Abstract
To verify the association between the air leak objectively measured intraoperatively (IAL) using the ventilator and the air leak duration after pulmonary lobectomy.
Prospective analysis on 111 patients submitted to pulmonary lobectomy (33 by video-assisted thoracic surgery). After resection, objective assessment of air leak (in milliliter per minute) was performed before closure of the chest by measuring the difference between a fixed inspired and expired volume, using a tidal volume of 8 ml/kg, a respiratory rate of 10 and a positive-end expiratory pressure of 5 cmH2O. A multivariable analysis was performed for identifying factors associated with duration of postoperative air leak.
Average IAL was 158 ml/min (range 0–1500 ml/min). The best cut-off (receiver-operating characteristics analysis) associated with air leak longer than 5 days was 500 ml/min. Nine patients had IAL >500 ml/min (8%). They had a longer duration of postoperative air leak compared with those with a lower IAL (mean values, 10.1 days, SD 8.8 vs 1.5 days, SD 4.9 P < 0.001). The following variables remained associated with days of air leak duration after multivariable regression: left side resection (P = 0.018), upper site resection (P = 0.031) and IAL >500 ml/min (P < 0.001). The following equation estimating the days of air leak duration was generated: 1.7 + 2.4 × left side + 2.2 × upper site + 8.8 × IAL >500.
The air leak measurement using the ventilator parameters after lung resection may assist in estimating the risk of postoperative prolonged air leak. An IAL > 500 ml/min may warrant the use of intraoperative preventative measures, particularly after video-assisted thoracic surgery lobectomy where a submersion test is often unreliable.
INTRODUCTION
The prolonged air leak (PAL), intended as a postoperative air leak from the alveoli or distal respiratory bronchioles for more than 5 days [1], is one of the most frequent complications after anatomic or non-anatomic lung resection [2].
This postoperative adverse event is usually associated with prolonged hospital stay and may contribute to the development of more severe complications such as empyema, atelectasis, pneumonia and respiratory failure [3, 4].
Several authors have tried to identify factors associated with PAL [5]. These factors include the preoperative characteristics of the patients [6, 7], some surgical variables such as the side and site of resection and even the grade of the air leak measured during the first postoperative hours [8, 9].
However, the association between an objective intraoperative measurement of air leak at the end of the parenchymal resection and the postoperative duration of air leak has never been assessed. The rationale of this analysis was to identify a threshold of intraoperative air leak associated with PAL, which could justify performing additional measures aimed at reducing the degree and duration of postoperative air leak.
Therefore, the objective of the present study was to verify the association between the air leak objectively measured intraoperatively using the ventilator parameters and the duration of air leak in patients submitted to pulmonary lobectomy or segmentectomy.
METHODS
In the present study, we prospectively enrolled patients subjected to anatomical lung resection for primary and secondary lung cancer at the Thoracic Surgery Unit of the Ospedali Riuniti of Ancona, Italy, from September 2012 to December 2013. We excluded from the analysis those patients treated with pneumonectomy or wedge resection, as well as the patients who needed mechanical ventilation at any time after surgery, since the assisted prolonged airways positive pressure (>24 h) could induce or influence the postoperative air leak independently from the findings verified intraoperatively.
Lung resection was performed using either the open or the video-assisted approaches. The open access consisted of an anterolateral muscle sparing and intercostal nerve-sparing thoracotomy. Video-assisted thoracoscopic procedures were performed through a 2-port anterior technique. Regardless of the surgical approach, all incomplete fissures were developed by mechanical staplers.
Once the lung resection and the lymphadenectomy were completed, a submersion test was performed in order to assess the presence and degree of the air leak, assigning a score to each patient [10]. In case of a large coalescent bubbling (Score 3), every attempt was made to minimize the presence of the air leak by applying sutures. No sealants or buttressing material or any other specific procedure to reduce the leakage were used in this series. The inferior pulmonary ligament was released after upper lobectomies only in case of large residual apical space.
After this phase and before closing the chest, an objective and standardized assessment of the air leak was performed by measuring the difference between a fixed inspired and expired volume using a tidal volume of 8 ml/kg (adapted to the compliance detected for the patient), a respiratory rate of 10 and a positive-end expiratory pressure of 5 cmH2O. The sum of the differences of 5 consecutive respiratory cycles was multiplied by 2 to obtain the air leak expressed as millilitre per minute.
All measurements were performed using the same ventilator, after full re-expansion of the isolated lung and were calculated automatically.
A single 24-Fr chest tube was placed in a mid-lateral position at the end of the operation to drain the pleural cavity.
