Introduction
Weaning patients from the ventilator remains a crucial issue. In 2,500 patients included in 6 large randomized trials, the incidence of weaning failure, which is defined as a failed spontaneous breathing trial (SBT) or the need for a re-intubation in the 48 hours following extubation, reached 31% [
1]. Weaning failure remains a clinically relevant challenge because it may result in significant morbidity (prolonged duration of mechanical ventilation, re-intubation) and may influence mortality [
2,
3].
Even if its actual incidence is unknown, cardiac dysfunction is a leading cause of weaning failure [
1]. Breathing in the context of weaning was described as a physical exercise [
4]. The abrupt cessation of positive pressure ventilation increases venous return and left ventricular (LV) afterload [
5], decreases LV compliance [
6], and may even induce cardiac ischemia [
7]. All these factors tend to increase LV filling pressure [
6,
8], and may subsequently result in cardiogenic pulmonary edema. Right heart catheterization has long been used in this clinical setting to detect an increase of the pulmonary artery occlusion pressure (PAOP) [
6,
9]. Nevertheless, PAOP may be difficult to precisely measure in the presence of large swings in intra-thoracic pressure, as observed in spontaneously breathing patients with increased inspiratory efforts [
10].
Although echocardiography allows the noninvasive assessment of cardiac function and LV filling pressures, its clinical value in the setting of ventilator weaning has yet to be determined. In this prospective, descriptive, bicentric study, we sought to evaluate the ability of transthoracic echocardiography (TTE) to detect the effects of SBT on central hemodynamics and to potentially identify indices that could help predicting weaning failure from cardiac origin.
Results
Of 142 eligible patients, 25 were excluded because of the absence of adequate apical four-chamber view (n = 7), atrial fibrillation (n = 13), paced cardiac rhythm (n = 4), and agitation precluding image acquisition (n = 1). Accordingly, 117 patients were studied (71 men, 46 women, age: 63 (58 to 67) years [median (25
th to 75
th percentile)]; SAPS II: 53 (47 to 58)). Sixteen patients (14%) had pre-existing chronic obstructive pulmonary disease (COPD). Reasons for intubation were acute respiratory failure (n = 20), severe sepsis or septic shock (n = 42), cardiogenic shock (n = 40), and neurological disorders and stroke (n = 15). Duration of mechanical ventilation was 5 (5 to 6.5) days and overall mortality in the ICU reached 10% (Table
1). Before the SBT, levels of pressure support and PEEP were 12 (10 to 12) cmH
2O and 3 (3 to 4) cmH
2O, respectively. Twelve percent of patients had already received diuretics in the 48 hours preceding the SBT.
Table 1
Characteristics of the study population
Age (years) | 63 (58-67) |
Gender (men/women) | 71/46 |
Height (cm) | 168 (166-170) |
Weight (kg) | 75 (71-79) |
SAPS II | 53 (47-58) |
Pre-existing cardiac disease, n (%) | 35 (30) |
Hypertension, n (%) | 26 (33) |
Pre-existing COPD, n (%) | 16 (14) |
Dobutamine infusion, n (%) | 10 (9) |
Dose μg/kg/min | 4.8 (3.5-5.0) |
Epinephrine-norepinephrine, n (%) | 15 (15) |
Dose μg/kg/min | 0.11 (0.08-0.14) |
Diuretic therapy before SBT, n (%) | 14 (12) |
Duration of mechanical ventilation (days) | 5.0 (5.0-6.5) |
Mortality, n (%) | 11 (10) |
SBT was unsuccessful in 11 patients (9%) for weaning-induced pulmonary edema. Among the 106 extubated patients, 12 (11%) were re-intubated for post-extubation cardiogenic pulmonary edema (n = 9), weakness and increased airway secretions (n = 2), and for stridor in the remaining patient. Overall, the incidence of weaning failure was 20% (23/117 patients), and related to a cardiac origin in 20 of 23 patients (87%). No episode of cardiac ischemia was documented, both during SBT and respiratory distress requiring reintubation. Three patients exhibited atrial fibrillation and two patients had marked sinus tachycardia (> 140 bpm) during the SBT, three of whom failing the weaning process.
