Left atrial strain correlates with severity of cardiac involvement in Anderson-Fabry disease

Objectives Cardiac involvement in Anderson-Fabry disease (AFD) results in myocardial lipid depositions. An early diagnosis can maximize therapeutic benefit. Thus, this study aims to investigate the potential of cardiac MRI (CMR) based parameters of left atrial (LA) function and strain to detect early stages of AFD. Methods Patients (n = 58, age 40 (29–51) years, 31 female) with genetically proven AFD had undergone CMR including left ventricular (LV) volumetry, mass index (LVMi), T1, and late gadolinium enhancement, complemented by LA and LV strain measurements and atrial emptying fractions. Patients were stratified into three disease phases and compared to age and sex-matched healthy controls (HC, n = 58, age 41 [26–56] years, 31 female). Results A total of 19 early-, 20 intermediate-, and 19 advanced-phase patients were included. LV and LA reservoir strain was significantly impaired in all AFD phases, including early disease (both p < 0.001). In contrast, LA volumetry, T1, and LVMi showed no significant differences between the early phase and HC (p > 0.05). In the intermediate phase, LVMi and T1 demonstrated significant differences. In advanced phase, all parameters except active emptying fractions differed significantly from HC. ROC curve analyses of early disease phases revealed superior diagnostic confidence for the LA reservoir strain (AUC 0.88, sensitivity 89%, specificity 75%) over the LV strain (AUC 0.82). Conclusions LA reservoir strain showed impairment in early AFD and significantly correlated with disease severity. The novel approach performed better in identifying early disease than the established approach using LVMi and T1. Further studies are needed to evaluate whether these results justify earlier initiation of therapy and help minimize cardiac complications. Key Points • Parameters of left atrial function and deformation showed impairments in the early stages of Anderson-Fabry disease and correlated significantly with the severity of Anderson-Fabry disease. • Left atrial reservoir strain performed superior to ventricular strain in detecting early myocardial involvement in Anderson-Fabry disease and improved diagnostic accuracies of approaches already using ventricular strain. • Further studies are needed to evaluate whether earlier initiation of enzyme replacement therapy based on these results can help minimize cardiac complications from Anderson-Fabry disease. Supplementary Information The online version contains supplementary material available at 10.1007/s00330-022-09183-7.


Introduction
Anderson-Fabry disease (AFD) is a rare, hereditary, X-linked, lysosomal storage disease caused by mutations in the ɑ-galactosidase A gene, resulting in an enzymatic deficiency of alphagalactosidase [1,2]. Accumulation of sphingolipids leads to myocardial lipid depositions. With disease progression, these structural changes may induce fibrosis, ultimately leading to heart failure and arrhythmia [1,3,4]. Although, enzyme replacement (ERT) and oral chaperone therapies are available their maximum therapeutic benefit is only achievable if initiated prior to irreversible organ damage or dysfunction [5][6][7][8][9].
Classic imaging biomarkers for advanced AFD are left ventricular (LV) hypertrophy and late gadolinium enhancement (LGE). T1 mapping on the other hand can directly visualize sphingolipid deposition in the myocardium, even in earlier disease stages [10,11]. In addition, the analysis of myocardial deformation, known as strain imaging, has recently gained more attention in the early diagnosis of AFD [6,[12][13][14][15][16].
Strain analysis can be performed by speckle tracking echocardiography (STE) and cardiac magnetic resonance imaging (CMR) [17][18][19][20][21]. While STE is highly user-dependent and may suffer from patient-specific factors such as pour acoustic windows, CMR has become the preferred method for strain assessment due to its excellent reproducibility and accuracy, especially in regard to the atria [22][23][24][25][26]. Analysis of atrial function using volumetric and deformation parameters may have the potential to provide insights into the early stages of impairment in AFD, as sphingolipid accumulation in the myocardium subsequently leads to myocardial stiffening and diastolic dysfunction. Both phenomena have been linked to disturbed atrial function [13,27].
Therefore, the goal of this study was to assess atrial function and deformation using CMR-derived total (TEF), passive (PEF), and active (AEF) atrial emptying fractions and strain analysis to address the hypothesis that these parameters allow to discriminate not only between AFD patients and healthy controls (HC) but are able to evaluate disease severity.

