Regional Variability in Longitudinal Strain Across Vendors in Patients With Cardiomyopathy Due to Increased Left Ventricular Wall Thickness

Introduction

See Editorial by Delgado and Ajmone Marsan

Echocardiographic analysis of left ventricular longitudinal strain (LS) has emerged as a sensitive and specific diagnostic and prognostic tool in various cardiomyopathies.^1–3 Several software programs exist to calculate strain values; some are vendor-specific while others are independent of the ultrasound vendor that was used to acquire the images. LS values can be reported by segment, by cross-sectional region (basal, mid, or apical), and by global across-segment average values (GLS). Ideally, LS values among software vendors should be consistent in all 3 of these domains. There has been an attempt to standardize GLS values among software manufacturers⁴ with moderate success.^5–7

Regional variations in LS patterns in disorders, such as cardiac amyloidosis and hypertrophic cardiomyopathy (HCM), have been described by our center and others, allowing for greater diagnostic accuracy in evaluating echocardiographic images in these patients.⁸ Between-vendor agreement of segmental and region-based strains is less clear and likely lower than agreement between global values.^9,10 Furthermore, segmental or regional strains have not been tested in specific cardiomyopathies with characteristic regional strain variation. This is especially important in multi-vendor echocardiographic laboratories where there is an obvious question related to the interchangeability of reconstructed images obtained on different ultrasound platforms. For this purpose, we sought to compare LS values between-vendors in specific cardiomyopathies that exhibit regional strain variation: cardiac amyloidosis, septal HCM, and apical HCM.

Methods

Study Sample

This was a prospective, cross-sectional study of patients at the Cleveland Clinic who underwent clinically indicated outpatient echocardiographic examinations between June 2016 and June 2017. To assess the regional differences in echocardiographic LS among vendors, patients were recruited with 3 common phenotypes of cardiomyopathy that have been previously described to have regional variation in LS values (amyloidosis, septal HCM, and apical HCM). Patients in the cardiac amyloidosis cohort carried this diagnosis based on endomyocardial biopsy or a noninvasive diagnosis centered around cardiac imaging. Patients without endomyocardial confirmation of immunoglobulin light chain (AL) amyloidosis had an extra-cardiac biopsy consistent with the disease as well as typical echocardiographic and cardiac magnetic resonance imaging features.¹¹ Patients without endomyocardial biopsy confirmation of transthyretin (ATTR) amyloidosis had technetium pyrophosphate nuclear scintigraphy¹² and cardiac magnetic resonance imaging features consistent with the disease. Patients in the HCM cohort were diagnosed based on history, examination, electrocardiography, and other multimodality imaging using the latest societal guidelines to diagnose HCM.¹³ Septal and apical variants of HCM define the region of maximal hypertrophy which has been shown to display differing regional variations in LS. Patients with apical aneurysms were excluded. A control group of patients was recruited without valvular, structural, or coronary artery disease and normal systolic and diastolic function. The study was approved by the institutional review board as well as the ethics committee, and all patients agreed to informed consent.

Measurement Techniques

A complete echocardiographic study was first performed on commercially available Vivid 7 or Vivid 9 ultrasound systems (GE Medical, Milwaukee). Septal thickness, LV end diastolic diameter, and posterior wall thickness were assessed by 2-dimensional measurements at the level of the mitral valve leaflet tips in diastole. Chamber volumes and LV ejection fraction were measured using Simpson’s method. Parameters of diastolic function were collected in the apical 4 chamber view using pulsed wave and tissue Doppler. All measurements were in accordance with the American Society of Echocardiography guidelines.¹⁴

In the same session, images were acquired for LS analysis sequentially using the aforementioned GE, Siemens (Acuson SC2000; Siemens Medical Solution USA, Inc, Malvern, Pennsylvania) and Philips (EPIQ 7C; Philips Medical Systems, Bothell, Washington) machines. Ultrasound probe location, image and zoom, and frame rate were consistent among the 3 sets of images. Zoomed images were acquired in the apical 4-, 2-, and 3-chamber views. Patients with poor overall image quality were excluded by the performing sonographer. Inadequate image quality was defined as a frame rate <40 frames per second or an inability to visualize or perform adequate tracking on >2 myocardial segments. Three studies were later excluded after image acquisition due to inability to generate reliable strain curves as deemed by the reviewing physician; all 3 studies were excluded due to QLab software. Peak systolic LS assessment was performed using EchoPAC (EchoPAC Version 113, Advanced Analysis Technologies, GE Medical Systems) on GE machines, Velocity Vector Imaging (Version 2.0, Siemens Medical Solutions) on Siemens machines, and QLab (version 10.2, Philips Healthcare) on Philips machines. Frame by frame tissue speckle tracking of the LV endocardium was performed, and endocardial borders were readjusted manually until satisfactory tracking was achieved. EchoPAC and QLab software also allow for visualization of the region of interest with respect to mid-myocardial and epicardial borders. These were similarly adjusted until optimal tracking was achieved. LS values were calculated by an experienced reader (Dr Sato) for each of the 3 software packages, blinded to the results of the same patient’s data using the other packages by using anonymized and randomized images. Segmental strain was calculated in each of the 18 myocardial segments. It is important to note that EchoPAC calculates LS using an 18-segment model (which was used for this analysis), but uses a 17-segment model to display LSs graphically. Regional strain was defined as the average LS values in each myocardial cross-sectional region (basal, mid, and apical). Relative regional strain ratio was defined as the average apical strains divided by the sum of the average basal and mid strains.² Inter- and intraobserver and test-retest variability are low in our lab as previously reported.^1,15

