Electrical impedance tomography during spontaneous breathing trials and after extubation in critically ill patients at high risk for extubation failure: a multicenter observational study

Background This study aims to assess the changes in lung aeration and ventilation during the first spontaneous breathing trial (SBT) and after extubation in a population of patients at risk of extubation failure. Methods We included 78 invasively ventilated patients eligible for their first SBT, conducted with low positive end-expiratory pressure (2 cm H2O) for 30 min. We acquired three 5-min electrical impedance tomography (EIT) records at baseline, soon after the beginning (SBT_0) and at the end (SBT_30) of SBT. In the case of SBT failure, ventilation was reinstituted; otherwise, the patient was extubated and two additional records were acquired soon after extubation (SB_0) and 30 min later (SB_30) during spontaneous breathing. Extubation failure was defined by the onset of post-extubation respiratory failure within 48 h after extubation. We computed the changes from baseline of end-expiratory lung impedance (∆EELI), tidal volume (∆Vt%), and the inhomogeneity index. Arterial blood was sampled for gas analysis. Data were compared between sub-groups stratified for SBT and extubation success/failure. Results Compared to SBT success (n = 61), SBT failure (n = 17) showed a greater reduction in ∆EELI at SBT_0 (p < 0.001) and SBT_30 (p = 0.001) and a higher inhomogeneity index at baseline (p = 0.002), SBT_0 (p = 0.003) and SBT_30 (p = 0.005). RR/Vt was not different between groups at baseline but was significantly greater at SBT_0 and SBT_30 in SBT failures, compared to SBT successes (p < 0.001 for both). No differences in ∆Vt% and arterial blood gases were observed between SBT success and failure. The ∆Vt%, ∆EELI, inhomogeneity index and arterial blood gases were not different between patients with extubation success (n = 39) and failure (n = 22) (p > 0.05 for all comparisons). Conclusions Compared to SBT success, SBT failure was characterized by more lung de-recruitment and inhomogeneity. Whether EIT may be useful to monitor SBT remains to be determined. No significant changes in lung ventilation, aeration or homogeneity related to extubation outcome occurred up to 30 min after extubation. Trial registration Retrospectively registered on clinicaltrials.gov (Identifier: NCT03894332; release date 27th March 2019). Electronic supplementary material The online version of this article (10.1186/s13613-019-0565-0) contains supplementary material, which is available to authorized users.


Background
Weaning is the whole process that leads patients to the discontinuation of mechanical ventilation and the removal of the endotracheal tube [1]. Weaning should be considered as early as possible to reduce the time spent on invasive mechanical ventilation (iMV), which is associated with morbidity and mortality [2,3]. To assess whether patients are eligible for extubation, a spontaneous breathing trial (SBT) is often performed. The outcome of SBT is evaluated by clinical assessment and acquisition of objective parameters, such as the arterial blood gases (ABGs) and the respiratory rate (RR) to tidal volume (Vt) ratio (RR/Vt) [1]. In the case of SBT success, the patient can be extubated. However, post-extubation respiratory failure may occur, in particular in patients at increased risk [1], who may benefit from prophylactic non-invasive ventilation (NIV) application immediately after extubation [4][5][6][7][8][9].
Electrical impedance tomography (EIT) is a non-invasive imaging technique that provides instantaneous monitoring of variations in overall lung volume and regional distribution of ventilation, as determined by variations over time in intrathoracic impedance, which is increased by air and reduced by fluids and cells. EIT allows for determining changes in end-expiratory lung impedance (EELI), a surrogate estimate of end-expiratory lung volume, assessing global and regional distribution of Vt, and obtaining indexes of the spatial distribution of ventilation, such as the global inhomogeneity index [10].
Because SBT failure and extubation failure are both major clinical problems, they have been repeatedly investigated, but the underlying mechanisms are not fully elucidated yet [1]. We reasoned that EIT could help comprehend the pathophysiology of SBT failure and extubation failure by providing additional insights. We, therefore, designed this observational study on patients at high risk of extubation failure, aimed at assessing variations in EELI (∆EELI) and Vt (∆Vt%), and the inhomogeneity index during SBT and after extubation.

Methods
The study was conducted from June 2015 to June 2016 in the intensive care units (ICUs) of three Italian hospitals ("Sant' Andrea" Hospital in Vercelli, "ARNAS Garibaldi" Hospital in Catania and "Infermi" Hospital in Rimini), in accordance with the principles outlined in the Declaration of Helsinki, after approval by the local Ethics Committees (Vercelli, protocol no AslVC.Rian.14.02, approved on 11/09/2014; Catania, protocol no 479/ CE, approved on 30/06/2015; Rimini, protocol no 1313, approved on 18/03/2015). Written informed consent was obtained from all patients according to national regulations.

