence for PS4. A full list of variants, corresponding ORs, and evidence strength is provided in Table 3. The burden analysis yielded strong evidence for 6 variants, moderate evidence for 3 variants, and supporting evidence for 1 variant. Variants that were reclassified from P/LP to VUS were not included in the burden analysis, as the majority were downgraded due to high population allele frequency.

Patient Level Reclassification

Overall, the variant reclassifications affected 49 of 227 patients (21.6%) (Figure 2B). Among the reclassifications, 15.5% of patients with only P/LP variants had 1 or more variants downgraded to a VUS (11 of 71 patients), 23.8% of patients with only VUS had 1 or more variants upgraded to P/LP (29 of 122 patients), and 26.5% of patients with a combination of P/LP variants and VUS had either a downgrade of their P/LP variants or upgrade of their VUS in the same patient resulting in a change in affected gene (9 of 34 patients).

Of note, only 8 reclassified variants (MYBPC3 c.3628‐41_3628‐17del, MYH7 c.1759G>A, p.Asp587Asn, MYH7 c.3667G>A p.Glu1223Lys, MYH7 c.4276G>A p.Glu1426Lys, MYH7 c.2683C>G p.Gln895Glu, NEXN c.1473+1G>T, ACTN2 c.784‐2A>G, and MYBPC3 c.442G>A p.Gly148Arg) in 12 patients had been previously reported by the clinical genetic testing laboratory at the time of our study. Among these variants, the MYH7 c.3667G>A p.Glu1223Lys reclassification was initiated by the clinical laboratory, and ACTN2 c.784‐2A>G reclassification was prompted by the patient's clinical team. For the remaining variants, the reasons triggering reclassification were not available.

Frequency of Variant Reclassification by Subgroups

The year of clinical testing was known for 358 variants. In this group, frequency of variant reclassification did not vary by year of clinical testing (before or after 2015; Figure 3A and 3B). Additionally, there were no significant differences in the frequency of variant reclassification based on cardiomyopathy phenotype (HCM versus DCM), ancestry (European versus non‐European descent), or sex (male versus female) (all adjusted P>0.05; Figure 4). Variants identified using single gene/single targeted panel testing, however, showed a higher likelihood of being downgraded from P/LP to VUS compared with variants identified using multiple panels or exome sequencing (8% versus 2%; adjusted P=0.0498).

Figure 3. Variant reclassification by year of clinical testing.

The graphs show the (A) frequency of variants reclassified by year of clinical testing (2005–2022); (B) percentage of variants reclassified by year of clinical testing. Year of testing was known for 358 variants. The average annual reclassification rate was 15%.

Figure 4. Proportion of variants reclassified by year of clinical testing, cardiomyopathy phenotype, ancestry, and sex.

Proportion of variants reclassified did not differ by year of testing (before and after 2015), cardiomyopathy phenotype (HCM vs DCM), ancestry (European vs non‐European descent), and sex (men vs women) (P>0.05). Variants identified using single gene/targeted panels showed a higher frequency of P/LP variants being downgraded to VUS compared with variants identified using multiple panels or exome sequencing (adjusted P=0.0498). Blue bar=% of VUS upgraded to P/LP; black bar=% of P/LP variants downgraded to VUS. DCM indicates dilated cardiomyopathy; HCM, hypertrophic cardiomyopathy; P/LP, pathogenic/likely pathogenic; and VUS, variant of unceratin significance.

DISCUSSION

We conducted a systematic reinterpretation of variants identified on clinical genetic testing in a real‐world cohort of 227 pediatric cardiomyopathy patients. The frequency of clinically actionable variant reclassification was 12.8% affecting 21.6% patients. This included downgrading of 10.9% P/LP variants to VUS and upgrading of 13.6% VUS to P/LP variants. Previous pediatric studies have reported a rate of variant reclassification ranging from 8.9% to 10.3% in 71 to 128 patients. ⁹ , ¹⁰ Our study with a larger cohort size reveals a somewhat higher (albeit not statistically different) reclassification rate and reinforces the importance of routine reinterpretation of variants on follow‐up.

A detailed description of evaluation of variant pathogenicity is important to guide confidence in pathogenicity predictions. ²⁸ In our study, we examined the criteria predominantly affecting the change in pathogenicity of the variants based on the ACMG evaluation methodology. The most common criterion for reclassification of a variant from P/LP to VUS was a higher‐than‐expected population allele frequency. The increase in genome sequencing including those of previously underrepresented populations has enabled more accurate assessment of variant frequencies in the general population. ²⁹ , ³⁰ Further, population‐specific databases are increasingly being used, ³¹ which continues to improve the accuracy of variant frequency assessments across different ancestries and will continue to enhance the precision of pathogenicity evaluation with time. The most common criterion for reclassification of a variant from VUS to P/LP was location of the variant in a mutational hotspot for the gene. Often the evaluation of a hotspot involves determining whether a variant is located within a constitutive domain of the protein or assessing the role of the domain in which the variant is located, based on literature and experimental data. In this study, reclassification was carried out using the hotspots defined by DiscoVari, the Cardiomyopathy Variant Curation Expert Panel from ClinGen, and published literature where hotspots had been identified. ¹⁹ , ²⁰ The number of genes with defined hotspots is still limited, and future efforts may enable the accurate assessment of variant hotspots in additional genes. Segregation analysis of the variant in family members and reevaluation of inheritance contributed to a change in pathogenicity in 10 out of 81 variants (12.3%). This constitutes a substantial proportion, and as recommended in the ACMG statement, ⁸ reclassification should include a reassessment of familial phenotype and variant inheritance on follow‐up.

