Left atrial strain by vendor-neutral echocardiography software in healthy subjects: vendor comparisons and associated factors

Abstract

Aims

Left atrial strains [reservoir (LASr), conduit (LAScd), and contractile (LASct)] are increasingly utilized, primarily in the areas of heart failure and valvulopathy. Commercially available software to measure strain are marketed as imaging machine vendor independent, although the normal ranges of their performance and external validation studies are lacking. We evaluated and compared LAS values, reference ranges, and associated factors using contemporary strain software.

Methods and results

Healthy subjects (n = 100) undergoing transthoracic echocardiography during January to April 2023 were studied, with equal number by age groups, sex, and GE vs. Philips equipment. LASs were quantified using TomTec version 51.02 (AutoStrain LA), EchoPAC version 206 (AFILA), velocity vector imaging (VVI) version 2.00, and Epsilon version 5.0.2.11295 software. Means and lower limits of normal [95% confidence interval (CI)] of LASr (%) were 44.1 (41.9, 46.3) and 30.3 (26.4, 34.1) for TomTec; 34.8 (33.6, 36.0) and 26.3 (24.2, 28.4) for EchoPAC (on GE scans only); 42.5 (40.2, 44.8) and 28.4 (24.4, 32.4) for VVI; and 37.0 (34.9, 39.1) and 25.2 (21.5, 28.8) for Epsilon. Factors significantly associated with variability in LASr measurements and their beta-coefficients (95% CI) were age −2.37 (−3.41, −1.33) per 10 years, EchoPAC (vs. TomTec) −9.63 (−12.75, −6.51), and Epsilon (vs. TomTec) −7.04 (−9.40, −4.67). Reference ranges and factors significantly associated with LAScd and LASct were reported. LAS measurements and normal ranges by strain software and associated clinical and echocardiographic factors were determined.

Conclusion

EchoPAC and Epsilon have lower magnitude mean LAS values compared with TomTec and VVI, and all software were vendor neutral except EchoPAC.

Graphical Abstract

Left atrial strain measurements: (A) standard strain curve; measured by four contemporary strain software (B) TomTec, (C) EchoPAC, (D) VVI, and (E) Epsilon; lower table shows means and lower limits of normal with 95% CIs of left atrial reservoir, conduit, and contractile strains in healthy patients. LAScd, left atrial conduit strain; LASct, left atrial contractile strain; LASr, left atrial reservoir strain; LLN, lower limits of normal.

Open in new tabDownload slide

Introduction

Left atrial assessment is an important component of the transthoracic echocardiography (TTE) evaluation of diastolic function and left ventricular filling pressure, with left atrial volume index (LAVi) to body surface area being part of the algorithms endorsed by current guidelines.^1–3 Recent studies have demonstrated the additive utility of left atrial strain (LAS) in this setting, leading to its incorporation into the recent European consensus statement for non-invasive diastology evaluation.^4,5 The prognostic roles of LAS have also extended to other cardiovascular conditions such as heart failure, ischaemic heart disease, valvular heart disease, and risk factors such as hypertension and diabetes.^6–10 For example, in heart failure, LAS has shown to be prognostic for all-cause mortality and heart failure hospitalizations, independent of left atrial volume and left ventricular systolic function.⁶ Post-myocardial infarction, LAS predicted the development of heart failure and atrial fibrillation, more reliably than traditional geometric parameters of the LA.⁸ Furthermore, another study of aortic stenosis patients found LAS to be independently associated with deaths and heart failure hospitalizations.⁹

LAS is typically measured using 2D speckle-tracking echocardiography and includes three components of reservoir (LASr), conduit (LAScd), and contractile (LASct) strain based on the cardiac cycle.¹¹ Despite the increasing clinical utilization, the reference ranges of each LAS component are not well established, with significant heterogeneity across studies in a few meta-analyses to date.^12,13 Furthermore, recent advances in strain software have allowed many to become applicable to different scanners (i.e. vendor neutral), although external validation studies are lacking. This study aims to evaluate LAS using contemporary strain software in healthy patients to determine reference ranges and identify factors that influence these measurements.