Patients
We included any ICU patient ≥ 18 years who (1) had received iMV for at least 48 h through an orotracheal tube, (2) was ready for the first SBT attempt [1] and (3) met at least one criterion for increased risk of extubation failure, as outlined in the first column of Additional file 1: Table S1 [4][5][6][7][8][9]11]. Patients were considered to be ready for SBT if they satisfied all of the criteria reported in the second column of Additional file 1: Table S1 [1,4,12]. Patients were excluded if they met one or more of the following criteria: (1) life-threatening heart arrhythmias or cardiac ischaemia, (2) pneumothorax, (3) pulmonary emphysema, (4) recent (1 week) thoracic surgery, (5) presence of chest burns, (6) pregnancy, (7) inclusion in other research protocols, (8) refusal of consent.
Inclusion and exclusion criteria were assessed daily in all patients.

Weaning protocol
Once able to trigger the ventilator, patients were ventilated with pressure support ventilation (PSV). Inspired fraction of oxygen (FiO 2 ) and positive end-expiratory pressure (PEEP) were set to maintain peripheral oxygen saturation (SpO 2 ) between 92 and 96%, while the inspiratory pressure support was titrated to generate a Vt of 6-8 ml/kg of ideal body weight with an active inspiration [13]. The SBT was conducted by setting the ventilator with 2 cm H 2 O in continuous positive airway pressure (CPAP) mode, with no inspiratory support for 30 min [12,14]. SBT failure was defined by the presence of at least one of the criteria listed in the third column of Additional file 1: Table S1 [1,15]. Patients who passed the SBT were immediately extubated and allowed to breathe spontaneously with a Venturi mask, while those who failed SBT were switched back to PSV with the same settings applied prior to SBT.

Data acquisition and analysis
After enrolment, a silicon 16-electrode EIT belt of proper size was placed around the patient's chest between the 4th and 6th intercostal spaces and connected to the EIT device (PulmoVista 500; Draeger Medical GmbH, Lübeck, Germany) [16,17]. Signal quality was determined through dedicated built-in software. The position of EIT belt was, therefore, marked on the skin with a dermographic pencil to avoid its displacement during the study period. PSV was applied with a ventilator (Infinity V500; Draeger Medical GmbH, Lübeck, Germany) connected to the EIT device through a RS232 interface.
We acquired 5-min EIT data records at baseline (during PSV), during the first (SBT_0) and the last (SBT_30) 5 min of SBT, and when the patient was extubated during spontaneous breathing immediately (SB_0) and 30 min (SB_30) after extubation.
EIT and ventilator (i.e., airway pressure, flow and volume) data were recorded at a sample of 20 Hz, downloaded and analysed off-line on a personal computer using dedicated software (EITdiag, Draeger Medical GmbH, Lübeck, Germany). Impedance changes were calibrated against the data acquired from the ventilator by the EIT device as previously described [18]. The last 3 min of each record were analysed.
The inhomogeneity index was computed as follows: GI = x,y lung DI xy − Median DI lung x,y lung DI xy where DI is the differential impedance, DI xy refers to the DI of pixels in a defined lung region and DI lung represents the DI of the whole lung [20]. ABGs were obtained at baseline, SBT_30 and SB_30. Extubation failure was defined by the occurrence of post-extubation respiratory failure, as characterized by one or more of the criteria outlined in the fourth column of Additional file 1: Table S1, in the 48 h following extubation [1,21]. Patients who failed extubation underwent ''rescue'' CPAP by means of a high-flow (> 45 l/min) generator (Flow-Meter, StarMed, Mirandola, Modena, Italy) in the case of normocapnic hypoxemia (i.e., PaO 2 / FiO 2 < 200 mmHg), or NIV, as delivered through NIV modality, in the case of respiratory acidosis (pH < 7.35 and PaCO 2 > 50 mmHg). Patients were re-intubated and iMV reinstituted if one or more of the criteria displayed in the fifth column of Additional file 1: Table S1 occurred [21].

Statistical analysis
Due to the lack of published data at the time of the study design, no power analysis could be performed. Therefore, Table 1 Patient's characteristics at ICU admission, risk factors for post-extubation respiratory failure, the reason for ICU admission and characteristics before SBT performance in the overall population and subgroups SBT: Spontaneous breathing trial; ICU: intensive care unit; SAPS-II: simplified acute physiology score II; SOFA: sequential organ failure assessment; PaCO 2 : arterial partial pressure of carbon dioxide; iMV: invasive mechanical ventilation; APACHE-II: acute physiology and chronic health evaluation II; ARF: acute respiratory failure; BMI: body mass index; PEEP: positive-end expiratory pressure; PaO 2 /FiO 2 : ratio between arterial partial pressure and inspired fraction of oxygen Receiving operating curves (ROC) were obtained for RR/Vt and PaO 2 /FiO 2 and EIT data. Cut-off values were obtained calculating the Youden index. The area under the curve (AUC), sensitivity, specificity, positive (LR +) and negative (LR −) likelihood ratios were also calculated.

