The relationship between myocardial and hepatic T1 relaxometry on cardiovascular magnetic resonance imaging and clinical outcomes late after tetralogy of Fallot repair

Abstract

Background

Cardiovascular magnetic resonance imaging (CMR) can estimate diffuse myocardial fibrosis using native T1 mapping and extracellular volume fraction (ECV) quantification.

Objectives

This study aimed to explore associations between, and prognostic value of, myocardial and hepatic parametric imaging measurements after tetralogy of Fallot repair (rTOF).

Methods

Adults with rTOF were prospectively enrolled along with healthy controls. Patients were categorized by residual hemodynamic load (pulmonary stenosis and/or pulmonary regurgitation). Available clinical data were recorded at study entry. Myocardial and hepatic native T1 were measured and ECV was calculated from pre/post-contrast T1 maps. The primary outcome was a composite of death, resuscitated sudden death, sustained ventricular tachycardia (VT), VT requiring invasive therapy, or heart failure requiring hospital admission. Multivariable models, adjusted for age and sex, examined the association between CMR measures and adverse cardiovascular events.

Results

In total, n=181 patients (56% male; median age 33 years [IQR 26–47]) and 20 healthy age and sex-matched controls were studied. Hepatic and myocardial ECV measures correlated modestly (r=0.31, p<0.001). Hepatic native T1 was higher in rTOF patients as compared with controls (572±102 versus 529±12 msec, p=0.006) with statistically significant differences between hemodynamic load type (p=0.044). Right ventricular (RV) native T1 was increased in rTOF as compared with controls (1078±43 versus 997±43 msec, p<0.001) with a trend towards significant differences when stratified by hemodynamic load (p=0.091). Pathologic late gadolinium enhancement (LGE) was seen in the RV in 46% (n=83) and in the left ventricle (LV) in 37% (n=66). At median follow-up 2.05 years (IQR 0.75–2.81), the composite primary outcome occurred in 9% (n=17). Independent predictors of adverse events included RV end-systolic volume (HR 1.43, [95% CI 1.09–1.99], p=0.009), RV native T1 (HR 1.15, [95% CI 1.07–1.25], p<0.001), and LV LGE (HR 3.18, [95% CI 1.07–10.23], p=0.038); hepatic parametric imaging was not associated with outcome.

Conclusion

In rTOF, myocardial and hepatic parametric imaging measures were modestly correlated and were increased as compared with controls. Diffuse myocardial fibrosis in the RV (native T1) and replacement fibrosis in the LV (LGE) were independently associated with major adverse cardiovascular outcomes, although hepatic relaxometry measures were not. These findings warrant further validation in larger populations with longer follow-up.

Graphical abstract

1. Introduction

Notwithstanding excellent surgical results in childhood, advancing age is associated with increased morbidity and mortality in adults with repaired tetralogy of Fallot (rTOF) [1]. Residual sequelae are prevalent despite successful repair, most commonly manifesting as chronic pulmonary regurgitation (PR), although isolated pulmonary stenosis (PS) can also be seen, albeit less frequently. Myocardial fibrosis has been implicated as an important driver of adverse clinical outcomes and has been associated with incomplete ventricular reverse remodeling, despite pulmonary valve intervention to restore valve competency [2], [3].

Advances in cardiovascular magnetic resonance imaging (CMR) technologies have enabled the detection of diffuse myocardial fibrosis using dedicated parametric imaging techniques. Measurement of myocardial native T1 by CMR has allowed for quantification of extracellular volume fraction (ECV), which in turn, has been correlated with histologic collagen content measured directly at the tissue level [4], [5], [6]. Compared with the clinical management algorithms for acquired heart disease, which now include ventricular T1 mapping assessment [7], less is known about parametric imaging in congenital heart disease (CHD), and this form of imaging has yet to be incorporated into the clinical care pathways of those with rTOF. Furthermore, concurrent hepatic parametric imaging at the time of myocardial mapping during CMR has not been directly studied in the rTOF population.

While the presence of a bidirectional influence between the heart and liver in adults with acquired heart disease and right heart failure has long been recognized, the prognostic implications of congestive hepatopathy and the potential for accurate estimation using CMR are only now emerging [8], [9]. Although rTOF is widely recognized as the archetypal right heart lesion in CHD characterized by chronically abnormal hemodynamics, the implications of pressure and/or volume sequelae on the liver in this population have not been well-elucidated to date. Therefore, in this study, we aimed to (1) explore a combined approach to imaging hepatic and myocardial parametric imaging by CMR and (2) determine if multi-organ parametric imaging, of the heart and liver, would enhance prediction of adverse clinical events in rTOF. We hypothesized that hepatic native T1 mapping and ECV measurements would be feasible at the time of myocardial parametric imaging assessment on CMR and that these measures would provide incremental value for risk stratification in rTOF.

2. Methods

2.1. Study population

Following an initial feasibility study at our site [10], adults with rTOF receiving care at our institution who were referred for clinical CMR, including gadolinium contrast injection and who met inclusion criteria were prospectively enrolled (2019–2024). Inclusion criteria were a diagnosis of TOF, age ≥18 years, and surgical repair in childhood. Informed consent was specifically obtained for the addition of parametric imaging sequences and bloodwork for hematocrit to enable ECV calculation. Exclusion criteria were inability to undergo CMR, contraindication to gadolinium injection, and known primary organic liver disease (including fatty liver disease, hepatitis, and cancer). For comparison, 20 age and sex-matched healthy volunteers without cardiovascular disease were recruited for comparison with 40 rTOF patients. All subjects underwent CMR study with parametric imaging sequences. The study was approved by our institutional research ethics board (study 14–7627) and informed consent was directly obtained from all participants.

2.2. Clinical data collection and adverse clinical outcomes

All pertinent clinical data were extracted from the electronic health record at study entry. Electrocardiography, echocardiography, cardiopulmonary exercise testing, and blood work within 6 months of the baseline CMR were preferentially reviewed. Pre-intervention hemodynamic load characteristics were documented based on the CMR and echocardiogram closest to the time of the procedure, although not exceeding 1 year. As previously defined in the literature, the primary clinical composite outcome was comprised of all-cause mortality, resuscitated sudden cardiac death, sustained ventricular tachycardia (VT) (>30 s) or VT requiring invasive therapy, and/or heart failure (HF) (hospital admission >24 h for intravenous diuretics) [11]. A secondary clinical outcome was defined as non-sustained VT (NSVT), VT stimulation during catheter electrophysiology study, and/or sustained atrial arrhythmia (>30 s of atrial flutter or fibrillation).