During the SBT, cardiac output and systolic arterial pressure significantly increased (Table
2). The increase in cardiac output was related to SBT-induced tachycardia because LV stroke volume remained unchanged. Mitral E/A significantly increased and E/E' tended to increase without reaching statistical significance, while DTE significantly decreased (Table
2). No change in RV size was observed, as reflected by the same median RVEDA/LVEDA ratio measured during PS/PEEP and SBT (Table
2). No moderate-to-severe mitral regurgitation was observed.
Table 2
Echocardiographic findings in the 117 patients during pressure support ventilation (PS/PEEP) and the spontaneous breathing trial (SBT)
HR (/min)
| 95 (90-99) | 98 (91-103) | 0.00001 |
SAP (mmHg)
| 138 (132-143) | 145 (136-152) | 0.019 |
CO (L/min)
| 5.8 (5.2-6.2) | 6.0 (5.4-6.7) | 0.004 |
SV (mL)
| 62 (57-67) | 64 (58-70) | 0.5 |
E/A
| 0.94 (0.82-1.05) | 1.00 (0.88-1.15) | 0.003 |
DTE (ms)
| 168 (150-187) | 147 (132-160) | 0.00001 |
E/E'
| 5.9 (5.4-6.4) | 6.5 (5.7-7.2) | 0.16 |
RVEDA/LVEDA
| 0.47 (0.44-0.50) | 0.47 (0.44-0.50) | 0.79 |
Weaning failure was observed in 10 of 58 patients (17%) with an LVEF above 50%, in 4 of 30 patients (13%) with an LVEF between 35 to 50%, and in 9 of 29 patients (31%) with an LVEF below 35% (
P < 0.05). At baseline (PS/PEEP), E/A was similar between groups, whereas DTE was significantly lower and E/E' significantly higher in patients with a LVEF below 35% (Table
3). During SBT, E/A significantly increased and DTE significantly decreased solely in patients with an LVEF of 50% or less. E/E' tended to increase during SBT in patients with LVEF of 35% or more without reaching statistical significance, whereas it remained elevated with no further increase in patients with a LVEF below 35% (Table
3). In patients who failed to be weaned from the ventilator due to a cardiogenic pulmonary edema (n = 20), both the median E/A and E/E' ratios increased significantly from PS/PEEP to SBT (1.05 (0.64 to 1.81) vs. 1.19 (0.71 to 3.18):
P < 0.01 and 8.0 (4.8 to 9.3) vs. 8.7 (5.3 to 10.8):
P < 0.05, respectively), whereas median DTE decreased from 105 ms (87 to 185) to 90 ms (75 to 133;
P < 0.05).
Table 3
Doppler echocardiographic findings according to baseline left ventricular ejection fraction (recorded under pressure support ventilation)
E/A
| | | | |
| PS/PEEP | 0.97 (0.81-1.09) | 0.82 (0.74-1.0) | 1.0 (0.78-1.47) |
| SBT | 0.99 (0.85-1.20) | 0.89* (0.76-1.32) | 1.0* (0.86-1.89) |
DTE
(ms) | | | | |
| PS/PEEP | 174 (152-193) | 187 (144-224) | 119
¶
(90-153) |
| SBT | 163 (155-181) | 140* (112-177) | 96* (80-137) |
E/E'
| | | | |
| PS/PEEP | 5.2 (4.7-6.0) | 6.0 (5.3-9.0) | 7.9
¶
(5.4-9.1) |
| SBT | 5.8 (4.9-6.6) | 6.9 (4.9-9.3) | 7.8 (6.7-9.7) |
In patients who could not be weaned from the ventilator, LVEF was significantly lower and E/E' was significantly higher than in those who underwent weaning success (Table
4). Despite a significantly higher heart rate, patients who failed to be weaned had a lower cardiac output due to a substantial reduction of LV stroke volume. E/A was similar between the two subsets of patients whereas DTE was shortened in the case of weaning failure, although not reaching statistical significance (Table
4).