Results
We enrolled 80 consecutive patients, as planned. Two patients, erroneously enrolled with no risk criteria for post-extubation respiratory failure, were excluded from the data analysis. The patients' flow of study enrolment is shown in Additional file 1: Figure S1, while the protocol flow is depicted in Fig. 1. None of the patients interrupted the study protocol. Table 1 shows the patients' characteristics at ICU admission and immediately before SBT, and the reasons for ICU admission and risk factors for post-extubation respiratory failure, for the overall population and separately for the SBT and extubation success and failure subgroups.

SBT success and failure
Seventeen patients (21.8%) failed SBT and were returned to iMV. At baseline, PEEP values were similar between groups (5.0 [4.9; 5.1] cm H 2 O and 5.0 [5.0; 5.0] cm H 2 O for SBT success and failure, respectively; p = 0.168). Table 2 reports ABGs and RR/Vt separately for SBT success and failure. ABGs were not significantly different between groups at both baseline and SBT_30. RR/Vt was also not different between the groups at baseline but was significantly greater at SBT_0 and SBT_30 in SBT failures compared to SBT success (p < 0.001 for both).
Examples of EIT images from three representative patients, two succeeding (panels A and C) and one failing (panel B) SBT, are depicted in Fig. 2; ∆EELI and the inhomogeneity index are shown at baseline, SBT_0 and SBT_30 (see figure legend for detailed information). Table 3 displays the values of ∆Vt%, ∆EELI and inhomogeneity index in SBT success and failure, respectively. The ∆Vt% was no different at baseline (p > 0.999), SBT_0 (p = 0.079) and SBT_30 (p = 0.054). Compared to SBT successes, SBT failures showed a greater reduction in ∆EELI at SBT_0 (p < 0.001) and SBT_30 (p = 0.001), and a higher inhomogeneity index at baseline (p = 0.002), SBT_0 (p = 0.003) and SBT_30 (p = 0.005).

Table 2 Arterial blood gases and the rapid shallow breathing index (RR/Vt) in patients with SBT success and failure
SBT, Spontaneous Breathing Trial; SBT_0, first 5 min after the beginning of the SBT; SBT_30, last 5 min of the SBT; PaCO 2 , arterial partial pressure of carbon dioxide; PaO 2 /FiO 2 , arterial partial pressure to inspired fraction of oxygen ratio; RR/Vt, respiratory rate to tidal volume ration //Data not present a Comparison between groups within baseline b Comparison between groups within SBT_0 c Comparison between groups within SBT_30. According to Bonferroni correction, the threshold for statistical significance is p < 0.025 for pH, PaCO 2

Extubation success and failure
Twenty-two of 61 (36.1%) patients extubated after SBT success met criteria for extubation failure: 15 for ABG worsening and respiratory distress and 7 for non-respiratory reasons (5 for neurological deterioration, one for inability to remove secretions and one for haemodynamic instability). Eleven patients were successfully treated by CPAP or NIV, while 11 required endotracheal intubation, 3 immediately and 8 after an unsuccessful attempt of CPAP or NIV. As shown in Additional file 1: Table S3, there were no differences between extubation success and failure in pH (p = 0.397 at baseline; p = 0.453 at SBT_30; p = 0.179 at SB_30), PaCO 2 (p = 0.176 at baseline; p = 0.467 at SBT_30; p = 0.757 at SB_30) and PaO 2 /FiO 2 (p = 0.932 at baseline; p = 0.636 at SBT_30; p = 0.139 at SB_30). RR/ VT was also not different among trials (p = 0.712 at baseline; p = 0.704 at SBT_0; p = 0.597 at SBT_30; p = 0.059 at SB_0; p = 0.185 at SB_30). Table 4 reports the ∆Vt%, ∆EELI and inhomogeneity index for extubation success and failure. No differences were observed for any of these variables between extubation success and failure.
The AUC, sensitivity, specificity, LR + and LR − of ∆Vt%, ∆EELI, inhomogeneity index, RR/Vt and PaO 2 / FiO 2 for extubation failure prediction are shown in Additional file 1: Table S4. In addition, as reported in Additional file 1: Tables S5 and S6, we found no difference in any EIT parameters between patients succeeding and failing NIV and between patients with post-extubation respiratory failure due to respiratory and non-respiratory causes.

