Left ventricular systolic function evaluated by strain echocardiography and relationship with mortality in patients with severe sepsis or septic shock: a systematic review and meta-analysis

A recent expert consensus updated the definition of sepsis as “life-threatening organ dysfunction caused by a dysregulated host response to infection” [1]. The response to the infective process can cause profound hemodynamic alterations, leading to organ dysfunction and accounting for high mortality and morbidity [2, 3]. The most severe manifestation of sepsis is septic shock, a condition characterized not only by vasoplegia but that can also present with myocardial depression. Septic cardiomyopathy is the widely used term for myocardial involvement in sepsis [4], and can present with left and/or right ventricular (LV and/or RV) impairment [5]. The majority of patients with sepsis and/or septic shock have underlying cardiac dysfunction as demonstrated by a post-mortem necropsy study that showed myocardial injury in more than half of patients with sepsis and/or septic shock [6], although such patients do not always exhibit signs and symptoms of myocardial ischemia [7].

While there is growing evidence of an association between LV diastolic dysfunction and mortality [8, 9], the influence of LV systolic dysfunction (LVSD) is less clear. Initial evidence suggested that mean LV ejection fraction (LVEF, the most commonly used index to define LVSD) is paradoxically higher in non-survivors of septic shock [10], although this association has not been subsequently confirmed [11]. Finally, a meta-analysis showed no association between LVEF and mortality in a septic population [12], and this has been confirmed by subsequent work [8].

Echocardiography is currently suggested as part of the first-line approach in the evaluation of patients with shock [13], but the use of conventional indexes of systolic function in patients with sepsis and/or septic shock may not accurately reflect the true systolic function. Traditional estimation of LVSD is problematic in sepsis since LVEF is based on significant geometric assumptions and is also highly dependent on loading conditions. Preload is highly variable according to fluid and vasoactive drug resuscitation, degree of endothelial insult and vascular leak. Afterload variations in sepsis (vasoplegia and vasoconstrictor use) causes changes in LVEF that are not necessarily related to true variations in intrinsic myocardial contractility [14].

An alternative echocardiographic modality, speckle-tracking echocardiography (STE) is emerging as a better marker of intrinsic LV function [15]. It was first described in 2004 as an angle-independent non-Doppler method [16], based on the generation of ultrasound echoes (“speckles”) representing discrete myocardial areas tracked throughout the cardiac cycle [17]. Strain represents the difference between the final length of each segment relative to its resting length and can be measured in different planes: longitudinal (from base to apex), radial (inward short axis), and circumferential (rotational short axis). Its assessment is performed during bedside echocardiography, and global longitudinal strain (GLS) is the most commonly reported strain measure, representing the ratio of the maximal change in the myocardial longitudinal length in systole to the original length in diastole. In short, more negative values of GLS indicate better LV systolic function. The conceptual difference between assessment of myocardial function with GLS or LVEF is illustrated in Fig. 1.

In this systematic review and meta-analysis we aim to investigate the association between values of strain and mortality in patients with severe sepsis and/or septic shock. We hypothesize that GLS, but not LVEF values, are associated with mortality in patients suffering from severe sepsis and/or septic shock.

This systematic review and meta-analysis was performed in accordance with preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines [18]. The review was registered with the international prospective register of systematic reviews (PROSPERO, number CRD 42016041712).

To our knowledge, this is the first meta-analysis investigating the role of STE in identifying patients with sepsis and/or septic shock at higher risk of mortality. We found a significant association between worse (less negative) GLS values and mortality, while the most commonly adopted conventional parameter - LVEF - was not associated with mortality when analyzing the results of the studies included in the meta-analysis, confirming previous findings [12].

Our findings are not entirely surprising since the value of GLS has been recognized in various settings. Strain imaging can detect subclinical myocardial dysfunction in experimental studies on septic animals. In a rabbit model, Li et al. showed that 2 h after injecting endotoxin, GLS declined before changes in LVEF were manifest [35]. Similar results have been reported in anesthetized pigs receiving infusions of Escherichia coli, where STE detected myocardial dysfunction before significant changes in LVEF and cardiac output [36].