2.3. Classification of RV hemodynamic load

Patients were categorized by hemodynamic load as either RV volume overload from isolated PR, RV pressure overload from PS with or without PR (PS±PR), or no hemodynamic load (neither PR nor PS); the cohort was further stratified by pulmonary valve replacement (PVR) status [4]. As previously described [12], [13], [14], we designated significant RV volume overload (≥moderate PR) if the PR fraction on CMR exceeded 25% and significant PS if RV systolic pressure (RVSP) exceeded 40 mmHg on echocardiography (Doppler derived echocardiographic data were used to estimate RV systolic pressure based on peak velocities of the tricuspid regurgitation jet and RV outflow tract velocities, as available; in cases where the Doppler assessment was felt to be unreliable, septal curvature in systole on echocardiogram or CMR was used as an estimate of RV pressure overload as compared with systemic blood pressure, as previously established) [15], [16].

2.4. CMR technique

All CMR studies were performed on a 1.5T scanner (Magnetom Avanto Fit, Siemens Healthineers, Erlangen, Germany) as previous [10], [14]. Retrospectively gated cine steady-state free precession images were obtained for assessment of ventricular volumes, mass, and systolic function by a stack of short-axis slices ensuring complete ventricular coverage. Free-breathing, ECG-gated, cine phase-contrast flow measurements were obtained at the main pulmonary artery to quantify PR fraction. Parametric imaging measurements of the heart and liver were obtained from the same image using modified Look-Locker inversion recovery (MOLLI) T1 mapping, which was performed pre-contrast and 12–15 min post-contrast in short-axis (basal, mid, and apical locations) and sagittal oblique RV outflow tract orientations (Fig. 1, Fig. 2) and analyzed by a single experienced observer. Hematocrit was determined from a venous blood sample obtained on the same day as the CMR study. Myocardial replacement scar was assessed by late gadolinium enhancement (LGE) imaging of the RV and LV; specifically presence or absence of pathologic LGE was recorded in the superior, anterior and inferior RV walls and in the LV at basal, mid and apical levels with exclusion of regions expected to enhance (such as surgical material involved in patching the RV outflow or ventricular septal defect closure) [10], [17].

Fig. 1.

Assessment of diffuse fibrosis in the myocardium. Short-axis images at mid-ventricular level which demonstrate ventricular T1 mapping pre-contrast (left panel) and post-contrast injection with gadolinium (right panel). Contours were drawn within the ventricular endocardial and epicardial border of the right ventricle and left ventricles, respectively

Fig. 2.

Assessment of diffuse fibrosis in the liver. Short axis mid-ventricular inline T1 parametric map, pre-contrast. Regions of interest are drawn in 3 locations in the liver parenchyma (adjacent to the diaphragmatic surface, in the region of the porta hepatis and at the tip) and hepatic blood vessel, as shown

2.5. CMR analysis of the myocardium

Analyses of cardiac volumes, systolic function, ventricular mass, PR flow, and T1 maps were performed using commercially available software according to an established protocol [10], [14]. Briefly, after inline, non-rigid motion correction of individual MOLLI images, an inline T1 map was generated using standard three-parameter fitting for analysis [10], [18]. Each region of interest (ROI) was meticulously drawn in the RV myocardium, by a single experienced observer blinded to clinical parameters, as previously described by our group, ensuring a minimum size of 0.15 cm² and inclusion of at least 1 pixel, with care taken to avoid LGE or artifact with exclusion of adjacent fat and blood pool [10], [19] (Fig. 1). The ROIs of the RV were only drawn in sufficiently thickened anterior, superior, and inferior segments at basal, mid and apical-ventricular level. If an adequately sized ROI could not be drawn reliably within the myocardium meeting these criteria, the T1 map was excluded from further analysis. The LV endocardial and epicardial borders were manually contoured on basal, mid, and apical ventricular slices using an endocardial-epicardial offset approach, as previously described, and as shown [20] (Fig. 1). Myocardial ECV values were calculated at each site based on pre- and post-contrast T1 values and the corresponding hematocrit level [21]. All pre-contrast T1 measures and calculated ECV values were averaged and expressed as a single value for the RV, LV, and liver, respectively. The thresholds for elevated RVECV and LVECV were defined as 34% and 30%, respectively [10], [22], [23], [24], [25]. A threshold value for normal was not set for liver parametric imaging given the exploratory nature of hepatic mapping on CMR.

2.6. CMR analysis of the liver

Hepatic T1 relaxation times were measured by tracing three ROIs on pre and post-contrast images to capture differing locations in the liver, including the tip, the central aspect in the region of the porta hepatis, and the component adjacent to the diaphragm, as previously reported [26] (Fig. 2). All ROIs were carefully traced to exclude the gallbladder, liver surface, and vasculature. Blood pool T1 relaxation ROIs were drawn on pre- and post-contrast images within the RV cavity and hepatic vasculature, respectively, with care taken to exclude myocardial trabeculations and hepatic tissue, respectively.

2.7. Inter and intra-observer variability

To assess inter-observer agreement of myocardial ECV measures, 30 patients were selected to represent the spectrum of hemodynamic loading conditions (specifically, 10 patients were randomly selected from each of the three hemodynamic subgroups). Pre- and post-contrast T1 maps were reanalyzed by a second expert cardiac imager blinded to patient identifying data, to the first set of measurements and to clinical outcomes. Similarly, to assess for intra-observer agreement of ECV measurements, 30 patients were selected (random selection of 10 patients from each of the hemodynamic subgroups). Pre- and post-contrast T1 maps were reanalyzed by the first imager blinded to patient-identifying data and to the first set of measurements after 2 months. A similar approach was applied to evaluation of inter- and intra-observer agreement for hepatic parametric imaging measurements.

2.8. Statistical analysis

Continuous variables are presented with mean (standard deviation) and median (interquartile range [IQR]) according to normality of distribution as assessed by the Shapiro-Wilk test. Categorical data are reported using frequency (%). Comparisons between 2 groups were made using the Student t-test or Wilcoxon rank sum test. Comparisons between three or more groups were made with the one-way ANOVA along with post hoc Bonferroni correction. For discrete variables, the chi-square or Fisher exact tests were used. For the relationship between ECV and continuous variables, the Pearson or Spearman correlation coefficient was used, as appropriate. Univariate analysis was used to inform multivariable logistic regression modeling for identification of independent predictors of adverse clinical outcomes. Exploratory multivariable models were constructed for right heart and left heart parameters, respectively, using hazard ratios (HR) and 95% confidence intervals [CI] with adjustment for age and sex, incorporating variables of prognostic relevance based on the univariate analyses and previously published literature. The potential for the confounding effect of anti-fibrotic medications on results was explored using sensitivity analyses. Measurement agreement between observers and within the same observer was assessed using the intra-class correlation co-efficient (ICC) with 95% CI. A p-value <0.05 was considered statistically significant. Analyses were conducted using JMP Pro. 16 statistical software package (SAS Institute, North Carolina, USA). Study data can be made available to researchers upon reasonable request.

