In our cross-sectional study, we included 54 (33 male) symptomatic subjects with LVNC phenotype and preserved LVEF (≥50%) and 54 (33 male) sex- and age-matched C individuals from a Caucasian population. The baseline characteristics are reported in Table 1.

In the LVNC group, symptomatic subjects with a persistent isolated hypertrabeculated phenotype fulfilling the Petersen (noncompact/compact myocardial layer ≥2.3) and Jacquier LVNC criteria (noncompact myocardial mass >20% of the total myocardial mass) and preserved LVEF (≥50%) were enrolled at our expert CMR center (4, 5). Patients with reduced LVEF (<50%); congenital, ischemic or valvular heart diseases; other or overlapping CMP; hypertension; relevant comorbidities (see details at 2.4 Clinical evaluation); exercise activity of >6 h/week or other cause of transient hypertrabeculation were excluded from our study (24). Other exclusion criteria included technical reasons, e.g., artifacts, implanted devices or contrast agent administration before segmentation (25).

The C group included healthy volunteers without known cardiovascular or extracardiac diseases and with exercise activity of <6 h/week (24).

All procedures performed in this study were in accordance with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Ethical approval was obtained from the Central Ethics Committee of Hungary, and all participants provided informed consent.

The CMR examinations were performed using 1.5 T magnetic resonance imaging (MRI) scanners (Magnetom Aera, Siemens Healthineers, Erlangen, Germany and Achieva, Philips Medical System, Eindhoven, the Netherlands). Retrospectively gated, balanced steady-state free precession (bSSFP) cine sequences were performed with short-axis (SA) and two-, three-, and four-chamber long-axis views from base to apex, covering the whole left and right ventricle (LV and RV). The slice thickness was 8 mm with no interslice gap, and the field of view was 350 mm on average, adapted to body size.

Contrast agent was administered when it was needed to 50 LVNC subjects: 44 received gadobutrol (0.1 ml/kg), 5 received gadobenate (0.2 ml/kg) and 1 received gadoterate (0.2 ml/kg). In four LVNC subjects [3 from the variant of uncertain significance (VUS) group and 1 from the benign (B) group] and healthy controls, contrast agent was not administered. To provide the best image quality for postprocessing analyses, contrast agent administration occurred after the acquisition of SA cine images (25).

For postprocessing analysis, we used Medis Suite software (Medis Suite QMass, version 4.0, Medis Medical Imaging Systems, Leiden, the Netherlands).

Semiautomatic tracing with manual correction from base to apex was used on end-diastolic and end-systolic SA cine images, and then the threshold-based (TB) algorithm (MassK module of the Medis Suite QMass program) was applied. Based on different signal intensities, the TB program classifies each voxel as blood or myocardium. The voxels identified as myocardium on the end-diastolic images within the epicardial contours represent the total myocardial mass (TM), and those within the endocardial border form the trabeculated and papillary muscle mass (TPM) (Figure 2). The threshold was set to the default of 50% (26). Manual correction of the threshold was not performed.

The LV and RV end-diastolic volume (EDV), end-systolic volume (ESV), stroke volume (SV), and ejection fraction (EF) were also calculated. All parameters were indexed to body surface area (i). As a reference, we used the Alfakih normal values for adults and the percentiles by Kawel-Boehm for children (27, 28).

For the feature-tracking (FT) analysis, commercially available software (QStrain, Medis Suite, version 4.0) was used. The LV endocardial contours were semiautomatically traced with manual correction on the SA view and manually traced on the 2-, 3-, and 4-chamber long-axis views. LV global longitudinal strain (GLS) and global circumferential strain (GCS) were measured and compared with the normal values presented by Peng et al. (29, 30). During the analysis, the strains were interpreted as absolute values: values closer to zero represented worse/poorer function, and values farther from zero represented better function.

Manual correction of the semiautomatically traced epi- and endocardial contours was performed by two observers with 11 years (ASZ) and 7 years (ARK) of experience.

The clinical data, including demographic characteristics, were evaluated by questionnaires at the time of cardiogenetic counseling in the LVNC subjects and at the time of CMR examination in the C group. Detailed family history data included hereditary cardiac diseases or arrhythmia, sudden cardiac death (SCD) or recurrent loss of consciousness, and implanted cardiac devices. Personal medical history included syncope; dizziness; atypical chest pain; palpitation; structural cardiac diseases and arrhythmias; SCD; thromboembolic events; implanted cardiac devices; myocardial infarction; and other comorbidities, such as diabetes mellitus; pulmonary, gastrointestinal, hepatic, kidney, endocrine, oncologic, rheumatic, autoimmune, neurologic, and psychiatric diseases; metabolic syndromes; acute infectious or current cardiac symptoms; and sports activity. LVNC subjects also participated in a personal clinical genetic counseling and genetic testing. Additional information was extracted from previous electronic medical reports.