In the clinical setting, GLS is helpful in the diagnosis and/or prognosis of early stages of heart failure [37], chemotherapy-related cardiotoxicity [38], hypertrophic cardiomyopathy [39], pregnancy-related myocardial dysfunction [40], cardiac amyloidosis [41] and subclinical cardiomyopathies in patients with chronic kidney disease [42]. A meta-analysis showed superior prognostic value of GLS as compared to LVEF in the prediction of major adverse cardiac events in a heterogeneous population with underlying cardiac conditions [43]. For such reasons, STE use has been included in updated American Society of Echocardiography and the European Association of Cardiovascular Imaging guidelines [44]. All together with its prognostic value, GLS has been shown to have good reproducibility and non-significant intra-observer error, in many cases outperforming most conventional echocardiographic parameters, including LVEF, although small but significant inter-vendor differences still persist [45].

The aforementioned guidelines suggest normal values of peak GLS in the range of − 20% [44]. In our meta-analysis none of the reported mean GLS values fell in this range, reinforcing the fact that a degree of systolic impairment may be present in the majority of patients with sepsis and/or septic shock, both in survivors and non-survivors. Therefore, GLS could be considered a better surrogate of intrinsic LV myocardial function contrary to LVEF. One animal study showed 100% of myocardial depression as assessed via the pressure-volume loop (gold standard method) [46], but this finding differs from other studies [47, 48], leaving uncertainties in this field. Nonetheless, GLS may become a very useful parameter in the early evaluation of septic cardiomyopathy. For instance, Boissier et al. elegantly showed impaired values of GLS in the first 24 h in patients developing secondary LV hypo-kinesia (defined as reduced LVEF on day 2 or day 3 after a normal value of LVEF on the first day in the ICU), confirming higher sensitivity of GLS in detecting sub-clinical cardiomyopathy. Moreover, in this study even patients with normal LVEF had mean GLS values worse than − 20%, highlighting the ability of GLS in detecting myocardial impairment not shown by LVEF. Moreover, only a few patients had good GLS values and all of them had normal LVEF [31].

Importantly, GLS values are different in other populations of critically ill patients, as shown by Dalla et al. [33] who reported a prevalence of depressed longitudinal LV function as high as 50% in patients with septic shock, while the prevalence was below 9% in controls (trauma patients without septic shock) [33]. Similarly, Ng et al. [34] observed worse GLS values in patients with septic shock as compared with those with sepsis only (− 14.5% vs − 18.3%, respectively) and significantly improved strain after recovery from septic shock in patients weaned off vasopressors within 72 h. However, Ng et al. [34] also demonstrated some improvement (not statistically significant) in GLS in a smaller group of patients remaining vasopressor-dependent [49]. On the contrary, another study in patients with septic shock found unchanged GLS over time despite normalization in LVEF and cardiac biomarkers and clinical recovery [27].

The use of STE may possibly overcome some limitations of LVEF, in relation to variability of loading conditions and geometric assumptions, indeed, as LVEF is a direct measure of the change in blood pool volume and thus dependent on volume and pressure load on the myocardium [50, 51], with systolic function increasing during vasodilatation and decreasing in states of vasoconstriction [51, 52]. It has been claimed that GLS is a reliable, reproducible, and sensitive modality for assessing cardiac systolic function [43, 53, 54]; however, it should be kept in mind that GLS is also dependent on LV loading conditions, especially afterload changes, with longitudinal fibers suffering from higher wall stress due to their orientation. This assumption is confirmed by animal [55] and clinical studies [56‐59]. A recent study by Nafati et al. demonstrated that in a heterogeneous population of preload-dependent critically ill patients, GLS is significantly affected (more negative values indicating improved LV function) after fluid resuscitation [60]. In critically ill patients with sepsis and/or septic shock it is possible that after the initial fluid resuscitation, afterload changes play a more important role than preload, particularly in terms of outcome. For instance, in patients with aortic stenosis with varying afterload, GLS was shown to be sensitive to these changes, while LVEF was not [61, 62]. It may be that this is also reflected in the value of GLS to predict mortality. Certainly, more research is warranted to understand the variation in GLS according to preload and afterload under stable hemodynamic and respiratory support in critically ill patients. Of course, LVEF variations are also related to changes in afterload more than in intrinsic myocardial contractility [14], and in this regard Vieillard-Baron et al. repeatedly confirmed the variability in the incidence of systolic dysfunction as evaluated by LVEF at different time points after the onset of septic shock [11, 51]. An independent association between hyperdynamic conditions (LVEF > 70%) and mortality was identified in a large retrospective cohort study of critically ill patients [63]. A further limitation in the estimation of systolic function by means of LVEF is the geometric assumptions. Wall motion measurements evaluating displacement cannot differentiate between active or passive movement of a myocardial segment with inward systolic movements dragging diseased segments, and thus the resulting LVEF is relatively unaffected by subtle systolic changes [64]. On the contrary, deformation analyses (i.e. GLS) allow discrimination between active and passive myocardial tissue movement and are less affected by measurement errors because they avoid geometric assumptions [65]. It is possible that a reduction in longitudinal strain is compensated by other factors and LVEF remains normal for such mechanisms. A recent study confirmed the hypothesis that strain better reflects systolic function as compared with LVEF, especially in patients with preserved LVEF, where the authors showed a flatter slope for GLS/LVEF correlation (slope − 0.6) as compared with patients with decreased LVEF (slope − 1.6). Such findings indicate that LVEF may be unaffected if a reduction in GLS is compensated by a change in other parameters (increase in circumferential strain and/or increase in LV wall thickness and/or reduction in LV end-diastolic volume), and that GLS can decrease earlier before effects on LVEF are manifest [66].