Methods

Study population

We performed a cross-sectional analysis of 100 healthy adult subjects ≥18 years old undergoing clinically indicated comprehensive TTE from January to April 2023 at our institution. The inclusion criteria are 50 male and 50 female subjects, 20 patients each in the age groups 18–29, 30–39, 40–49, 50–59, and 60+ years, and 50 scanned with GE Vivid 7 or E9 (GE Medical, Milwaukee, WI, USA) and 50 scanned with Philips EPIQ 7C (Philips Medical Systems, Bothell, WA, USA) echocardiography machines, in order to achieve equal split of patients for each category by sex, age group, and scanner. Patients were also required to have adequate windows of the atrium by apical four-chamber and apical two-chamber views as adjudicated by two independent echocardiography readers. The exclusion criteria to arrive at 100 healthy subjects include history of any cardiovascular diseases (including coronary heart disease, heart failure/cardiomyopathy, valvular heart disease at least moderate in severity, arrhythmias, congenital heart disease, pericardial diseases, cardiac medications, surgery, interventions, and devices), hypertension, diabetes, stroke, chronic lung, kidney, liver, vascular and inflammatory diseases, malignancies, chemotherapy, and radiation therapy. Relevant patient clinical characteristics were recorded. The study was approved by our Institutional Review Board (IRB 23-207) with patient consent waiver. The data underlying this article will be shared on reasonable request to the corresponding author.

Echocardiography

TTE scans were acquired on GE Vivid 7 or E9 (GE Medical, Milwaukee, WI, USA) and Philips EPIQ 7C (Philips Medical Systems, Bothell, WA, USA) machines. The apical four-chamber and apical two-chamber views were utilized for LA evaluation from the complete TTE study, ensuring LA was not foreshortened and endocardial definition could be visualized throughout the entire cardiac cycle, adjudicated independently by staff cardiologists T.K.M.W. and Z.B.P. Standard left ventricular and left atrial indexed volumes and function parameters were measured. LAS was analysed for all patients utilizing strain vendors TomTec version 51.02 (AutoStrain LA), EchoPAC version 206 (AFI-LA), velocity vector imaging (VVI) 2.00, and Epsilon version 5.0.2.11295 software to calculate LASr, LAScd, and LASct (Graphical Abstract). Comparisons of LAS are described as lower or higher in magnitude indicating lower or higher absolute values, respectively, regardless of positive or negative strain. Twenty randomly chosen patients had repeated LAS measured on all four strain vendors by the same author (T.D.) to assess intra-reader variability and by other authors (A.A., A.D.A., J.E.D., and M.M.) to assess inter-reader variability, and test–retest variability was also assessed, using standard error (SE) of the mean (SEM) technique.

Statistical analyses

Mean ± standard deviation and frequency (percentage) were used to report baseline continuous and categorical variables, respectively. LAS means, SEs, and 95% confidence intervals (CIs) were calculated for the overall cohort by each strain vendor software, along with each age group, sex, and TTE machine categories. The lower limit of normal (LLN) for LAS was defined as the 2.5th percentile strain value 1.96 SD from the mean in the 100-patient cohort at the lower end of magnitude of strain values, for example less positive for LASr strain and less negative for LAScd and LASct strains. The SE for the LLN is calculated using the previously reported formula SE_LLN = √(SD_mean² × [1/n + 2/(n − 1)],¹⁴ which is then used to derive the 95% CI of the LLN. Intra- and inter-reader variabilities were evaluated using SE of the means for each LAS component and strain software. Linear mixed model multivariable regression longitudinal analyses were used to identify clinical and echocardiographic factors associated with LASr, LAScd, and LASct measurements. SPSS (version 24, IBM, Chicago, IL, USA) and Prism (version 8, GraphPad, San Diego, CA, USA) software were used for statistical analyses and graphing. P < 0.05 was deemed statistically significant, and all tests were two tailed.

Results

Clinical and TTE characteristics of this 100 healthy subjects’ cohort are listed in Table 1. Mean age is 45 ± 15 years, 50 (50%) were female, and 83 (83%) were white. Main physiological parameters include body surface area 1.92 ± 0.26 m², systolic blood pressure 122 ± 17 mmHg, and heart rate 70 ± 13 bpm. Mean TTE left heart parameters include left ventricular end-diastolic, end-systolic, and stroke volume indexed of 50 ± 11, 19 ± 6, and 31 ± 7 mL/m², respectively. Mean left ventricular ejection fraction was 62 ± 5% and mean LAVi 24 ± 7 mL/m².