Table 4
Patients' characteristics prior to SBT, according weaning success or failure
SAP (mmHg)
| 139 (133-147) | 132 (115-149) | 0.28 |
HR (bpm)
| 92 (86-97) | 110 (95-120) |
0.007
|
SV (mL)
| 63 (57-70) | 60 (39-66) | 0.06 |
CO (L/min/m
2
)
| 5.8 (5.2-6.3) | 5.4 (3.3-6.5) | 0.19 |
LVEF (%)
| 51 (43-55) | 36 (27-55) |
0.04
|
E/A
| 0.94 (0.82-1.03) | 0.88 (0.68-1.65) | 0.7 |
DTE (ms)
| 170 (150-189) | 138 (98-195) | 0.07 |
E/E'
| 5.6 (5.2-6.3) | 7.0 (5.0-9.2) |
0.038
|
RVEDA/LVEDA
| 0.47 (0.44-0.51) | 0.48 (0.43-0.52) | 0.99 |
Discussion
In the present study, we showed that TTE can accurately depict changes in central hemodynamics induced by SBT and potentially select patients at high risk of cardiac-related weaning failure. In our patients, SBT resulted in a significant increase in heart rate and in cardiac output well-reflecting the greater work of breathing which has been compared with a true exercise [
4]. We also observed a rise in systolic arterial pressure, which was consistent with greater LV afterloading. Although not statistically significant, the increase in E/E' during the SBT is in keeping with an increase in LV filling pressure induced by the shift from positive pressure ventilation to spontaneous breathing. More significant, both the increase of E/A and shortened DTE during the SBT suggest that altered LV diastolic properties may also potentially contribute to the rise in LV filling pressure [
14]. Similar results were found by Ait-Oufella and colleagues [
18] in 31 patients who were successfully weaned from the ventilator. In the present study, median RV/LV end-diastolic area was similar prior to and at the end of the SBT. This presumably reflects the absence of SBT-induced pulmonary hypertension in our study population, which comprised a low proportion of COPD patients.
Under PS/PEEP, E/A was not a discriminating Doppler parameter between subgroups of patients based on baseline LVEF. In contrast, DTE was significantly reduced in patients with severely depressed LV systolic function when compared with other subsets of patients, and E/E' ratio significantly increased with the deterioration of LV systolic function. ICU patients with LV systolic dysfunction who enter the process of weaning from the ventilator usually have underlying heart disease with associated LV diastolic dysfunction, reduced compliance and increased filling pressure [
19]. In our patients with LVEF below 35%, shortened DTE presumably reflected underlying LV diastolic dysfunction, while the gradual increase in E/E' ratio across the three subsets of patients was consistent with the progressive elevation of cardiac filling pressure which was associated with the deterioration of LV compliance [
20]. Interestingly, a statistically significant increase of E/A ratio and shortening of DTE during SBT was only observed in the subsets of patients with LV dysfunction. This suggests further deterioration in LV diastolic properties induced by SBT, which may be attributed to decreased LV compliance [
6] or even silent cardiac ischemia [
7], as previously suggested in ICU patients with associated LV systolic dysfunction [
18]. In our patients with LVEF below 35%, median E/E' at baseline (under PS/PEEP) was as high as 7.9 and failed to further increase during the SBT. We previously showed in ventilated ICU patients that a lateral E/E' ratio above 8.0 predicted a PAOP of more than 18 mmHg with a 83% sensitivity and a 88% specificity [
21]. Similarly, Lamia and colleagues recently reported that the conjunction of an E/A ratio above 0.95 and E/E' ratio above 8.5 at the end of the SBT in a selected population of patients difficult to wean allowed predicting a PAOP of 18 mmHg or more with a 82% sensitivity and a 91% specificity [
22]. Interestingly, no significant change in E/E' was observed during SBT in our patients, regardless of LVEF. In 102 ICU patients, Mekontso-Dessap and colleagues reported that circulating brain natriuretic peptide (BNP), a biomarker correlated with LV filling pressure, failed to increase at the end of a 60-minute SBT [
23], whereas Grasso and colleagues showed that NT-proBNP increased only in those patients who developed acute cardiac dysfunction during the SBT [
24].