3. Results

3.1. Patient population

A total of 201 patients were enrolled into this study, with 181 patients ultimately entered into the final study analysis (20 patients were excluded due to insufficient CMR image quality [n=12] or early termination due to patient claustrophobia [n=8]) (Supplementary Figure 1). Of the 181 patients included, 56% were male with median age at enrollment of 33 years (IQR 26–47). The median follow-up duration was 2.05 years (IQR 0.75–2.81). Of the 181 patients with myocardial parametric imaging sequences, hepatic measurements were successfully completed in 172 patients from the same cardiac image (n=9 excluded due to insufficient image quality for analysis). The control group comprised 20 healthy age- and sex-matched adults with contemporary CMR studies (Supplementary Table 1).

Baseline demographic, clinical, and CMR data pertaining to the study population (n=181) are shown in detail (Table 1). Close to half of the study population identified as racially white (46%, n=83) based upon self-report by questionnaire. Medications with potential anti-fibrotic effects were used by 18% (n=33) of the rTOF population (including angiotensin converting enzyme inhibitors, angiotensin receptor blockers, mineralocorticoid receptor antagonists, and statins) (Table 1). Serum measures of brain natriuretic peptides were available for review in only a small subset of patients (n=6) with a median value that was slightly elevated at 114 pg/mL (IQR 31–194) (upper limit of normal 99 pg/mL). Serum measures of liver function were obtained in 18% of patients (n=33) as part of routine clinical practice, including bilirubin, aspartate transaminase, alanine transaminase (ALT), and gamma-glutamyl transferase, with an abnormality in at least one liver parameter observed in 42% of those studied.

Table 1.

Demographic, clinical and imaging characteristics of the rTOF cohort (n=181)

	rTOF (n = 181)
Age at CMR (y)	33 (26–47)
Male	102 (56%)
BSA (m2)	1.87±0.27
Race/Ethnicity
White	83 (46%)
Black	8 (4%)
Asian	19 (10%)
Southeast Asian	5 (3%)
First Nation	4 (2%)
Arab	4 (2%)
Jewish	1 (1%)
Latin American	1 (1%)
Other/Not Reported	56 (31%)
TOF Anatomy
Simple TOF	158 (87%)
Complex TOF	23 (13%)
Surgical Repair
Prior shunt	54 (30%)
Age at full repair (years)	2 (1–6)
CPB time (min)	111 (88–121)
Ventriculotomy	72 (40%)
Repair Strategy
Transannular patch	81 (45%)
Valve sparing	37 (20%)
RV-PA conduit	13 (7%)
Surgical notes not available	50 (28%)
Anti-fibrosis Medications	33 (18%)
ACEi/ARB	19 (11%)
Spironolactone	11 (6%)
Statin	13 (7%)
CMR
RVEDVi (mL/m2)	123 (105–144)
RVESVi (mL/m2)	66 (55–82)
RVEF (%)	46 (40–50)
RV mass indexed (g/m2)	27 (21–34)
RV mass:volume ratio (g/mL)	0.22 (0.19–0.25)
RA area indexed (cm2/m2)	10 (9–12)
LVEDVi (mL/m2)	77 (68–90)
LVESVi (mL/m2)	34 (28–41)
LVEF (%)	56 (52–60)
LV mass indexed (g/m2)	48 (42–56)
LV mass:volume ratio (g/mL)	0.61 (0.53–0.72)
LA area indexed (cm2/m2)	9 (8–11)
PR fraction (%)	19 (3–37)
RVECV increased ≥34%	70 (38)
LVECV increased ≥30%	27 (15)
RVECV (%)	32.6±5.2
RV T1 (msec)	1066±66
LVECV (%)	26±3
LV T1 (msec)	1021±45
RV LGE*	83 (46)
LV LGE	66 (37)
Hepatic T1 (msec)	586±69.34
Hepatic ECV (%)	29 (26–32)
Echocardiography
RVp (mmHg)	32 (25–43)
Peak RV outflow tract gradient**	18 (13–28)
Cardiopulmonary exercise testing (n=71)
Peak aerobic capacity (ml/kg/min)	23.7 (18.2–23.7)
Peak aerobic capacity (% predicted)	67% (59–78%)
Anaerobic threshold (% predicted)	42% (37–50%)
VE/VCO2 at anaerobic threshold	34 (31–38)

Data are numbers of cases with percent (%) and medians with IQR (interquartile range).

ACEi angiotensin-converting enzyme inhibitor, ARB angiotensin II receptor blocker, BSA body surface area, CMR cardiovascular magnetic resonance, CPB cardiopulmonary bypass time, ECV extracellular volume fraction, EDVi end-diastolic volume indexed, EF ejection fraction, ESVi end-systolic volume indexed, IQR interquartile range, LA left atrium, LGE late gadolinium enhancement, LV left ventricular, PA pulmonary artery, PR pulmonary regurgitation, RA right atrium, rTOF repaired tetralogy of Fallot, RV right ventricle, RVp right ventricular pressure, RVSP right ventricular systolic pressure, SD standard deviation, TOF tetralogy of Fallot, VCO₂ carbon dioxide production, VE ventilation

Data are numbers of cases with percent (%) and medians with IQR (interquartile range).

^*Expected sites of RV LGE, ventricle hinge points or sites of surgical patching were not considered pathologic enhancement

^**There were no patients with pulmonary hypertension in this cohort (peak RVOT-PA gradient reflects extent of pulmonary stenosis)

3.2. Myocardial and hepatic parametric imaging characteristics

Measures of diffuse and replacement myocardial fibrosis in the RV and LV are shown (Table 1). Specifically, RVECV was elevated in 38% (n=70), and LVECV was elevated in 15% (n=27); pathologic RV LGE was observed in 46% (n=83) and LV LGE in 37% (n=66) of the study cohort. Global and segmental ECV values for the RV and LV were measured and global values were entered into further analysis (Fig. 1). Although the RV had nine potential segments for analysis, there were no patients with contours achievable in all nine segments. The majority of patients had adequate image quality for contouring in 1, 2, or 3 segments (35%, 30%, and 20% of the population, respectively). In contrast, only one patient had reliable contouring in seven segments and no patients in eight or more segments.

Correlations with ECV measurements are demonstrated (Table 2). A moderately strong correlation was observed between LVECV and RVECV (r=0.56, p<0.001). Native T1 measurements and ECV measurements were moderately correlated for the RV (r=0.49, p<0.001), LV (r=0.44, p<0.001) and liver (r=0.52, p<0.001), respectively. A statistically significant correlation of moderate strength was observed between hepatic and myocardial ECV in the LV (r=0.31, p<0.001), but no association was apparent between hepatic and myocardial ECV in the RV (r=0.11, p=0.154). Hepatic relaxometry measurements did not differ according to sex or age.

Table 2.