Table 2 presents the means and LLNs of LAS by strain software and their intra- and inter-reader SEMs. EchoPAC could only accurately measure LAS on GE (n = 50) and not Philips scans due to technical incompatibility. LASr (%) means and LLNs (95% CIs) were 44.1 (41.9, 46.3) and 30.3 (26.4, 34.1) for TomTec; 34.8 (33.6, 36.0) and 26.3 (24.2, 28.4) for EchoPAC; 42.5 (40.2, 44.8) and 28.4 (24.4, 32.4) for VVI; and 37.0 (34.9, 39.1) and 25.2 (21.5, 28.8) for Epsilon.

Figure 1 illustrates the Bland–Altman plots of pairwise vendor software comparisons of LASr for TomTec, EchoPAC, VVI, and Epsilon software. The magnitude of LAS measurements appears to be lower for EchoPAC and Epsilon compared with TomTec and VVI. Mean LAS parameters by subgroups of age, sex, and scanner are listed in Table 3. The magnitude of LASr and LAScd appears to reduce with older age groups and higher for females than males except for Epsilon.

Results of the linear mixed model multivariable regression longitudinal analyses are presented in Table 4. Age (per 10-year increase), EchoPAC (vs. TomTec), and Epsilon (vs. TomTec) were independently associated with lower magnitude of all three components of LAS. LAVi (per 10 mL/m² increase) was independently associated with lower magnitude of LASct only, with beta-coefficient (95% CIs) of 1.19 (0.10, 2.29). Sex, body mass index, heart rate, systolic blood pressure, frame rate, Philips vs. GE scanners, and VVI vs. TomTec strain software were not significantly associated with variabilities in LAS measurements.

Discussion

This cross-sectional study evaluating LAS across four strain software in healthy subjects has several key findings. Firstly, we determined normal means and LLNs with their 95% CIs for LASr, LAScd, and LASct for each of the four commercially available strain software in our cohort of 100 healthy subjects that require external validation. Secondly, EchoPAC and Epsilon had lower magnitude mean LAS than the TomTec and VVI. Thirdly, EchoPAC was also unable to measure LAS on Philips scans because of technical incompatibility, while the other software were vendor independent. Furthermore, normal values for LAS by age, sex, and scanner categories were also determined. Lastly, clinical and TTE parameters significantly associated with each LAS component were also identified, with strain software and age being associated with all three.

The clinical roles of LAS have been mainly focused on the non-invasive evaluation of diastolic heart failure by echocardiography, with recent studies and guidelines advocating a threshold of LASr < 18% to indicate elevated filling pressure especially when diastolic function is otherwise indeterminate.^4,5 In healthy subjects, however, the normal ranges of all LAS components are not well established. A prior meta-analysis of 40 studies and 2542 healthy subjects undergoing 2D speckle-tracking TTE (mostly with EchoPAC software) reported pooled means (95% CIs) of 39.4% (38.0%, 40.8%) for LASr, −23.0% (−20.7%, −25.2%) for LAScd, and −17.4% (−16.0%, −19.0%) for LASct, similar to what we reported; however, LLNs (95%CIs) were not reported.¹² Another meta-analysis focusing on 3D speckle-tracking TTE of five studies and 316 healthy subjects (most with Toshiba software) reported LASr of 27.5% (25.2–29.8%), indicating significantly lower magnitude than that reported in our study, and again this study did not report LLN.¹³ On the other hand, the World Alliance Societies of Echocardiography study reported not only mean ± SD of 2 and 3D LAS using TomTec but also the LLN-upper limit of normal, separate for both sexes at age groups of 18–40, 41–65, and >65 years. LLNs reported were 21–29% for LASr, −19 to −9% for LAScd, and −9 to −2% for LASct, similar to our findings.¹⁵

Another cohort study from 22 European hospitals and 371 healthy subjects and TomTec software reported also reported LLNs ± SEs of 26.1 ± 0.7% for LASr, −12.0 ± 0.5% for LAScd, and −7.7 ± 0.3% for LASct, also similar to our findings.¹⁶ Further studies and meta-analyses focusing on the LLNs for LAS are necessary to validate these reference ranges separately for different strain software and evaluate their prognostic implications.