The use of GLS probably offers advantages over assessment of the systolic pulsed wave (s’) obtained with tissue Doppler imaging (TDI). The s’ TDI of the base of LV wall is another index of longitudinal LV systolic function and has been validated against LVEF in patients with cardiac disease [67]. In critically ill patients, TDI has been predominantly used in the assessment of LV diastolic function [9], but there is growing literature on s’ TDI [25, 26]. However, s’ provides an angle-dependent, unidirectional (longitudinal) and unidimensional (one LV segment) assessment of myocardial motion, thus being exposed to underestimation or overestimation (regional wall motion abnormalities, tethering, etc.). In this context, GLS has advantages over TDI, likewise assessing all the LV segments and being less angle-dependent [68, 69], but on the other hand better imaging definition is required compared with TDI [70]. Possible M-mode surrogates of GLS may be longitudinal wall fractional shortening, the curved anatomical M-mode fractional shortening and the mitral annulus systolic plane excursion. Such parameters have recently been shown to correlate strongly with GLS, with longitudinal wall fractional shortening being the best unbiased strain predictor. Given the simplicity of M-mode measurements, this may have significant clinical and practical implications for future critical care echocardiography investigations [71].

From a clinical perspective, obtaining a more accurate identification of underlying septic cardiomyopathy has not only prognostic but also therapeutic value. For instance, tachycardia is one of the hallmarks of sepsis and it is associated with worse clinical outcomes in critically ill patients [72]. However, it is still difficult to distinguish between tachycardia as an adaptive response to low preload conditions and tachycardia related to a persistent hyperadrenergic state (in turn responsible for septic cardiomyopathy). Lanspa et al. identified an association between impaired GLS and tachycardia [28]. Despite this, controversy remains surrounding the use of beta blocker therapy for treatment of tachycardic patients with sepsis and/or septic shock [73, 74], and it is advisable that future studies on beta blockade in sepsis evaluate LVSD by STE for the identification of patients who might benefit from pharmacological heart rate control.

Furthermore, with the wider use of echocardiography we believe that RV function measured with STE should be a focus of future research. The RV has a preponderance of longitudinal fibers and can be assessed with longitudinal strain [75] and there is growing evidence with regard to the prognostic value of RV strain, for instance in patients with pulmonary hypertension [76]. The utility of RV strain in sepsis has not been assessed in our meta-analysis because of the small number of available studies, but interesting findings have been reported by Orde et al., where higher RV strain was identified in sepsis survivors despite similar values of LV strain [29]. However, RV function - more than LV - is affected not only by preload but also by mechanical ventilation, making its assessment challenging in mechanically ventilated patients with septic shock.

Worse values of global longitudinal strain are associated with higher mortality in patients with severe sepsis or septic shock, while such an association is not valid for left ventricular ejection fraction. More research is warranted to elucidate such an association, which could be related to the ability of speckle-tracking echocardiography in demonstrating underlying intrinsic myocardial disease as opposed to left ventricular ejection fraction.