Transthoracic echocardiography

Transthoracic echochardiography (TTE) was performed within 12 hours after admission and daily for a 7-day observation period or until discharge from ICU by one of four experienced echocardiographers (LB, MC, PG, MD). Images were acquired using a Hewlett- Packard Sonos 5500 (Andover, Mass, U.S.A) scanner and a 3 MHz transducer. Two-dimensional (2D) imaging examinations were performed in the standard apical four- and two- chamber views (2C- and 4C views). Tissue harmonic imaging was used to enhance 2D image quality. LV ejection fraction (LVEF) was assessed by visual estimation of EF, based on “eyeball” ejection fraction. M-mode images were obtained at the LV septal, lateral, anterior, and posterior borders of the mitral ring [18] in the apical 2C- and 4C views, and an average mitral annular plane systolic excursion (MAPSE) value was calculated. Pulsed-wave (PW) tissue Doppler recorded the peak systolic velocity (TDIs) of the LV septal wall at the level of the mitral annulus in the apical 4C view [26]. Transmitral velocities were measured with PW Doppler in the 4C view. For LV diastolic function, we used the mitral inflow profile, the E- and A-velocity and calculated the E/A ratio. PW tissue Doppler recorded the diastolic velocities (é) of the LV septal wall at the level of the mitral annulus in the apical 4C view. The E/é ratio, an index of LV filling pressure, was calculated and é (septal) < 8cm/s indicated diastolic dysfunction [27]. All TTE studies were recorded over three consecutive cardiac cycles independently of the respiratory cycle and averaged. In patients with non-sinus rhythm measurements were collected over 5–10 heartbeats. Analyses of the measurements were made in Phillips digital storing and analysis software Xcelera (Best, the Netherlands) offline.

Statistical analysis

Data are presented as median (lower quartile: upper quartile), percentages or absolute values. For not normally distributed variables we used non-parametric test exclusively. For correlation between two variables, Spearman’s rank correlation was used and for differences between two groups we used Mann-Whitney’s U-test. Categorical data were analyzed with Fisher’s exact test. HsTNT and BNP were log transformed with natural logarithm due to skewed distribution. Discrimination analysis was performed using receiver operating characteristics (ROC) curve under the area using multiple logistic regression predictions. Our aim was to investigate how 28-day mortality can be predicted by more than one explanatory variable measured early during ICU stay. Since we did not have any censored data during this period and odds ratio was the outcome of interest, logistic regression was chosen to be the most suitable method [28]. Multivariate (backward stepwise selection method with probability for the removal of 0.10) logistic regression analyses were used to determine the association of variables with 28-day mortality. The Hosmer and Lemeshow test of goodness of fit was used to indicate if the model provides an adequate fit to the data. Odds ratios (OR) were calculated. We have used the method described by Hoaglin and Iglewicz [29] to test for outliers. The intra- and inter-observer variability of echocardiographic parameters was measured by the coefficient of variation (CV). CV was defined as the ratio of the standard deviation to the mean multiplied by 100. All probability values are two-tailed and significance was set at p < 0.05. The analyses were performed using SPSS 18.0 (SPSS, Chicago, IL, U.S.).

Patient characteristics

Mitral annular plane systolic excursion (MAPSE) and relation to other echocardiographic parameters

The intra- and inter-observer variability for MAPSE was 4.4% and 5.3% respectively, and for other echocardiographic markers as reported previously [13, 25].

Mortality

MAPSE (p= 0.028), SOFA score (p= 0.007), age (p= 0.033) and hsTNT (p= 0.014) were identified as univariate predictors of 28-day mortality (Table 3). A multivariate logistic regression analysis including these variables identified a model with MAPSE (p= 0.015) and SOFA (p= 0.018) as independent predictors of 28-day mortality. The Hosmer and Lemeshow test for the model was non-significant (p=0.998) indicating adequate fit to the data. The adjusted OR for MAPSE and SOFA score were 0.6 (95% CI 0.5- 0.9) and 1.6 (95% CI 1.1- 2.3) respectively (Table 3). With regards to 28-day mortality the area under the curve (AUC) at day 1 for MAPSE was 0.709 (95% CI 0.548- 0.870, p = 0.028) with 69% sensitivity and 68% specificity for a cut-off value of 8 mm and AUC for SOFA score was 0.733 (95% CI 0.592- 0.874, p = 0.014) with 69% sensitivity and 68% specificity for a cut-off value of 12. When adding MAPSE to SOFA score, the AUC for the combined predictor increased to 0.831 (95% CI 0.711- 0.952, p <0.001) (Figure 1).

