Funding Sources: This work was supported by grants to DRL from the Friedreich Ataxia Research Alliance.
Longitudinal Strain in Friedreich Ataxia: A Potential Marker for Early Left Ventricular Dysfunction
Article first published online: 9 JUL 2013
© 2013, Wiley Periodicals, Inc.
Volume 31, Issue 1, pages 50–57, January 2014
How to Cite
- Issue published online: 3 JAN 2014
- Article first published online: 9 JUL 2013
- Friedreich Ataxia Research Alliance
- Friedreich ataxia;
- myocardial strain;
- speckle tracking echocardiography;
- left ventricular hypertrophy;
- heart failure
Friedreich's ataxia (FRDA) is a neurodegenerative disorder resulting from deficiency of frataxin, characterized by cardiac hypertrophy associated with heart failure and sudden cardiac death. However, the relationship between remodeling and novel measures of cardiac function such as strain, and the time-dependent changes in these measures are poorly defined.
Methods and Results
We compared echocardiographic parameters of cardiac size, hypertrophy, and function in 50 FRDA patients with 50 normal controls and quantified the following measures of cardiac remodeling and function: left ventricular (LV) volumes, mass, relative wall thickness (RWT), ejection fraction (EF), and myocardial strain. Linear regression analysis was used to identify significant differences in echocardiographic parameters in FRDA compared with normal subjects. In analyses adjusted for age, sex, and body surface area, significant differences were observed between parameters of remodeling (LV mass, RWT, and volumes) and function in FRDA patients compared with controls. In particular, longitudinal strain was significantly decreased in FRDA patients compared with controls (−12.4% vs. −16.0%, P < 0.001), despite similar and normal left ventricular ejection fraction (LVEF). Over 3 years of follow-up, there was no change in strain, LV size, LV mass, or LVEF among FRDA patients.
Longitudinal strain is reduced in FRDA despite normal LVEF, indicative of subclinical cardiac dysfunction. Given late declines in LVEF in FRDA, longitudinal strain may provide an earlier index of myocardial dysfunction in FRDA.
Friedreich's ataxia (FRDA) is an autosomal recessive neurodegenerative disorder associated with hypertrophic cardiomyopathy, scoliosis, and diabetes mellitus.[1-4] The genetic defect in FRDA consists of expansion of a tri-nucleotide repeat (GAA) in intron 1 of the frataxin (FXN) gene, located on chromosome 9q13.[1-4] This expanded repeat leads to deficiency of frataxin, a small mitochondrial protein involved in iron sulfur cluster synthesis, oxidative phosphorylation, and production of adenosine tri-phosphate. The number of GAA repeats in the shorter allele correlates with age of onset, neurologic severity, the severity of cardiomyopathy, and age at death.[1, 3, 5-7]
The exact prevalence of cardiac dysfunction in FRDA is unknown. FRDA is typically marked by significant cardiac hypertrophy, which is thought to represent a compensatory process to mitochondrial dysfunction. Whether this hypertrophy is associated with left ventricular (LV) dysfunction is not clear. Furthermore, the changes in these parameters over time are also incompletely defined. Recent studies have only been able to develop broad staging of cardiomyopathy over 4 grades, with substantially overlapping features between grades. On the basis of a study examining strain in a small, old FRDA cohort, we hypothesized that strain echocardiography (longitudinal circumferential, radial) could detect subtle abnormalities of regional and global LV function despite normal left ventricular ejection fraction (LVEF) even in early stages.[9-11] The aims of this study were (1) to determine the differences in cardiac size, geometry, and function in FRDA compared with normal controls; (2) to define the changes in myocardial strain and remodeling over time in FRDA; and (3) to assess the relationships among genetic severity, left ventricular hypertrophy (LVH), and myocardial strain in FRDA.
Fifty FRDA patients were recruited from a natural history study of FRDA and from the Friedreich's Ataxia Parents Group.[7, 12, 13] All FRDA patients had clinically indicated echocardiograms 3 years apart to assess the time-dependent changes in LV size and function. Fifty normal subjects recruited from our database served as controls for LV size, wall thickness (WT), LVEF, and myocardial strain, and were free from relevant comorbidities. This group was chosen to provide specific reference ranges for the lab for disease-free individuals, with further analyses included to account for the inherent age difference. The study protocol was approved by the Institutional Review Board.
