Determinants of Left Ventricular Dysfunction and Remodeling in Patients With Corrected Tetralogy of Fallot

Background The aim of this study was to identify in asymptomatic patients with repaired tetralogy of Fallot the prevalence and determinants of impaired left‐sided cardiac function and adverse ventricular remodeling and the relation of left ventricular (LV) dysfunction and remodeling with cardiopulmonary exercise capacity. Methods and Results In a cross‐sectional study, 103 patients with tetralogy of Fallot (median age, 16.3 years) in New York Heart Association class 1, with surgical repair at a median age of 1.1 years, and 63 age‐matched controls were studied. LV, right ventricular function and geometry, LV myocardial extracellular volume (n=57), and left atrial function were quantified with cardiac magnetic resonance. Peak oxygen consumption was measured by a standardized cardiopulmonary exercise test (n=70). Patients with tetralogy of Fallot had lower LV ejection fraction (P=0.001; 49% below age‐adjusted fifth percentile for controls), lower LV mass index (P=0.003), lower LV mass/volume ratio (P<0.01), and impaired left atrial function. Right ventricular mass/volume ratio was the best predictor for LV systolic dysfunction and for a lower LV mass/volume ratio. Compared with controls, LV extracellular volume was higher (P<0.001), particularly in female patients, and associated with subnormal peak oxygen consumption (P=0.037). A peak oxygen consumption below the third percentile reference level was more likely with decreasing LV ejection fraction (P=0.008), and lower LV mass index (P=0.024), but independent of right ventricular ejection fraction. Conclusions In New York Heart Association class 1 patients with tetralogy of Fallot, frequent impaired systolic and diastolic LV function, LV adverse remodeling with LV atrophy, a decreased mass/volume ratio, and extracellular matrix expansion suggest cardiomyopathic changes. The best predictor for LV systolic dysfunction was the right ventricular mass/volume ratio. The subnormal peak oxygen consumption indicates that monitoring of LV status may be important for long‐term prognosis.

I n tetralogy of Fallot (ToF), the main clinical focus after surgical repair is directed to the right ventricle (RV) (eg, to the degree of pulmonary regurgitation and RV dilatation). 1 Left ventricular (LV) dysfunction and remodeling in patients with RV overload has received much less attention, although some studies indicate that it may in the long-term significantly contribute to left heart failure. 2,3 Indeed, a recent study identified LV systolic dysfunction as an outcome predictor for death and ventricular tachycardia in ToF. 4 However, the interrelationship between RV and LV remodeling has not been studied in detail in clinically asymptomatic patients with ToF.
Cardiac magnetic resonance (CMR) is optimally suited to simultaneously assess adverse RV and LV volumes and function. 5 Emerging CMR techniques allow tissue phenotyping within the heart, including the assessment of diffuse interstitial myocardial fibrosis, by measuring the expansion of the extracellular volume (ECV). 6 The goals of the present study were as follows: (1) to determine the prevalence of LV dysfunction in NHYA class 1 patients with ToF and to identify associations between right heart markers of pathologic and LV remodeling; and (2) to determine, in a subset of patients (n=70) with cardiopulmonary exercise testing, if LV dysfunction and remodeling have an impact on age-and sex-adjusted peak exercise capacity, a potentially important risk predictor for adverse events in ToF. 7

Methods
The data that support the findings of this study are available from the corresponding author on reasonable request.

Patients and Controls
The cohort for this prospective cross-sectional study consisted of consecutive patients with ToF without overt symptoms, who underwent CMR imaging between January 2012 and August 2016. Patients with clinical symptoms of New York Heart Association functional class >1 as well as patients taking cardiac medications, like angiotensin-converting enzyme inhibitors and b blockers, were excluded. The Institutional Review Board of our hospital approved the study (protocol No. A168/07), and informed written consent was obtained from either the patients or their guardians. Agematched healthy controls were recruited for comparison with the ToF cohort among patients scheduled to undergo a clinically indicated, contrast-enhanced neurological magnetic resonance imaging or an magnetic resonance imaging study to rule out a vascular abnormality within the thorax. All individuals in the control group had a normal CMR examination result and no history of left-sided heart disease or cardiomyopathy.

