Defining Normal Reference Ranges of Left Ventricular Global Longitudinal Strain Among Different Vendors: Are We There Yet?

Strain and strain rate imaging have evolved significantly since their introduction two decades ago, providing an objective means to quantify both regional and global myocardial function [1]. Initially developed using tissue Doppler strain, strain imaging has largely transitioned to speckle-tracking echocardiography (STE), a technique that is angle-independent and offers improved feasibility and reproducibility. Left ventricular global longitudinal strain (LVGLS) has become an established measure of left ventricular (LV) systolic function with numerous clinical applications. However, despite its promise, LVGLS has not yet been widely adopted in real-world clinical practice. One major barrier is the variability across different echocardiographic vendors, which introduces inconsistency in measurements and complicates standardization efforts. Traditionally, left ventricular ejection fraction (LVEF) has been the most commonly used parameter for assessing LV systolic function and predicting outcomes. However, despite its widespread use, LVEF has notable limitations, including significant intra- and inter-observer variability, dependence on geometric assumptions, and susceptibility to technical artifacts such as endocardial dropout and foreshortening of the apex. Similar challenges exist for LVGLS, where accurate endocardial contour delineation and motion tracking are crucial for reliable assessment. Reproducibility of LVGLS using different vendors’ algorithm software remains controversial and has been described as acceptable by some [2] or shown to have significant discrepancies by others [3-5]. Advancements in STE, including machine learning-driven algorithms, have improved strain analysis. However, a key limitation of strain imaging is that measurements remain dependent on proprietary software algorithms, leading to variability between vendors. Moreover, despite abundant literature, defining universal reference ranges remains a challenge, as different software platforms yield slightly different results. A proposed solution has been the use of intra-institutional vendor-independent software capable of obtaining off line strain measurements. These vendor-independent strain software packages have evolved but have not undergone rigorous external validation studies. The study by Arockiam et al. [6] addresses this gap by evaluating normal reference ranges for LVGLS using contemporary, vendor-neutral strain software in a cohort of healthy individuals. The authors examined 100 subjects across different age groups, using echocardiographic scans from General Electric (GE) and Philips systems. Four strain software packages—TomTec, EchoPAC, VVI, and Epsilon—were used to quantify LVGLS, enabling direct comparative and regression analyses. Their findings indicate that mean LVGLS values varied slightly across the four software packages, with EchoPAC producing the most negative strain values and VVI yielding the least negative values. Regression analyses identified sex, heart rate, LVEF, and the choice of strain software (particularly EchoPAC and VVI vs. TomTec) as significant contributors to LVGLS measurement variability. Despite differences across the four software platforms, the study provides reference ranges that may aid in defining normal, abnormal, and borderline LVGLS values for clinical application. Based on these results, the authors conclude that LVGLS measurements were feasible across all four strain software on both GE and Philips scans in this study. While acknowledging differences among the four strain software packages, by reporting LLNs and their 95%CI the authors provide helpful reference ranges to define normal, abnormal and borderline LVGLS values to enable clinical application. The results led the authors to cautiously recommend using the same strain software and their own respective reference ranges for interpreting and comparing LVGLS measurements. Importantly, the authors demonstrate that the inter-vendor variability of LVGLS is now comparable to, or even smaller than, the variability seen with LVEF. This marks a significant step toward the routine clinical application of LVGLS as an alternative or adjunct to LVEF. An increasing number of studies have suggested that GLS is superior to EF as a measure of LV function and as a predictor of mortality and cardiac events [7-10]. Previous reports have suggested that an absolute GLS of −12% represents severe systolic dysfunction and carries an adverse prognosis, and −15%–16% identifies patients with relatively preserved EF at higher risk for future events. However, LVGLS reference range values associated with risk tiers have yet to be firmly established, in part due to lack of established lower limit of normal values. While the study represents a significant contribution to the field, some limitations should be acknowledged. The cohort predominantly comprised younger subjects, which may not reflect the typical age distribution seen in clinical practice. Since older patients often have more comorbidities affecting myocardial mechanics, the findings may not be fully generalizable. Most participants were white, limiting the applicability of the findings to ethnically diverse