Ventricular dyssynchrony late after the Fontan operation is associated with decreased survival

All patients with a Fontan circulation who had at least one available CMR after 7/1/05 through 9/23/21 were screened for eligibility. The study workflow is illustrated in Fig. 1. Patients with a pacemaker were excluded as it is a relative contraindication to CMR and rarely undergo CMR examinations at our center. Ventricular morphology was designated as either LV or RV, as appropriate, if the non-dominant ventricle was ≤ 20% of the combined end diastolic volume (EDV) and mixed type if the non-dominant ventricle was > 20% of the combined EDV [16]. When multiple studies were available per patient, the earliest CMR was analyzed. Patients with inadequate image quality for FT analysis were excluded. In the instance of poor image quality, a subsequent CMR was used when available. A comparison group was identified as individuals referred for CMR for suspected cardiomyopathy or congenital heart disease (CHD) but whose studies were subsequently interpreted as normal and at the time of study acquisition or interim follow-up did not have known systemic or genetic disease. The comparison group was age-matched ~ 1:12 using age in years in a random order due to the limited availability of normal studies. Demographic and clinical data were extracted from electronic medical records.

Fontan imaging was conducted according to standard practice at our center, as has been previously described [17]. Briefly, studies were performed on a 1.5T scanner (Achieva, Philips Healthcare, Best, the Netherlands) using surface coils appropriate for patient size. A ventricular short-axis balanced steady-state free precession (bSSFP) cine stack with breath-holding and ECG-gating was used for volumetric and FT analysis. Typical slice thickness was 8–10 mm, spatial resolution was 1.7–2 × 1.7–2 mm and temporal resolution was 30–40 ms with 30 reconstructed phases per cardiac cycle. Ventricular volumes and blood flow were measured using commercially available software (cvi42, Circle Cardiovascular Imaging Inc., Calgary, Alberta, Canada; and QMass, Medis Medical Imaging Systems, Leiden, the Netherlands). The following conventional measurements were recorded: indexed end-diastolic volume (EDV_i), indexed end-systolic volume (ESV_i), indexed stroke volume (SV_i), EF, indexed ventricular mass (Mass_i), and ascending aortic flow as a measure of cardiac output. When two ventricles contributed to the systemic circulation (i.e. mixed type ventricles), their mass and volumes were combined. For the comparison cohort, only the LV measurements were considered.

Feature tracking analysis was performed on the short-axis cine stack of images as previously described [18]. Briefly, the endocardial and epicardial borders were manually traced at end-diastole for all slices from the apex to base of the dominant ventricle in the case of a single LV or single RV. For mixed-type ventricles, borders were traced around the free wall of both ventricles (excluding the ventricular septum). A commercial feature tracking program was used for the analysis (cvi42, Circle Cardiovascular Imaging Inc., Calgary, Alberta, Canada). Adequacy of feature tracking was assessed by visual inspection and when not satisfactory, attempts were made to improve tracking by adjusting the baseline contour. If significant time elapsed (typically 5 min) without achieving reliable tracking, the case was excluded. The apical slice was defined as the most apical slice with blood pool through the entire cardiac cycle. The basal slice was defined as the most basal slice with a full rim of myocardium through the entire cardiac cycle. A minimum of 4 slices was required. Studies with artifact at the base of the heart that precluded accurate volumetric analysis were included if the most basal slice with circumferential myocardium was free of artifact and suitable for FT. For the comparison cohort, the same analysis was performed on the LV.

