Pleural effusion during weaning from mechanical ventilation: a prospective observational multicenter study

Weaning initiation was defined as the day of first SBT. The first SBT used a T-piece trial in three centers and a low-level pressure support (7–10 cm H₂O) with zero end-expiratory pressure in two centers, as per usual care. Failure of the SBT was based on predefined criteria (see the online supplement, Additional file 1). Extubation failure was defined as death or reintubation within the 7 days following extubation; this delay was used instead of 48–72 h because prophylactic noninvasive ventilation may postpone reintubation [11]. Indications for prophylactic noninvasive ventilation included patients older than 65 years and those with underlying cardiac or respiratory disease [12]. According to the International Consensus Conference [10], weaning success was defined as a first successful SBT followed by successful extubation. Failure of the weaning process was defined [10] as failure of the first SBT or extubation failure. Because some patients could not be classified with this definition, weaning was also categorized according to the WIND definition [13] as follows: short when the first SBT resulted in a successful termination of the weaning process or death within 1 day after the first SBT; difficult in case of successful weaning or death after more than 1 day but in less than 1 week after the first SBT; prolonged if weaning was still not terminated 7 days after the first SBT. Ventilator-free days at day 28 were computed as days without invasive mechanical ventilation during the 28 days following first SBT; patients who died before day 28 or were dependent on mechanical ventilation for more than 28 days after the first SBT had zero ventilator-free days [14]. Other definitions (e.g., Mac Cabe classification, ARDS, ventilator-associated pneumonia, failure of SBT) and data collection process are reported in the online supplement (Additional file 1).

Lung ultrasonography was performed on the day of first SBT and repeated on the 2 days following a SBT failure and on the day of extubation, if applicable. Maximal end-expiratory interpleural distance, sonographic patterns of effusion (homogeneously anechoic, complex nonseptated, complex septated, or homogeneously echogenic) [15], and of lung parenchyma (condensation or atelectasis) [16] were assessed on each side with the patient in the semirecumbent position. A moderate-to-large pleural effusion was defined as a maximal interpleural distance ≥ 15 mm (predicting an effusion volume of 300 mL or more) [17]; a large pleural effusion was defined by a maximal interpleural distance ≥ 25 mm [4, 17]. A pleural effusion was deemed drainable if the maximal interpleural distance was ≥ 15 mm, and the effusion was visible over at least three intercostal spaces [18]. When possible, a transthoracic echocardiography was also performed to assess left ventricle ejection fraction (see the online supplement, Additional file 1). In patients with SBT failure, attempts at depletion (by diuretics or ultrafiltration) and fluid balance were collected during the 2 days following the SBT. There was no mandatory depletive strategy for the management of pleural effusion.

The primary endpoint was the prevalence of pleural effusion at weaning start. The sample size was calculated by hypothesizing a prevalence of pleural effusion of 40% [1‐3], and considering a precision of 8%. The study required a minimum of 170 patients (for an alpha risk of 5%, i.e., a confidence interval of 95%) and a maximum of 260 patients (for an alpha risk of 1%, i.e., a confidence interval of 99%). Continuous data were expressed as medians [25th–75th centiles] unless otherwise specified, and were compared using the Mann–Whitney test. Categorical variables, expressed as percentages, were compared using the Chi-square test or Fisher exact test. To evaluate independent factors associated with the presence of moderate-to-large pleural effusion at weaning start or with failure of the weaning process, significant or marginally significant (p < 0.10) bivariate risk factors (using the above mentioned tests) were examined using univariate and multivariable backward stepwise logistic regression analysis. Among related univariate factors, only the most statistically robust (yet clinically relevant) was entered into the regression model in order to minimize the effect of colinearity. The selection process was guided by consistency (less than 5% missing values) and maximal imbalances between groups (as estimated by absolute standardized differences, which are independent of the sample size and variable unit) [19]. Coefficients were computed by the method of maximum likelihood. The calibration of models was assessed by the Hosmer–Lemeshow goodness-of-fit statistic (good fit was defined as p value > 0.05), and discrimination was assessed by the area under the receiver operating characteristics curve (with a value of 1 indicating perfect discrimination, and a value of 0.5 indicating the effects of chance alone). Correlations were tested using the Spearman’s method. Two-tailed p values < 0.05 were considered significant. Data were analyzed using the IBM SPSS Statistics for Windows (Version 19.0, IBM Corp Armonk, NY, USA).