For a better assessment of the clinical picture, based on the risk stratification recommendation from available literature data, risk factors and negative clinical endpoints were considered “red flags” for clinical evaluation (11–14). These red flags included family history of CMP or SCD in first-degree relatives (subsequently referred to as positive family history), increased LVEDVi on CMR images, ventricular tachycardia or ventricular fibrillation, left bundle branch block (LBBB) or inverted T waves on ECG, medical history of unexplained syncope or thromboembolism or SCD, and nonischemic, mid-myocardial or subepicardial late gadolinium enhancement (LGE) on CMR (11, 12, 14, 19, 27, 28, 31, 32). In previous clinical management, a minor temporary decrease in LVEF (max. to 45%) that normalized after initiation of drug therapy was also considered a red flag in risk stratification (11).

After genetic counseling, a peripheral venous blood sample was taken from LVNC subjects for genetic testing.

The samples were analyzed using a next-generation sequencing (NGS) method with a 174-gene panel (TruSight Cardio Sequencing Kit, Illumina, CA, USA) containing genes previously associated with cardiac diseases (covering 571,897 nucleotides and 3,251 exons). This method can identify single-nucleotide variations and insertions or deletions in small genetic segments without detecting large CNVs.

During the sequencing reaction, paired-end reads 150 nucleotides in length were synthesized. The fastq files converted from optical information were quality controlled using FastQC (v0.11.9) and MultiQC (v1.9) software. The obtained genetic information was compared to the GRCh37.p13 assembly version of reference genome hg19 with BWA software (v0.7.12). Thereafter, to detect mutations, we used Broad Institute GATK software (v4.1.7.0), followed by manual variant screening.

Variant categorization and clinical relevance were determined using the Franklin, NCBI - ClinVar and VarSome databases, with the variants being classified according to the American College of Medical Genetics and Genomics (ACMG) recommendations. To assess previously determined mutation-associated cardiac and extracardiac diseases, we used the ClinGen and OMIM clinical genetic databases.

During genetic data interpretation, pathogenic (P), likely pathogenic (LP), VUS, likely benign and benign mutations were distinguished according to international ACMG guidelines and recommendations (33). In our study, subjects were divided into the following genetic subgroups according to CMP-related mutations: subjects with P and LP variants were combined into the P subgroup, individuals with VUS were classified in the VUS subgroup, and subjects without these mutations were classified in the B subgroup.

Furthermore, CMP-related P and VUS mutations were divided into two subtypes: directly LVNC-related, when the mutation is directly associated with “excessive trabeculation” according to international databases (e.g., Franklin), and other CMP-related, which are mutations that are not directly associated with “excessive trabeculation” but, due to the genetic overlap with CMPs, could have a pathogenic effect (15).

Continuous parameters are described as the mean and standard deviation (SD), and discrete parameters are described as numbers and percentages. The homogeneity of variances was assessed with Levene's test, and the normal distribution was assessed with the Shapiro–Wilk test. The differences in baseline and CMR parameters between the LVNC and C groups were assessed with independent t tests when normally distributed and Mann‒Whitney U-tests when nonnormally distributed. The continuous parameters among the P, VUS and B subgroups were compared with one-way analysis of variance (ANOVA) and Tukey's post hoc test in normally distributed variables with equal variances, Welch test and Games-Howell post hoc test in variables with unequal variances, and Kruskal–Wallis test with Bonferroni correction in nonnormally distributed data. The chi-square test and Fisher's exact test were used to compare discrete data. Correlations were assessed with the Pearson correlation coefficient. The multinomial logistic regression model was used to further analyze the connection between red flags and genotype. The interobserver agreement of the two observers was tested using the intraclass correlation coefficient (ICC). A p-value <0.05 was considered indicative of statistical significance. IBM SPSS Statistics (Version 28.0) was used for the statistical analyses.

Although the functional CMR parameters of the total LVNC population were within the normal ranges, the LV volumetric and muscle mass parameters were significantly higher, and the LVEF and strain values were significantly lower in the LVNC group than in the C group. These results are consistent with literature data by Zemrak et al. and Kiss et al. (7, 34, 35). In terms of the RV, Kiss et al. reported significantly higher RV volumes and RVTMi in the LVNC group than in controls; however, our results were comparable between the two groups (35). This could be explained by our smaller study population.

In contrast to our results, where the CMR parameters showed no significant differences among the genetic subgroups, previous investigations described the relationship of pathogenic genotype with LV systolic dysfunction and raised a connection with the extent of trabeculation; however, other studies reported no prognostic impact of trabecular mass (7, 11, 19, 21, 36–38). Notably, all these studies were conducted on both preserved and reduced LVEF LVNC populations with a high number of HF patients, and to the best of our knowledge, no data were available about these associations in LVNC subjects with preserved LVEF.