Cardiac Magnetic Resonance
All CMR studies were performed with a 1.5-T scanner (Achieva; Philips Medical Systems, the Netherlands) using a phased-array coil for cardiac imaging. In patients aged <7 years, sedation with propofol and midazolam was necessary. Heart rate, oxygen saturation, and noninvasive blood pressure were assessed with a CMR-compatible monitor (Invivo Precess 3160; Invivo, Orlando, FL). Dedicated CMR software (ViewForum release 6.3 [Philips Medical Systems, the Netherlands]; Mass CMR software [Medis, Leiden, the Netherlands) was used to analyze the images.

LV Function and Volumes
Short-axis cine imaging was performed to measure LV and RV volumes at end systole (ES) and end diastole (ED), mass at ED, and ejection fraction (EF), per standard guidelines. 8 Volumes and masses were all indexed to body surface area (BSA). Indexed quantities are denoted by an "i" at the end (eg, LV EDVi for LV ED volume index). Global ventricular remodeling was assessed by the ratio of myocardial mass (LV or RV), divided by the EDV of the LV or RV, respectively. Eccentric remodeling is characterized by a mass/volume ratio that is smaller than in normal controls. 9,10 Cine imaging was performed with steady-state free precession (field of view, 2809224 mm; voxel size, 1.8891.9496 mm; repetition time/echo time=4.4/2.5 ms; 25 cardiac phases; no breath hold; averages=2; total scan duration, 3-6 minutes) with 8mm interslice distance and full LV and RV coverage. Axial cine acquisitions were used to measure left atrial (LA) volumes. The LA volumes were also indexed to BSA and are reported for 3 time points during the cardiac cycle: at maximal LA volume before mitral valve opening (LA max ), minimal LA volume at mitral valve closure (LA min ), and LA volume immediately before atrial contraction (LA bac ). LA volumes and LA functional parameters (including passive, active, and total LA EF) were calculated as described previously. 11 Total LA emptying volume was defined as the difference between LA max and LA min , and was divided into LA passive emptying volume and LA contractile volume. LA contractile volume was the difference between LA bac and LA min , and LA passive emptying volume was the difference between LA max and LA bac . LA

Clinical Perspective
What Is New?
• Impaired systolic and diastolic left ventricular (LV) function, adverse LV remodeling with extracellular matrix expansion, reduced cardiomyocyte mass (atrophy), and a decreased mass/volume ratio have been reported in young to middleaged patients with tetralogy of Fallot at a median of 14 years after repair, suggesting LV cardiomyopathic changes, despite the absence of any limitations to normal physical activity (New York Heart Association class 1). • Markers of LV systolic dysfunction and adverse tissue remodeling (reduced cardiomyocyte mass), measured by cardiac magnetic resonance imaging, were associated with subnormal peak oxygen consumption during cardiopulmonary exercise testing, an important risk predictor in tetralogy of Fallot.
What Are the Clinical Implications?
• The presence and location of LGE was identified qualitatively by signal enhancement within the myocardium. Some patients or their guardians either declined to receive contrast or woke up from sedation and did not receive contrast agent (36%). Heart-healthy control subjects did not receive any contrast agent, except for 12 subjects who were scanned for a neurological indication.