An important strength of our analysis was the ability to leverage several additional independent and diverse cohorts with detailed clinical annotation not routinely available in ClinVar and other public databases that helped us confirm that several of the reclassified VUS were in fact enriched in cases compared with controls on burden analysis (PS4 criterion). Therefore, it is crucial to flag variants with the potential for reclassification and to document the criteria that have the potential to change over time. It is equally important to revisit not just variant pathogenicity but also the availability of updated gene testing panels that may capture disease‐associated genes that were not captured at the time of previous genetic testing. The ClinGen curation of the literature and new evidence serve an important role in informing the size and nature of gene panels.

In our study, the MYBPC3 c.3628‐41_3628‐17del (Δ25bp) intronic deletion was reclassified from pathogenic to a VUS. This 25‐base pair deletion is located within intron 32 of the MYBPC3 gene and is observed in 4% to 8% of individuals of South Asian ancestry. ³² , ³³ Given its allele frequency, which exceeds the incidence of pediatric cardiomyopathy, this variant was reclassified to a VUS (BS1 criterion). The reclassification of a MYBPC3 variant observed in 6 South Asian individuals underscores the importance of comparing ancestry‐specific allele frequency in variant interpretation. Variants that are rare globally but common in certain populations may lead to misclassification if universal thresholds are applied. Therefore, consideration of population‐specific data is essential, particularly for genes known to harbor founder variants in specific ethnic groups. It is now known that this variant alone without MYBPC3 c.1224‐52G>A as part of a haplotype is not sufficient to produce a phenotype, ³⁴ although in the adult population, it has been associated with myocardial diastolic dysfunction. ³⁵

There were no significant differences in the proportion of patients reclassified before and after 2015, when the ACMG/AMP guidelines were first published. Of note, 2 of 11 variants (18%) identified on clinical testing in 2022, the most recent year included, showed a change in pathogenicity within 2 to 3 years of clinical reporting. The downgrading of RRAGD c.289_303del p.Thr97_Cys101del from LP to VUS was based on the application of in silico tools (PP3), which indicated that the variant was not deleterious. Conversely, the upgrading from VUS to LP of MYH7 c.1423C>A p.Gln475Lys was because of evidence of a mutational hotspot at the codon (PM1) and a differential missense change at the same codon (MYH7 c.1425G>T p.Gln475His), which had been previously reported as pathogenic (PM5) in 2022, further supported by in silico analysis using AlphaMissense issued in 2023 (PP3). In this regard, there remains a gap in regular reinterpretation of genetic variants by testing laboratories. Only 8 of the 49 variants reclassified by us were reported back to the clinical team by the original testing laboratory. The ACMG and the other reports recommend that clinical laboratories reevaluate variants at least every 3 to 5 years and the costs of recontacting physicians can be built into original testing costs. ⁸ , ³⁶ , ³⁷ The Canadian College of Medical Geneticists recommends that laboratories reanalyze variants when a health care provider initiates the request. ³⁸ This is important because laboratories do not always have access to updated family phenotype information. This emphasizes that periodic reinterpretation should, where feasible, be built into the ongoing clinical care of patients by their health care providers. Based on our findings, we propose that reinterpretation at least every 2 to 3 years may be more appropriate. The ACMG in its most recent policy statement emphasizes that “re‐contact is a shared responsibility,” ³⁹ indicating that both clinicians and clinical laboratories should take on an increasing role in assessing and communicating variant reinterpretations to patients. In addition, patients and families should be advised to notify their physician if family genotype or phenotype information changes on follow‐up.

To date, there are no large‐scale, publicly available data sets quantifying the uptake of variant reevaluation in clinical practice. Despite increasing recognition of its importance, implementation of variant reevaluation remains limited. Potential barriers for clinicians include lack of awareness of the importance of variant reevaluation, time, and knowledge constraints for clinicians or limited access to genetic counselors. Addressing these challenges through education, system‐level infrastructure, and policy initiatives will be essential for equitable implementation.

The present analysis revealed that variants identified using single gene/single targeted panel showed a significantly higher reclassification rate compared with multiple panels, primarily driven by a greater proportion of variants that were downgraded from P/LP to VUS. Given the high degree of genetic heterogeneity with overlap in genetic pathogenesis between different cardiomyopathies, this finding suggests that comprehensive genetic testing is more likely to identify true P/LP variants with fewer false positives compared with targeted gene panels. In the future, it may be useful to consider developing a framework that considers the type of genetic test performed when determining which cases should be prioritized for variant reinterpretation.

