Cardiac magnetic resonance arrhythmia prediction model in Ebstein’s Anomaly: A multicenter study

Background: Ebstein’s anomaly (EA) is a rare congenital heart disease, in which patients are at increased risk of arrhythmia. Cardiac magnetic resonance (CMR) is the gold standard for volumetric and functional indexes; anatomical dissaragement parameters have been studied in CMR; however, data regarding arrhythmia prediction in EA is scarce. We aim to predict arrhythmia occurrence using a CMR-derived model.

Methods: : Multicenter retrospective study of EA patients who underwent CMR at two tertiary centers (Cardioinfantil and Children’s Hospital of Philadelphia). A total of 147 EA patients from both institutions were evaluated, after excluding patients with pre-existing arrhythmias, complex concomitant congenital heart disease, previous tricuspid valve repair or replacement, or insufficient data for CMR measurements or medical records, 93 patients were included in the final analysis. Three patients were included in the analysis. Anatomical, volumetric, and functional CMR parameters were obtained, and new arrhythmia events after CMR were confirmed from medical records. Continuous variables were compared using the t-test/Wilcoxon test, and categorical variables were compared using the Chi-square/Fisher test. Given the large number of correlated variables, the least absolute shrinkage and selection operator (LASSO) regression was applied to identify the most relevant predictors.

Results: Ninety-three patients were included (mean age 24.7±14.9 years, 52% female); 16 (17.2%) developed arrhythmias, predominantly supraventricular tachycardia (68.8%) and 9.1% sustained ventricular tachycardia. Patients with outcome had larger right ventricular end systolic volume (141 vs. 72 ml/m², p=0.025), greater tricuspid annulus size (59.4 vs. 47.5 mm, p=0.001), and lower right ventricular ejection fraction (RVEF) (27.1 vs. 32.5%, p=0.135). LASSO regression selected tricuspid annulus size (OR 1.12, 95% CI 1.02–1.28, p=0.04) and RVEF (OR 0.91, 95%CI 0.82–1.00, p=0.07) as the most relevant predictors. The final model showed very good performance (AUC=0.85) and good calibration.

Conclusion: In this multicenter EA cohort, the size of the tricuspid annulus, as measured by cardiac magnetic resonance (CMR), emerged as the strongest predictor of arrhythmia. However, given the limited data on ventricular arrhythmias, this analysis primarily pertains to atrial arrhythmias. The hypothesis includes an enlarged tricuspid annulus, which may result in electro-anatomical disarray that contributes to arrhythmias through various mechanisms, including increased tricuspid regurgitation, greater apical displacement, and reduced functional ventricular capacity. Further research is needed to draw more definitive conclusions on this topic.

Table 1: Patient baseline characteristics on each center

	Fundación Cardioinfantil			Children's Hospital of Philadelphia
	Arrhythmia	No Arrhythmia	p value	Arrhythmia	No Arrhythmia	p value
Patient general data
Age	26.1 (19.7)	23.5 (14.0)	0.70	15 (2-17)	21 (16-32.75)	0.03
Male	7 (63.6)	14 (45.2)	0.48	10 (76.9)	19 (41.3)	0.03
BSA	1.7 (1.6-1.7)	1.6 (1.5-1.8)	0.58	1.46 (0.97-1.72)	1.8 (1.38-1.8)	0.47
CMR variables
Heart rate	73.0 (66.5-82.5)	66.0 (58.0-86.5)	0.46	87.8 ± 15.6	78.5 ± 19.4	0.227
RVEDVi	232.4 (132.9-282.1)	138.4 (112.2-158.1)	0.06	217.15 (90.47-352.15)	219.15 (170.35-271.02)	0.93
RVESVi	141.1 (72.5-163)	71.7 (54.4-81.0)	0.03	98.7 (38.15-173.65)	98.95 (73.82-129.95)	0.93
RVSVi	91.3 (59.8-116.4)	67.3 (56.0-77.1)	0.14	114.65 (57.1-153.15)	115.15 (105.75-143.73)	0.739
RVEF	43.0 (7.0)	49.3 (8.1)	0.02	51.33 ± 12.37	55.5 ± 7.61	0.456
Aortic net flow	44.0 (16.9)	50.3 (18.1)	0.33	0.2 (0.12-1.73)	1.3 (0.4-2.12)	0.05
LVEF	47.9 (38.8-56.9)	55.1 (53.2-58.5)	0.10	59 ± 8.05	55.14 ± 7.78	0.31
True tricuspid annulus size	59.4 (8.8)	46.3 (11.2)	< 0.01	48.52 ± 11.63	52.13 ± 9.2	0.356
Ebstein angle	57.1 (9.3)	49.4 (12.5)	0.05	48.05 ±14.03	47.71 ± 12.94	0.943
Apical displacement index	49.4 (16.8)	34.0 (15.3)	0.02	45.2 (22.25-53.45)	35.4  (26.25-50.1)	0.974
Net pulmonary flow	24.3 (7.9)	33.7 (9.9)	< 0.01	1 (0.34-1.84)	0.45 (0.1-1.25)	0.284
Outcomes
Non sustained VT	0 (0.0)	NA		1 (1.72)	NA
Sustained VT	1 (2.3)	NA		0 (0.0)	NA
Supraventricular tachycardia	8 (19.0)	NA		8 (13.7)	NA
Atrial flutter	2 (4.7)	NA		4 (6.9)	NA
Atrial fibrillation	1 (2.3)	NA		4 (6.9)	NA
High grade AV block	1 (2.3)	NA		4 (6.9)	NA

Figure 1: Lasso regression model for arrhythmia prediction A. Receiver operating characteristic (ROC) curve showing the performance of the LASSO regression model in predicting new onset arrythmia B. Four-chamber view by CMR imsging demonstrating the direct measurement of the true tricuspid annulus C. Forest plot depicting the odds ratio (OR) and 95% CI for three independent predictors of arryhtmia: Right ventricular ejection fraction (RVEF), left ventricular stroke volume index (LVSVi) and true tricuspid annulus dimension