All patients included in the present analysis were managed in the postoperative period following standardized pathways of care. No routine chest X-ray was performed immediately after surgery. This precluded quantification of possible residual apical spaces immediately after resection to be used as potential variable to adjust the analysis.
In particular, postoperative chest tubes were managed using an electronic device featuring a regulated suction and continuous monitoring of air leak (Thopaz, Medela Healthcare). In all cases, a regulated pressure of −8 cmH2O was constantly applied throughout the period in which the drain was maintained. A chest X-ray was obtained in the first postoperative day for all patients, unless otherwise clinically indicated. The chest tube was removed if the following criteria were satisfied: presence on an air leak of 20 ml/min or less for at least 8 h without spikes and a pleural effusion <400 ml/24 h. For the purpose of this study, an air leak was considered prolonged if lasting more than 5 days.
The present study project has been developed taking into account the ethical principles of good clinical practice. The authors followed the ethical rules for the medical research involving humans, as stated in the Declaration of Helsinki by the World Medical Association. The present study was approved by the local institutional review board.
Statistical analysis
The normal distribution of the numeric variables was assessed by the Shapiro–Wilk normality test. Numeric variables with normal distribution were compared between patients with and without high intraoperative air leak (IAL) by the unpaired Student’s t-test while those without normal distribution by the Mann–Whitney test. Categorical variables were compared between patients with and without high IAL using χ2 test or the Fisher’s exact test (in case the number of observations was <10 in at least one of the cells).
In addition to the volume of IAL measured at the end of surgery, the following variables were tested for a possible association with postoperative duration of air leak (numeric variable used as dependent variable): age, sex, percentage forced expiratory volume in 1 s, percentage carbon monoxide lung diffusion capacity, body mass index, side and site of resection, presence of pleural adhesions and surgical approach [video-assisted thoracic surgery (VATS) versus thoracotomy].
These variables were used as independent factors in a multivariable linear regression analysis with backward stepwise elimination. Variables with a P-value of < 0.05 were retained in the final model. The stability of the regression analysis was tested using bootstrap resampling technique. One thousand samples of 111 patients were generated with sampling with replacement and the multivariable regression repeated in each of these samples. The variables in the final model were judged to be stable if they occurred in more than 50% of the bootstrapped samples [11, 12].
The difference in IAL between the bubbling scores was assessed by analysis of variance.
A threshold effect was searched using receiver-operating characteristics analysis to determine the best cut-off of IAL associated with prolonged air leak. Calculation of the Juden index was used to select the best cut-off value.
All tests were performed on Stata 12.0 statistical software (Stata Inc., College Station, TX, USA).
RESULTS
We analysed 111 patients (65 males and 46 females) submitted to pulmonary lobectomy or segmentectomy (31 right upper lobectomies, 25 right lower lobectomies, 20 left upper lobectomies, 17 left lower lobectomies, 6 middle lobectomies, 5 bilobectomies and 7 anatomic segmentectomies). Thirty percent (33 patients) of the resections were performed by VATS. Considering the entire cohort, 11 patients (10%) experienced a PAL in the postoperative period.
Table 1 describes the general characteristics of the population.
Baseline characteristics of the patients
| Variables | |
|---|---|
| Age (years) | 67.9 (10.1) |
| Male sex (n, %) | 65 (59%) |
| FEV1 | 86.1 (20.5) |
| FEV1/FVC | 0.7 (0.1) |
| DLCO | 76.6 (19.8) |
| BMI | 26.5 (4.8) |
| Upper site of resection (n, %) | 66 (60%) |
| Right side of resection (n, %) | 68 (61%) |
| Hospital stay (days) | 5.4 (3.1) |
| Chest tube duration (days) | 4.4 (5.7) |
| Air leak duration (days) | 2.2 (5.8) |
| Variables | |
|---|---|
| Age (years) | 67.9 (10.1) |
| Male sex (n, %) | 65 (59%) |
| FEV1 | 86.1 (20.5) |
| FEV1/FVC | 0.7 (0.1) |
| DLCO | 76.6 (19.8) |
| BMI | 26.5 (4.8) |
| Upper site of resection (n, %) | 66 (60%) |
| Right side of resection (n, %) | 68 (61%) |
| Hospital stay (days) | 5.4 (3.1) |
| Chest tube duration (days) | 4.4 (5.7) |
| Air leak duration (days) | 2.2 (5.8) |
Results are expressed as means and standard deviations (numeric variables) or as count and percentages (categorical variables).
FEV1: forced expiratory volume in 1 s; FVC: forced vital capacity; DLCO: carbon monoxide diffusing capacity; BMI: body mass index.