Cardiac biomarkers

Results at day 1 showed that hsTNT was significantly higher in non-survivors (152 [IQR 80–501] ng/L) than in survivors (77.5 [IQR 18.6-125.3] ng/L) (p= 0.014) whereas there was no difference in BNP (Table 1). In patients with diastolic dysfunction (é< 8cm/s) both hsTNT and BNP were significantly higher in non-survivors compared to survivors (p= 0.020 and p= 0.039 respectively); this was not seen in those with systolic dysfunction (LVEF≤ 50%). hsTNT showed a significant negative correlation with MAPSE but not with LVEF or TDIs (Table 3). In patients with diastolic dysfunction (é< 8cm/s) hsTNT and BNP showed a significant negative correlation with MAPSE (r= −0.478, p= 0.033; borderline for BNP r= −0.441, p= 0.051) but there was none with TDIs.

Daily measurements over 7 days showed that in non-survivors MAPSE was significantly lower in most days (day 1 p=0.028, day 2 p=0.003, day 3 p=0.060, day 4 p= 0.036, day 5 p=0.026, day 6 p=0.017, day 7 p=0.075) (Figure 2) whereas LVEF was not significant different (p>0.1) at any day (Figure 3).

MAPSE and prognosis

In critically ill patients echocardiography has gained popularity as a tool for assessing LV function [5, 30]. LVEF is probably the most commonly used and accepted method of measuring LV systolic function in this setting [1] however its usefulness in predicting mortality has produced conflicting results [2, 5, 31]. In patients with septic shock studies measuring the LV longitudinal function by tissue Doppler imaging (TDI) have moved into focus during the recent years identifying mainly diastolic TDI indices as prognostic markers whereas systolic TDI parameters seem to be less consistently related to mortality [31–34]. Interestingly in none of these studies MAPSE was used as a marker of LV systolic function.

MAPSE is a simple, easily obtained parameter and may contribute to the evaluation of systolic function. MAPSE was obtainable in all patients, and showed inter- and intra-observer variability of 4.4% and 5.3% [13]. It is less well investigated than its right ventricular counterpart, tricuspid annular plane systolic excursion (TAPSE), and has received considerably less attention than TDI variables. In critical care settings where acoustic windows are often suboptimal, MAPSE seems to be an attractive parameter. A decreased MAPSE is known to be associated with conditions affecting LV function such as myocardial infarction, heart failure, atrial fibrillation and age [15–17] and its relation to mortality has been described by several studies in patients with cardiovascular disease [15, 19–21].

We found that MAPSE on day 1 was significantly lower in non-survivors compared to survivors and could together with SOFA score be identified as independent predictors of 28-day mortality. Further, combining MAPSE and SOFA score seemed to increase the risk of death. These results are strengthened by the finding that MAPSE was significantly decreased in non-survivors compared to survivors in most days of the 7-day observation whereas LVEF was not. LVEF, on the other hand, being near normal could not be identified as a prognostic marker and we speculate if this, like in patients with cardiovascular disease and preserved ejection fraction, could be due to LVEF being less sensitive in uncovering subtle myocardial changes [35, 36].

MAPSE in relation to other markers of LV function and myocardial injury

Previous studies in patients with cardiovascular disease have suggested MAPSE as a surrogate measurement for LVEF with both normal and reduced LV function [12, 14]. A mean value for MAPSE of > 10 mm was linked with preserved EF (≥ 55%) and values < 8 mm with reduced EF (< 50%) [6, 19, 37]. Our results are in line with this, with MAPSE correlating significantly with LVEF. MAPSE was 11 [11–12.8] mm in patients preserved EF and in those with reduced EF slightly higher than 8 mm (MAPSE 9 [7.3-12.3] mm).

Although MAPSE and LVEF may be related, they are not entirely interchangeable [13, 17]. MAPSE is suggested to be primarily representative of subendocardial, longitudinally oriented, myocardial fibres compared to the subepicardial, circumferential fibres measured by LVEF, and is known to detect more subtle abnormalities of LV function [7, 18]. This is seen in patients with increasing age, myocardial hypertrophy or diastolic dysfunction with preserved ejection fraction (HFpEF) where long axis function of the heart is already impaired while the radial function can be preserved or even increased [18, 35]. Thus by using LVEF the long axis function of the heart is not necessarily considered.

Similar to MAPSE, tissue Doppler imaging is described to be superior to conventional echocardiography in detecting abnormalities of LV function [38] and its correlation with MAPSE has been described previously [39]. In a recent study (Wenzelburger), TDI indices of both LV diastolic and systolic function correlated significantly with MAPSE in patients with HFpEF [35] illustrating the close relationship between systolic and diastolic LV function. LV systolic torsional (twisting) deformation is one mechanism by which potential energy is stored. After systole the heart relaxes or untwists, an energy releasing process, and aids to early LV filling by suctioning [40]. Thus a decrease in atrioventricualar plane motion will result in less energy stored during systole and hence reduced LV diastolic mechanics.

The relationship between MAPSE and TDI indices is supported by our results where MAPSE correlated significantly with é, a diastolic marker, and showed a negative significant correlation with E/é, a surrogate marker for LV filling pressure. Additionally we found a significant association between diastolic dysfunction (é< 8cm/s) and MAPSE, but none with LVEF. Of note, in our previous study we showed a significant correlation between MAPSE and the systolic marker of tissue Doppler imaging (TDIs) (r= 0.427, p< 0.01) [13].

Notably, MAPSE and TDI are not interchangeable. The two parameters describe different vector components of longitudinal systolic motion. Although TDIs has been described to correlate well with MAPSE in studies with healthy individuals [39, 41], they differ in some important aspects [42] as MAPSE measures the entire systolic phase including isovolumetric contraction [15] in contrast to TDIs. In addition a blurred TDI signal can be a confounding factor when measuring the velocity [43] and in these situations MAPSE can be a valuable. Overall we believe that MAPSE and tissue Doppler measurements are important echocardiographic parameters in assessing LV function.

Finally, we sought to investigate if there was a relationship between cardiac biomarkers (hsTNT and BNP) and MAPSE and found that hsTNT but not BNP was significantly higher in non-survivors and correlated significantly with MAPSE but not LVEF. This is in line with a recent study in septic neonates were hsTNT was significantly higher in non-survivors and correlated with longitudinal LV systolic function measured by TDI but not with fractional shortening [44]. Landesberg et al. [32] found that hsTNT was significantly higher in patients with decreased é and LVEF. This is similar to our findings where hsTNT in patients with diastolic dysfunction showed a significant negative correlation with MAPSE. Although we found no relationship between LVEF and cardiac troponin T our findings are generally in support of these studies and we speculate if MAPSE may be more sensitive than LVEF in detecting early myocardial changes in critically ill patients with shock.

Limitations

Firstly, we used eyeballing to measure LVEF. Simpson’s biplane method is the currently accepted standard. We and others have previously shown that eyeball EF was as good as the Simpson’s method [13, 45], and was more easily obtained in ICU patients [13]. Although unlikely, we cannot exclude that using Simpson’s method for measuring LVEF could have influenced our results. Secondary analysis (data not shown) in 44 patients where good-quality Simpson’s EF could be obtained showed no relationship to mortality. Secondly, we did not screen our patients for specific conditions affecting MAPSE measurements such as localized wall motion abnormalities or mitral annular calcifications. Thirdly, no dynamic fluid responsiveness tests were used limiting our results, as MAPSE in analogy with tissue Doppler measurements, is affected by changing fluid conditions. Finally, the multivariate analysis was limited by the small number of patients in this study, and we cannot exclude other confounding factors. Nevertheless, we clearly demonstrate a relationship between MAPSE and 28-day mortality.