Myocardial and hepatic ECV correlations

Demographic and imaging characteristics	RVECV (n=181)		LVECV (n=181)		Hepatic ECV (n=172)
	r-value	p-value	r-value	p-value	r-value	p-value
Age at CMR	0.12	0.122	0.09	0.221	0.03	0.735
Age at repair	0.05	0.515	0.15	0.049	0.02	0.837
CPB time	-0.02	0.858	-0.03	0.805	-0.03	0.741
CMR
RVEDVi (mL/m2)	0.12	0.103	0.15	0.050	0.10	0.189
RVESVi (mL/m2)	0.07	0.357	0.10	0.194	0.08	0.315
RVEF (%)	0.06	0.460	0.06	0.440	0.02	0.820
RV mass (g/m2)	0.06	0.415	0.17	0.026	0.16	0.042
RV mass/volume (g/mL)	-0.06	0.393	0.06	0.445	0.08	0.315
RV LGE	0.15	0.043	0.05	0.503	0.06	0.473
RAAi (cm2/m2)	0.10	0.161	0.18	0.017	0.06	0.459
LVEDVi (mL/m2)	0.09	0.217	0.13	0.074	0.16	0.041
LVESVi (mL/m2)	0.03	0.665	0.11	0.125	0.16	0.043
LVEF (%)	0.10	0.189	-0.01	0.883	-0.14	0.079
LV mass (g/m2)	-0.12	0.105	-0.18	0.016	0.001	0.957
LV mass/volume (g/mL)	-0.27	<0.001	-0.38	<0.001	-0.23	0.003
LV LGE	0.12	0.117	0.03	0.714	-0.06	0.461
LAAi (cm2/m2)	0.23	0.002	0.32	<0.001	0.09	0.270
PR (%)	-0.02	0.783	0.01	0.847	-0.05	0.486
RVECV	-	-	0.56	<0.001	0.11	0.154
LVECV	0.56	<0.001	-	-	0.31	<0.001
Hepatic ECV	0.11	0.154	0.31	<0.001	-	-
LV native T1	0.25	0.001	0.44	<0.001	0.13	0.088
RV native T1	0.49	<0.001	0.31	<0.001	0.10	0.186
Hepatic native T1	-0.02	0.798	0.20	0.009	0.52	<0.001
Echocardiography
RVSP (mmHg)	-0.28	<0.001	0.001	0.872	0.11	0.184
CPET (n=71)
VO2 (mL/min/kg)	-0.05	0.693	-0.12	0.307	-0.11	0.367
VO2% predicted	0.08	0.505	0.25	0.037	0.11	0.377
VO2% AT	0.06	0.624	0.16	0.195	0.02	0.892
VE/CO2 at AT (mL/min)	-0.01	0.905	0.25	0.035	0.17	0.172

Ai area indexed, AT anaerobic threshold, CMR cardiovascular magnetic resonance, CPB cardiopulmonary bypass time, CPET cardiopulmonary exercise testing, ECV extracellular volume fraction, EDVi end-diastolic volume indexed, EF ejection fraction, ESVi end-systolic volume indexed, LA left atrium, LGE late gadolinium enhancement, LV left ventricular, PA pulmonary artery, PR pulmonary regurgitation, RA right atrium, RV right ventricle, RVSP right ventricular systolic pressure, VCO₂ carbon dioxide production, VE ventilation, VO2 oxygen consumption

Numbers shown are correlations (r) with corresponding p values.

As compared with controls, patients with rTOF had increased RV native T1 measures (1078±43 versus 997±43 msec, p<0.001), increased RVECV (33.4±6.3 versus 26.2±2.1%, p<0.001), increased LV native T1 measures (1021±45 versus 994±41 msec, p<0.001), and increased hepatic native T1 measures (572±102 versus 529±12 msec, p=0.006) (Supplementary Table S1). There was no statistically significant difference in LVECV between study and controls (26.2±3.3 versus 24.9±4.0%, p=0.324). The number of controls with post-contrast T1 hepatic imaging sufficient for analysis were too few to allow for meaningful comparisons of hepatic ECV between controls and rTOF patients. Correlations between demographic/imaging characteristics and T1 measures in the myocardium and liver, respectively, are shown (Supplementary Table 2). Native T1 and ECV measures in the myocardium and liver did not differ according to type of surgical repair (transannular patch, valve sparing, or conduit) or by surgical approach (ventriculotomy versus atriotomy)(Supplementary Table 3).

3.3. Hemodynamic load

Among the 181 patients included in the study, 112 (62%) had a native RV outflow tract and the remaining, 69 (38%) had a PVR, of which 6 were transcatheter and 63 were surgically implanted valves. The non-PVR subgroup was further stratified according to the predominant hemodynamic load resulting in three categories: PR (n = 53), PS±PR (n = 28), and no significant hemodynamic load (n = 31)(Fig. 3). When ECV values for the myocardium and liver were compared across the three subgroups no statistically significant differences were observed.(Table 3) However, when native T1 values were compared in myocardium and liver with further stratification by subgroup, a significant difference was observed between the hepatic T1 measurements between those with PR (553 msec [524–611]) versus those with PS (590 msec [551–658]) versus those with no load (586 [538-622)], p=0.044. Invasive hemodynamic measurements from cardiac catheterization were only available in 6 patients; in this modest subset there were no significant correlations between hepatic measurements (ECV or native T1 time) and various right heart hemodynamic parameters (including right atrial [RA] pressure, RV pressure, mean pulmonary artery pressure, and pulmonary capillary wedge pressure).

Fig. 3.

T1 mapping in various loading conditions. Patient with mixed PS and PR shown in the top panels (pre-contrast map, panel A; post-contrast map panel B) and PR only shown in the bottom panels (pre-contrast map, panel C; post-contrast map panel D). In the patient with mixed PS/PR, the RV native T1 value was 1189 msec and the RV extracellular volume fraction (ECV) was 45%; the indexed RV end-diastolic and end-systolic volumes (RVEDVi and RVESVi) were 134 mL/m² and 69 mL/m², respectively, the RV EF was 48% and the indexd RV mass was 34 g/m2. In the patient with PR only, RV native T1 value was 985 msec and the RVECV was 32%; the RVEDVi and RVESVi were 162 mL/m² and 101 mL/m², respectively, the RVEF was 36% and the indexed RV mass was 34 g/m². PS pulmonary stenosis, PR pulmonary regurgitation, RV right ventricular, EF ejection fraction

Table 3.