The inclusion period lasted from 2 to 12 months depending on centers, between June 2015 and May 2016. Four hundred seventy-seven patients mechanically ventilated for more than 24 h were screened (Fig. 1). Sixty-seven patients died before weaning start, and 161 patients were excluded because of either a do-not-reintubate decision at time of inclusion (n = 72), unavailability of pleural ultrasound (n = 63), or drainage of pleural effusion before inclusion (n = 26). Thus, the present study comprises 249 patients assessed with lung ultrasonography at weaning initiation. Median duration of mechanical ventilation before weaning was 4 [2–7] days. The weaning trajectories are summarized in Fig. 1. Two hundred and three patients succeeded the first SBT, and 200 of them were extubated (the remainder three patients were not extubated despite the success of the first SBT because of borderline cough, and experienced a novel complication leading to death before any extubation attempt). Forty-six patients (18%) failed the first SBT; 41 of them succeeded a subsequent SBT and were extubated latter in the course of weaning, while the remainder five patients died before any extubation attempt. Reasons for SBT failure were respiratory rate > 35 breaths/min with increased accessory muscle activity (n = 17), SpO₂ < 90%, while on FiO₂ ≥ 0.5 (n = 6), systolic blood pressure < 90 mmHg or > 180 mmHg (n = 2), or a combination of those reasons (n = 22).

A moderate-to-large pleural effusion was detected in 81 of 249 patients assessed at weaning initiation, for a prevalence of 33%, 95% confidence interval: 27–39%. Most of pleural effusions were homogeneously anechoic (n = 74, 93%) and associated with pulmonary condensation or atelectasis (n = 68, 85%) (see the online supplement, Table e1, Additional file 1). Seventy-six (31%) patients had a bilateral pleural effusion. The maximal interpleural distance was equally located either on the left (n = 41, 51%) or right side (n = 40, 49%). Patients with moderate-to-large pleural effusions at weaning initiation were older, had more baseline comorbidities and more organ failures before weaning as compared to their counterparts (see the online supplement, Table e4). In multivariable analysis, older age, McCabe class 2, cardiac disease, acute respiratory failure as cause of intubation, and need for dialysis before the first SBT were the five independent factors associated with a moderate-to-large pleural effusion at initiation of weaning (Table 1).

According to the International Consensus Conference, the 249 patients were classified as follows: 173 (69%) weaning successes (a first successful SBT followed by successful extubation); 76 (31%) weaning failures (including 46 who failed the first SBT and 27 who succeeded the first SBT but failed extubation); three unclassifiable patients (despite the success of the first SBT, they were not extubated because of borderline cough, and experienced a novel complication leading to death before any extubation attempt) (Fig. 1). According to the WIND classification, 161 (65%) patients had a short weaning, 60 (24%) had a difficult weaning, and 28 (11%) had a prolonged weaning. The presence of a moderate-to-large pleural effusion at weaning initiation was associated with more failures of the first SBT [27 (33%) vs. 19 (11%), p < 0.001], more weaning failures [37 (47%) vs. 36 (22%), p < 0.001], less ventilator-free days at day 28 (21 [5–24] vs. 23 [16–26], p = 0.01), and a higher mortality at day 28 [14 (17%) vs. 14 (8%), p = 0.04] (Table 2, Fig. 2). All variables associated with weaning failure are shown in Table 3 and Table e5. In multivariable analysis, PaO₂/FiO₂ ratio, chronic obstructive pulmonary disease, a longer duration of mechanical ventilation prior to weaning, and the presence of moderate-to-large pleural effusion at weaning initiation were the four independent factors associated with weaning failure (Table 4). In sensitivity analyses, the association of pleural effusion with weaning failure also persisted after adjustment on SAPS II, in selected centers using the T-piece trial, or in those using a low-level pressure support, and when considering pleural effusions deemed drainable (as defined by a maximal interpleural distance ≥ 15 mm with the effusion visible over three intercostal spaces) or those considered large (as defined by a maximal interpleural distance ≥ 25 mm) [4, 17] (Table e3). A moderate-to-large pleural effusion was detected in 60 (28%) of 218 patients assessed on the day of first extubation attempt. The extubation failure rate was higher in patients with a moderate-to-large pleural effusion on the day of extubation as compared to their counterparts [14 (23%) vs. 19 (12%), p = 0.04]; of note, this extubation failure rate was similar in patients with or without pleural effusion assessed earlier, at weaning initiation [24 (15%) vs 15 (20%), p = 0.31]. As compared to patients without effusion (n = 168), those with a unilateral (n = 21) or bilateral (n = 60) moderate-to-large pleural effusion had similarly altered weaning outcomes, including SBT failure [19 (11.3%) vs. 8 (38.1%) vs. 19 (31.7%), p < 0.001] and weaning failure [37 (22.0%) vs. 11 (52.4%) vs. 28 (46.7%), p < 0.001].

