Discussion
This meta-analysis confirms the prognostic significance of GLS when compared to LVEF in patients with different underlying cardiac abnormalities. A 1 SD change in GLS was a stronger predictor of all-cause mortality when compared with the same change in LVEF in both univariable and multivariable models. A similar HR was maintained for composite endpoints.
Use of GLS and EF for Assessment of Global LV Function
Estimation of LV systolic function is the mainstay of risk evaluation and management of cardiac diseases. Historically the most common means of obtaining LVEF have been echocardiography and nuclear ventriculography, with the more recent addition of cardiac CT, cardiac MR (CMR), and 3D echocardiography. Although the performance of echocardiography has shown prognostic benefit in patients with underlying cardiac diseases, and the simplicity and ready availability of echocardiography have made this test the instrument of choice for LVEF assessment, this technique carries a number of potential limitations. The use of standard 2D echocardiography to examine 3D cardiac structure is limited by geometric assumptions, foreshortening and difficulties in proper delineation of the endocardial borders. Moreover, all assessment of LVEF is affected by load, heart rate, and beat to beat variability (eg, atrial fibrillation). Disturbance of haemodynamic load may contribute to the transition into HF, and potentially this transition may be hidden in the setting of reduced afterload—a situation where a more sensitive index of systolic performance may facilitate recognition of LV deterioration.
In these and other situations, subclinical LV impairment may be identified by reduced longitudinal function—perhaps because subendocardial fibres are most susceptible to injury. GLS is a semi-automated tool to assess multidimensional myocardial mechanics which is more reproducible, unaffected by tethering effects and non-reliant on geometric assumptions. CMR measurement of EF showed a better correlation with GLS (r=−0.69, p<0.0001) than did 2D EF (r=0.58, p<0.0001). GLS is well validated as a marker for the measurement of LV longitudinal deformation, which has emerged as a sensitive and specific marker to detect early and subtle myocardial dysfunction. GLS can identify patients with early stages of HF, subclinical chemotherapy related cardiotoxicity, early stages of cardiomyopathies or infiltrative heart diseases, and hence provide vital guidance about diagnosis and management.
Specifically, while echocardiographic EF is capable of differentiating between severe, moderate and mild LV impairment, its high test-retest variability has led to reclassification of normal to mild LVEF in up to 13% of patients. The limitations of near normal LVEF likely account for the better correlation of GLS and EF when EF is <35% (r=0.74, p<0.001) than with EF>35%.
GLS and Risk Evaluation
It is well known that patients with impaired LVEF pose a high risk of all-cause and operative mortality, but this effect is less apparent in patients with EF>45%. The inherent limitations of EF may limit its ability to identify mild degrees of LV systolic impairment. GLS has also been proven to offer incremental predictive value of prognosis on multiple logistic regression models when compared to LVEF and WMSI, and a number of studies have shown impaired GLS in the setting of normal LVEF (Table 2).
Limitations
As with many meta-analyses of observational studies, nonuniform design and variations in the inclusion criteria, follow-up periods and endpoints are all potential sources of heterogeneity among studies (Table 1). Nonetheless, both measured heterogeneity and publication bias appear limited. For this meta-analysis we used 2D speckle tracking imaging and not tissue Doppler based parameters due to their inherent limitations of poor signal-to-noise ratio, angle dependency and alignment issues in patients with ischaemic cardiomyopathy. We did not add WMSI in our meta-analysis because only a few studies reported these findings. Likewise, although we provided subanalyses according to clinical presentation, the subgroups are small. In some situations, GLS (rather than LVEF) was shown to be an independent predictor of primary and secondary outcome —for example, after acute myocardial infarction and in moderate to severe aortic stenosis. On the other hand, there were only four HF studies, and the analysis was underpowered to show differences in the performance of EF and GLS in this setting.
Additional potential limitations of GLS are its dependency on high quality 2D images and appropriate imaging settings. 2D speckle tracking seems to work best at frame rates between 50– 70/s, and tachycardia may result in under-sampling. Increasing the frame rate risks compromising spatial resolution. Like EF, strain is dependent on both preload and afterload, and alterations in loading can result in increasing and reducing strain, irrespective of myocardial status. Lack of reproducibility and out of plane motion, although much less compared with tissue Doppler based strain, remains a potential problem. However, while both EF and GLS are potentially limited by technical factors, the interclass correlation coefficients between and within observers with EF (0.67 and 0.80) are inferior to GLS (0.92 for both). Similarly, Munk et al have shown that GLS has better intra-observer and inter-observer reproducibility when compared with LVEF and end systolic volume index (ESVI) for predicting infarct size. The one specific limitation of GLS is that there is a lack of standardisation between multiple vendors, causing inter-vendor variability.