Association of cardiovascular magnetic resonance diastolic indices with arrhythmia in repaired Tetralogy of Fallot

We conducted a single-center retrospective investigation of patients evaluated at Vanderbilt University Medical Center, Nashville, Tennesee, USA. The study was approved by the Vanderbilt University Institutional Review Board. The electronic medical record was searched using an online tool developed at Vanderbilt that allows researchers to customize searches for ICD-9, ICD-10, CPT codes, and keywords. Our search for patient identification included ICD10 Q21.3, ICD9 745.2, and CPT codes associated with CMR (75,552–75,558, 75,560, and 75,565). The inclusion criteria were adults > 18 years with rTOF with pulmonary stenosis (PS), pulmonary atresia (PA), or absent pulmonary valve who had undergone at least one CMR after the age of 15 years and had at least one clinical encounter with a cardiologist. No gender-based differences were present. Individual medical charts were then reviewed by the research team. Study data were collected and managed using REDCap (Research Electronic Data Capture), an electronic data capture tool hosted at Vanderbilt University Medical Center [20]. The study cohort consisted of 167 patients with rTOF and were compared to healthy controls over 15 years old from a cohort of healthy subjects who had previously consented for non-contrasted CMR imaging as part of a separate project. The study was approved by the institutional review board with a waiver of consent for retrospective enrollment.

Chart review was performed to collect demographic and clinical data, including date of birth, gender, height, weight, age at time of CMR, original cardiac anatomy, date and age of each surgical procedure, type of surgical procedures, arrhythmia outcomes and mortality. The QRS duration was reviewed from the closest available electrocardiogram (ECG) from the time of CMR. Original anatomy was classified as TOF with PS, TOF with PA, TOF with PA and major aortopulmonary collaterals (MAPCAs), TOF with atrioventricular septal defect (AVSD) or TOF with absent pulmonary valve. Types of repairs documented were transannular patch, transannular patch with monocuspid valve, valve sparing, or RV to pulmonary artery conduit. Chart review was used to identify patients with a history of AA and VA, as well as the age of the patient at the occurrence of each arrhythmia. AA was defined as supraventricular tachycardia, atrial fibrillation, atrial flutter, and atrial tachycardia. VA was defined as non-sustained or sustained ventricular tachycardia, aborted sudden cardiac death, and sudden cardiac death. Non-sustained VA was defined as greater than 3 consecutive beats on cardiac monitoring. Sustained VA was defined as greater than 30 s. Only spontaneous arrhythmias were included.

Patient characteristics and CMR data associations were compared using Pearson’s Chi-squared for categorical variables and Wilcoxon rank sum for continuous variables. Data is presented as bootstrapped means with 1000 repetitions (95% confidence interval [CI] of the mean). Univariate logistic regression was performed to identify CMR and clinical predictors associated with arrhythmia. Odds ratios are reported with 95% CIs. Stepwise forward selection was then used to fit a multivariable model that maximized outcome prediction with factors considered to be significant at the 0.05 level. Receiver-operating characteristic (ROC) curves were generated with a calculated Area Under the Curve (AUC) to quantify arrhythmia prediction. For time to event analysis, patients with arrhythmia following CMR were analyzed with Cox proportional hazard regression models to estimate associations of each predictor with time of onset of arrythmia. Survival analyses were performed for VA using LV volumes, and LV diastolic variables (PER and tPER). Survival curves were compared using the log-rank test. Due to small sample size of patient developing arrhythmia following CMR analysis in this study, the time-to -event analysis was felt to be underpowered and is included as a supplement to the main manuscript. A linear regression analysis was also completed correlating pulmonary regurgitation and QRS duration with diastolic predictors to analyze if factors affecting ventricular septal position may affect diastolic variables. A commercially available statistical software package was used for data analysis (STATA, version 17.0, College Station, Texas, USA).

One hundred sixty-seven rTOF and 58 healthy control subjects were included. Table 1 summarizes baseline characteristics rTOF and healthy control group subjects at the time of CMR. The bootstrapped mean age of rTOF patients was 32.4 years (95% CI [30.4–34.4]) and 50.3% were male. The original anatomy of rTOF patients included 81.4% categorized as TOF with PS, 13.8% with PA, 2.4% with PA with MAPCAs, 1.8% with absent pulmonary valve, and 0.6% with TOF with AVSD. Most patients (75%) underwent rTOF with trans-annular patch, 12% underwent valve-sparing repair, and 13% were repaired with RV-Pulmonary Artery conduit. The mean age of repair was 59 months [47–71]. At the time of analysis 49% had undergone at least one pulmonary valve replacement (PVR) at a mean age of 26.8 years [26.8–32.4]. rTOF patients with ECG near the time of CMR (N = 162) also had a mean QRS duration of 141 ms [137–147].

Conventional volumes, LA parameters and LV diastolic data of rTOF patients versus control subjects are shown in Table 2. Compared to controls, rTOF patients had lower LV volumes (mean LVEDV index (LVEDVI) 67.2 ml/m² [64.1–70.3] v. 84.6 [81.0–88.0], p < 0.001), larger RV volumes (mean RVEDV index (RVEDVI) 123.8 ml/m² [116.6–131.0] v. 86.3 [82.3–90.4], p < 0.001), and lower RV ejection fraction (RVEF) (48.1% [46.5–49.7] v. 57.6% [55.8–59.3], p < 0.001). LV ejection fraction (LVEF) did not differ (p = 0.997). Total and passive LA function were lower in rTOF patients (51.1% [59.8–63.4], p < 0.001 v. 61.7% [59.8–63.4], p < 0.001 for total), (26.5% [24.7–28.3], p < 0.001 v. 39.5% [37.2–41.8], p < 0.001 for passive). Active LA function did not significantly differ (p = 0.074). LV diastolic indices showed that rTOF subjects had lower PER (311 ml/s [297–325] v. 386 [364–409], p < 0.001), PFR (297 ml/s [281–314] v. 427 [403–451], p < 0.001) and PFR/EDV (2.69 s⁻¹ [2.57–2.80] v. 3.18 [3.03–3.33] p < 0.001) compared to healthy controls.