Among the 46 patients who failed the first SBT, lung ultrasonography was repeated 24 and 48 h later in 41 and 31 patients, respectively. Patients in whom diuretics and/or ultrafiltration were used had a lower fluid balance as compared with their counterparts (− 484 [− 1210–330] vs. 858 [205–1806] mL after 24 h, p < 0.001), but this depletive strategy did not alter the interpleural distance (see the online supplement, Table e2, Additional file 1). Fluid balance was not significantly correlated with changes in interpleural distance (ρ 0.13, p = 0, 17, see Figure e1 of the online supplement, Additional file 2). Pleural effusion was drained in only four patients during weaning.

In our study, one-third of patients had a moderate-to-large pleural effusion at the initiation of the weaning process. This prevalence is higher than that of 13% reported by Dres et al. [9]. This discrepancy may be explained by differences in definitions used. Volume of pleural fluid was estimated in our report according to interpleural distance, which may be more sensitive than the classification of the British Thoracic Society [20] used in the latter study; indeed, patients with moderate-to-large pleural effusion in our study had a median interpleural distance inferior to the value found in the Dres’ study (27 [20–41] vs. 45 [30–60] mm). Other differences between these two studies include patient’s comorbidities (with more patients included with cardiac diseases in our report) and/or timing of inclusion (with less patients excluded because of prior pleural drainage before SBT in our study).

Risk factors for pleural effusion found in our study are in accordance with previous reports [1, 2]. Congestive heart failure is one of the leading factors associated with the occurrence of pleural effusion in ICU [1]. All patients intubated for acute respiratory failure had acute cardiac failure or pneumonia, two common risk factors for pleural effusion [1]. Our study suggests that diastolic dysfunction may be of importance in the association of cardiac failure with pleural effusion. Acute renal failure has also been previously reported as a risk factor for nonmalignant pleural effusions, an association possibly mediated by fluid overload [21]. The association of Mc Cabe class (i.e., a rapidly fatal underlying disease) with pleural effusion may be driven by other comorbidities like liver cirrhosis, cancer, and hypoalbuminemia [22, 23].

Our study is the first to show an association between moderate-to-large pleural effusion on the one hand and worse weaning outcomes and survival on the other hand. Dres et al. found similar weaning outcomes in patients with or without pleural effusion, but the limited number of patients with pleural effusions in their report (n = 18) weakened their conclusions [9]. Our findings are consistent with a previous report by Mattison et al., suggesting an association between pleural effusion and a longer duration of mechanical ventilation [2]. Although pleural effusion in patients with early ARDS do not seem to significantly influence lung physiology and gas exchange [3], its role in the latter stages of mechanical ventilation seems more relevant [4]. Physiological alterations associated with pleural effusion may worsen the respiratory load during the weaning process. Indeed, physiological studies showed that pleural effusion increases chest wall volume and decreases the length of inspiratory muscles and thus their efficiency and power [5, 6]. Umbrello et al. recently showed that during weaning, drainage of a unilateral pleural effusion improves diaphragmatic contractile activity [24]. This improvement could decrease dyspnea and mitigate weaning failure. Pleural fluid accumulation may also result in relaxation atelectasis of the adjacent lung. In our study, most pleural effusions (85%) were associated with condensation or atelectasis.

There was no significant association between fluid balance and the evolution of pleural effusion during the 48 h following SBT failure. However, the negative fluid balance achieved was modest in the depletive group and this limitation precludes any definite conclusion. Pleural drainage with ultrasonography guidance has a low risk of complication under mechanical ventilator support [7] and may improve oxygenation, respiratory mechanics [4], and diaphragm performance [6, 24]. Removal of pleural fluid may therefore decrease the work of breathing and increase the ability of patients to succeed weaning. Further studies are needed to test whether a strategy of aggressive diuretic management or drainage of pleural effusions, in mechanically ventilated patients entering the weaning process, with others risk factors of weaning failure or a SBT failure, has the potential to decrease its duration [25].

Strengths of our study include the large sample size, the prospective and multicentric design, and the use of ultrasound, which is currently considered the most sensitive method to detect pleural effusion at bedside. Our study has several limitations. First, only 249 of the 477 screened patients were included, a fact that may alter the external validity of our prevalence estimation. Second, the inter- or intra-observer agreement for pleural ultrasonography was not evaluated, but several reports previously demonstrated an excellent agreement for the measurement of left or right maximal interpleural distance [26]. Third, the physicians in charge of the patient were not fully blinded to the ultrasound examination results, and this may have theoretically influenced extubation decision and outcomes. However, criteria for SBT result were defined a priori and independent from ultrasound findings. Fourth, no estimation of respiratory drive (e.g., with airway occlusion pressure) nor respiratory muscle strength (e.g., with maximal inspiratory pressures) was performed. Last, a significant decrease in pleural effusion during difficult weaning may have required more time, and/or more intense depletive fluid management [27].