Statistical Analysis

The authors will not make the data from this article publicly available, though the methods and materials used are included in this article. Continuous variables are expressed as a mean±SD, and categorical data are presented using percentage and frequencies. Differences between groups were compared using one-way ANOVA for continuous variables (Mann-Whitney U test for variables with non-normal distribution) and the Fisher exact test for categorical variables. Pearson correlation coefficients were determined to assess correlations between LS values. Correlations were made with EchoPAC as the standard for comparison. To assess how well the naturally occurring strain gradient between different left ventricular regions (ie, basal, mid, or apical) is captured by different software packages, a linear mixed model analysis was used. LS values were treated as the dependent variable with myocardial region as a covariate, diagnosis and software type as fixed effects, and patients as random effects. Finally, receiver operating curves were fit to model the ability of strain values (stratified by myocardial region) to diagnose cardiac amyloidosis based on the different software packages using area under the curve. Data assembly and statistical analyses were performed with Stata (version 13, College Station, Texas), and SPSS 21.0 software (SPSS Inc, Chicago, IL). All statistical tests were 2-sided, and P values <0.05 were considered statistically significant.

Results

The study cohort included 69 patients: 18 patients with amyloidosis (9 with AL and 9 with ATTR), 30 patients with septal HCM, 6 patients with apical HCM, and 15 control patients. Demographic and standard echocardiographic data are included in Table 1. Patients with amyloidosis were more likely to be older with lower ejection fractions, higher LV mass index, lower average mitral valve e`, and higher E/e` values. Patients with HCM had a predominantly thickened anteroseptum as compared with the posterior LV wall, while those with apical variant HCM did not have significant thickening in either the basal anteroseptum or posterior walls.

Table 2 details LS values in the cohort by software package. Overall, GLS was highly correlated between EchoPAC and velocity vector imaging (VVI; r=0.874, P<0.001), EchoPAC and QLab (r=0.853, P<0.001), and VVI and QLab (r=0.802, P<0.001). However, QLab tended to report higher GLS values as compared with EchoPAC and VVI (Figure 1). This is also noted on Bland-Altman plots, which show the mean differences among software packages and the trends in mean difference in normal versus abnormal strain values (Figure 2). All patients with cardiomyopathy had more abnormal strain values as compared with controls, with patients with amyloidosis having the most abnormal values.

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

When assessing regional differences in strain, patients with amyloidosis demonstrated a gradient of improving LS values progressing from base to apex, while those with apical HCM had the opposite pattern. This is represented by higher relative regional strain ratio values in the amyloidosis cohort and lower in the apical HCM cohort (Table 2, Figure 3). With respect to vendor software, VVI was comparable to EchoPAC after controlling for myocardial regions and diagnosis in a mixed model (β, −0.25; 95% CI, −1.14 to 0.64; P=0.413; Figure 4A). However, QLab showed significantly higher strain values as compared with EchoPAC (β, −3.81; 95% CI, −4.71 to −2.91; P<0.001) which was predominantly driven by higher strain values in apical segments (β, −6.18; 95% CI, −6.93 to −5.43; P<0.001; Figure 4B). In fact, patients with apical variant HCM had normal apical strain values when measured by QLab (mean −21.8%). Representative examples of echocardiography with LS in amyloidosis (Figure 5A), septal HCM (Figure 5B), and apical HCM (Figure 5C) are depicted.

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

Receiver operating curves were then fit to model the effect of software type on the association between regional strain values and the diagnosis of amyloidosis. The AUC was significantly higher when using EchoPAC software (0.932) as compared with VVI (0.883) or QLab (0.814, P<0.001; Figure 6).

• Download figure
• Download PowerPoint

Discussion

In this study, we compare LS values derived from 3 different strain software packages in patients with cardiomyopathies leading to an increase in LV wall thickness. Two dimensional echocardiographic and strain parameters of patients with amyloidosis, HCM, and apical HCM were characteristic of the respective diseases. GLS was highly correlated among the 3 software packages and most abnormal in patients with amyloidosis. QLab software tended to produce more negative strain values, driven by overestimation of apical strains, and failed to identify the typical pattern of apical HCM. EchoPAC had the greatest ability to discriminate patients with amyloidosis using regional strain values as compared with VVI and QLab. These results have clinical implications when assessing LS in patients with cardiomyopathies that produce regional strain abnormalities.