ECV Fraction
Cardiac T1 measurements were performed using a cine Look-Locker technique 6,12 for one short-axis slice in the mid-LV. This technique uses a non-slice-selective inversion pulse after the R wave, followed by segmented gradient echo cine acquisition for 20 cardiac phases over 2 to 3 cardiac cycles, representing 20 TI. Parameters are as follows: TI increment, 40 milliseconds; slice thickness, 8 mm; repetition time >3 RR intervals. T1 measurements were performed once before contrast, and at least once after contrast, with the postcontrast acquisition done at least 10 minutes after contrast administration. Look-Locker images were segmented along the endocardial and epicardial borders of the LV, which was further divided into 6 standard segments. The mean signal intensity in each segment was plotted against TI, and T1 determined with a nonlinear least-squares algorithm, and correction for radiofrequency pulse effects. The relaxation rate R1 (=1/T1) in the myocardium was plotted against the R1 in the blood pool. The partition coefficient was calculated as the slope of the least-squares regression line when the number of T1 measurements was >2, or by calculating the ratio of the precontrast/postcontrast R1 change in myocardium, and the R1 change in blood, when only one precontrast and one postcontrast measurement were made. The ECV (for each segment) was calculated as the product of the partition coefficient and (1-hematocrit), as previously described. 6,13,14 We used a global ECV for the mid-LV slice, calculated as the average of the segmental ECV values.

Cardiopulmonary Exercise Testing
Peak oxygen consumption (VO 2 max) was measured by a standard exercise test 15,16 within 6 months before or after the date of magnetic resonance imaging. VO 2 max was obtained in mL kg À1 min À1 , and expressed as age-and sexadjusted percentile ranking on the basis of previously published cardiopulmonary exercise testing (CPET) reference data from 548 female and 647 male patients within 4 to 75 years of age. 15 Subnormal exercise capacity was defined as VO 2 max lower than the third percentile levels for a normal reference population. 15,17 Statistical Analysis As most LV parameters in normal controls show a significant association with age, we defined their normal ranges as a function of age. Smooth curves, corresponding to the 5th and 95th percentile levels for LV parameters measured in healthy volunteers, were calculated with generalized additive models for location scale and shape, using a spline for age as single predictor (package "gamlss" in R). As all controls who had ECV measurements in our study were aged <18 years, we used published referenced data 19 for ECV in adults, and estimated the normal range from means and SDs, to determine the proportion of patients with ToF and ECV above the normal range in their age group (<18, 18-40, 40-60, or >60 years).
Multivariate regression models were built for LV EF and LV mass/volume ratio, to determine which measures of rightsided dysfunction were associated with LV dysfunction and LV remodeling. Age, sex, BSA-indexed RV volumes at ED and ES, RV mass, RV mass/volume ratio, LV/RV mass ratio, and RV peak pressure from a contemporaneous Doppler-echocardiography examination were considered as predictors. A best model was chosen by step-wise forward and backward predictor selection, using the Akaike information criterion. Associations between the prevalence of eccentric LV remodeling and continuous RV parameters were analyzed by logistic regression analysis. A multivariate model for VO 2 max (in mL/ kg per minute) included age, sex, LV EF, and cardiomyocyte mass index as predictors (ie, the univariate predictors, other than age and sex, that had a significant effect on abnormal exercise capacity by univariate analysis). (LV mass was not included as simultaneous predictor because of a relatively strong collinearity with cardiomyocyte mass index.) Preliminary analysis suggested that some of the relationships with VO 2 max appeared to be nonlinear. Therefore, we used a generalized additive model and represented age, LV EF, and cardiomyocyte mass index by cubic regression splines to account for these nonlinearities. Multivariate logistic regression analysis was used to identify predictors of abnormally low VO 2 max.