When compared with patients who successfully completed the weaning process, patients under PS/PEEP who failed had significantly lower LVEF and higher E/E', and tended to have shorter DTE. In addition, the subset of patients who failed to be weaned from the ventilator exhibited a significant increase of both E/A and E/E' and a shortened DTE at the time of the SBT, when compared with PS/PEEP. In keeping with our results, Mekontso-Dessap and colleagues found that circulating BNP was significantly increased in patients under PS/PEEP who finally failed the weaning process, as a result of an overloaded LV [
23]. TTE appears ideally suited to routinely screen patients at risk of weaning-related pulmonary edema (e.g., chronic obstructive pulmonary disease, heart failure) prior to the SBT. Patients with a severely decreased LVEF (< 35%) should be considered at high risk of cardiac-related weaning failure, particularly when exhibiting shortened DTE and elevated E/E'. These TTE indices are simple yet robust and fairly reproducible [
17]. This screening strategy could potentially help the intensivist to better select patients for tailored therapy as an attempt to facilitate the weaning process (e.g., diuretics, control of systolic blood pressure, non invasive pressure support ventilation after extubation). In the present study, a large proportion of patients with predisposing heart or lung disease would have been eligible for such a clinical approach which still remains to be validated.
In this clinical setting, TTE must be performed by an experienced intensivist with an advanced level in critical care echocardiography [
25], because image acquisition and interpretation is frequently challenging. Intensivists with competence for basic level critical care echocardiography have the ability of qualitatively assessing LVEF, but are not adequately trained to precisely evaluate LV diastolic properties and filling pressures [
25]. Such training requirement is undoubtedly a substantial limitation of this TTE-based approach. Only four of our patients (5%) were excluded because of poor echogenicity because all examinations have been performed by the same experienced operators. In contrast, Grasso and colleagues reported in COPD patients a 26% rate of inadequate four-chamber view [
24]. This apparent discrepancy is presumably explained by the markedly lower proportion of COPD patients in our study and by the measurement of LVEF which requires an optimal visualization of the endocardial border in the study by Grasso and colleagues. We only used pulse-wave Doppler during the SBT which remains possible to acquire and accurate even in the presence of suboptimal two-dimensional image quality.
Our study has several limitations. First, the physiologically interesting changes in E/A and DTE during SBT cannot be used in clinical practice to select patients at high risk of cardiac-related weaning failure. Indeed, the fairly large inter-patient variability observed in absolute variations of Doppler parameters induced by the SBT precluded the determination of a clinically useful threshold value. However, we found that the echocardiographic profile during PS/PEEP was more relevant to select those patients at high risk of cardiac-related weaning failure. Second, we could not clearly identify using TTE whether LV diastolic dysfunction or increased LV filling pressure was the leading cause of weaning failure in our patients. As these two entities are closely linked [
19,
20], a combined effect is presumably operant in the setting of weaning process from the ventilator. Third, up to 15% of our patients had received diuretics during the 48 hours preceding the SBT. This may have introduced a relevant bias of selection in our study and led to soften the effects of SBT on central hemodynamics, especially in patients with severely reduced LV function. This may also have reduced the proportion of weaning pulmonary edema and might explain the lower incidence of weaning failure in our series (20%) when compared with previous larger cohorts of patients with a mean 31% rate of weaning failure [
1]. Fourth, silent myocardial ischemia cannot be confidently excluded in our patients because a conventional 12-lead ECG was not recorded prior to and at the end of SBT. Fifth, we purposely excluded patients with non sinus rhythm because Doppler indices are more challenging to precisely measure in this setting. Nevertheless, DTE and E/E' have also been validated to evaluate LV filling pressure in patients with atrial fibrillation [
26,
27]. Finally, we did not investigate the additional value of combining a biological marker to echocardiography to better evaluate the cardiovascular system during weaning.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
VC, JBA, CC, GB, AVB, PV participated to the elaboration of the study project, the enrollment of patients and performance of echocardiographic examinations, and they participated to data analysis. VC, AVB and PV contributed to the preparation of the manuscript.