In this study, we incorporated AlphaMissense in addition to Combined Annotation Dependent Depletion and Rare Exome Variant Ensemble Learner to support variant interpretation. Whereas Combined Annotation Dependent Depletion and Rare Exome Variant Ensemble Learner are Ensembl prediction scores, AlphaMissense uses deep learning‐based protein structure prediction through AlphaFold and provides residue‐level estimates of pathogenicity. The tool, published in 2023, ⁴⁰ demonstrated strong performance with a sensitivity of 92%, specificity of 78%, positive predictive value of 97%, and negative predictive value of 53%. However, its performance may be limited for variants located in proteins with large intrinsically disordered regions, large protein size, or genes with a high background rate of benign missense variants, such as TTN and MYBPC3. Accordingly, TTN missense variants were not considered pathogenic in our analysis.

In our cohort, 3 variants (MYBPC3 c.442G>A, p.Gly148Arg; MYH7 c.1759G>A, p.Asp587Asn; and LMNA c.868G>A, p.Glu290Lys) were reported with differing classifications across families by clinical laboratories. These discrepancies may be attributed to differences in interpretation frameworks among laboratories due to use of proprietary in‐house data, varying burden analysis strategies, varying selection or thresholds of in silico prediction tools, and varying access to detailed family history and phenotypic information. To improve consistency and reproducibility in variant interpretation, it is essential to establish standardized classification protocols and promote transparent data sharing among laboratories through public reporting of classification criteria.

The use of artificial intelligence in variant evaluation offers a promising way to improve efficiency and scalability. ²⁵ , ⁴¹ Artificial intelligence methods can automate the integration of updated databases, guidelines, and prediction tools and help identify variants needing reassessment. However, clinical interpretation in pediatric cardiomyopathy often relies on phenotype, family history, and follow‐up data, which current artificial intelligence systems may not fully capture. Thus, expert judgment remains essential to ensure clinically meaningful reevaluation.

It is important to acknowledge as a limitation that we were not privy to the exact rationale and criteria for the original variant classification used by the clinical laboratory (especially for ACMG criteria that were published only in 2015) or their threshold for recontacting physicians with updated reclassifications. The testing laboratory may not have had access to complete family history and phenotype at the time of initial testing, making the evaluation of de novo status or segregation analysis difficult, and may also have been limited by the unavailability of comprehensive gene curation data from ClinGen, absence of well‐established variant classifications in ClinVar, lack of population‐specific allele frequency information, incomplete or outdated in silico prediction tools, or insufficient genotype–phenotype correlation data at the time of analysis. In our reinterpretation, the absence of complete family history and detailed phenotype information in some cases may have limited our ability to evaluate de novo status or perform segregation analysis, potentially leading to missed opportunities for variant reclassification. Another limitation is related to statistical analysis. As some patients harbored multiple variants, the assumption of independence required for Fisher's exact test may not have been fully satisfied.

Clinical Significance

The findings of a meaningful change in variant pathogenicity in 21.6% patients has clinical implications. The 14 patients in whom P/LP variants were downgraded will require all previously genotype‐negative family members to return for ongoing clinical screening. For the 35 patients in whom VUS were upgraded to P/LP, family members can now be offered cascade genetic testing to determine who is at risk and needs ongoing clinical surveillance and who is genotype negative and can be discharged from follow‐up. Providing timely feedback to patients regarding changes in variant classification can lead to changes in follow‐up protocols and may improve prognosis for other family members. A reevaluation system for a broad range of variants, including those that emerge as children transition into adulthood is essential given the long‐term prognosis of pediatric patients.

CONCLUSIONS

One in 5 patients with childhood onset cardiomyopathy had a clinically actionable change in variant pathogenicity on follow‐up that affects clinical and genetic family screening recommendations. This highlights the importance of serial reevaluation of variant pathogenicity by clinical providers as well as by clinical genetic testing laboratories.

Sources of Funding

This project received support from the Canadian Institutes of Health Research's Canadian Heart Function Alliance Network Grant (HFN 181992) (Seema Mital), the Ted Rogers Centre for Heart Research (Seema Mital, Rebekah Jobling), and the Heart & Stroke Foundation of Canada/Robert M Freedom Chair of Cardiovascular Science (Seema Mital). Takanori Suzuki was funded by the Japan Heart Foundation Research Grant, Canadian Heart Function Alliance, and Philip Witchel Research Fellowship in Heart Failure at the Hospital for Sick Children. Christoph Sandmann was funded by the German Cardiac Society, the German Centre for Cardiovascular Research, and the Dr Rolf M. Schwiete Foundation. Perundurai S. Dhandapany was funded by the Department of Biotechnology (BT/PR45262/MED/12/955/2022). Vinay J. Rao was funded by ICMR‐SRF (3/1/1 (8)/CVD/2020‐NCD‐1).

Disclosures

Seema Mital is on the Scientific Advisory Board of Bristol Myers Squibb, Rocket Pharmaceuticals, and Tenaya Therapeutics. The remaining authors have no disclosures to report.

Supporting information

Tables S1–S3