Baseline Characteristics
A total of 103 patients with ToF (43 female patients, 60 male patients; all New York Heart Association class 1) were studied at 15.1AE8.7 years (median, 14.3 years; range, 1.3-39 years) after surgical repair of ToF. The patient characteristics are summarized in Table 1 and compared with 63 healthy controls. Twenty patients initially had a shunt implantation (17 Blalock-Taussig shunts and 3 Waterston shunts). In 51 patients (49%), a transannular patch was used for RV outflow tract reconstruction. A total of 25 patients (24%) underwent a reoperation of pulmonary insufficiency. Of those patients, 14 (56%) received a valved RV pulmonary artery conduit and 11 (44%) underwent a reconstruction of the pulmonary valve.
Twelve patients (11%) had a QRS duration >180 milliseconds. There were no differences in age, weight, height, BSA, and blood pressure between patients and healthy subjects ( Table 1). CPET was performed in 70 patients (68%). A total of 20 (29%) of 70 patients with ToF had an abnormally low VO 2 max value. Figure 1 shows a 4-chamber end-diastolic view from a cine acquisition, with dilated right and LVs, borderline LV EF, and LV mass index below the normal reference range. 20 Overall, RV mass index, RV volumes at ED (RV EDVi) and ES (RV ESVi), and RV mass/volume ratio were significantly higher in patients with ToF compared with healthy controls ( Table 2). LV EF, LV mass index, LV mass/volume ratio, and LV/RV mass ratio were decreased in patients with ToF, and LV ESVi was higher than in healthy controls. LV EDVi did not differ between the ToF group and the control group. LV EDVi correlated with RV EDVi (r=0.36, P<0.001). Patients with ToF and controls had similar heart rates (P=0.115; Table 1). In patients with ToF and healthy controls, LV volumes and mass correlated strongly with age, and the correlations remained moderate to strong if these LV parameters were indexed by BSA. For each LV parameter, Table 3 lists the proportion of  21 we observed no cases of severe dysfunction (LV EF <35%), 23% of cases showed moderate dysfunction (35% <LV EF <45%), 41% had mild dysfunction (45% <LV EF <55%), and 36% had normal LV EF. Figure 2A shows how the degree of LV systolic dysfunction was associated with remodeling of the RV, quantified by the RV mass/volume ratio (P=0.002 for Jonckheere nonparametric test for increasing RV mass/volume ratio across ordered categories of LV systolic dysfunction). The association between LV systolic dysfunction and RV mass/volume ratio remained significant in a multivariate model for RV mass/volume ratio, with LV EF and age (P=0.33 for age effect) as simultaneous independent predictors. Figure 2B shows how LV systolic dysfunction is associated with RV ESVi, a marker of systolic function that is potentially less prone to confounding by RV volume overloading than RV EF. Nevertheless, RV EF (not shown in Figure 2) also decreased between normal through mild to moderate LV dysfunction (P=0.002 for Jonckheere nonparametric test). The probability of less than mild LV dysfunction decreased by 0.38:1 for a 0.1 increase of the RV mass/volume ratio (P=0.009).

Ventricular Volumes and Function
LV mass/volume ratio, a parameter characterizing LV remodeling, was significantly lower in patients with ToF than in healthy controls (0.58AE0.13 versus 0.65AE0.13 g/mL; P=0.001). LV mass/volume ratio correlated with RV mass/ volume ratio (r=0.54; P<0.001), as shown in Figure 3, and RV mass index (r=0.42; P<0.001). Approximately 28% of all patients with ToF had a LV mass/volume ratio below the agespecific fifth percentile level in healthy controls, as illustrated in Figure 4.  The probability of abnormally low LV mass/volume ratio decreased by 0.53:1 for a 10-g/m² increase of RV mass index (P=0.013; Figure 5A) and also by 0.39:1 for each 0.1-g/mL increase in RV mass/volume ratio (P=0.007; Figure 5B). There was no significant association between abnormal LV mass/volume ratio and RV EF (P=0.225; Figure 5C).

Atrial Volumes
LA volumes and function were decreased compared with controls ( Table 2). LA max , a reflection of LV filling pressures, correlated with age, age at surgical repair, and LV volume indexes (P<0.01). Total and passive LA EFs were negatively associated with age, unlike in healthy controls, in whom they remained nearly constant.

Focal Fibrosis and ECV Fraction
LGE imaging was performed in 65 patients: 62 (95%) of those had LGE at the site of ventricular septal defect patching, 30 (46.1%) showed LGE in the RV outflow tract, and 9 (13.8%) had small spots of LGE that were detected in other areas of the heart but not the LV free wall.
Myocardial ECV was obtained in 57 (55%) of 103 patients with ToF, from the time when the T1 sequence was available on the scanner. It was significantly elevated compared with controls (0.32AE0.05 versus 0.26AE0.01; P<0.01). Female patients with ToF showed a significantly higher ECV than male patients (P=0.033). Even when the patients with ToF and LGE in the LV were excluded, ECV was still significantly higher compared with controls (0.32AE0.06 in patients with ToF versus 0.26AE0.01 in controls; P<0.01). For patients with ToF and abnormally low LV EF, ECV was significantly higher in the septum, compared with patients with ToF and LV EF within the normal range (P=0.02). The LV mass index correlated negatively with ECV (r=À0.32; P=0.016).