populations. A number of studies have provided reference ranges for specific ethnic groups. In the 2012 JUSTICE (Japanese Ultrasound Speckle Tracking of the Left Ventricle) study [5] of 817 healthy mostly male volunteers with average age 36 years, the overall mean full thickness, peak systolic GLS using GE equipment was reported as −21.3% ± 2.1%. In a 2014 Italian study [11] of 260 Caucasian healthy mostly female volunteers with average age 44 years, the mean full-thickness, peak systolic GLS using GE equipment was reported as −21.5% ± 2.0% (lower limits of normal or average: 2 SD was 16.9% for men and 18.5% for women). Different values were obtained in a 2009 multicenter (Australian, European and American) study [12] of 242 healthy mostly female volunteers with average age 51 years in which mean full-thickness, peak systolic GLS using GE equipment was reported as −18.6% ± 0.1%, with no significant differences among geographical regions. Given known ethnic differences in cardiac structure and function, further studies incorporating broader demographic diversity are needed. The study does not provide data on whether LVGLS measurements obtained from different ultrasound machines (GE vs. Philips) in the same patient yield consistent results across all vendor-neutral software platforms. This remains a crucial question for laboratories that use multiple echocardiographic systems. Additionally, the study focused solely on LVGLS and did not assess global circumferential strain, global radial strain, or strain rate measurements. These parameters are less commonly used in clinical practice due to their greater variability, but further investigations into their standardization remain necessary. Similarly, GLS is highest in the endocardium and lowest in the epicardium. To assess inter-vendor global strain differences, the study results only apply to endocardial GLS, the only parameter that could be provided by all vendors. Results may not be valid if region of interest for GLS is set in the midwall, epicardial, or full thickness. Indeed, most companies now include software capable of mid/full wall tracking which preliminary data have shown to have similarly good inter-vendor bias and reproducibility as endocardial GLS [13]. There is currently a lack of reference values for segmental strain measurements in part due to suboptimal reproducibility and large inter-vendor measurement variability. The current study does not provide information regarding quantitative assessment of the magnitude of regional deformation. This leaves open the possibility that, inter-vendor differences for segmental strain values may be considerably higher than that reported for global values. The extent of spatial smoothing aimed at reducing noise in the regional tracking of speckles may vary in the different algorithms used by each vendor and may contribute to inter-vendor variability. This gap in knowledge relegates regional strain measurements to assessment of regional differences in polar strain maps rather than precise definition of numerical segment-specific strain values. While reductions in local strain may correctly identify areas of inflammation or fibrosis, the availability of reference values might enhance the ability to quantify regional abnormalities as compared to other imaging techniques. Thanks to continued collaboration between vendors and potential sharing of proprietary software, this has resulted in further reduction in inter-vendor variability. A recent report of the European Association of Cardiovascular Imaging (EACVI) Strain Standardization Task Force [13] aimed at comparing the current inter-vendor variability and reproducibility to the findings from their previous 2013 results evaluated 372 echocardiographic examinations performed in 62 subjects with a wide range of LV function (ejection fraction from 30% to 64%) using ultrasound systems from six manufacturers. The results showed that both endocardial and mid/full-wall GLS measurements were comparable and within a very narrow range (maximum inter-vendor bias of 0.9% strain units) whereas in 2013, the maximum absolute difference was 3.7 % strain units, leading the authors to conclude that in contrast to the situation 10 years earlier, a substantial improvement in inter-vendor bias has been accomplished. In addition, most companies now allow mid/full-wall tracking, which had similarly good inter-vendor bias and reproducibility as endocardial GLS. The study by Arockiam et al. [6] represents an important step toward the standardization of LVGLS measurements across different software platforms. By providing vendor-neutral reference ranges, the authors offer valuable insights that may facilitate the integration of strain imaging into routine clinical practice. However, persistent variability among independent vendors, albeit reduced, underscores the need for continued refinement in strain software and external validation in larger, more diverse populations. Moving forward, widespread adoption of vendor-neutral strain software, along with further multicenter studies, will be crucial for achieving true standardization. In the interim, clinicians should remain aware of the potential measurement discrepancies among software packages and ensure consistency by using the same strain analysis tool for serial patient evaluations.