In a previous cross-sectional screening study of 5,169 middle and high school students (mean age, 13.1 ± 1.78 yr) in which we estimated the prevalence of high-risk cardiovascular conditions associated with sudden cardiac death, we incidentally detected by cardiac magnetic resonance (CMR) 959 cases (18.6%) of left ventricular noncompaction (LVNC) that met the Petersen diagnostic criterion (noncompaction:compaction ratio >2.3). Short-axis CMR images were available for 511 of these cases (the Short-Axis Study Set). To determine how many of those cases were truly abnormal, we analyzed the short-axis images in terms of LV structural and functional variables and applied 3 published diagnostic criteria besides the Petersen criterion to our findings. The estimated prevalences were 17.5% based on trabeculated LV mass (Jacquier criterion), 7.4% based on trabeculated LV volume (Choi criterion), and 1.3% based on trabeculated LV mass and distribution (Grothoff criterion). Absent longitudinal clinical outcomes data or accepted diagnostic standards, our analysis of the screening data from the Short-Axis Study Set did not definitively differentiate normal from pathologic cases. However, it does suggest that many of the cases might be normal anatomic variants. It also suggests that cases marked by pathologically excessive LV trabeculation, even if asymptomatic, might involve unsustainable physiologic disadvantages that increase the risk of LV dysfunction, pathologic remodeling, arrhythmias, or mural thrombi. These disadvantages may escape detection, particularly in children developing from prepubescence through adolescence. Longitudinal follow-up of suspected LVNC cases to ascertain their natural history and clinical outcome is warranted.
During development, the left ventricle (LV) of the heart contains bundles of muscle called trabeculae. Compaction of these trabeculae eventually transforms the heart muscle from spongelike to smooth and solid. Left ventricular noncompaction (LVNC), a rare congenital cardiomyopathy, results when compaction does not occur. In some cases, it may coincide with another myocardial disease (hypertrophic, dilated, or restrictive cardiomyopathy).
Left ventricular noncompaction can be diagnosed at any age but in many individuals goes undiagnosed until later in life. Its prevalence in the general population is unknown but ranges, according to various qualified estimates, from <0.3% to 1.26% in young children1,2 and from 0.02% to 0.17% in adults.3,4
The Texas Heart Institute recently completed a large population-based screening study (the Screen to Prevent [S2P] Study) of 5,169 middle and high school students from across the Harris County (Houston), Texas, area to estimate the prevalence of congenital conditions associated with sudden cardiac death in the young.5,6 Screening included cardiac magnetic resonance (CMR) imaging, a resting electrocardiogram (ECG), and a focused questionnaire.
Various CMR-based diagnostic criteria for LVNC have been proposed (Table I).7–18 Because these criteria are based on small samples, various assumptions, and no accepted standard, their reliability remains in question. Applying the Petersen criterion15 (noncompaction:compaction [NC:C] ratio >2.3) to CMR data from the S2P Study population identified 959 (18.6%) cases of LVNC.5,6 This observational finding, which is at odds with published estimates of LVNC prevalence,1–4 prompted us to examine information from the S2P Study more closely by using CMR-based metrics to determine whether the cases of LVNC detected by applying the Petersen criterion were truly abnormal or just normal anatomic variants.
Patients and Methods
Data for the present study were obtained from the S2P Study database. The 5,169 voluntary participants in the S2P Study (mean age, 13.1 ± 1.78 yr) were a representative sample of the general population of children 11 to 18 years old in Harris County (Houston), Texas. Of the 959 participants in the S2P Study identified as having LVNC, 511 had short-axis CMR imaging results available in the database. This cohort, called the Short-Axis Study Set, was the subject of the current analyses.
Data collected in the S2P Study included demographics, vital signs, brief cardiac medical histories, CMR-based screening data, resting ECGs, and results of overall screening evaluations by cardiology experts. The CMR imaging protocol for the S2P Study, which was designed for primary screening and not full-fledged clinical imaging, included acquisition of long-axis views (2–chamber, 4–chamber, and LV outflow tract), acquisition of short-axis views (for many but not all cases), and sequences for detecting coronary artery origins and proximal courses. Sixteen-lead ECGs were acquired and reviewed by an experienced electrophysiologist. These data were acquired, evaluated, deidentified, and used as provided for in the institutional review board-approved protocol for the S2P Study.5,6
Short-Axis Image Measurements
The current analysis relied on the use of long– and short-axis CMR images acquired in the S2P Study. Long-axis CMR imaging enables rapid detection of a distinct layer of excessive ventricular trabeculation. Short-axis imaging permits more accurate delineation of the compact LV myocardial layer and the contained ventricular cavity; when present, the additional layer of ventricular trabeculae can also be distinguished from the compact layer. Of note, however, measurements derived from both long– and short-axis images are subject to bias and error due to imaging plane location and orientation; in addition, nonperpendicularity relative to the long axis of short-axis images may introduce systematic errors that might be accentuated in the apical region.