Baseline characteristics of the patients
| Variables | |
|---|---|
| Age (years) | 67.9 (10.1) |
| Male sex (n, %) | 65 (59%) |
| FEV1 | 86.1 (20.5) |
| FEV1/FVC | 0.7 (0.1) |
| DLCO | 76.6 (19.8) |
| BMI | 26.5 (4.8) |
| Upper site of resection (n, %) | 66 (60%) |
| Right side of resection (n, %) | 68 (61%) |
| Hospital stay (days) | 5.4 (3.1) |
| Chest tube duration (days) | 4.4 (5.7) |
| Air leak duration (days) | 2.2 (5.8) |
| Variables | |
|---|---|
| Age (years) | 67.9 (10.1) |
| Male sex (n, %) | 65 (59%) |
| FEV1 | 86.1 (20.5) |
| FEV1/FVC | 0.7 (0.1) |
| DLCO | 76.6 (19.8) |
| BMI | 26.5 (4.8) |
| Upper site of resection (n, %) | 66 (60%) |
| Right side of resection (n, %) | 68 (61%) |
| Hospital stay (days) | 5.4 (3.1) |
| Chest tube duration (days) | 4.4 (5.7) |
| Air leak duration (days) | 2.2 (5.8) |
Results are expressed as means and standard deviations (numeric variables) or as count and percentages (categorical variables).
FEV1: forced expiratory volume in 1 s; FVC: forced vital capacity; DLCO: carbon monoxide diffusing capacity; BMI: body mass index.
In the entire cohort, the mean air leak duration was 2.2 days [median 0, interquartile range (IQR) 0–2], while the chest tube was removed after 4.4 days (median 2, IQR 2–4), on average. The mean postoperative hospital stay for the entire population was 5.4 days (median 4, IQR 3–6).
We were able to assess the bubbling score in 92 patients (83%). For the remaining 19 patients, all submitted to an anatomic pulmonary resection using the VATS technique, IAL assessment using the bubbling score was deemed not feasible or reliable.
As shown in Fig. 1, 52 patients had a score of 0, 28 of 1, 9 of 2 and 3 of 3. The average IAL in the 4 different classes was as follows: Score 0 = 45 ml/min, Score 1 = 239 ml/min, Score 2 = 311 ml/min and Score 3 = 667 ml/min (analysis of variance test P < 0.001). The mean IAL in the entire population was 158 ml/min (SD 271, range 0–1500 ml/min).
Correlation between bubbling score and intraoperative air leak measurement.
Figure 2 displays the scatterplot of the IAL and duration of air leak.
Scatter plot of the intraoperative air leak (milliliter per minute) and postoperative duration of air leak (days).
The receiver-operating characteristics analysis demonstrated that the best cut-off value associated with a PAL (air leak >5 days) during the postoperative period was a leakage of 500 ml/min assessed intraoperatively. Nine patients had an IAL > 500 ml/min (8%). Compared with those with lower IAL, they were older (mean age 75.5, SD 6.7 vs 67.2, SD 10.0, P = 0.01), had lower carbon monoxide lung diffusion capacity (mean values 66.6, SD 15.7 vs 77.5, SD 19.9, P = 0.06) and lower forced expiratory volume in 1 s (mean values 77.2, SD 19.2 vs 86.9, SD 20.5, P = 0.11). They developed a significantly longer duration of postoperative air leak compared with those with a lower IAL (10.1 days, SD 8.8, median 10, IQR 2–17 vs 1.5 days, SD 4.9, median 0, IQR 0–0, P< 0.001). Five of the 9 patients with high IAL developed a PAL postoperatively (55%) versus 6 of the 102 patients with lower IAL (5.9%, P < 0.001; Fig. 3).
Flow chart of the air leak distribution.
Table 2 shows the results of the multivariable analysis performed using as dependent variable the postoperative air leak. The preoperative and intraoperative factors independently associated with a postoperative air leak were a left side resection (P = 0.031), an upper side resection (P = 0.018), and an IAL > 500 ml/min (P < 0.001). The bootstrap analysis demonstrated that an IAL >500 ml/min was the more stable predictor of postoperative air leak.
Results of the multivariable analysis for postoperative air leak predictors
| Predictors | Coefficients | Standard error | 95% Confidence interval | P-value | Bootstrap% |
|---|---|---|---|---|---|
| Intercept | 1.7 | 0.9 | −0.25 to 3.55 | ||
| Left side | 2.4 | 1.0 | −4.41 to − 0.43 | 0.031 | 68 |
| Upper site | 2.2 | 1.0 | 0.20 to 4.14 | 0.018 | 62 |
| Intraoperative air leak >500 ml/min | 8.8 | 1.8 | 5.34 to 12.45 | <0.001 | 97 |
| Predictors | Coefficients | Standard error | 95% Confidence interval | P-value | Bootstrap% |
|---|---|---|---|---|---|
| Intercept | 1.7 | 0.9 | −0.25 to 3.55 | ||
| Left side | 2.4 | 1.0 | −4.41 to − 0.43 | 0.031 | 68 |
| Upper site | 2.2 | 1.0 | 0.20 to 4.14 | 0.018 | 62 |
| Intraoperative air leak >500 ml/min | 8.8 | 1.8 | 5.34 to 12.45 | <0.001 | 97 |
Only the variables retained in the final model after backward stepwise elimination are displayed.