Myocardial and hepatic ECV values stratified by hemodynamic load (n=112)^*

	All*	PR only (n=53)	PS±PR (n=28)	No hemodynamic load (n=31)	p-value
RV T1 (msec)	1057 (1025–1115)	1053 (1025–1112)	1074 (1008–1123)	1064 (1028–1125)	0.091
RVECV (%)	30.3 (28.9–35.2)	32.8 (29.9–37.0)	29.9 (27.3–34.7)	33.3 (28.4–35.2)	0.896
LV T1 (msec)	1019 (994–1047)	1021 (998–1043)	1030 (1001–1057)	1012 (992–1042)	0.163
LVECV(%)	24.9 (23.5–27.0)	26.4 (23.7–29.6)	25.8 (23.9–28.1)	24.9 (23.4–27.0)	0.207
Hepatic T1 (msec)	593 (538–623)	553 (524–611)	590 (551–658)	586 (530-622)	0.044
Hepatic ECV(%)	28.9 (26.1–32.5)	27.9 (26.2–30.5)	28.7 (27.1–33.0)	28.9 (26.1–32.5)	0.360

ECV extracellular volume fraction, LV left ventricle, PR pulmonary regurgitation, PS pulmonary stenosis, RV right ventricle

Numbers shown are median with IQR (interquartile range) and corresponding p values.

^*patients with PVR excluded from analysis

Our population was also stratified according to presence or absence of hemodynamically relevant TR (defined as moderate or severe insufficiency on echocardiography). While patients with important TR (n=14) had longer native hepatic T1 values and higher hepatic ECV as compared with those without significant TR (native T1 623 msec [542–686] versus 570 [533-630], p=0.046; ECV 33.0 [30.7–35.2] versus 28.1 [26.2–32.5], p<0.001), RV fibrosis measures did not differ between the groups (native T1 1089 msec [1031–1142] versus 1054 [1017–1114], p=0.243 and ECV 32.8 [30.1–37.0] versus 32.2 [28.7–35.8], p=0.371).

3.4. Adverse cardiovascular events

The primary composite clinical outcome was observed in 9.4% (n=17) of rTOF patients. Non-mutually exclusive events which comprised the primary outcome included all-cause mortality or resuscitated sudden death (n=3), sustained VT or VT requiring invasive therapy (n=10), and HF requiring hospital admission (n=5). Of the deaths, one was attributable to complications related to endocarditis, one was a result of rupture of a pulmonary artery patch during surgical PVR, and one was related to metastatic gastroesophageal cancer.

The secondary composite clinical outcome was observed in 13.3% (n=24) patients. Non-mutually exclusive events which comprised the secondary outcome included NSVT (n=10), inducible VT at the time of electrophysiology study (n=4), and sustained atrial flutter or fibrillation (n=15) (Table 4). For the purposes of multivariable model construction, a combined outcome was defined as the occurrence of mutually exclusive primary or secondary outcome events, which occurred in 17% of the study population (n=31).

Table 4.

Adverse cardiovascular events shown as mutually exclusive events and by event type

Primary or secondary (mutually exclusive events)	31
Primary outcomes (mutually exclusive events)	17
Death	3
Ventricular tachycardia (sustained or requiring invasive therapy)	10
Heart failure (requiring admission for diuresis)	5
Secondary outcomes (mutually exclusive events)	24
Sustained atrial arrhythmia (flutter or fibrillation)	15
Non-sustained ventricular tachycardia	10
Inducible ventricular tachycardia at the time of EP study	4

Data shown are number of clinical events.

3.5. Univariate and multivariable analyses

Univariate analyses were performed independently for the primary clinical outcome and the combined (primary and secondary) outcome, incorporating demographic, imaging, and exercise test parameters (Table 5). Separate multivariable models were created for the right and the left heart to enable exploration of the predictive value of CMR measures for occurrence of adverse clinical outcomes with adjustment for age and sex. Sensitivity analysis with adjustment for patient treatment with anti-fibrotic medications did not affect model results.

Table 5.

Univariate and multivariable analyses

5a. Univariate analysis
	Univariate
	Primary outcome (n=17)			Combined (primary and secondary) outcome (n=31)
	HR	95% CI	p-value	HR	95% CI	p-value
Demographics
Age at CMR (y)	1.08	1.04–1.12	<0.001	1.07	1.04–1.10	<0.001
Female	0.24	0.05–0.78	0.015	0.38	0.15–0.86	0.021
Male	4.13	1.29–18.4	0.015	2.65	1.16–6.67	0.021
Age at repair (y)	1.03	0.97–1.08	0.250	1.03	0.98–1.07	0.219
CPB time (min)	1.00	0.98–1.02	0.973	0.99	0.98–1.01	0.435
Post repair PVR	0.87	0.29–2.42	0.800	0.61	0.25–1.37	0.245
Right heart variables
RVEDVi (mL/m2) per 10 mL ↑	1.31	1.15–1.52	<0.001	1.30	1.16–1.49	<0.001
RVESVi (mL/m2) per 10 mL ↑	1.33	1.15–1.57	<0.001	1.36	1.19–1.60	<0.001
RVEF (%) per 10% ↑	1.97	1.14–3.46	0.016	2.21	1.38–3.83	0.003
RV mass indexed (g/m2)	1.09	1.03–1.16	0.002	1.08	1.04–1.14	<0.001
RV systolic pressure (mmHg) ↑	1.03	0.99–1.06	0.101	1.02	0.99–1.04	0.317
RAAi (cm2/m2)	1.20	1.06–1.37	0.005	1.29	1.15–1.48	<0.001
RV LGE (%)	2.82	0.98–9.29	0.055	1.67	0.76–3.75	0.202
RV native T1 per 10 msec ↑	1.15	1.07–1.25	<0.001	1.12	1.06–1.19	<0.001
RVECV (%)	1.07	0.97–1.18	0.151	1.03	0.96–1.11	0.366
PR (%)	1.00	0.97–1.02	0.860	1.01	0.99–1.03	0.426
Left heart variables
LVEDVi(ml/m2) per 10 mL↑	1.39	1.10–1.76	<0.01	1.29	1.07–1.57	0.009
LVESVi(ml/m2) per 10 mL ↑	1.47	1.04–2.06	0.03	1.60	1.20–2.19	0.001
LVEF (%) per 10% ↑	1.33	0.64–2.65	0.43	2.14	1.23–3.77	0.007
LV mass indexed (g/m2)	1.07	1.03–1.13	<0.01	1.05	1.02–1.09	0.003
LAAi (cm2/m2)	1.18	1.00–1.40	0.05	1.25	1.09–1.46	0.001
LV LGE (%)	3.21	1.13–9.88	0.03	2.31	1.04–5.17	0.039
LV native T1 per 10 msec ↑	1.11	0.98–1.26	0.12	1.08	0.98–1.20	0.115
LVECV (%)	1.04	0.90–1.20	0.61	1.01	0.91–1.14	0.820
Liver variables
Hepatic native T1 per 10 msec ↑	1.03	0.96–1.11	0.360	1.00	0.95–1.06	0.896
Hepatic ECV (%)	1.10	0.98–1.24	0.123	1.04	0.95–1.13	0.437
Ai area indexed, CI confidence interval, CPB cardiopulmonary bypass time, CMR cardiovascular magnetic resonance, ECV extracellular volume fraction, EDVi end-diastolic volume indexed, EF ejection fraction, ESVi end-systolic volume indexed, HR Hazard ratio, LA left atrium, LGE late gadolinium enhancement, LV left ventricular, PR pulmonary regurgitation, PVR pulmonary valve replacement, RA right atrium, RV right ventricle, RVSP right ventricular systolic pressure, PR pulmonary regurgitation