Materials and methods
The protocol of this retrospective single-center study was approved by the responsible institutional ethics committee.

Study cohort
A total of 58 AFD patients (age range 8-77 years; 5 patients under the age of 18) were enrolled and stratified into three disease severity phases based on CMR findings [4] (Table S1), as follows: phase I included patients with known disease-causing mutation but without signs of disease; phase II comprised patients with an increased myocardial mass and/ or pathologically shortened T1; and phase III involved patients with shortened T1 and highly abnormal myocardial mass along with pathological LGE. Evaluation and patient stratification were performed by a radiologist with > 10 years of experience in CMR and sex-specific cut-offs were based on the healthy controls' mean values ± 2 standard deviations (SD) for phase II and mean values ± 3 SD for phase III findings. In case of the presence of pathological LGE with a typical pattern, the patient was moved to phase III by default. In case of conflicting results from T1 and LVMi measurements, cases were advanced to the higher of the phases in question.
A group of 58 healthy subjects who were matched 1:1 with AFD patients on a basis of age and sex served as a control group. Based on their clinical history, none of the subjects have had any cardiovascular events or symptoms or cardiovascular risk factors (e.g., hypertension or diabetes). All healthy controls (HC) had normal LV volumes, normal myocardial T1 and T2 relaxation times according to institutional reference ranges, and no LGE.
For contrast enhancement, 0.2 mmol/kg gadoteric acid (Dotarem, Guerbet) was administered. LGE images were acquired 10 min after contrast injection using a segmented T1weighted inversion recovery ultrafast spoiled gradient echo sequence in long-and short-axis views identical to those used for cine imaging. The following pulse sequence parameters were applied: TR 2 × RR-interval, TE 4.38 ms, in-plane resolution 1.4 × 1.8 mm 2 , slice thickness 8 mm, inter-slice gap 2 mm, and flip angle 25°. The inversion time (TI) ranged between 260 and 320 ms and was defined based on a separately acquired TI scout. TI was increased by 10 ms for every minute of acquisition to optimally null the signal from the normal myocardium.
Image analysis and post-processing Image analysis was performed by a specifically trained reader. For post-processing, a dedicated cardiovascular software (cvi42, V5.11, Circle) was used. Atrial strain analysis was semi-automatically performed as recently described [28] and manually adjusted if necessary.
Volumetric analysis of left atrial (LA) time-volume curves was semi-automatically performed using the biplane arealength calculation technique, which is the method of choice if the Simpsons method is not practical [29]. As the current standard, LA volumes were indexed to body surface area (BSA) [21,30], then TEF, PEF, and AEF were derived from the time-volume curves according to current literature (Table S2) [27,[31][32][33].
To calculate LA strains, endocardial and epicardial borders were manually traced in a single slice at end-diastole in the 2-, 3-, and 4-chamber-views (Fig. 1). The LA appendage, the pulmonary veins, and the inferior and superior vena cava were excluded, according to standard practice [34][35][36]. The software subsequently propagated the contours throughout the entire cardiac cycle and thus calculated longitudinal LA strain automatically.
Using the LV end-diastolic frame as a zero-reference, strain curves were then analyzed to identify peaks corresponding to the following phases of the LA cycle: (1) reservoir-passive LA filling due to inflow from pulmonary veins, (2) conduitpassive flow through the LA after the opening of the mitral valve in the early LV filling phase, and (3) booster-active LA contraction during the late LV filling phase [33,37]. In doing so, reservoir, conduit, and booster strains were derived (Fig. 2).