Cardiopulmonary Exercise Testing
CPET results were available for 70 patients with ToF (68%) who reached the anaerobic threshold. VO 2 max was   subnormal (lower than the third percentile for reference values in normal population) in 20 patients (29%), and there was no significant difference in the proportion of abnormal VO 2 max between male and female patients with ToF (21% for male patients and 39% for female patients; P=0.16).
In patients with ToF and abnormally low VO 2 max, impaired LV resting systolic function and adverse LV remodeling were more frequent than in patients with VO 2 max above the third percentile level for controls. In univariate logistic regression models, the likelihood of an abnormally low VO 2 max increased with lower LV EF (odds ratio of 1.13 per 1% decrease of LV EF; P=0.008) and a lower LV mass index (odds ratio of 1.9:1 for each 10-g/m² decrease of LV mass index; P=0.024). LV EF and LV mass index remained significant predictors of an abnormally low VO 2 max, when RV mass/ volume ratio, RV ESV index, and RV EF were each added as additional predictor in these 2 models, suggesting that LV dysfunction increases the likelihood of impaired exercise capacity, independent of the degree of RV systolic dysfunction and remodeling. A comparison of the patients with ToF and abnormally low VO 2 max and normal VO 2 max showed that they had a lower LV EF ( Figure 6A), a lower LV mass index ( Figure 6B), and a lower cardiomyocyte mass index, resulting from a higher ECV fraction ( Figure 6C). To further corroborate our results, we calculated 2 ECV averages for each patient: one for the intraventricular septum and one for the LV free wall (the latter is definitely representative of ECV in segments without any nearby LGE). Herein, we see also a significant (P=0.02) difference for the free wall ECV average between patients with ToF with normal and abnormal VO 2 max.
In a multivariate generalized additive model for VO 2 max (in mL/kg per minute), the cardiomyocyte mass had a significant association with VO 2 max (P=0.014), with simultaneous adjustment by sex (P=0.06), age (P=0.25), and LV EF (P=0.057). The predicted dependence of VO 2 max on cardiomyocyte mass index is shown in Figure 7 as continuous lines for male and female sex, with gray-shaded 95% confidence bands around each line.

Discussion
In this study, clinically asymptomatic patients with repaired ToF frequently show left-sided diastolic and systolic dysfunction and a lower than normal LV mass/volume ratio that is consistent with a previously described pattern of eccentric LV remodeling. 9,10,22 Adverse changes in the RV, such as a lowerthan-average RV mass, RV dilation (ie, higher RV EDVi), and a lower RV mass/volume ratio, are associated with a lower LV mass/volume ratio. The diminished LV mass observed in ToF, in combination with the expansion of the ECV in the LV, implies a loss of total cardiomyocyte mass indicative of LV atrophy. LV systolic dysfunction and the loss of LV mass correlated with impaired exercise tolerance, independently of the RV parameters that predict LV remodeling (RV mass/volume ratio) and LV dysfunction (RV ESVi and RV EF). In a multivariate analysis, LV atrophy had a stronger effect on VO 2 max than LV EF, suggesting that LV cardiomyocyte mass may be an important parameter for monitoring LV remodeling in patients with ToF, with the advantage, compared with LV EF, that it is not dependent on hemodynamic loading conditions. In light of previous studies, 4,23,24 showing that adverse changes in the LV may contribute to the overall risk of adverse clinical outcomes in ToF, our results suggest that surveillance of LV remodeling in ToF, starting in clinically asymptomatic patients, may provide an early therapeutic window to prevent progression to clinically overt LV dysfunction and failure. 24