5b. Multivariable analysis
	Multivariable*
	Primary outcome (n=17)			Combined (primary and secondary) outcome (n=31)
	Right heart model			Right heart model
	HR	95% CI	p-value	HR	95% CI		p-value
Right heart variables
RVESVi (mL/m2) per 10 ml ↑	1.43	1.09–1.99	0.009	1.45	1.11–1.96		0.004
RVEF(%) per 10% ↑	0.61	0.19–1.75	0.363	0.64	0.25–1.54		0.323
RAAi (cm2/m2)	1.14	0.97–1.33	0.116	1.23	1.08–1.43		0.002
RV native T1 per 10 msec ↑	1.15	1.07–1.25	<0.001	1.12	1.07–1.21		<0.001
	Left heart model			Left heart model
	HR	95%CI	p-value	HR		95%CI	p-value
Left heart variables
LVESVi (mL/m2) per 10 ↑	1.58	0.87–2.98	0.140	1.15		0.71–1.85	0.560
LVEF (%) per 10% ↑	0.52	0.14–1.53	0.251	1.43		0.62–3.23	0.391
LAAi (cm2/m2)	1.13	0.92–1.39	0.244	1.21		1.02–1.46	0.029
LV LGE (%)	3.18	1.07–10.23	0.038	2.11		0.90–4.98	0.086
Ai area indexed, CI confidence interval, CMR cardiovascular magnetic resonance, ECV extracellular volume fraction, EDVi end-diastolic volume indexed, EF ejection fraction, ESVi end-systolic volume indexed, HR Hazard ratio, LA left atrium, LGE late gadolinium enhancement, LV left ventricular, PR pulmonary regurgitation, PVR pulmonary valve replacement, RA right atrium, RV right ventricle, RVSP right ventricular systolic pressure

Date shown are hazard ratios (HR) with 95% confidence interval (CI) and associated p values.

The right heart multivariable models included 4 variables: indexed RV end-systolic volume index (RVESVi), RV ejection fraction (RVEF), RV native T1, and indexed RA area. Although variables with the highest hazard ratios (HRs) were preferentially selected from the univariable analysis for inclusion in the multivariable models, in the case of RV volumes, RVESVi was selected instead of (not in addition to) RV end-diastolic volume, given slightly higher HR of the former, coupled with the knowledge that RVESVi is a more robust parameter for outcome prediction as an expression of both ventricular size and systolic function. While RVESVi and native RV T1 were independently predictive of both the primary and combined (primary and secondary) outcome (RVESVi primary outcome HR 1.43 [1.09–1.99] p=0.009 and combined outcome HR 1.45 [1.11–1.96] p=0.004; RV native T1 primary outcome HR 1.15 [1.07–1.25] p<0.001 and combined outcome HR 1.12 [1.07–1.21] p<0.001), indexed RA area was independently predictive of the combined outcome only (HR 1.23 [1.08–1.43] p=0.002). Hepatic native T1 time and ECV were not found to be statistically significant on univariate analysis and were therefore not entered into the multivariable models.

The left heart modeling included LVESVi, LVEF, indexed left atrial (LA) area, and LV LGE in the primary outcome and in the composite combined (primary and secondary) outcome models, respectively. Similar to the RV, LVESVi was selected ahead of LV end-diastolic volume given the higher HR of the former on univariate analysis along with the recognition that ESVi reflects both volume and systolic function. For symmetry of variable selection among models and guided by prior published literature establishing its strength as a prognostic marker in the rTOF cohort, LVEF was entered into primary and combined multivariable outcome models (although significance was only observed on univariate analysis for combined outcome prediction). The only independent predictor for primary outcome in the left heart model was LV LGE (HR 3.18 [1.07–10.23] p=0.038) and the only predictor for the combined outcome was indexed LA area (HR 1.21 [1.02–1.46] p=0.029).

In the univariate analysis exploring associations with compromised cardiopulmonary exercise testing, CMR measurements were not associated with decreased aerobic capacity or impaired ventilatory efficiency. While PS$\pm$PR was associated with increased risk of decreased aerobic capacity (functional capacity <70% predicted) (HR 3.43, 95% CI 1.11–12.02, p=0.031), PR alone was not associated with compromised exercise tolerance.

3.6. Inter-and intra-observer agreement

The intra-class correlation co-efficients were excellent within and between observers for ventricular measurements (volumes, function, and mass) as well as atrial measurements and PR fraction, as previously published [27] and as shown (Supplementary Table 4). For myocardial measurements, inter-observer agreement was excellent for both native RV T1 and RVECV (ICC 0.92 [0.84–0.96] and ICC 0.92 [0.85–0.97], respectively) and similarly strong for native LV T1 and LVECV (ICC 0.92 [0.84–0.96] and ICC 0.94 [0.87–0.97], respectively). Intra-observer agreement was also excellent for native RV T1 and RVECV (ICC 0.94 [0.88–0.97] and ICC 0.96 [0.91–0.98], respectively) and as well for native LV T1 and LVECV (ICC 0.97 [0.93–0.98] and ICC 0.99 [0.98–0.99], respectively).

For hepatic measurements, inter-observer agreement was excellent for native hepatic T1 and good for hepatic ECV measurements (ICC 0.94 [0.89–0.97] and 0.87 [0.74–0.94], respectively). Similarly, intra-observer agreement was excellent for native hepatic T1 and good for hepatic ECV measurements (ICC 0.95 [0.90–0.97] and 0.82 [0.64–0.91], respectively).

4. Discussion

This single-centre prospective study was designed to explore the prognostic value of CMR myocardial and hepatic parametric imaging in adults late after TOF repair. Salient findings of our study include the following: (1) concurrent acquisition and analysis of hepatic and myocardial mapping at the time of CMR are feasible in the vast majority of patients; (2) native T1 measurements in the liver were higher in rTOF as compared with healthy matched controls; and (3) while diffuse fibrosis in the RV (native myocardial T1) and replacement fibrosis in the LV (LGE) were associated with adverse cardiovascular events in rTOF patients, hepatic measurements were not.