Statistical analysis
Statistical analyses were performed using SPSS Statistics (V23, IBM Corp). The Kolmogorov-Smirnov test was used to assess the normal distribution of the continuous data. Continuous variables are reported as means with standard deviations in brackets if normally distributed, and as median and the respective interquartile ranges in case of non-normal distribution. Categorical variables are reported as absolute frequencies and proportions. Differences between AFD phases and HC were compared using two-way ANOVA with post hoc testing and Bonferroni correction. Binary logistic regression models were created to evaluate multiparametric approaches.
To evaluate differences in the ability to discriminate between HC and AFD patients, receiver operating characteristic (ROC) curve analyses were performed and areas under the curve (AUC) were evaluated. Cut-off values were determined by the highest Youden's index from the comparison between all AFD patients and all HC, as this most closely resembles the real-world scenario where it is not known what disease phase a patient is in ahead of the CMR. These cut-offs were subsequently used for phase-specific evaluation of sensitivity, specificity, and diagnostic accuracy. Pairwise DeLong comparisons were performed to compare different approaches. A p value < 0.05 was considered significant.

Study population
The overall study population consisted of 58 patients (mean age 40 years  LA emptying fractions, LVMi, and global T1 relaxation times did not show significant differences between the groups (all p > 0.05). Further details for all parameters can be found in Table 1.
In the ROC-curve analyses, LA reservoir strains reached the highest discriminatory power between phase I patients and matched HC (AUC 0.88). Using the cut-off at 35.9 %, a sensitivity of 89 % and a specificity of 75 % were reached. LVMi (AUC 0.51) and T1 (AUC 0.52) performed significantly lower when compared using the DeLong test (all p < 0.02). In multiparametric testing, the addition of LA reservoir strain significantly increased the AUC compared to the combination of LVMi and T1 alone (0.52 vs. 0.90) as well as LVMi, T1, and LV GLS (0.79 vs. 0.90) for differentiating phase I AFD from HC. Comparison between mono-and multiparametric approaches revealed that LA reservoir strain alone was noninferior to all tested multiparametric approaches (all p > 0.211) (Fig. 3). Detailed results for all AUC analyses are reported in Table 2.

Phase II AFD patients
As seen in Fig. 4, mean T1 relaxation times were significantly shortened in phase II in comparison to HC (p = 0.022) and phase I patients (p = 0.012). LVMi was significantly increased in comparison to matched HC (68.9 g/m 2 [    The comparison between multiparametric approaches showed a significant increase of the AUC when LA reservoir strain was added to the combination of LVMi and T1 (AUC 0.68 vs. 0.96) as well as LVMi, T1, and LV GLS (0.79 vs. 0.96). Comparison between mono-and multiparametric approaches revealed that LA reservoir strain alone was superior to the combination of LVMi and T1 as well as non-inferior to all tested multiparametric approaches (all p ≥ 0.088) (Fig. 3 Figure 4 shows significantly shortened T1 relaxation times compared to HC and phase I (p < 0.001); however, there were no significant differences between phases II and III (p = 0.706).
In ROC-curve analyses, T1 reached an AUC of 0.85 (sensitivity 95 % and specificity 80 %) for the differentiation between phase III patients and HC, while LVMi reached an AUC of 0.99 (sensitivity 95 % and specificity 93 %).

Sex-specific results
As AFD is an x-linked disease, results must also be viewed as split by sex. Comparisons between male and female patients showed a significantly smaller LA TEF for phase I male patients, significantly younger male patients in phases II and III, and a significantly higher increase in LVMi for male patients in both phases II and III. Significantly lower (yet still normal) LV EF was found in male phase II patients and significantly higher LA PEF was observed in male phase III patients. The detailed comparison of sex-specific baseline characteristics for AFD patients can be found in Table 3.
In addition, Table 4 shows detailed sex-specific cut-off points and diagnostic accuracies for the comparison of all AFD patients with all HC. In short, LA reservoir and conduit strains as well as LV GLS showed only small differences in optimal cut-offs between women and men (< 1 %), while other parameters had slightly higher sex-specific differences.