Baseline clinically indicated transthoracic two-dimensional (2D) echocardiograms with simultaneous electrocardiogram in all participants were obtained through an ongoing natural history study in FRDA, and analyzed retrospectively. Ultrasound machines used included Philips (IE-33; Best, The Netherlands), General Electric (Vivid-7; Fairfield, CT, USA), and Siemens (Malvern, PA, USA). Echocardiograms were digitized and analyzed at the Echo Core Laboratory at the University of Pennsylvania by a senior echocardiographic technician (TP) with >25 years of experience performing quantitative echo analysis. Analysis of all echocardiograms was performed blinded to subject identity, demographics, genetic, clinical, or previous echo data. We also collected medication history including the use of idebenone.
LV Dimensions, Volumes, and Ejection Fraction
Linear measurements were made of LV dimensions at end-diastole (LVIDd) and at end-systole (LVIDs) using 2D parasternal images as recommended by the American Society of Echocardiography (ASE). 2D echocardiograms were digitized to obtain LV end-diastolic and end-systolic LV volumes using Simpson's method of discs, from which LVEF was calculated.
Left Ventricular Hypertrophy and Cavity Shape
Linear measurements of interventricular septal and posterior LV WT were made from parasternal images. Relative wall thickness (RWT) was computed as the ratio of 2LV WT/LV cavity diameter at end-diastole, and was used as an indicator of LV mass (LVM) and architecture. LVM was calculated at end-diastole as described previously, and indexed left ventricular mass index (LVMI) to body surface area (BSA). LV cavity shape was assessed as a sphericity index defined as the ratio of LV volume and the volume of a sphere with the diameter equal to LV cavity length at end-diastole and end-systole.
The severity of mitral regurgitation (MR) was assessed visually in multiple planes and also quantified as the average area of the Doppler color-encoded mitral regurgitant jet divided by the left atrial area (MR jet area/(LA) area).
Strain (S) was assessed by speckle tracking and defined as the change in myocardial segment length (ΔL) divided by resting segment length (L0): S = ΔL/L0. Speckle tracking depends upon the temporal and spatial tracking of naturally occurring intra-myocardial reflectors of ultrasound (speckles) within the 2D echocardiographic images of the LV walls. Displacement of these speckles allows estimation of deformation[9, 16] from which myocardial strain can be calculated; longitudinal strain from the LV long axis and radial and circumferential strains from LV short-axis images obtained at mid-cavity level. Automated calculation of myocardial strain was performed using a TomTec Imaging System (Unterschleissheim, Germany) that is independent of ultrasonoscope vendor. Estimates of myocardial strains by echo speckle tracking have been validated in man by cardiac magnetic resonance (CMR) with tissue tagging and in animals by sonomicrometry.[16, 17] In this study, peak myocardial systolic strains were recorded simultaneously from the inter-ventricular septum and from the lateral LV wall in 3 segments (apical, mid, and proximal) from the apical four-chamber view. The frame rate selected for the speckle tracking strain analysis was 30 frames/sec. Strain analysis was done retrospectively. Radial and circumferential strains were measured in the LV short axis at mid-cavity level. The time period from onset of QRS to peak strain was recorded. Strain rate was also recorded but not systematically analyzed due to apparent confounding by spatial and temporal variability. The time period from onset of QRS to peak strain was also recorded in each segment.
The reproducibility of LV linear dimensions, volumes, and LVEF in the Echo Core Laboratory has been published previously.[18, 19] For strain measurements, intraobserver concordance was on the order of 0.8 with an intraobserver coefficient of variation (CV) for longitudinal strain of 10.4%, and an interobserver CV of 15.2%.
A paired t-test was used to identify significant changes in echocardiographic parameters between baseline and follow-up in FRDA patients with measurements at each time -point, with a P-value <0.05 as the criterion for statistical significance. Similarly, parametric tests were used to compare mean echocardiographic parameters between FRDA patients versus normal subjects measured at baseline and follow-up. The Wilcox-on rank sum test was also applied to confirm that the results were not sensitive to any potential violations of the assumptions for the t-test, including an assumption of normality.
After making unadjusted comparisons of means between groups, we fit linear regression models to adjust for gender, BSA, and age. The estimated regression models were then used to obtain fitted values for each group, evaluated at the average value of age, gender, medication use, and BSA of patients in the estimation sample. The fitted values allow for comparison of predicted echocardiographic parameters between groups.