4.1. Confirmatory and novel findings of this study

Several CMR variables were found to be independent predictors of adverse cardiovascular events in this cohort. The strongest independent predictor of outcome in the right heart was RVESVi, a measure which is widely recognized as one of the highest fidelity CMR measures for outcome prediction in rTOF, as this is a combined expression of both ventricular size and systolic function [28], [29]. In our study population, there was an approximately 40% increase in risk of an adverse outcome with each 10 ml increase in RVESVi (HR 1.43 [1.09–1.99], p=0.009 for the primary outcome and HR 1.38 [1.08–1.84], p<0.01 for the combined [primary and secondary] outcome). Indexed RA area was associated with a 20% increase in risk of the combined outcome, likely reflecting the incorporation of atrial arrhythmias into this composite endpoint (HR 1.20 [1.03–1.17], p<0.01). Based on recent data from our group and others, RA size as measured on CMR appears to be a variable of emerging prognostic relevance [30], [31].

The findings from this cohort underscore the importance of myocardial fibrosis estimation for risk prediction in rTOF. Specifically, we observed that replacement fibrosis in the LV and diffuse fibrosis in the RV were independent predictors of adverse cardiovascular events. Patients in our population with LV LGE had a three-fold increased risk of an adverse event on multivariable analysis, in keeping with previous publications [32], [33]. In the right heart, for each incremental increase in RV native T1 of 10 msec, there was a 15% increase in risk of the occurrence of the primary outcome and a 10% risk of occurrence of the combined outcome (HR 1.15 [1.07–1.25], p<0.01 for the former and HR 1.10 [1.03–1.17], p<0.01 for the latter). Worthy of note is the magnitude of risk associated with increased native RV T1, our findings are virtually identical to a recently published meta-analysis which determined that the cumulative risk based on a review of 16 studies describing ECV in rTOF (random effects model HR 1.15 [1.04–1.27] with test for heterogeneity I²=0% and p=0.46) [34]. To the best of our knowledge, ours is the first and the largest prospective study to date to identify native T1 in the RV as a promising marker of cardiovascular risk in rTOF.

While previous publications have preferentially focused on RVECV in rTOF, based on the knowledge that there is greater comparability between ECV estimates (across differing scanner manufacturers, field strengths and patient populations) [10], [12], [35], [36], the belief that ECV is a more robust measure has recently come into question given the fact that native T1 measurements are strongly correlated with histologic collagen content (possibly even more so than ECV) [37], do not require a hematocrit value for calculation, are less affected by motion and the need for co-registration to avoid errors, and do not require contrast injection (avoiding factors such as dosing, selection of optimal timing of imaging post injection, and potential for false ECV elevation in the event of biliary excretion of gadolinium) [38]. Indeed, the salient finding of RV native T1 as a strong prognostic marker in our population is in alignment with wider current clinical practice in acquired forms of heart disease.

4.1.1. Implications for clinical practice: myocardial fibrosis

Over recent decades, parametric techniques have matured considerably, such that tissue characterization by parametric imaging is now considered an essential component of the work-up for acquired and inherited cardiomyopathies given critical implications for diagnosis and prognosis [39], [40]. In fact, recent consensus guidelines have designated incorporation of native T1 myocardial measurements (rather than myocardial ECV) as a class 1 recommendation in the evaluation of some forms of cardiomyopathy in the adult [41]. Although not presently included in the guidelines for CMR imaging of patients with CHD, parametric mapping, particularly native T1 measurements, may be a reasonable sequence to consider adding into routine clinical practice given the relatively fast acquisition time without the need for additional contrast injection or phlebotomy - particularly in those subsets of patients known to be at risk of high fibrosis (some histologic studies have suggested that interstitial fibrosis is present in 65% of patients with rTOF after the initial repair in childhood and fibrosis extent may be one of the key factors determining extent of reverse remodeling following PVR) [2], [42]. It is worth noting that there is excellent agreement between our fibrosis measures and those in the reported literature; specifically the native T1 and ECV measurements in our study are nearly identical to the 103 rTOF patients described by Cochet and colleagues (i.e., RV native T1 1023$\pm$72 msec, LV native T1 1002$\pm$48 msec, RVECV 33.8$\pm$4.5%, and LVECV 27.4$\pm$3.3%) [18].

4.1.2. Implications for clinical practice: hepatic T1 relaxometry

Hepatic congestion is a known consequence of right heart disease. Hepatic parametric mapping measurements have been associated with extent of liver cirrhosis as characterized by the Child-Pugh classification (specifically, hepatic native T1 relaxation time and ECV measures were 519±48 msec and 25.9± 4.5% in control patients, compared with Child-Pugh class A 621±79 msec and 33.0±6.0%; class B 667±73 msec and 39.6±4.9%; class C 828±91 msec and 52.8±1.2%) [43]. Furthermore, based on studies in patients with pulmonary hypertension, elevated hepatic T1 values have been correlated with increased mean RA pressures (≥10 mmHg) and elevated brain natriuretic peptide (BNP) levels [44].

Parametric liver imaging is increasingly being recognized as a powerful biomarker for outcome prediction in acquired heart disease. Specifically, hepatic native T1 values have been observed to predict outcomes after myocardial infarction (MI) [9]. Bergamaschi and colleagues reported that hepatic T1 values demonstrated good diagnostic performance for detection of complications following MI in 142 adults and higher hepatic native T1 (>605 msec) was specifically associated with greater extent of RV disease [45]. In a large CMR cohort of 1075 patients, the risk of an adverse outcome in all-comers (defined as death, non-fatal MI, and HF hospitalization) was associated with hepatic T1 threshold >610 msec [8]. In comparison with these ischemic heart disease studies, much less is known about the extent of, and implications of, hepatic congestion in rTOF, with no dedicated rTOF studies published to date (small numbers of rTOF patients have been described in comparison with Fontan patients in studies primarily focused on Fontan associated liver disease, specifically 20 rTOF patients in one study and 12 patients with either rTOF or Ebstein anomaly in another) [46], [47]. Our study of parametric imaging exclusively in a rTOF population comprised of 172 adults is a substantial contribution to the literature and our findings expand the current knowledge base in several ways.

We demonstrated that acquisition of liver maps was feasible in the majority of patients at the time of acquisition of standard cardiac T1 maps obtained in sagittal oblique views (n=172, 95%). We expect that abnormalities in hepatic T1 relaxometry in our study reflect underlying liver congestion in the context of right heart disease as reported by others, while recognizing that chronic hepatic venous congestion can stimulate hepatic cells to deposit collagen leading to fibrosis [45], [48], [49], and we were unsurprised to observe that hepatic ECV was correlated, albeit modestly, with myocardial ECV (r=0.31, p<0.001) (Table 2). Although hepatic parametric imaging measurements were not associated with adverse outcomes per se, native T1 measures were found to be significantly higher in rTOF as compared with controls (572±102 versus 529±12 msec, p=0.006, Supplementary Table S1) with variations noted according to hemodynamic load type (lower in PR as compared with PS, Table 3). In keeping with a recent report which associated hepatic mapping values with TR severity, we observed that patients with moderate or severe TR had longer native T1 values and higher hepatic ECV, albeit from a modest population (n=14). Although the prognostic value of hepatic ECV in patients with TR has been observed in older adults with acquired heart disease, the performance of this biomarker in our cohort is limited by the small sample size in this subset and we suggest that further study is warranted in larger rTOF populations [49].