From this, the following six deformation measurements were obtained: circumferential strain (CS), circumferential strain rate (CSR), circumferential displacement (CD), radial strain (RS), radial strain rate (RSR), radial displacement (RD). These deformation measurements were used to calculate quantify dyssynchrony using 4 methods: (1) standard deviation of time-to-peak (SDTTP) for circumferential and radial deformation measurements for all segments, (2) maximum opposing wall delay (MOWD) as the maximum difference in the average time-to-peak for 3 opposing wall pairs, and (3) base-to-apex delay (BAD) as the time difference between peak deformation (average of 6 segments) of the most basal slice and the most apical slice (BAD1) as well as (4) the difference between the peak deformation for the 2 most basal and apical slices, respectively (BAD2). As such, 24 dyssynchrony indices were derived from the FT data (Fig. 2). Global deformation measures were recorded as global circumferential strain (GCS), global circumferential strain rate (GCSR), global radial strain (GRS), and global radial strain rate (GRSR). QRS duration was captured as a marker of electric dyssynchrony from the ECG closest to the CMR from all available ECGs, with no intervening surgical or catheter-based procedure having been performed. Heart rate was recoded from the CMR report and maximal heart rate (MHR) was calculated by using the Tanaka equation (MHR = 208 − 0.7 * [age at CMR]). Heart rate was standardized as percent MHR (%MHR) by dividing the heart rate by MHR [19].

The primary outcome was a time-to-event composite outcome of all-cause mortality or heart transplant listing (D/HTx). For time-to-event analysis, follow-up was measured from the date of CMR to the composite outcomes (earliest event in case of multiple) or last known documented follow-up in the medical record. If the first occurrence of transplant listing was prior to the CMR, it was excluded as an outcome.

Data are presented as a median with interquartile ranges (IQR) as appropriate for continuous variables and as a frequency (percentage) for categorical variables. Continuous variables were compared between groups using a Mann–Whitney U test while proportions were compared using a Fisher’s exact test. Correlation between continuous variables was quantified using the Pearson’s correlation coefficient (r). Using a set of 11 a priori chosen predictors, bivariate and multivariable Cox regression analyses were performed to assess relationships between predictors and time to composite outcome (time to D/HTx, whichever comes first, or censored at time of last follow-up without these events). Approximately 7% of patients’ CMR examinations (36 out of 503) were missing data. Schafer [20] found that a rate of missing data of 5% or less likely results in minimal bias, while Bennett [21] found that a missing rate of 10% or more could lead to bias. As we were between these two rates, missing data for the predictors in the Cox regression model were addressed with multiple imputation to prevent bias [22, 23]. Based on the number of the composite outcome events, an a priori threshold of at most 5 predictors to be included in multivariable Cox model (i.e. maximum of 1 predictor for 10 outcomes) was set. Predictors chosen for the final Cox regression model were based on a forward selection Akaike Information Criterion (AIC). Kaplan Meier survival curves with log rank tests were constructed to compare freedom from the composite outcome between groups. The optimal cut point for a continuous covariate in predicting the outcomes was obtained as the value that maximizes the score statistic (equivalent to the log rank statistic for the survival outcome) [24]. Because this is an exploratory analysis, we did not adjust for multiple testing. A two-sided p-value of ≤ 0.05 was considered statistically significant. Adjustment multiple testing for Tables 1 and 2 by the Benjamini–Hochberg method did not change significance threshold level of 0.05. Statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC) and SPSS version 27 (IBM Corp, Armonk, NY).

A total of 576 patients with a Fontan circulation and CMR were screened for eligibility and the final study cohort included 503 patients with a median age of 15.2 (IQR 10.3, 21.3) years and 42 comparison group patients with median age of 15.7 (IQR 11.0, 19.7) years. A total of 73 patients were excluded from after screening due to inadequate imaging. Common reasons for inadequate imaging were lack of full short-axis image stack (51 patients), inability to obtain reliable feature tracking due to motion artifact or blurring (10 patients), metallic artifact (9 patients), imaging performed at a lower than standard 30 phases/cardiac cycle temporal resolution (3 patients). Ventricular morphology was RV in 190 (38%), LV in 157 (31%), and mixed type in 156 (31%) of the patients with a Fontan circulation. The type of Fontan operations were lateral tunnel (69%), extracardiac conduit (20%), RA-PA (10%), and RA-RV (1%). Compared to the comparison group, Fontan patients had a lower BSA and were more frequently male (Table 1). Compared to the LV group, patients in the RV group were younger, had a lower BSA, and had a higher heart rate. The indications for CMR in the comparison group were family history of cardiomyopathy (N = 24, 57%), concern for arrhythmogenic right ventricular dysplasia (N = 3, 7%), family history of a bicuspid aortic valve (N = 3, 7%), family history of sudden cardiac death (N = 3, 7%), and concern for other abnormality on prior echocardiogram (N = 9, 21%). The patients undergoing evaluation for cardiomyopathy had a normal CMR as well as normal clinical and genetic evaluations (when available) at most recent follow-up.