Although current imaging technology does not enable full resolution of the complexities of ventricular trabeculae, the trained observer can distinguish gross features, including the location, local apparent thickness, and distinctive layering of trabeculae against the compact myocardial walls and their irregular boundary with the ventricular cavity. Therefore, the conventional tools and methods for quantifying imaging-derived myocardial variables in the normal heart were used to evaluate LV trabeculation, especially the volume of the trabeculated region.
The short-axis imaging stacks used for screening in the S2P Study typically included 8 slices, each 8 mm thick and separated by 1 mm. The slices were acquired from just below the mitral valve to near the apex. Additional technical details of the CMR imaging procedure are available elsewhere.5,6
Clear and consistent short-axis images of the LV cavity, along with clear delineations of the trabeculated layer and compact myocardium, were obtained for the Short-Axis Study Set at both end-diastole and end-systole. The epicardial and endocardial contours were drawn by 2 experienced cardiac radiologists (CU and BYC). To quantify areas of excessive trabeculation, borders were drawn between the compact myocardium and the trabeculated layer, as well as between the trabeculated layer and the open ventricular cavity (Fig. 1).
Epicardial and endocardial contours were analyzed to generate volume estimates for the LV cavity, trabeculated region, and compact myocardium according to established standards.19 From these measurements, values for other structural and functional variables were derived. Contouring and measurement of CMR images were done with use of cvi42 image-postprocessing software (Circle Cardiovascular Imaging Inc.). Conventional image-processing software was used to generate volume and mass estimates for the various structures delineated by the contours described above and displayed in Figure 1. For example, the green and red contours were analyzed with the cvi42 software to estimate the volume and mass of the compact LV myocardium. The green and blue contours were analyzed to obtain aggregate volume estimates of compact and trabeculated (noncompact) LV myocardium. The trabeculated LV volume was then obtained by difference. Last, estimates of trabeculated LV volume at end-diastole and end-systole were obtained and used to assess changes in trabeculated LV myocardium across the cardiac cycle.
Trabeculated Left Ventricular Mapping
For each case, the short-axis images were used to map the trabeculated LV as described above and to measure its areal extent and the thicknesses of noncompact (trabeculated) and compact myocardial tissue at end-diastole. Each short-axis image slice was subdivided into 6 segments for this purpose. The fraction of all segments in which trabeculation met the Petersen criterion (NC:C ratio, >2.3) served as a proxy for the interior LV surface area covered with trabeculae meeting that criterion.
As a first step toward better characterizing the effects of excessive LV trabeculation, we used conventional volumetric measurements derived from short-axis LV images to generate a 2-dimensional, quantitative description of LV trabeculation in terms of the following:
EFe = LV ejection fraction (LVEF) estimated with trabeculae and papillary muscles excluded from the blood pool;
VdT = volume of trabeculae at end-diastole as estimated from LV images;
VsT = volume of trabeculae at end-systole estimated from LV images; and
LVMc = mass of LV compact myocardium estimated from short-axis LV images.
To ensure consistency and to avoid systematic errors, all imaging-based volumetric measurements were subjected to the following logic checks:
The aggregate volume of LV trabeculae and compact myocardium at end-diastole was greater than at end-systole.
The aggregate volume of LV trabeculae and compact myocardium at end-systole was greater than that of the compact myocardium alone.
The LV cavity volume with trabeculae included in the blood pool was greater than the LV cavity volume with trabeculae excluded from the blood pool, at both end-diastole and end-systole.
The LV stroke volume estimated from contours that included trabeculae in the blood pool was greater than the LV stroke volume estimated from contours that excluded trabeculae from the blood pool.
Net Relative Trabeculation Mass
We used short-axis LV images to estimate more accurately the trabeculated LV mass, specifically excluding the mass of the blood contained in the intertrabecular luminal spaces. This variable was defined as the net relative trabeculation mass (NRTM).
Estimation of Left Ventricular Noncompaction Prevalence
Measurements of LV variables were indexed to body surface area (BSA) (computed by the Mosteller formula) where applicable. Continuous data were summarized as mean ± SD. Statistical comparisons were performed with use of t tests. P <0.05 was considered statistically significant. Pearson and intraclass correlation coefficient analyses and Bland-Altman analyses were performed to evaluate measurement repeatability. Stata version 14.0 software (StataCorp LLC) was used for all analyses.