The fit of the regression models was assessed via the R-squared criterion, which is the sample correlation between the outcome variable and fitted values from the regression. The assumption of normality for the regression models was assessed via graphical checks that included constructing histograms and quantile-quantile (QQ) plots for the residuals. In addition, the Shapiro-Wilk test was applied to test the normality of the residuals. Only LVM Index had a departure from normality that could not be corrected by standard transformations that include the natural log, square root, and Box-Cox transformations; this is noted as a limitation of the results for this echo measure, which is not the major focus of this study.
The relationship between echo parameters and GAA repeat length was also assessed by linear regression, accounting for age and sex. Statistical analysis was conducted using Stata version 12.0 (StataCorp, College Station, TX, USA).
The study population consisted of 50 patients with clinically genetically confirmed FRDA ranging in age from 2 to 42 years (mean 17.9 ± 0.9; 42% male). Fifty normal subjects ranging in age from 15 to 71 (mean age 36.4 ± 15.6; 25% male) served as controls for LV size, WT, LVH, LVEF, and myocardial strain. Mean BSA was 1.42 ± 0.35 m2 for FRDA subjects, 1.77 ± 0.19 m2 for controls. Patients with FRDA were smaller in size as compared with controls (mean height 153 ± 20 cm vs. 162 ± 26 cm and mean weight 49 ± 19 kg vs. 69 ± 18 cm), resulting in lower BMIs in the FRDA subjects. Heart rate was faster in FRDA subjects (78 ± 12 vs. 68 ± 11).
Left Ventricular Dimensions and Ejection Fraction
LV dimensions are shown in Table 1. To account for differences in age, comparative analyses of echocardiographic parameters between populations were only made after adjusting for age, sex, and BSA. In adjusted analyses, comparison between FRDA and normal subjects showed that LV dimensions and volumes were smaller in FRDA (Table 2). However, LVEF was similar between the 2 groups (53 ± 8% in FRDA vs. 55 ± 2% in the normal subjects; P > 0.05).
|Echo Parameter||FRDA Baseline (n = 50)||FRDA F/u (n = 50)||Normal (n = 50)|
|IVSTd (cm)||1.07 ± 0.25||1.15 ± 0.31||0.80 ± 0.15|
|RWT (cm)||0.51 ± 0.13||0.52 ± 0.14||0.35 ± 0.04|
|EF (%)||53.0 ± 8.1||53.3 ± 10.9||55.11 ± 2.1|
|LVMI||98.8 ± 41.1||85.0 ± 39.0||72.4 ± 9.5|
|Diastolic LV volume (cc)||90.1 ± 29.4||92.5 ± 30.6||116.3 ± 19|
|Systolic LV volume (cc)||43.12 ± 17||44.88 ± 25||52.2 ± 9.2|
|Diastolic LV volume indexed||61.2 ± 10.3||56.6 ± 12.8||65.06 ± 8.8|
|Systolic LV volume indexed||27.9 ± 5.9||25.8 ± 8.3||29.2 ± 4.3|
|Diastolic sphericity index||0.33 ± 0.08||0.32 ± 0.08||0.35 ± 0.06|
|Systolic sphericity index||0.25 ± 0.07||0.22 ± 0.07||0.26 ± 0.04|
|Ave. long strain (4ch)||−12.5 ± 3.5||−12.3 ± 3.7||−16.1 ± 2.3|
|Ave. time to peak L strain (4ch) (msec)||351.6 ± 65||366.3 ± 49||382.5 ± 43|
|Ave. circ. SAX strain||−24.9 ± 7.1||−24.9 ± 6.0||−26.7 ± 3.7|
|Ave. time to peak C strain SAX (msec)||318 ± 53||341 ± 65||376 ± 44|
|Ave. rad. SAX strain||32.0 ± 14.2||32.9 ± 12.3||38.3 ± 13.9|
|Ave. time to peak R strain SAX (msec)||327 ± 73||346 ± 61||388 ± 58|
|Echo Outcome||Predicted Values for Baseline FA||Predicted Value for Normal Controls||Beta Coefficient||P-Value||R- Squared|
|LV mass index||97.87||74.46||−23.41||0.007||0.21|
|LV diastolic volume (cc)||96.63||107.36||10.73||0.012||0.73|
|LV systolic volume (cc)||44.15||48.03||3.88||0.093||0.60|
|Average longitudinal strain||−12.58||−15.89||−3.32||<0.001||0.41|
Left Ventricular Hypertrophy
Although only 40% of FRDA subjects had LVH at baseline (Fig 1), mean LVM (148 ± 58 gm), and LVMI (66 ± 24 gm/m2) were elevated at baseline in FRDA patients (Table 1). In the majority of FRDA patients, septal and free WT were similar. Asymmetric septal hypertrophy occurred in 4 of the 50 FRDA patients, but was not associated with LV outflow tract obstruction. RWT was greater in FRDA patients than in normal subjects (0.52 ± 0.14 vs. 0.35 ± 0.04; P < 0.001) (Tables 1 and 2).