Our study adds to the current literature to both assist in standardization of normal ranges determination of LAS measurements and emphasize inter-vendor variability even for vendor-neutral strain software. Our reported LLNs and their 95% CIs in Table 2 can be interpreted as dividing LAS values to being normal, abnormal, or borderline. We propose that if the LAS measurement lies within the 95% CI of the LLN, there is uncertainty if the value is normal or abnormal and therefore should be classified as borderline. If the LAS measurement is higher in magnitude than the upper limit of the 95% CI for the LLN, it should be considered normal, while it should be considered abnormal if it is lower in magnitude than the lower limit of the 95% CI.^17,18 For example, using TomTec LASr, values of 24.2–28.4% would be borderline, >28.4% would be normal, and <24.2% would be abnormal. Having these thresholds for different LAS parameters and different software especially allows for LAS value categorization for routine clinical use and decision-making, although external validation and prognostic studies remain necessary to broaden clinical applications. It is important to note that currently EchoPAC strain software cannot determine LAS measurements on Philips scanners, unlike the other three software that can be used on both GE and Philips scanners, although this may be improved in the future.

It is important to identify factors significantly associated with LAS measurements, as this may change how the LAS value measured should be interpreted and compared. The most important factor affecting LAS measurements is strain software, where there are significant inter-vendor differences in LASr, LAScd, and LASct values and LLNs, including between the two most commonly used software clinically TomTec and EchoPAC, whereas scanner vendor did not affect strain values by TomTec, Epsilon, and VVI. Therefore, we advocate implementing separate reference ranges by strain software and using the same software for serial monitoring in the same patient for more accurate comparisons. We also found age to be negatively associated with the magnitude of LASr and LAScd but positively associated with the magnitude of LASct and LAVi being only negatively associated with LASct magnitude. The World Alliance Societies of Echocardiography study similarly found older age to be associated with lower magnitude of LASr and LAScd but increase in magnitude of LASct reflecting changes in the proportions of passive and active filling during diastole with aging similar to our findings, while sex differences in LAS were relatively small; nevertheless, age- and sex-specific reference ranges for LAS were reported.¹⁵ The meta-regression of a prior meta-analysis also reported heart rate to be positively associated with LASr, which is physiologically plausible, while body surface area was negatively associated with LASr.¹² However, these associations were not identified in our study or the World Alliance Societies of Echocardiography study, likely related to differences in patient demographics, strain software and versions, and factors adjusted in multivariable regression.

Our study has several limitations. It is a single-centre observational cohort study that has intrinsic biases, albeit prospectively studied. It was performed at a tertiary referral cardiology centre in Cleveland, OH, USA, so results may not be applicable to populations with varying ethnicities, especially outside the USA. Study sample size of 100 healthy subjects meant modest power including for subgroup and regression analyses. Each strain software has real-world limitations regarding measurement errors, systematic biases, and imperfect intra- and inter-reader variability reflected by the ‘ranges of normality’, which may limit clinical utility especially for VVI software being the least robust with the highest SEMs; however, inter-vendor differences in LAS values are still important and not fully accounted by measurement variability. Furthermore, EchoPAC software was able to analyse GE scans but could not analyse Philips scans due to technical incompatibility. Test–retest variability was unfortunately not tested. Lastly, we do not focus on outcomes during follow-up to be able to investigate the prognostic utility of the reference ranges we identified.

In conclusion, we determined the means and LLN with their 95% CIs of LASr, LAScd, and LASct in healthy subjects for all four contemporary strain software, enabling categorization into normal, borderline and abnormal LAS. All software except for EchoPAC demonstrated vendor-neutral property with adequate reproducibility. Factors significantly associated with each component of LAS were identified, most importantly age and strain software, which affected all three LAS components. Our findings require external evaluation in both healthy subjects and with cardiovascular conditions to determine their clinical applications and prognostic value.

Clinical perspectives

1. LAS measurements in healthy individuals vary depending on vendor-neutral software and clinical factors including age.

2. Depending on vendor and age, values for LAS should be categorized as normal, borderline, and abnormal as defined by the 95% LLN.

3. Future studies should validate our cut-offs and study the prognosis of those with borderline and abnormal strain values as compared with their normal counterparts.

Consent

Ethical approval was obtained prior to the commencement of the study with a patient consent waiver.

Funding

None declared.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.

Lead author biography

Tiffany Dong is an advanced cardiac imaging fellow at Cleveland Clinic Foundation. Her interests include structural imaging as well as nuclear, computed tomography, and cardiac magnetic resonance.

References

Author notes