The Short-Axis Study Set (N=511) was comparable to the full S2P Study cohort (N=5,169) with regard to age, race, ethnicity, and sex distribution (Table II). Ninety-four (18.4%) participants in the Short-Axis Study Set reported symptoms: shortness of breath with exercise in 60 (11.7%), chest pain in 22 (4.3%), syncope with exercise in 7 (1.4%), and syncope at rest in 5 (1%). Thirteen participants had cardiac disorders: hypertrophic cardiomyopathy in 2; dilated cardiomyopathy in 3; anomalous origin of coronary arteries in 3; and the ECG abnormalities Wolff-Parkinson-White syndrome in 2, prolonged QT interval (QTc >470 ms) in 2, and fixed ventricular bigeminy in one.
Left Ventricular Volume Estimates
Left ventricular end-diastolic volume (LVEDV) estimates were affected by the presence of trabeculae and their inclusion in or exclusion from the blood pool. Supplemental Table I presents data from 2 sample cases to demonstrate how we derived our estimates of LV trabeculation-related variables from conventional imaging-derived volumetric measurements. Supplemental Table II presents BSA-indexed summary statistics for those variables for the entire Short-Axis Study Set.
The mean indexed LVEDV was significantly greater when trabeculae were included in the blood pool (76.9 ± 12.2 vs 49.2 ± 9.6 mL/m2; P <0.001) (Fig. 2). The trabeculated LV region had a spectrum of presentations and a measurable volume at end-diastole (mean, 25.5 ± 8.1 mL/m2). The volume of blood contained in intertrabecular luminal spaces was therefore dependent on the extent and structure of the trabeculated region (Fig. 3).
The trabeculated region was apparently compressed across the cardiac cycle. The mean reduction in its volume at end-systole was 52.5% ± 17.2%, reflecting primarily compression of the intertrabecular luminal spaces (Fig. 4). Thus, blood ejected from the intertrabecular luminal spaces at end-systole necessarily became part of the total LV stroke volume, and the wide range of trabeculated region compression suggests a varied gross structure (that is, a spectrum of intertrabecular luminal space volumes).
Including papillary muscles and trabeculae in the LV blood pool resulted in the systematic underestimation of LVEF (Fig. 5).
The repeatability of imaging-derived volumetric measurements was evaluated in a randomly selected sample of 28 (5.5%) cases to check the consistency of the initial contouring done as described above. In brief, repeat measurements of 7 distinct variables derived from primary volumetric estimates were generated by the image-postprocessing software from the contours as initially drawn at baseline. Baseline and repeat measurements were made approximately one year apart. As expected, and as shown by correlation and bias analyses, measurements at end-systole were more susceptible to error (Table III). These errors were propagated because LV stroke volume estimates were affected by the presence of trabeculae.
Estimated Prevalence of Left Ventricular Noncompaction
After applying the criterion of Jacquier and associates11 to the Short-Axis Study Set, we identified 482 cases as abnormal, a prevalence of 17.5%. In contrast, after applying the criterion of Choi and associates,8 we identified 205 cases as abnormal, a prevalence of 7.4%. This mismatch between these 2 estimates of LVNC prevalence is shown in Figure 6. In addition, our trabeculation modeling indicated that 163 (17%) of cases in the Short-Axis Study Set met the Petersen criterion15 but had trabeculae of no significant mass (that is, open, lace-like trabeculae that on long-axis images appeared to be very thick). Thus, relatively few of the LVNC cases in the Short-Axis Study Set that met the Petersen criterion also had high NRTM values (Fig. 7).
We also applied the CMR mass-based criteria proposed by Grothoff and associates9 and our own method for estimating trabeculated LV mass in terms of NRTM (Fig. 8). This identified 35 (3.7%) cases in the Short-Axis Study Set as abnormal, corresponding to a prevalence of 1.3% when this finding was extrapolated to the larger S2P Study cohort.
Our detailed analysis of 511 CMR-detected cases of LVNC, originally identified according to the Petersen criterion,15 suggests that many were normal anatomic variants. After we applied the criteria of Grothoff and colleagues9 to our cohort, the estimated prevalence of LVNC was much lower than that in our original S2P Study. The Petersen criterion is too broad, but other criteria (for example, those proposed by the Grothoff,9 Jacquier,11 and Choi8 groups) may be no more reliable, as we have shown in this analysis.
Left ventricular noncompaction is a poorly understood cardiomyopathy20 with various reported phenotypes (for example, isolated or concomitant with hypertrophic or dilated cardiomyopathy) that may change over time.21 Its cause is uncertain.7,20,21 Multiple genetic markers have been identified, and potential familial involvement is often cited.22–34 As annotated in the ClinVar public archive of genetic variants35 and reported elsewhere,22–25,27–36 more than 70 pathogenic or likely pathogenic variants in more than 20 genes have been deemed to be of clinical significance in LVNC. Not all of those genetic variants may cause disease, and mutations in several LVNC-linked genes are also linked with other cardiomyopathies.