Correlation of LA strain and functional parameters with disease severity
Significant positive correlations were observed between LA reservoir strain and T1 (Pearson's r = 0.40 (female) and r = 0.28 (male), p ≤ 0.05). In addition, there were negative correlations between LA reservoir strain and LVMi for both women (r = −0.47, p < 0.001) and men (r = 0.42, p = 0.001) (Fig. S4).

Discussion
In this study, various LA parameters were analyzed in a multiparametric CMR diagnostic workup of AFD patients. Disease severity and progression were evaluated based on novel atrial, as well as already established parameters, such as functional markers, LV strain, and myocardial T1. The results demonstrated that LA reservoir strain reliably differentiates between  early-phase AFD patients and HC when atrial volumetric parameters and T1 mapping fail to do so. In addition, this study found a significant correlation between LA strains and disease severity in AFD.
Previous studies have shown that abnormalities in myocardial T1 occur before LV hypertrophy develops [10][11][12]38]. Therefore, T1 mapping complements myocardial volumetry and LGE imaging for the detection of cardiac involvement and the monitoring of disease progression. While the diagnosis and treatment of classic AFD have become standardized and straightforward, certain patients, such as women and those with rarer gene mutations may experience a challenging path to establish their diagnosis. This is due to the X-linked inheritance pattern, reduced penetrance, and variable expression [39,40]. Therefore, it remains crucial to evaluate new parameters for their potential to detect myocardial involvement as early as possible.
The single CMR-based atrial strain study available in the literature was designed to compare AFD patients with severe LV hypertrophy (equivalent to phase III patients in this study) and patients with hypertrophic cardiomyopathy and thus had a different scope [14]. However, to the extent a comparison is warranted, the ranges of LA reservoir strain in phase III AFD patients from this study were within one standard deviation of the above-mentioned study population (21.2% [11.7-30.7] in this study vs. 25% [19][20][21][22][23][24][25][26][27][28][29][30][31] in Moroni et al [14]).

LA volumetry
An STE-based study has shown that abnormalities of LA functional parameters occur in AFD patients before LV hypertrophy becomes evident [13]. Interestingly, in this study, LA volumetric parameters had no diagnostic impact in the early disease phases. LA emptying fractions (TEF, PEF, and AEF) showed pathologically altered values only in patients with advanced disease (phase III). This is most likely due to the fact that impairment of deformation patterns, therefore abnormalities in strain, do not implicate functional impairment in the early phases.
In addition, this study confirms prior literature findings that AEF proved to be the least diagnostically accurate in comparison to TEF and PEF [41,42]. This may be explained by the nature of this parameter and its respective phase of the cardiac cycle, as small errors in contouring have a higher impact on parameters that rely on the detailed acquisition of small values.