We also intended to analyze the genetic subgroups and their clinical manifestations. Previous studies about the genetic background of LVNC conducted on the heterogeneous spectrum of LVNC populations described similar or slightly higher rates of pathogenic mutations in the same genes as our results, which could be explained by the previously described relationship between LV systolic function and pathogenic genotype (1, 18, 19, 39). Moreover, in patients with severe HF resulting in heart transplantation, this rate was even higher (87.5%) (23). Similarly, according to our results and those of previous studies, the prevalence of P mutations can also be higher in cases of childhood diagnosis (19, 22, 40, 41).

Based on the available literature reviews by Vergani et al. and Negri et al, which emphasize the importance of using red flags in the risk stratification of LVNC subjects and previous studies describing the prognostic effect of clinical characteristics, we developed a new red flag model presented in Methods (11, 12, 14, 19, 27, 28, 31, 32). The novelty of our model is the described statistical association between red flags and genotype. Here, we would like to analyze this revised red flag risk stratification model.

In the literature, major cardiovascular events of LVNC, such as thromboembolic events, arrhythmia and HF symptoms as a part of the red flag system, usually accumulate with systolic dysfunction, which can be associated with the presence of pathogenic genotypes (11, 12, 14, 18, 21, 22). However, complications could occur despite a preserved LVEF, as we found a significantly higher prevalence of thromboembolic events and positive family history (CMP or SCD) in subjects with P and VUS mutations, while SCD occurred in only one case with the P genotype. Interestingly, no previous literature data were found concerning the genotype-phenotype association in the preserved LVEF group.

In addition to the abovementioned clinical entities, the presence of unexplained syncope, ventricular tachycardia or fibrillation, inverted T waves, LBBB, LGE, elevated LVEDVi and temporary decrease in LVEF were also considered parts of this model. Although these issues did not show a significant relationship with P and VUS genetic mutations, probably due to the small sample size, their importance in long-term outcome and risk stratification was described previously (11, 12, 14, 19, 27, 28, 31, 32). Syncope was reported as an early warning sign of ventricular arrhythmias; moreover, some studies revealed the importance of the P genotype in the development of different types of ventricular and supraventricular arrhythmias (12–14, 18, 22, 31, 42). It was also suggested that repolarization abnormalities indicated by inverted T waves and LBBB also have connections with cardiovascular events (12, 43–45). Interestingly, Van Waning et al. reported a higher occurrence of LBBB in sporadic LVNC than in genetically determined LVNC (19). Further red flags, such as elevated LVEDVi and temporarily decreased LVEF, were identified previously as predictors of negative outcomes in LVNC patients (11). Similarly, a recent follow-up study identified LGE as the strongest independent predictor of LVNC-specific complications. Other investigators also described the association between LGE and the P genotype, which is comparable with our results, as we detected LGE only in subjects with CMP-related mutations (11, 32, 36, 37). Interestingly, atypical chest pain was more frequent in subjects without CMP-related mutations, and the prevalence of dizziness and palpitation were comparable among the three genetic subgroups, raising questions regarding their importance in risk stratification.

In contrast to evaluating each clinical parameter separately, some reviews have suggested the importance of a more complex risk stratification as the red flags model, which enables a better characterization of the entire clinical picture. However, no literature data were found concerning its relationship with genotype and its usability in clinical practice (12, 14). Thus, our present study is unique in determining a relationship between genetic mutations and red flags. In this study, we also described not only the presence but also the number of red flags per person: subjects with multiple red flags have a higher probability of P mutations; thus, in these cases, genetic testing can be recommended. In contrast, LVNC subjects with no or only one red flag have a lower possibility of CMP-related mutations, so they might require a more flexible follow-up during clinical management (1). These results are supported by the “Expert Consensus Statement on the state of genetic testing for cardiac diseases” by Wilde et al., in which similar clinical circumstances are considered indicators of the need for genetic testing (46). Interestingly, we found significantly more red flags in subjects with directly LVNC-related mutations than in individuals with other CMP-related mutations. This may support specific aspects of diagnosing the “excessive trabeculation” phenotype and its follow-up management.

Finally, the importance of the VUS subgroup in the genetic classification must be mentioned, as in our study, subjects with these mutations usually had one or two red flags per person. According to the ACMG guidelines, VUS mutations should not be considered in clinical decision-making, but additional monitoring may be necessary during the follow-up of these subjects, which is also underlined by our results (2, 33). Furthermore, the interpretation of VUS can be difficult in daily medical practice, as new clinical data can modify the current ACMG database.