Demographic, ECG, CMR, and dyssynchrony parameters for the study cohort are presented in Tables 1 and 2. Median number of days between ECG and CMR was 0 (IQR 1, 4 ) days. In contrast to the comparison cohort, the Fontan patients had larger systemic ventricular volumes and mass, and lower EF, indexed stroke volume, GCS, GRS, and longer QRS interval. Details of all 24 CMR-derived dyssynchrony indices for the Fontan cohort and the comparison group are presented in Additional file 1: Table S1. Figure 3 shows the distributions for EDV_i, EF, SDTTP-CS, and QRS interval across ventricular morphology types and comparison subjects. Within the Fontan cohort, the RV and mixed groups have higher volumes, lower EF, longer QRS duration, and higher SDTTP-CS. SDTTP-CS for the LV group was similar that of the comparison LV group. Ventricular size and function measurements demonstrated modest correlation with QRS duration and SDTTP-CS (Fig. 4; Additional file 1: Figure S1). Among all the evaluated dyssynchrony metrics, SDTTP-CS, SDTTP-RS, and SDTTP-RSR had the highest correlation with measures of ventricular function (GCS, GRS, and EF; correlation coefficients ranging from 0.3 to 0.6; Additional file 1: Table S2).

With a median follow-up period of 4.3 (IQR 1.3, 7.8) years, 57 (11.3%) of the patients met the composite primary outcome with 46 deaths, 8 heart transplantations, and 18 heart transplant listings. In the RV group, 28 (14.7%) patients had the outcome compared to 19 (12.2%) in the mixed group, and 10 (6.4%) in the LV group. On bivariate Cox regression analysis, D/HTx was associated with RV or mixed ventricular morphology, lower blood pressure, higher heart rate, longer QRS, ventricular dilation, lower EF, and higher dyssynchrony indices (Table 3). A multivariable logistic regression model found that RV morphology (HR ratio 2.3; 95% CI 1.1, 4.9; p-value 0.026), EDV_i per 10 ml (HR 1.1; 95% CI 1.1, 1.2; p-value < 0.001) and percent MHR (HR ratio 1.5; 95% CI 1.2, 2.0; p-value 0.003) were independently associated with D/HTx, however, dyssynchrony metrics were not on multivariable analyses (Table 4). Figure 5 depicts Kaplan Meier plots for each ventricular morphology type stratified by cut off values for EDV_i and SDTTP-CS. The estimated 3-year percent D/HTx was 7.9% (95% CI 3.7, 12.0) for single RVs, 3.5% (95% CI 0.4, 6.4) for mixed typed ventricles, and 1.4% (95% CI 0.0, 3.2) for single LVs (p-value = 0.005). Using the cut off point of 73 ms for SDTTP-GCS that maximizes the log rank statistic, those with a SDTTP-CS of  > 73 ms had a higher Kaplan Meier estimate of 3-year percent D/HTx (12.5%; 95% CI 3.4, 15.5) compared to those below the cut off (3.3%; 95% CI 1.3, 5.1). Patients with EDV_i > 146 ml/m² had a higher rate of the primary outcome than those with EDV_i ≤ 146 ml/m² (17% [95% CI 6.4, 23.11) vs. 2.8% [95% CI 0.95, 4.6]; p-value 0.002).