Given the relatively young age of patients in our cohort (median age 33 years) compared with the published literature describing hepatic abnormalities of hepatic T1 relaxometry in older adults with pulmonary hypertension (mean age 55 years) [50], with ischemic heart disease (mean age 58–59 years) [8], [45] and with tricuspid valve disease (mean age 66 and median age 68 years) [48], [49] we speculate that hepatic disease may become more apparent as the rTOF cohort ages. We hypothesize that hepatic parametric imaging may be of greater value if measurements are completed serially over time, specifically given the well-recognized association between advancing age and increased burden of morbidity and risk of mortality in rTOF. We note that the reproducibility of measurements within and between observers is better for hepatic native T1 as compared with hepatic ECV (ICC 0.94 and 0.95 compared with 0.82 and 0.87, respectively) which may reflect the technical limitations inherent in ECV measurement, as described above, or the difficulties with image analysis post contrast, particularly with regard to identification of the boundaries of the liver vasculature (Fig. 2). We therefore suggest that hepatic native T1 may be preferable to hepatic ECV for reproducible assessment of liver disease; such observations are in keeping with variability reported by others (for instance, in a relatively large study which identified Fontan associated liver disease in 61 patients, de Lange and colleagues reported smaller co-efficients of variation within and between observers for native hepatic T1 as compared with hepatic ECV [intra/inter-observer values were 2.3 and 2.8 for T1 respectively, compared with 10.7 and 10.9 for ECV, respectively] [26], [51].

5. Conclusions

In rTOF, combined myocardial and hepatic mapping revealed abnormalities in parametric imaging measures, in both the heart and liver. Diffuse myocardial fibrosis in the RV (native T1) and replacement fibrosis in the LV (LGE) were independently associated with adverse cardiovascular outcomes, alongside other established markers, such as indexed RVESV and RA area. Hepatic T1 measures varied according to the type of hemodynamic load on the heart but were not associated with clinical outcomes. Native RV T1 measurements may serve as a promising non-invasive tool for myocardial fibrosis screening in this population, with the potential to guide treatment strategies and timing of intervention, however this speculation is worthy of focused future study.

6. Limitations

There are several limitations of our study that are worthy of mention, in terms of overall study design as well as CMR technique. This was a single-centre study which may affect external validity (generalizability). Parametric imaging occurred only at a single timepoint which precluded evaluation of serial measures over time. The follow-up duration was relatively short which may have implications for the number of adverse events noted and extent of hepatic disease observed. Future studies would ideally involve multiple centers, serial fibrosis assessments and longer follow-up. While the study population reflects only those referred for CMR which may impose an element of selection bias, there is no alternative non-invasive approach for myocardial fibrosis assessment in current clinical practice. Given the observational nature of the study design (with management directed by the clinical cardiologist providing care for the patient), the biochemistry profile (namely BNP and hepatic function measures) had many missing values which were unavoidable and imposed limitations on statistical analyses; similarly invasive measures obtained during cardiac catheterization were small in number. The value of established and novel serum biomarkers (including circulatory measures of fibrosis) in rTOF has been identified as an avenue of future research for our group. The potential impact of anti-fibrotic medications on diffuse fibrosis as assessed by CMR was incompletely evaluated given the small number of patients on medical therapy in this study (18%). Indeed, an integrated approach to the study of patients with rTOF, including advanced imaging, serologic markers, and invasive hemodynamic assessment may allow for a more comprehensive and refined understanding of the long-term consequences of right heart dysfunction in this expanding yet particularly vulnerable population.

Although we are reassured that our RV T1 relaxometry and ECV results are aligned with previous publications [34], [52] and are in keeping with our previous pilot study [10], we recommend caution if applying our approach to other populations (such as younger rTOF patients with less RV hypertrophy as compared with the adult cohort or other forms of right-sided congenital heart disease or acquired heart disease with thinner RV walls) [53]. We achieved excellent reliability (test-retest measurements both within and between observers, as above) and suggest that this is a reflection of several features of our research protocol which were implemented in order to mitigate potential error which could arise from contamination from adjacent blood pool or epicardial fat. Not only did we apply a strict minimum area for each region of interest while contouring, we also had a low threshold to discard a segment of the RV or exclude a patient if we determined that results may be compromised by suboptimal contouring technique.

While our study demonstrates the overall feasibility of a combined myocardial and hepatic approach to parametric imaging in a prospective cohort of adults with rTOF (which has not been previously described), our results should be interpreted with a degree of caution for a variety of reasons. First, our study is exploratory in nature and it should be recognized that, to date, there is no published validation of this approach in the rTOF population. Future research should compare results from a combined imaging approach to that of a dedicated liver imaging protocol. Second, although we excluded patients with known fatty liver disease or cirrhosis which could contribute to skewed mapping results, we did not have abdominal imaging in all patients entered into the study and we may have therefore inadvertently included patients with undiagnosed fatty or fibrotic liver disease. Finally, the impact of technical limitations (i.e., the association between increased distance from the isocentre of imaging and magnetic field inhomogeneities, differences in magnetization transfer effects between myocardial and hepatic tissue, partial-volume effects from thicker slices, and the absence of correction for fat or iron) remains uncertain and is worthy of further study. Nevertheless, we believe that our meticulous methodologic approach (as described above) has mitigated these potential drawbacks to as great an extent as possible. We suggest that these observations can be incorporated into the growing body of publications examining the potential impact of T1 mapping in the liver for prediction of outcomes in patients with acquired forms of right heart disease, including patients with high-risk tricuspid valve disease, ischemic heart disease, and pulmonary hypertension [45], [48], [49], [50].

Funding

Canadian Institutes of Health Research Operating Grant (MOP 119353) to RMW.

Author contributions

Ayako Ishikita: Methodology, Formal analysis, Data curation. Lais Marques: Writing – review & editing, Writing – original draft, Methodology, Investigation, Data curation. Subin Thomas: Validation, Investigation, Data curation. Isaac Begun: Investigation, Data curation. Jacques DuPlessis: Investigation, Formal analysis, Data curation. Romina DSouza: Writing – review & editing, Data curation. Soshun Asakawa: Investigation, Data curation. Gauri R. Karur: Writing – review & editing, Investigation, Data curation. Kate Hanneman: Writing – review & editing, Data curation. Darren A. Yuen: Writing – review & editing. Rajiv R. Chaturvedi: Writing – review & editing. Mark K. Friedberg: Writing – review & editing. Slava Epelman: Writing – review & editing. S. Lucy Roche: Writing – review & editing, Conceptualization. Rachel M. Wald: Writing – review & editing, Project administration, Methodology, Investigation, Data curation, Conceptualization.

Declaration of competing interests

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary material

Supplementary material