ECV and Loss of Cardiomyocyte Mass
Histological studies have shown increased interstitial collagen content in both the RV and LV walls of patients with ToF. 25,26 ECV is a noninvasive surrogate marker of diffuse interstitial fibrosis, which, similarly to the histological studies, indicates that diffuse interstitial fibrosis is an important pathophysiological feature of myocardial tissue remodeling in ToF. 12,27 In our study, female sex was an independent predictor of elevated ECV in ToF, consistent with the previous finding 28 that female patients with ToF are more likely than male patients to show a cardiomyopathic phenotype with LV dilatation.
ECV was negatively associated with the LV mass index, suggesting that interstitial fibrosis is linked with LV atrophy in ToF. The negative correlation in ToF of ECV with the LV mass index is opposite to a positive correlation observed in pathological hypertrophy resulting from hypertensive heart disease 29 or aortic stenosis, 13  represented by gray and dashed red lines, respectively, for healthy controls, and were determined with a generalized additive model for location, shape, and scale. With age-specific percentile levels, 28% of patients with tetralogy of Fallot (ToF) had an LV mass/volume ratio <5th percentile in healthy controls (red filled circles), and define a subgroup with eccentric remodeling. Patients with ToF with LV mass/volume ratio >5th percentile level in controls are shown as blue filled circles. A similar analysis was performed for other cardiac magnetic resonance-derived LV parameters to determine the proportions of patients with ToF with values outside the 90th percentile range, summarized in Table 3. One data point for a patient aged 55 years is not shown in A because the data for normal volunteers lacked sufficient density above %50 years of age to determine a 95 th percentile range. B, Linear regression analysis showed that, in healthy controls, LV mass/volume ratio increased significantly with age (P=0.007), whereas in patients with ToF, there was no significant association with age. The lines represent the predicted mean from the linear regression model, and the gray bands represent their 95% CIs. C, The median and 5th and 95th percentile levels for LV ejection fraction (EF) in healthy controls, shown as gray and dashed red lines, respectively, changed significantly with age. With age-specific percentile levels, 49% of patients with ToF had an LV mass/volume ratio <5th percentile in healthy controls (red circles). Patients with ToF and LV EF >5th percentile level in controls are shown as blue filled circles. ED indicates end diastole; EDV, ED volume.
cases ECV is similarly increased. The key differentiator between these 2 forms of extracellular remodeling is the loss of LV cardiomyocyte mass, which can be estimated from the product of (1-ECV), the cardiomyocyte fraction of the tissue, and LV mass. An increase of the ECV fraction, or analogously a decrease of (1-ECV), if not compensated by an increase of LV mass, results in a net decrease of the LV cardiomyocyte mass. In patients with ToF, a below normal LV mass and an ECV expansion both contribute to a net loss of cardiomyocyte mass. The functional and potential clinical significance of these findings is highlighted by the fact that ECV was higher in the patients with ToF with less exercise tolerance ( Figure 6A through 6C). The underlying mechanisms for a loss of total cardiomyocyte mass or LV atrophy in ToF are not well understood but may be related to reduced LV diastolic filling rates observed in patients with ToF. [30][31][32][33] Analogously, LV atrophy was observed as a correlate of impaired diastolic filling in an animal model of RV failure 34 and in patients with RV failure, secondary to chronic thromboembolic pulmonary hypertension. 35