The average peak systolic longitudinal strain in FRDA patients at baseline was lower than that in normal subjects (P < 0.001) (Table 1); this difference remained significant after adjusting for age, sex, and BSA. Strain was not associated with BSA (P = 0.63) or age (P = 0.88) in our data, though previous literature suggests a decline in longitudinal, circumferential, and radial strain with increasing age in control subjects (Table 2). In contrast, the average peak radial strain and circumferential strain in FRDA patients were not different from the normal patient cohort. LV volume measurements were not available in 7 FRDA patients at baseline and 10 at follow-up; longitudinal strain was not assessed in 23 FRDA patients at baseline and 19 at follow-up, due to an inability to capture adequate apical echocardiographic images. The group with missing longitudinal strain values, usually reflecting the lack of detail in clinically obtained echocardiograms, was younger (mean age 13 ± 7 years) than those with measureable longitudinal strain values (21 ± 9).
Correlation of Strain with Echocardiographic Metrics
Although patients with FRDA on average had LVH and abnormal longitudinal strain, these abnormalities were not necessarily directly connected. FRDA subjects with septal thickness interventricular septal thickness in diastole (IVSd) >1.2 cm had average longitudinal strain values (−12.4 ± 2.4%) was similar to those with IVSd <1.2 cm, (−12.7 ± 3.5%). In addition, the time from the onset of the electrocardiographic QRS complex to peak longitudinal, peak circumferential, and peak radial strains within patients were similar in all FRDA subjects and within a normal period of 65 msec, indicating a lack of dyssynchrony or temporal contractile delay. Interestingly, this time to peak strain was shorter in the FRDA patients as compared with controls (P < 0.005). Within the FRDA cohort, longitudinal strain and heart rate correlated poorly (r = 0.21), showing that heart rate did not directly lead to the abnormal longitudinal strain values noted in FRDA. Furthermore, these differences could not be explained by age, as mean IVSd (1.06 for adults vs. 1.08 children), longitudinal (−12.2% vs. −12.6), and radial (−31.7% vs. −32.1) strain were similar between adults and children with FRDA. Circumferential strain differed slightly between adults (−26.1%) and children (−21.9%) with FRDA.
Forty-six percent of FRDA patients had no MR, 40% had trace MR, 4% had mild MR, 4% had mild-to-moderate MR, and 6% had no baseline Doppler recordings of MR. The severity of MR in FRDA patients did not change from baseline to 3 years.
Baseline versus Follow-Up
Comparison of paired baseline and 3-year follow-up echocardiographic parameters in the FRDA patients showed no significant changes in LV volumes, LVMI, RWT, or LVEF (data not shown). Unadjusted analysis showed increased septal thickness (P = 0.022), and a change in LV systolic sphericity (LV shape) (P = 0.012) (Table 1). However, when adjusted for age, sex, and BSA as fixed effects and time as random effect, only the decrease in systolic sphericity remained significant (P = 0.016). Use of idebenone (n = 19) was not associated with any change in LV size, geometry, or function at 3 years. Interestingly, when comparing the myocardial strain values in patients at follow-up to normal controls, there were significant differences not only in longitudinal strain (−11.93 vs. −16.07, P < 0.001) but also in circumferential strain (−24.76% vs. −27.38%, P = 0.02).
