To the best of our knowledge, we conducted the largest study investigating the correlation between mitral S’ and LVEF in a population of critically ill patients. Our study had two important findings. First, we found only a moderate correlation between mitral S’ and LVEF among patients admitted to the medical ICU with sepsis or septic shock, and this finding was consistent in several prespecified subgroup analyses. Second, mitral S’ was independently associated with a linear increase in 28-day mortality from ICU admission in septic patients, therefore suggesting a statistically significant prognostic value. Conversely, the relationship between LVEF and 28-day mortality from ICU admission was U-shaped.

Cardiac function assessment is integral to the management of sepsis and septic shock. LVEF using the Simpson method is routinely used to assess LV systolic function parameter. However, LVEF measurement often requires an optimal image for measurement, which may be limited in critically ill patients. Mitral annular plane systolic excursion (MAPSE) and TDI-derived LV systolic velocity (mitral S’), both represent regional measurements of LV longitudinal systolic function, have been suggested as good surrogates for the LV systolic function. Both parameters are conceptually simple, do not rely on geometric assumptions, are easy to obtain and highly reproducible even when performed by practitioners with limited experience [30]. Studies have found mitral S’ and MAPSE to have excellent correlation and concordance. Mitral S’ is routinely performed as a part of systolic assessment of the LV, while MAPSE is still not part of comprehensive TTE. The correlation between mitral S’ and LVEF is good-to-excellent among stable cardiac outpatients [6, 11,12,13, 31]. However, studies in critically ill patients remains limited in sample size, and evidence so far suggests that mitral S’ values are not associated with prognosis in septic patients [14, 15], which was confirmed by a recent meta-analysis [32]. Such meta-analysis included 13 studies and 1200 patient, which is less than half of our sample size. In addition, the sample size varied in the included studies (from 21 to 262). The larger sample size in our study provided more power to detect any relationship between mitral S’ and mortality, and this is a possible explanation for different findings.

Surprisingly, even though mitral S’ has been repeatedly promoted as a good surrogate for LVEF, our study showed these parameters cannot be used interchangeably in critically ill septic patients. We just found a moderate correlation between mitral S’ and LVEF, findings that were consistent across different subgroup analyses with correlation ranging between 0.29 (those with LVEF > 45%) and 0.50 (septic shock with high dose of NEE). The correlation of mitral S’ was low regardless the values of LVEF. The mitral S’ values in reduced LVEF group were lower compared to normal LVEF group, however utility of mitral S’ to diagnose sepsis cardiomyopathy remains limited. These results are comparable to a smaller prospective study by Bergenzaun et al. of 50 patients with septic shock, who underwent TTE every 24 h until 7 days or death with an overall correlation of r = 0.473 [14]. Similarly, Furian et al. also demonstrated a moderate correlation (r = 0.49; p = 0.003) among 45 patients with severe sepsis [15]. Our larger cohort with varied severity and co-morbidities not only validates but conclusively proves that mitral S’ and LVEF are non-interchange entities among critically ill patients.

The moderate correlation between LVEF and mitral S’ in septic patients are likely impacted by several factors. LVEF assessment includes both the radial and longitudinal components of LV systolic contraction. Normally, longitudinal shortening contributes approximately 75% to cardiac contractility and overall stroke volume [33]. As short axis shortening (radial function) gets impaired with various disease states, the heart compensates by increasing contribution from the longitudinal component to maintain cardiac function. This adaptation may partially explain why our study and previous evidence demonstrated only a moderate correlation observed between mitral S’ and LVEF. Another explanation is the impact of loading conditions on LVEF and mitral S’. LVEF is often reflective of the coupling between LV contractility and its afterload [34, 35]. Therefore, it is affected by both preload and afterload changes, the latter being particularly reduced in patients with septic shock. Thus, septic patients with reduced intrinsic LV contractility may show a preserved LVEF in the setting of severely reduced afterload [30]. Conversely, mitral S’ seems influenced by afterload to a lesser extent and to depend mostly on changes in preload [36,37,38,39]. We observed the LVEF to sepsis mortality curve to be U-shaped, while it increases linearly with mitral S’. We hypothesize the different shape of the curve stems from difference in parameters, which are assessed and we believe these may be mainly due to influence of loading conditions on each of these parameters, especially the afterload that highly affects LVEF and possibly to a lesser extent for s’. For instance, the higher mortality of hyperdynamic LVEF can be attributed to under-resuscitation, severe vasoplegia, and/or sympathetic overstimulation, whilst values of s’ can greatly vary. However, despite this possible physiological interpretation, such hypothesis is not supported by the subgroup analyses where we assessed correlation according to the LVEF and to vasopressor dosages”. Notably, the same moderate correlation was reported by two smaller studies that we discussed. Hence, our findings are in the same direction. The reasons why correlation is lower in critically ill patients as compared to other groups of stable patients deserves prospective evaluation.

We acknowledge that recent data has shown that global longitudinal strain (GLS) can potentially identify early myocardial dysfunction, often missed by the conventional indexes of systolic function [40] (as LVEF). GLS is dependent on LV loading conditions, but the largest influence on this parameter seems due to afterload changes [37, 38, 41,42,43]. The possibly lower degree of dependence of GLS on preload as compared to LVEF and TDI variables makes GLS an exciting prognostic variable for critically ill patients. However, strain echocardiography is not widely available for clinical use [44]. Mitral S’ has been shown to detect myocardial contraction impairment before clinical deterioration. This would potentially make mitral S’ an attractive alternative to GLS to effectively study the full spectrum of LV systolic dysfunction.

The biggest strength of our study is the large number of patients included in the final analysis and the use of homogeneous criteria for the diagnosis of sepsis and septic shock. We also assessed for intra-operator reliability of mitral S’ measurement to reduce measurement errors influencing our results. However, our study remains a single-center retrospective cohort study, and as a result, we cannot eliminate selection bias completely. Including all consecutive patients who met the study criteria has mitigated some risks of selection bias. Second, the time window for echocardiograms in this study was three days after medical ICU admission. The loading conditions and vasopressor dosage can change significantly during the first three days after ICU admission. Most patients (77.3%) underwent an echocardiogram within 24 h of admission to ICU, coinciding with the onset of sepsis and septic shock. Pulmonary artery catheter, central venous pressure, and central venous oxygen saturation are no longer frequently assessed in routine critical care. The unavailability of more precise loading parameters and markers of tissue perfusion other than MAP, fluid balance, and serum lactate in our sample limits our ability to understand the impact of loading conditions and tissue perfusion on mitral S’ and mortality. The retrospective nature of the study limits also our ability to completely abide by requirements of PRICES guidelines. Another limitation, is the exclusion of patients where peak annular velocity was not available, particularly from the lateral mitral annulus. However, in all patients and subgroup analyses, average mitral S’ had similar or better correlation compared to either lateral or septal velocities (Additional file 1: Table S1).