Discussion
In this physiological study aimed at describing, in a general population of critically ill patients at the first SBT attempt, changes in lung aeration, ventilation distribution and inhomogeneity occurring during the weaning process, we found the following: (1) compared to SBT success, SBT failure is characterized by more lung derecruitment, as suggested by the higher loss in ∆EELI and by the concomitant PaO 2 /FiO 2 worsening, and greater lung inhomogeneity, as indicated by higher inhomogeneity index value; and (2) no EIT variables were significantly different between patients succeeding and failing extubation.
All patients considered ready-to-wean were considered eligible for inclusion in this study, and we did not focus only on patients with prolonged weaning, who represent only approximately 15% of the overall ICU population [1]. We found an SBT failure rate of 21% which was not different from those previously reported [22][23][24], while in keeping with Ferrer et al. [6], the rate of extubation failure in our population at high risk of post-extubation respiratory failure amounted to 35%. Finally, consistent with recently published data in patients recovering from hypoxemic acute respiratory failure [25], 50% of all patients experiencing post-extubation respiratory failure required reintubation, 42% of whom after failing CPAP or NIV. The application of EIT during weaning has been recently reported only in small populations of patients characterized by prolonged weaning or prolonged mechanical ventilation [26][27][28]. These studies report that, compared to patients undergoing a successful SBT, patients experiencing SBT failure are characterized by greater loss of EELI at the end of the SBT. While confirming this finding in a less selected patient population, we find that the loss of EELI occurs after only 5 min.
In the present study, the inhomogeneity index appears to be greater in SBT failure, as opposed to SBT success, which is also consistent with previously published data. In patients who had received prolonged iMV, Zhao et al. [27] also found that weaning was more likely to be successful when the ventilation distribution was more homogeneous at decreasing support levels. In 31 patients with prolonged weaning, Bickenbach et al. [26] found the inhomogeneity index to be greater in SBT failures, as opposed to successes. In our study, the inhomogeneity index was also significantly higher at baseline, while RR/Vt showed significant differences only in the course of SBT. In addition, after 5 min of SBT, we observed a remarkable increase in the inhomogeneity index, without important changes at SBT completion.
We also investigated potential variations in EIT parameters related to the outcome of extubation but found no differences in any of the EIT parameters considered between patients who succeeded and failed extubation. This finding may suggest, on the one hand, that the process leading to failure takes more than the 30-min period we observed in the present study, while on the other hand, that factors other than alterations of ventilation and homogeneity intervene in the pathophysiology of post-extubation respiratory failure. Indeed, in 7 patients (32%), the primary causes of extubation were either unmanageable secretions or haemodynamic instability or neurologic deterioration. In addition, we failed to observe differences in any EIT parameters between patients who succeeded and failed NIV as rescue treatment of overt post-extubation respiratory failure.
Our study has some points of strength. First, in contrast to previous investigations [26][27][28], this is a multicentre study. Second, to reduce the risk of bias, the data analysis was performed in only one centre by a single investigator unaware of SBT and extubation outcomes. Third, we were extremely careful in maintaining the position of the EIT belt throughout the study period by marking the skin with a dermographic pencil, to prevent the risk of belt displacement, which is a well-known source of bias [10].
Our study also includes some limitations. We did not calculate the sample size because of the lack of published data at the time this study was designed and initiated. We did not register the trial prior to patient enrolment because the registration was not mandatory for observational studies when the study was initiated. We did not record the amount of fluids administered in the course of the study protocol. Fluid overload alters the measurement of EELI, mimicking a reduction in end-expiratory lung volume. One study in pigs showed that a rapid infusion (33 ml/kg/h) of fluids caused a significant reduction in overall relative impedance [29]. In another animal study, after a rapid (approximately 6 min) infusion of 500 ml of crystalloids the authors observed a decrease in EELI, corresponding to a mean virtual reduction in end-expiratory lung volume of 227 ml [30]. Conversely, a sudden increase in urine output up to 1200 ml consequent to a bolus of diuretics [31] or a loss of volume during haemodialysis [32] may determine EELI to increase. It should be noted, however, that a stable cardiovascular Table 3 EIT data in patients with SBT success and failure SBT: Spontaneous Breathing Trial; SBT_0: first 5 min after the beginning of the SBT; SBT_30: last 5 min of the SBT; ∆Vt%: change from baseline of the tidal volume in percentage; ∆EELI: change from baseline of the end-expiratory lung impedance a Comparison between groups within baseline b Comparison between groups within SBT_0 c Comparison between groups within SBT_30. According to Bonferroni correction, the threshold for statistical significance is p < 0.017