Relation of Echo Parameters and Genetic Severity
GAA repeat length (accounting for sex and age) was significantly associated with IVSd in FRDA (P = 0.008; R2 = 0.26). Sex also predicted IVSd (P = 0.01). When strain values in FRDA subjects were analyzed in a linear regression model using age, sex, and GAA repeat length, decreased longitudinal strain (P = 0.088) and radial strain (P = 0.083) trended to a longer GAA repeat length. This suggests that greater genetic severity might be associated with lower strain values. When we assessed relationships at follow-up, IVSd was not predicted by GAA repeat length (P = 0.26). However, longitudinal (P = 0.0014; R2 = 0.47), circumferential (P = 0.038; R2 = 0.36), and to a lesser extent radial strain (P = 0.086; R2 = 0.25) were associated with GAA repeat length. The changes over time show that with progression, abnormal longitudinal strain values become more indicative of genetic severity, while IVSd values are more variable.
This study demonstrates that average peak longitudinal strain is consistently and significantly decreased in FRDA patients compared with normal controls, even though LVEF is normal. This extends results from a smaller older cohort[11, 22] Abnormal strain, particularly longitudinal strain, appeared to be associated with greater genetic severity. After 3-year follow-up of FRDA, repeat echocardiographic analysis, adjusted for age, sex, and BSA showed no significant differences in LV volumes, LVMI, RWT, or strain from baseline, providing no cumulative evidence for substantive progressive LV remodeling over this time period. There were, however, some differences in both the longitudinal and circumferential strain in the FRDA at follow-up compared with normal subjects. These observations suggest that longitudinal strain changes early in FRDA and may be a useful marker for early myocardial involvement in FRDA. The difference in circumferential strain in unadjusted analyses and in follow-up compared with control suggests that it may also provide a marker of progression, albeit later and less robust.
Material Properties and Interstitial Fibrosis
The differences in myocardial strain in FRDA patients compared with controls presumably reflect genetically determined alterations in the properties of the myocardium perhaps due to accumulation of interstitial fibrosis. Interstitial fibrosis is ubiquitous in autopsied hearts from FRDA patients[2, 23, 24] and is associated with reduced myocardial shortening and decreased longitudinal strain. The significantly lower average peak longitudinal strain in FRDA could not be reconciled by differences in LV WT or LV geometry. These data suggest a systematic difference in properties of the myocardium in FRDA. Alternatively, FRDA is a dysregulation of myocardial bioenergetics that is associated with adaptive ventricular hypertrophy. We speculate that the abnormalities in strain identified here may reflect impaired mitochondrial bioenergetics in the heart. This is consistent with the improvement in strain observed with idebenone treatment in a smaller, more hypertrophic cohort. The precise reason for abnormal strain values in FRDA does not appear to reflect cardiac hypertrophy directly, as in this study only 40% of FRDA patients had LVH, and there was a relative paucity of FRDA patients with intermediate or severe LVH. Indeed, the highly variable nature of hypertrophy in FRDA was also shown by the varying relationship of GAA repeat length and IVSd at baseline and follow-up. The large proportion of FRDA patients with mild or absent LVH was concordant with a recent report that again demonstrates the difficulty in using hypertrophy as a measure of the severity of heart disease in FRDA.
Raman and colleagues have implicated microvascular ischemia due to myocardial hypo-perfusion as a cause for reduced myocardial shortening and abnormal strain with delayed gadolinium enhancement by CMR imaging. This mechanism would explain the greater effect of FRDA on longitudinal strain rather than radial or circumferential strain, because hypo-perfusion would first affect the subendocardial myocardium that is composed of longitudinally oriented myocardial fibers before involving the circumferential myocardial fibers that comprise the LV mid-wall. Reduction in longitudinal strain did not result in any temporal or spatial dyssynchrony of myocardial contraction either at baseline or at 3-year follow-up of our FRDA cohort. Thus, dyssynchrony is not the causal mechanism for the progressive deterioration in LV function that typifies FRDA.
Earliest Echocardiographic Changes
In our FRDA cohort, the most reproducible echocardiographic changes in the LV were hypertrophy and reduced average peak strain in the longitudinally oriented myocardial fibers. These echocardiographic findings were present before any changes occurred in either LV size or EF. We hypothesize that these changes characterize the onset of LV dysfunction in FRDA and that adaptive LV hypertrophy increases slowly over time to involve first the subendocardial longitudinally oriented LV muscle bundles and then the circumferential/spirally arranged myocardial muscle bundles. Later in the course of the disease, we speculate that there is involvement of circumferential fibers with progressive reduction in circumferential and radial strains in a pattern that is specific to the genetic defect in FRDA. This hypothesis has also been proposed by others using tissue Doppler imaging.