Eighty-three rTOF subjects had an arrhythmia, with 63 experiencing an AA and 47 experiencing VA. Of these, 42 (51%) had arrhythmia present prior to date of CMR, while 41 (49%) developed arrhythmia following CMR and one developed arrhythmia on the date of CMR. The mean time of arrhythmia diagnosis prior to CMR was 5.3 years [95% 3.0–7.6]. Older age at the time of repair was associated with increased AA (OR 1.01 per month, p = 0.005) and TA (OR 1.01 per month, p = 0.010) (Table 3). When analyzing conventional CMR indices, increased RVEDVI was associated with increased AA (OR 1.08 per 10 mL/m², p = 0.032) and VA (OR 1.10 per 10 mL/m², p = 0.029). Increased RVESV index (RVESVI) was associated with increased AA (1.16 per 10 mL/m², p = 0.009), VA (1.18 per 10 mL/m², p = 0.010), and TA (OR 1.14 per 10 mL/m², p = 0.016). Increased RVEF was associated with decreased AA (OR 0.95, p = 0.005), VA (OR 0.95, p = 0.007), TA (OR 0.96, p = 0.009), while LVEF did not have significant association with arrhythmia (p = 0.673 for TA) (Table 3).

Analysis of diastolic indices showed all LA volume measurements were positively associated AA, VA, and TA (Table 3). The most significant associated measure in this group was indexed LA minimum volumes, with OR between 2.30–2.85 per 10 mL/m² for each subtype of arrhythmia. Total LA function was associated with AA (OR 0.97, p = 0.046). The remainder of LA function measurements were not significantly related to arrhythmia. LV diastolic indices showed that tPFR (OR 1.10 per 10 ms, p = 0.007 for TA) and PFR/EDV (OR 0.56, p = 0.006 for TA) were associated with all types arrhythmia, with worsening of diastolic dysfunction conferring increased association of arrhythmia (Table 3).

In patients who had contrast CMR performed, LGE (not including the RV insertion points) was also assessed. Forty-two patients underwent contrast CMR and 15 (36%) of these subjects had LGE. On univariate analysis, the presence of LGE was not associated with AA (OR 1.14, p = 0.874), VA (OR 1.90, p = 0.388), or TA (OR 1.19, p = 0.804).

There were 11 subjects with known mortality in the rTOF cohort. PFR/EDV was decreased and tPFR was increased in rTOF subjects with mortality as compared to those without mortality (PFR/EDV 1.94 s⁻¹ [1.57–2.30] v. 2.74 [2.62–2.86], p < 0.001), (tPFR 207 ms [143–271] v. 141 [134–148], p = 0.020) (Table 4). Given the small sample size of patients, multivariable analysis was not performed for mortality.

Linear regression of diastolic variables with factors that may affect septal position including degree of pulmonary insufficiency and QRS duration are shown in Additional file 1: Table S1 and Additional file 2: Table S2. The analysis with pulmonary insufficiency demonstrated poor correlation with all diastolic variables. For QRS correlates, there were a few statistically significant relationships with low r-squared values suggesting unclear clinical significance.

An initial multivariable model was made using conventical CMR and clinical variables that were significant in univariate analysis and that have previously been described as predictive of clinical outcomes [5, 9, 24]. These included QRS duration, RVEDVI, RVESVI, and RVEF as continuous variables for each type of arrhythmia (AA, VA, and TA). QRS duration was the only variable that significantly differentiated in all arrhythmia subsets (Table 5). A subsequent multivariable analysis was then performed using diastolic markers that were significant in univariate analysis as well as QRS duration. By multivariable logistic regression analysis, the best model for predicting AA and TA included QRS duration, LA indexed minimum volume, and TPFR. For VA prediction, the best model included QRS duration, LA indexed minimum volume, and PFR/EDV. In all subset analyses, this model for arrhythmia prediction demonstrated improved AUC statistics compared to the model using conventional parameters (Fig. 3). To account for age at the time of CMR in the models given risk for diastolic dysfunction with increasing age, an additional multivariate model that includes age at the time of CMR was assessed with comparison to classic predictors (Additional file 1: Table S3) with similar results to our diastolic predictor models without age.

In patients with arrhythmia following CMR, there were 28 AA and 24 VA. The mean follow-up period from initial CMR was 7.5 years [6.9–8.2]. Time to arrhythmia analysis was limited due to small sample size of patients with arrythmia following CMR and was felt to be underpowered. Cox proportional hazard ratios to assess risk of onset of all arrhythmias with conventional, LA, and LV diastolic CMR parameters are shown in Additional file 1: Table S4. For time to new AA, there was an associated risk with increased PER (HR 1.005, p = 0.032). LVEDVI (HR 1.014, p = 0.013), LVESV index (LVESVI) (HR 1.017, p = 0.022), PER (HR 1.005, p = 0.022), and tPER (HR 1.021,p = 0.001) were associated with new onset of VA. Overall new arrhythmia risk was associated with tPER (HR 1.012, p = 0.007).