Current diagnostic criteria for LVNC are unreliable,7–18,23 and treatment is driven by phenotype and symptoms and aimed at preventing or treating heart failure.37–40 There is no cure. Neonates and very young children with more severe phenotypes have poor outcomes and low survival rates.2,39,41 Some young people with isolated LVNC may have a better prognosis.23,42
It is unclear whether poor outcomes in the young are caused by LVNC, or whether LVNC is secondary to another cause of those outcomes.43,44 Case reports repeatedly point to the clinical complexity of LVNC cases,45–57 and numerous other studies document the limited understanding of this condition.7,50,58,59 Asymptomatic LVNC is often identified incidentally, and its clinical management typically includes follow-up with prophylactic measures, such as anticoagulation and antiarrhythmia therapies, and restrictions on physical activity.37–40 The effect of these restrictions on lifestyle choices for children is complex and not fully understood.60,61 The change in some LVNC phenotypes over time is of concern for children as they develop from prepubescence to early adolescence. Evolving LVNC phenotypes require adaptive clinical surveillance and treatment strategies.20
This work was based on cross-sectional CMR-based data obtained at primary screening during the S2P Study and focused on analyzing objective measurements derived from those data. Longitudinal measurements may provide more insight. Participants in the S2P primary screening study took part voluntarily and, therefore, do not constitute a rigorous sample of the general population. That said, the demographics of the S2P Study population compare favorably with those of the general population in the greater Houston, Texas, area.5
In analyzing short-axis images from the 511 participants in the Short-Axis Study Set, we considered only the nonapical regions of the heart and thus may have underestimated trabeculated LV mass. We made no attempt to exclude the papillary muscles from the regional volume measurements of LV trabeculae. Although this may have introduced a systematic but limited bias into our estimates of trabeculated LV mass, we think that this bias had little effect on estimates of the volume of nonapical intertrabecular luminal spaces.
No clinical assessments were performed to confirm LVNC in study participants. In addition, we did not acquire any longitudinal follow-up data that would help to detect changes or provide information on clinical outcomes.
The cases of LVNC that we identified during screening in the S2P Study included several cardiac disorders associated with LVNC; however, confirmatory clinical evaluations and longitudinal follow-up are needed to ascertain their clinical course. One-time screening does not allow detection of changes in a cardiac disorder over time, and small sample size impedes reliable estimation of the underlying prevalence of each of those disorders.
Observer-related variability and measurement error, especially regarding the visual delineation of the irregular features of ventricular trabeculae, markedly affect the imaging-derived quantification of cardiac variables. Automated image quantification will improve measurement accuracy and precision and help provide information that could be used to reliably evaluate patients presenting with excessive trabeculation. Quantification of LV function and mass by contour drawing is widespread, and there is good understanding of measurement variability62 and the usefulness of CMR-based measurement in defining significant changes.63 We acknowledge that automation of CMR image analysis, although not routine, is advancing. Examples include automated endocardial border detection,64–66 selection of short-axis slices for automated quantification,67 LV segmentation and finite element modeling,68,69 application of “machine learning” to contouring,70 and fractal analysis of myocardial trabeculation.71
Our findings suggest that many of the 511 suspected cases of LVNC identified in our previous S2P Study may have been normal anatomic variants and that many could be reclassified as normal on the basis of our current findings. The remaining cases, including those that were largely asymptomatic, may have involved subclinical changes in LV structure and function. Therefore, the true prevalence of pathologically excessive LV trabeculation in our 11- to 18-year-old cohort in the S2P Study is probably <1%, while the severity and clinical significance of potential LVNC remain unknown. Young people presenting with excessive LV trabeculation, even if asymptomatic, may exhibit distinct structural features that are physiologically disadvantageous. Whether aggravated by exertion (for example, by participation in sports) or by normal development from puberty into early adolescence, the eventual cumulative effect may be reduced ventricular function, with its associated complications and risks. Excessive LV trabeculation, even if initially asymptomatic, may serve as a substrate for malignant arrhythmias or thrombus formation. In light of our findings based on a population sample, longitudinal follow-up of individuals with suspected LVNC is warranted in general. A more detailed analysis of other CMR-based metrics in the Short-Axis Study Set will be presented in the future.
Raja Muthupillai, PhD, provided the CMR imaging protocol and guidance regarding imaging quality control. Veronica V. Lenge De Rosen, MD, reviewed the CMR screening data to help identify potential abnormalities. Stephen N. Palmer, PhD, ELS, contributed to the editing of the manuscript.
Supplemental tables for this article are available at 10.14503_thij-19-7157.s1.pdf.
Funding for the Screen to Prevent (S2P) Study was generously supplied by the Kinder Foundation, the Brown Foundation in Houston, and the Texas Heart Institute's general funds.