Systolic and Diastolic Dysfunction
Approximately two thirds of patients with ToF had mild-tomoderate LV systolic dysfunction, and in nearly half of them the LV EF was below the age-adjusted fifth percentile for    (Table 3). This is a significantly higher proportion than the 21% prevalence reported in a previous echocardiographic study 21 that did not use age-adjusted percentile curves (this may result in an underestimation of the prevalence of impaired LV function in younger patients with ToF, in whom the fifth percentile level for LV EF in healthy controls reaches 55%, rather than the customary adult cutoff of 50% or 45%). 36,37 The subclinical LV abnormalities reported in our study (eg, increase of ECV and LV ESVi and eccentric remodeling) may contribute to the reported risk of later clinical left-sided heart failure and cardiac death in ToF. 4,21,23,38 LV systolic dysfunction was associated in our patients with ToF with impaired exercise capacity (Figure 6), independent of markers of RV systolic dysfunction (RV ES volume and RV mass/ volume ratio). Specifically, with the statistical model that predicts an abnormal exercise capacity from LV EF, any RV parameter added to the model did not have a significant effect and, furthermore, did not change significantly the effect LV EF. This suggests that LV dysfunction may not be simply a mechanical corollary of RV dysfunction, like a correlation between RV EF and LV EF, but is compounded by cardiomyopathic processes in the LV, characterized by atrophy and an increase of ECV. Figure 7 illustrates that the decline of VO 2 max sets in for LV cardiomyocyte mass indexes <%39 g/m 2 in male patients and <%30 g/m 2 in female patients, indicative for LV atrophy. For comparison, assume a healthy 20-year-old man with a myocardial ECV fraction of 0.25 (eg, the study by Broberg et al, 6 with an ECV of 24.8AE2.0% for healthy controls) and a 95% CI for LV mass index from 60 to 93 g/m 2 : the resulting 95% confidence range for LV cardiomyocyte mass index, calculated as Figure 7. Peak oxygen consumption (VO 2 max), measured in patients with tetralogy of Fallot (ToF) during cardiopulmonary exercise testing (in mL/kg per minute), was modeled with a generalized additive model (GAM) as a function of age, sex, left ventricular (LV) ejection fraction, and LV cardiomyocyte mass index (CMi) (ie, the univariate predictors, other than age and sex, that had a significant effect on abnormal exercise capacity) (see Figure 6). The dependence of VO 2 max on CMi (P=0.015) predicted by this GAM is shown for male and female sex by the continuous lines for male and female patients with ToF, respectively. VO 2 max declined for CMi <%39 g/m 2 in male patients and CMi <%30 g/m 2 in female patients, demonstrating the effect of LV atrophy in ToF on exercise capacity, independent of age, sex, and systolic LV function. Age, LV ejection fraction, and cardiomyocyte mass were represented as cubic regression splines. The GAM model for VO 2 max explained 51% of the observed deviance. The gray bands show the 95% CIs.
(1ÀECV)9LV mass index, would range from %45 to 70 g/m 2 . The observed decline of VO 2 max around a level of 39 g/m 2 is, therefore, consistent with the expected threshold for LV atrophy on the basis of published reference values for LV mass index. A higher ECV, indicative of diffuse fibrosis, would lower this threshold further. The investigated parameters of LV dysfunction (eg, LV ESVi and LV EF) may, therefore, have a functional significance for predicting impaired exercise capacity that goes beyond the changes in RV volumes and function. In the statistical analysis, the significance of LV EF for prediction of impaired exercise capacity remained with adjustment by RV parameters, equally suggesting that LV EF in this context should not be interpreted as surrogate for RV markers of dysfunction. In contrast to previous studies, 3,38,39 only a moderate but highly significant (P<0.001) correlation between RV EF and LV EF was observed in our study (q=0.41 versus q=0.58 in Geva et al 38 ). The stronger correlation between RV and LV EFs in the study by Geva et al 38 may be caused by the relatively high proportion of patients with ToF in functional class of New York Heart Association >1 (52%), whereas our cohort only included asymptomatic patients with ToF. In adults with acquired heart disease and compromised RV EF, 40 the correlation between RV and LV EF is noticeably stronger, compared with patients with normal biventricular function. Similarly, Broberg et al showed that LV systolic dysfunction is more prevalent in patients with ToF and moderate to severe RV dysfunction, compared with those with mild RV dysfunction. 21 As RV EF is prone to confounding by pulmonary insufficiency and volume overloading in ToF, 28,41 we determined which other RV parameters correlated with left-sided heart remodeling and dysfunction: both RV mass/volume ratio and RV ESVi, a marker of RV systolic function, were predictors of LV remodeling (LV mass/volume ratio) and systolic dysfunction (LV EF), independent of sex and age.
Valente et al previously investigated in patients with ToF the association between markers of RV and LV remodeling/ dysfunction and outcome (death and ventricular tachycardia). 4 Their best multivariate predictive model for death and ventricular tachycardia within a median follow-up time after magnetic resonance imaging of 1.9 years included as predictors poor LV systolic function (LV EF z score <À2.0), an RV mass/volume ratio ≥0.3 g/mL, and history of atrial tachyarrhythmias; however, more important, RV EF was not a predictor. These findings in an older population (median age, 24 versus 16 years) with signs of overt heart failure underscore the potential importance of LV systolic dysfunction for adverse outcomes. Bokma et al 42 more recently presented a prediction model from a retrospective multicenter study of 575 patients with ToF (mean age, 31 years) from which LV EF and RV EF emerged as the 2 strongest outcome predictors. The results suggest that the at-risk LV in ToF should be monitored for subclinical signs of adverse remodeling and dysfunction.
Diastolic dysfunction may be one of the earliest manifestations of adverse LV remodeling in ToF. We examined LA function by CMR cine volumetry, 11 as a surrogate marker for LV diastolic function. Passive and total LA EFs, as well as the LA volumes, were decreased in our ToF cohort (Table 2), which is in line with the echocardiographic findings of Koenigstein et al 31 in a cohort of 25 asymptomatic patients with ToF. In another CMR study of 20 patients with ToF, Riesenkampff et al 33 also found diminished LA volumes as a sign of pulmonary underfilling and lower available preload.