Three Year Follow-Up
Direct comparison of echocardiographic parameters at baseline and 3-year follow up in FRDA using unadjusted analysis showed a significant increase in septal thickness and change in LV cavity shape. However, after adjusting for age, sex, and BSA, the only remaining significant difference at follow-up was a minor change in LV shape. Similarly, average peak longitudinal, circumferential, and radial strain values at 3 years were not different from baseline values although both longitudinal and circumferential strain remained significantly different in the FRDA patients at 3 years as compared with controls. This change relative to controls, along with the greater association of strain data with genetic severity suggest that even though strain values are abnormal early in FRDA, there appears to be little change over time, suggesting a slow course to ventricular remodeling. However, the impact of abnormal longitudinal strain on prognosis and the precise time course of these changes over time will need to be delineated through longer term prospective observational and intervention studies.
Although the study design was retrospective, the study population was well characterized genetically, morphologically, and functionally with detailed quantitative measures of cardiac function. In our retrospective approach, longitudinal strain values could not be readily obtained in some FRDA patients, particularly younger subjects. This could reflect technical difficulty in obtaining high quality apical images in such subjects due to their thin chest wall. As individuals with earlier onset disease progress more rapidly, this might explain a relative lack of change in longitudinal strain. Analyzed images were archived at a frame rate of 30 frames per second, which, although acceptable, may result in loss of data. The follow-up of 3 years for this genetically determined chronic disease involving the heart was too short to identify major changes in LV geometry or myocardial strain from baseline that either predicted or provided insight into clinical outcome. However, these findings do provide insight into the natural history of the time course of remodeling and demonstrate the slow evolution of heart disease in FRDA.
The cohort of FRDA subjects was modest in size, reflecting the low prevalence of this disease. There was heterogeneity in the age of our cohort, and we were unable to match the group of FRDA patients with normal controls according to age, gender, or body size. However, we did adjust for age, BSA, and gender in our multivariable analyses to account for the potential confounding effects by these measures. In other studies, control values of longitudinal strain in children are comparable to those obtained in the controls of our study, and cross sectional data suggest that values for parameters such as strain rate reach adult values by adolescence. The group below adolescence was also less likely to have measureable strain values in our retrospective analysis. For these reasons, it is unlikely that differences in age between controls and FRDA subjects contribute significantly to our results on longitudinal strain. Still prospective analysis of strain including the youngest FRDA patients would be useful to understand the earliest aspects of strain measurement in this disease.
There is wide variation in ventricular morphology and function from normal to a small proportion of FRDA patients with massive hypertrophy that may be concentric or asymmetric. LVH was present in only 40% of FRDA patients when the mean age was 18 years. To our knowledge, this is the first study of a moderately sized cohort of diverse FRDA patients in whom myocardial strains were measured in 3 planes at baseline and again at 3-year follow-up. In our young cohort of FRDA patients, there was consistent and significant reduction in longitudinal strain that was evident before traditional measures of LV function became abnormal, indicating that longitudinal strain may be a marker for early LV dysfunction in FRDA. Changes in longitudinal strain in our FRDA cohort could not be explained by the degree of LVH alone. This may relate to dysregulation of bioenergetics caused by reduced FXN that is genetically determined and specific for FRDA. GAA repeat length predicts genetic severity and is associated with decreased longitudinal and circumferential strains. Long-term follow-up studies are necessary to define the time-dependent changes in cardiac function and identify triggers for heart failure and sudden death.
We thank Justine Shults for performing statistical calculations and aiding in the design of the statistical plan.
- 14A report from the American Society of Echocardiography's Nomenclature and Standards Committee and the Task Force on Chamber Quantification, developed in conjunction with the American College of Cardiology Echocardiography Committee, the American Heart Association, and the European Association of Echocardiography, a branch of the European Society of Cardiology. J Am Soc Echocardiogr 2005;18:1440–1463., , , et al:
- 18Sustained reverse left ventricular structural remodeling with cardiac resynchronization at one year is a function of etiology: Quantitative Doppler echocardiographic evidence from the Multicenter InSync Randomized Clinical Evaluation (MIRACLE). Circulation 2006: 113:266–272., , , et al:
- 22Marked variation in the cardiomyopathy associated with Friedreich's ataxia. J Am Soc Echocardiogr 2011;24:782–789., , , et al: