Transthoracic Echocardiography with Doppler Tissue Imaging predicts weaning failure from mechanical ventilation: evolution of the left ventricle relaxation rate during a spontaneous breathing trial is the key factor in weaning outcome

Cardiogenic pulmonary edema has been recognized as a frequent cause of weaning failure [1‐3]. The definitive diagnosis remains difficult; in this situation right heart catheterization was proposed as the reference tool [3, 4]. Lemaire et al. demonstrated that a spontaneous breathing trial (SBT) induced an elevation in the left ventricular filling pressure (LVFP), which plays a key role in weaning failure [3]. Since the introduction of a pulmonary artery catheter remains an invasive and potentially harmful procedure, there is growing evidence to suggest that transthoracic echocardiography (TTE) should be used to identify the cardiac origin of respiratory weaning failure [4]. Tissue Doppler is a novel technique that directly measures myocardial velocities [5, 6]. The early diastolic mitral annulus velocity (Ea) has been shown to be a relatively load independent measure of myocardial relaxation [6‐8]. When Ea is combined with pulsed Doppler mitral flow in early diastole E, the resulting E/Ea ratio closely correlated with the measured invasive LVFP [6, 9]. Lamia et al. showed that the combination of E/A mitral inflow and E/Ea at the end of a SBT predicted an increase in the LVFP, with a good sensitivity and specificity, in a selected population of patients who were difficult to wean [4]. More recently, Caille et al. found a significantly higher E/Ea ratio at baseline in failed patients [10]. These studies excluded patients with atrial fibrillation, which represent 20 to 30% of patients hospitalized in intensive care units [11]. Nevertheless, despite this arrhythmia, the LVFP could be assessed accurately with measurement of the E/Ea ratio [12].

The primary objective of our study was to test the ability of the mitral E/Ea ratio, measured at baseline and shortly after initiation of SBT, to determine the risk of weaning failure in a non-selected population, including patients with atrial fibrillation. The second aim was to determine whether the depressed left ejection fraction and/or diastolic dysfunction participated in weaning outcome.

This observational prospective study was conducted from November 2009 to April 2011 in a 10-bed medical intensive care unit. No change of standard care was introduced for the needs of this study, which was, therefore, accepted as a descriptive study by the clinical research committee of the university hospital of Nice, in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki. Because it is a descriptive study, written informed consent was waived but all patients involved in our study or their next of kin were informed and accepted to participate.

We screened patients under mechanical ventilation for more than 48 hours before they were considered ready to undergo an initial SBT. Patients were included if they met the following criteria: improvement of the underlying cause of acute respiratory failure, body temperature < 39°C, hemoglobin level > 7G/dl, PaO2 > 60 mmHg, FIO2 < 40%, positive end expiratory pressure (PEEP) under or equal to 8 cm H2O, respiratory rate less than 35 breathes/minute, systolic arterial pressure > 90 mmHg and < 160 mmHg without need for vasoactive drugs, no sedation and a stable neurological status [13].

Non-inclusion criteria included tracheostomy, active neuromuscular disease, pregnant women, a poor echocardiography window, heart rate faster than 140 beats per minute at baseline, and patients with atrioventricular conduction abnormalities. Patients with severe mitral regurgitation, mitral stenosis or a prosthetic mitral valve were also excluded because E/Ea had not been validated in such patients for evaluation of LVFP.

From November 2009 to April 2011, 82 patients were screened, but 14 were excluded for the following reasons: suboptimal echocardiography images (n = 7), neuromuscular disease (n = 3), and tracheostomy (n = 4).

Sixty-eight patients were included. The reasons for intubation were septic shock (n = 20), cardiac arrest (n = 5), cardiogenic pulmonary edema (n = 6), acute respiratory distress syndrome (n = 10), post-operative period (n = 7) pneumonia (n = 12) and coma (n = 8).

Twenty patients failed to wean. Sixteen patients failed the SBT, 14 patients with cardiogenic pulmonary edema and 2 with respiratory acidosis without hypoxemia. The cause of these two respiratory failures was under diagnosed ICU acquired weakness. These two patients underwent tracheotomy, one died in intensive care while the other one was delivered from mechanical ventilation four weeks later.

Four patients failed in the post extubation period. The time of occurrence of SBT failure was 24 minutes [17‐33]. The reasons for extubation failure were septic shock (n = 2) and excessive bronchorrhea (n = 2).

Fifteen patients presented with atrial fibrillation (22.1%) at the time of inclusion, among these 15 patients 9 (60%) failed the SBT and all presented with weaning induced pulmonary edema. Six others patients succeeded the SBT and none needed re-intubation.

The major finding of this study shows that serial measurement of E/Ea accurately predicts weaning failure. E/Ea prior to SBT was significantly higher in the failed group; however, diagnosis in predicting weaning failure, as estimated from the ROC curve was poor. Interestingly, E/Ea measured 10 minutes after starting the SBT showed better diagnostic performance; the cut-off value of 14.5 predicted weaning failure with a sensitivity of 75% and a specificity of 95.8%.

Moreover, the TTE detected changes in the central hemodynamics induced by the SBT. The failed group showed a marked increase in E/Ea, whereas no significant change was observed in the successful group. These results are suggestive of an increase in LVFP during the SBT in the sole failed group. These findings are in line with previous studies; pulmonary artery catheterization showed an increase in the pulmonary artery occlusion pressure (PAOP) during the SBT, only in patients who failed the weaning process [18].

In our study, no significant difference in the E mitral was observed at baseline between the two groups. This is explained by the fact that the E velocity or even E/A is not only determined by LVFP but also depends on ventricular factors, such as the relaxation rate, age of patients, compliance and systolic function [19]. Correlation between E/A mitral and LVFP were too weak in giving an accurate estimation [5]. The E/A mitral should not be used alone and must be considered in the context of the clinical picture, especially with respect to age and systolic function. These difficulties in interpreting the E/A ratio limit its use in clinical practice. In contrast, E/Ea closely correlated with the measured invasive LVFP, even in patients under mechanical ventilation [9].

As in our study, Caille et al. found a significantly higher E/Ea prior to SBT in patients who fail, but their cut-off values for E/Ea were lower than in the present study [10]. This discrepancy is explained by a gray zone when using E/Ea alone for the estimation of the LVFP. Ommen et al. showed that an E/Ea < 8 accurately identified patients with a normal LVFP, and E/Ea > 15 for those with an elevated LVFP [20]. For patients with intermediate values of E/Ea, between 8 and 15, it can be difficult to accurately estimate the LVFP using only E/Ea [20]. Lamia et al. demonstrated that the combination of E/A > 0.95 mitral and E/Ea > 8.5 predicted an elevation in PAOP at the end of the SBT and could then identify patients at risk of weaning failure [4]. Nevertheless, patients with atrial fibrillation exhibited a single mitral flow. Some patients present with partial or complete merging of the E and A wave due to tachycardia, despite a normal sinus rhythm. E/A could not be used in these cases and this limits its utilization for clinical practice at the bedside. Nagueh et al. demonstrated that E/Ea could be used to assess the LVFP with good accuracy in sinus tachycardia [21], even with complete merging of the E and A waves, and in the same way in patients with atrial fibrillation [12]. Patients with atrial fibrillation cannot be excluded from this study because they represent 20 to 30% of patients hospitalized in intensive care unit. Our study is the first to include such patients. For all these reasons, E/Ea seems to be the measure of choice to predict weaning failure at the bedside, especially in those with a preserved or slightly reduced LVEF, atrial fibrillation or in those with single mitral flow due to sinus tachycardia.

The second major finding of this study shows that the systolic function assessed by LVEF was not associated with weaning outcome. This is in accordance with previous studies. Mekontso et al. [22] showed, in a subgroup of patients who underwent TTE, that the LVEF did not differ between patients in the failed or in the successful group. Similar data were obtained by Lamia et al. [4] and more recently by Zapata et al. [23]. This apparent paradox is explained by the fact that in studies into cardiology, 40 to 50% of patients with typical signs of heart failure have a normal or slightly reduced LVEF [24‐27]. So the LVEF is not a good predictor of clinical disability and suggests that congestive symptoms are more closely related to the diastolic properties of the ventricle than to the systolic properties [24, 25]. The diagnosis of diastolic abnormalities can be confirmed when Doppler with tissue imaging provides evidence of impaired myocardial relaxation [7, 28, 29]. The Ea velocity constitutes a non-invasive and relatively load independent index of LV relaxation and diastolic function. Relaxation impairment is defined by an Ea velocity below or equal to 8 cm/sec [7]. In our study, in comparison to the successful group, patients in the failed group had a lower Ea velocity and exhibited more frequent impairment in relaxation. Abnormal relaxation resulted in a prolongation of the isovolumic relaxation time with an upward shift and to the left of the diastolic pressure/volume curve (Figure 2) [24]. In such patients, a relatively small increase in the end diastolic volume may lead to an exaggerated increase in LVFP despite a normal end diastolic volume [24]. This increased sensitivity to volume changes could explain the occurrence of pulmonary edema during the SBT because an inspiratory drop in the intra-thoracic pressure tends to increase the systemic venous return and LV preload.

In our study, the evolution of the relaxation rate was also important. The failed group showed a reduced Ea before the SBT, and the Ea did not vary during the SBT, whereas in the successful group the Ea increased during the SBT. This finding explains why E wave increased during SBT in both groups. This change in mitral flow reflects an enhancement in the LV relaxation rate for the successful group while it reflects a major increase in LVFP for the failed group [30]. This is in agreement with previous studies that demonstrated that in healthy individuals the Ea velocity increased with exercise; however, in patients with abnormal relaxation and a reduced Ea velocity, the Ea did not vary with exercise or with loading conditions [31‐33]. During exercise, like during the SBT, the LV filling rate could increase dramatically [31]. Normally, this increased filling occurred because the left ventricle early diastolic pressure decreased with enhanced myocardial relaxation. Thus a substantial increase in filling could occur without an increase in LVFP [32]. This normal response to exercise was lost in heart failure in patients with diastolic dysfunction who were not able to enhance relaxation with exercise [32]. In patients with impairment in relaxation, increases in LV filling in response to exercise were dependent on increases in LVFP [32]. Because the cardiovascular response to weaning from mechanical ventilation is similar during exercise [34], these considerations could explain the marked increase in LVFP in patients with diastolic dysfunction during the SBT. To our knowledge, this is the first study demonstrating that relaxation impairment and the inability to enhance LV relaxation during the SBT were closely associated with the risk of weaning failure.

Our study holds several limitations. In our study, diastolic dysfunction was associated with weaning failure; however, we included a large proportion of patients with chronic obstructive pulmonary disease [35], diabetes [36], hypertensive history, and atrial fibrillation [37]; all these clinical disorders are known to be associated with diastolic dysfunction. At present, this particular population precludes generalization of our findings to all populations of intensive care units. Another limitation is the poor echocardiographic windows in some patients under mechanical ventilation; seven patients were excluded because of suboptimal images.

The diagnosis of weaning-induced pulmonary edema was made using a clinical approach; we did not assess invasive PAOP. However, the diagnosis of cardiac dysfunction during weaning is still difficult and the clinical presentation is often non-specific. In contrast, the diagnosis of weaning failure is clearly defined; this is why we chose to evaluate the utility of serial measurement of E/Ea to predict weaning failure and not the risk of cardiac dysfunction.

Transthoracic echocardiography with measurement of the E/Ea ratio can identify patients at risk of weaning failure at the bedside. A reduced LVEF did not seem to influence the weaning outcome, whereas diastolic abnormalities with impaired relaxation and no ability to enhance relaxation during the SBT are closely associated with weaning failure.