The assessment of regional strains in cardiomyopathies with increased LV wall thickness is important and may lead to a dramatic change in diagnosis, prognosis, and treatment. The initial studies using tissue Doppler-derived strain in amyloidosis demonstrated impaired basal peak systolic LS and strain rate,¹⁶ which was associated with cardiac and all-cause death.¹⁷ Current generation LS using speckle tracking provides an angle-independent strain analysis which is more reproducible and accurate at detecting abnormal motion. Our group first described a regional gradient of impaired strains in basal and mid segments using speckle tracking LS with EchoPAC software in amyloidosis^1,8 which was also associated with prognosis.² EchoPAC has become the standard of comparison to assess regional LS variation in this cardiomyopathy. Speckle tracking LS in septal HCM was also initially described with EchoPAC¹⁸ and has been shown to be associated with adverse events with both EchoPAC^19,20 and VVI³ software. In apical HCM, speckle tracking LS demonstrates a characteristic impairment in apical strains as compared with hypertensive controls using EchoPAC.²¹

Ideal LS software would be highly accurate and reproducible for global, regional, and segmental values. Additionally, ideally there would be a high correlation among software vendors in all 3 of these domains. Each vendor uses proprietary software to produce LS curves. Attempts have been made, with the more recent software versions, to standardize GLS values among software manufacturers and reduce intervendor variability of LS measurements based on global task force guidelines.⁴ Studies have examined intervendor variability in 2D LS measurements^5,7,22 showing general agreement between-vendor–specific and vendor-independent software packages. However, there is enough variability in GLS with limits of agreement of 3% to 4.5% that the authors suggest using the same ultrasound machine and software when measuring GLS in serial and cross-sectional studies. Another group found that GLS between EchoPAC and TomTec demonstrated good reproducibility, while reproducibility was moderate to poor for circumferential and radial strains.⁶ Our group has previously demonstrated more negative GLS values with EchoPAC as compared with VVI on the same images in normal subjects with similar interobserver variability.⁹ Of note, older software versions were used in the aforementioned studies.

A pair of very detailed investigations from the European Association of Cardiovascular Imaging and the American Society of Echocardiography attempted to elucidate the between-vendor differences in global and segmental LS strain values.^10,23 While GLS correlates moderately among vendors as detailed above, segmental and regional strain values are less reliable. A reproducibility analysis done in normal control patients demonstrated significant differences in global and regional longitudinal, circumferential, and radial strains among 3 vendor softwares.²² It is well known that segmental strains are less consistent among vendors, and global LS compensates for this due to the aggregation of all segmental strains. The correlation between regional strain (ie, basal, mid, and apical) is likely somewhere in between the agreement among global and segmental values. Based on our current data, there appears to be agreement with the vendor-independent VVI software and EchoPAC, but a systematic bias towards higher (more negative) LS values when using QLab software, particularly in the apical region. Of note, QLab software was not studied in the EACVI/ASE investigation of segmental strains.

It is important to be aware of the correlation in regional LS values among vendors for several reasons. First, cardiomyopathies with increased LV wall thickness demonstrate characteristic LS patterns. These strain patterns provide diagnostic information; for example, cardiac amyloidosis has an apical sparing pattern while apical HCM has a regional impairment in apical strains. Second, regional strain has been found to be prognostic in conditions such as amyloidosis.^2,17 Lastly, cutoffs of global or regional LS used for diagnosis and prognosis in these diseases are not likely to be interchangeable among software vendors. We agree with published guidelines that the same vendor machine and software should be used when comparing serial global or regional strains. Additionally, new vendors should strive for accuracy and reproducibility of global, regional, and segmental LS and standardization with other vendors.

Limitations

This study used a limited sample size, particularly in the apical variant HCM cohort. Additionally, not all patients had a pathological diagnosis of their respective cardiomyopathy. However, patients were diagnosed based on guideline documents and had echocardiographic exams consistent with the disease. The impact of region of interest size on LS values was not tested in this analysis, which may be important in patients with left ventricular hypertrophy. Future versions of these LS softwares with an altered strain algorithm may change the correlation among vendors.

Conclusions

Speckle tracking LS values derived from 3 different strain software packages (EchoPAC, VVI, and QLab) were compared in patients with cardiomyopathies with increased LV wall thickness: amyloidosis, septal HCM, and apical HCM. GLS was highly correlated among the 3 software packages and most abnormal in patients with amyloidosis. EchoPAC had the greatest ability to discriminate patients with amyloidosis using regional strain values as compared with VVI and QLab. QLab software tended to produce more negative strain values, driven by overestimation of apical strains and failed to identify the typical pattern of apical HCM. These results have clinical implications when assessing LS in patients with cardiomyopathies producing regional strain abnormalities. Serial studies examining regional LS differences should be performed with the same vendor machine and software package.

Footnotes