LA strain analysis
Prior research demonstrated that ventricular strain analysis by CMR could show AFD-related strain abnormalities even before sphingolipid deposition can be detected by mapping sequences [16]. The results corroborate these findings as LA strain parameters were able to confidently differentiate phenotypically negative AFD patients from HC. Phase I AFD patients showed similar impairment of atrial deformation patterns as patients with abnormal myocardial T1 (phase II). In addition, LA reservoir strain correlated with disease severity and allowed for the stratification of AFD patients with advanced disease (phases II and III). In contrast, T1 is better suited for the detection of early disease, while LVMi is more useful in advanced AFD. While T1 relaxation times are abnormally short in early disease (phase II), they can pseudonormalize in the presence of myocardial fibrosis (phase III), therefore reducing diagnostic confidence. To an extent, this can also be seen in this study's data, as there were no significant differences in T1 relaxation times between phases II and III, and diagnostic confidence from T1 mapping alone was Fig. 4 Boxplots showing a gradual reduction in LA reservoir strain (left) and T1 relaxation times (right) over all three AFD phases. Significant differences between groups are marked with brackets and asterisks. LA, left atrial; ns, not significant inferior to LVMi in phase III patients. LVMi, on the other hand, can be mostly seen pathologically increased in advanced disease and is not able to confidently detect early disease (phases I and II). As previously reported, LA booster strain had the lowest diagnostic performance among atrial strain parameters [21,43,44]. Importantly, ROC analyses also proved that atrial strain assessment has an incremental value when compared to multiparametric approaches including ventricular strains. From a pathophysiological viewpoint, the early impairment and gradual deterioration of LA strain with advancing disease could be explained by an accumulation of multiple pathways with abnormal atrial deformation patterns as their common outcome [45,46]: (1) Deposition of sphingolipids in the atrial myocardium Histopathological studies have proven that sphingolipid deposition does occur in the atria [47]. Unfortunately, as the atrial myocardium is extremely thin, T1 mapping cannot be performed in order to correlate LA reservoir strain with the amount of fatty deposition. Atrial hypertrophy is not commonly found in AFD patients, which sets AFD apart from other storage diseases such as amyloidosis and makes primary atrial involvement preceding ventricular involvement less likely to be the main cause for the findings.
(2) Ventricular diastolic dysfunction The strain has recently been highlighted in diseases accompanied by diastolic dysfunction [26,30,[48][49][50]. In addition, it has been shown that myocardial lipid depositions can lead to diastolic dysfunction which ultimately progresses into restrictive cardiomyopathy [51]. The results of this study suggest that this mechanism may occur at earlier disease phases than previously thought. Corroborating this thesis is that LA reservoir strain was the most accurate predictor of cardiac involvement in this study. Its nature and corresponding phase of the cardiac cycle make it a parameter both reliant on atrial mechanics itself, as well as the mitral annular plane excursion, and therefore suggests high sensitivity for diastolic dysfunction [52].
Overall, this study complements the current literature in regard to atrial strain analysis and involvement in AFD by demonstrating its diagnostic performance. It provides novel insights into atrial deformation in AFD patients, suggesting latent diastolic dysfunction pre-detectable accumulation. In addition, it was able to show a correlation between disease severity and LA reservoir strain. The careful evaluation of the atria could aid in further understanding the different pathophysiological pathways in which the disease impairs cardiac function and potentially help guide therapeutic decisions in the future [53].

Limitations
This study has several limitations. Despite the balanced sex ratio and efforts to manage an equal distribution of HC and patients with AFD, this remains a single-center, retrospective study. Several studies have already validated the reproducibility of atrial strain analyses; however, there are still significant differences between various vendors. Thus, there exists no current reference standard [26,28,54]. Due to its retrospective nature, this study cannot provide outcome data. Further studies are needed to investigate the connection between pathological atrial strain values on disease progression. Only with that knowledge, the benefits of starting ERT at the earliest time possible can be weighed against the challenges of a lifelong bi-weekly infusion regimen with its innate risk of immune responses limiting later use of the ERT [55]. Because of the retrospective nature of the study, patients did not undergo endomyocardial biopsy as ground truth for cardiac involvement in AFD. However, as a surrogate, thorough genetic analyses were performed, and all mutations with their respective nucleotides and protein changes are reported in Table S3.

Conclusion
The analysis of atrial deformation patterns by CMR featuretracking strain analysis revealed that LA reservoir strain reliably differentiates between early-phase AFD patients and HC whereas atrial volumetric parameters and T1 mapping fail to do so. In addition, atrial strain analysis provides incremental value in the diagnosis of early AFD when compared to multiparametric approaches including ventricular strain analysis. This study also found a significant correlation between LA strains and disease severity in AFD measured by T1 and LVMi. These results suggest that the evaluation of LA deformation patterns should be an integral part of the diagnostic work-up in AFD with questionable cardiac involvement. Further studies are needed to evaluate whether earlier initiation of ERT based on these results can help minimize cardiac complications from AFD.