Limitations
The results of this study should be viewed in the context of its design. Given that this is a cross-sectional study, we cannot relate causes and effects in the interplay of LV function and structure, intrinsic tissue-level alterations (by ECV), RV-LV interdependence, diastolic properties (as reflected by LA function), or vascular stiffness. Patients aged <7 years (15% of ToF cohort) were sedated during the CMR studies, which may have depressed cardiac function. The statistical significance of the main findings (eg, Figures 2 through 7) did not materially change if the analysis was repeated after exclusion of CMR studies with sedation. The potential implications of our study in terms of optimal timing for surgical intervention or medical therapy for prevention of heart failure, arrhythmias, and sudden cardiac death need to be investigated in larger, longitudinal studies. Further limitations are missing data for LGE (40%) and ECV (45% of patients with ToF), as patients or their guardians declined contrast, and T1 mapping became available on the scanner only during the course of the study. Cardiopulmonary exercise testing results were not available for 33 (32%) of patients, mostly because of young age, refusal of the test, or not reaching the anaerobic threshold.
The availability of CPET for only a subset of 70 (68%) of patients with ToF may have introduced a selection bias in the analysis of the associations between abnormally low VO 2 max and LV parameters. Patients with ToF with CPET were older (19.8AE10.4 versus 11.4AE8.85 years for patients with ToF without CPET) and had a lower body mass index (18.4AE4.82 versus 21.9AE4.7 kg/m 2 ), but showed no difference in sex distribution.

Conclusions
This multiparametric CMR study on LV remodeling in asymptomatic patients with ToF at a median of 14 years after surgical repair revealed the following: (1) frequent (49%) impaired systolic function over the full age range, abnormal diastolic relaxation, a decreased LV mass, and interstitial matrix expansion (as reflected by ECV), which are all evidence of adverse LV remodeling and a cardiomyopathic phenotype (the best predictor for LV systolic dysfunction was the RV mass/volume ratio); and (2) CMR markers of LV systolic dysfunction and adverse tissue remodeling (reduced cardiomyocyte mass) correlated with subnormal VO 2 max, an important risk predictor in ToF.
Therefore, early surgical reintervention for pulmonary insufficiency if present, or medical heart failure therapy, both guided by CMR-guided surveillance of the LV, may delay progression of adverse remodeling and improve the long-term prognosis.