Results of the multivariable analysis for postoperative air leak predictors
| Predictors | Coefficients | Standard error | 95% Confidence interval | P-value | Bootstrap% |
|---|---|---|---|---|---|
| Intercept | 1.7 | 0.9 | −0.25 to 3.55 | ||
| Left side | 2.4 | 1.0 | −4.41 to − 0.43 | 0.031 | 68 |
| Upper site | 2.2 | 1.0 | 0.20 to 4.14 | 0.018 | 62 |
| Intraoperative air leak >500 ml/min | 8.8 | 1.8 | 5.34 to 12.45 | <0.001 | 97 |
| Predictors | Coefficients | Standard error | 95% Confidence interval | P-value | Bootstrap% |
|---|---|---|---|---|---|
| Intercept | 1.7 | 0.9 | −0.25 to 3.55 | ||
| Left side | 2.4 | 1.0 | −4.41 to − 0.43 | 0.031 | 68 |
| Upper site | 2.2 | 1.0 | 0.20 to 4.14 | 0.018 | 62 |
| Intraoperative air leak >500 ml/min | 8.8 | 1.8 | 5.34 to 12.45 | <0.001 | 97 |
Only the variables retained in the final model after backward stepwise elimination are displayed.
The residuals plot indicated a well-fitted model without any patterns of the residuals against the fitted values.
The following equation estimating the air leak duration in days was generated: 1.7 + 2.4 × left side + 2.2 × upper site + 8.8 × IAL >500 ml/min (adjusted R2 statistics 0.14).
We repeated the regression analysis by entering the IAL variable as a continuous one and obtained the following equation to estimate the air leak duration: 1.3 + 2.4 left side + 2.2 upper site +0.007 × IAL (expressed in millilitre per minute).
DISCUSSION
Background and objectives
Despite the development of increasingly sophisticated surgical instruments and the constant rise of minimally invasive techniques, the postoperative air leak remains one of the most common adverse events after pulmonary lobectomy.
As reported from the last ESTS Database Annual Report the incidence of PAL (>5 days) after lobectomy and segmentectomy is still 8.9 and 6.4%, respectively (http://www.ests.org/_userfiles/pages/files/ESTS%202016Silver_Book_FULL_FINAL_14.50.pdf).
Several studies have shown that the occurrence of PAL could lead to the onset of other complications and is associated with prolonged hospital stay and increased hospital costs [3, 4, 13]. As an example, in 2006, we published a study on a group of 726 patients submitted to open pulmonary lobectomy or bilobectomy for lung cancer. Within this cohort, 57% of patients showed an air leak during the first postoperative day and 17% experienced a PAL (defined in that article as a postoperative air leak longer than 7 days). Using a propensity score analysis to match the patients with and without PAL, we found that those with PAL had a higher rate of postoperative empyema (8.2% vs 0%, P = 0.01) and a longer hospital stay (16.9 vs 9 days, P < 0.0001) compared with those without it [4].
In order to reduce the incidence of PAL, the identification of patients with a higher risk of experiencing this adverse event would allow to design and adopt preventative strategies. These could be represented by a more rationale management of chest tubes in the postoperative period as well as by the adoption of specific technical precautions during the operation, such as creating a pleural tent, using a fissureless technique or applying sealants [14–18]. With the aim of selecting those individuals at higher risk of PAL, some authors have analysed the correlation between patients’ and surgical-related characteristics and the development of postoperative air leak [6, 19, 20]. Nevertheless, until now, in the literature, there is poor evidence about the identification of reliable risk factors for a persistent air leak [7, 9, 21]. Moreover, considering the rapid transition of the surgical technique to a minimally invasive approach, the currently available parameters for predicting postoperative air leak may be ineffective.
The aim of the present study was to assess whether the air leak objectively measured at the ventilator after completion of the parenchymal resection was associated with the duration of postoperative air leak (in turn objectively measured with a digital chest drainage system). Moreover, we verified the ability of the IAL measurement to identify patients at high risk for a PAL after a lobectomy or a segmentectomy.
Main finding
We were able to objectively measure the IAL assessed at the ventilator using a standardized procedure on 111 patients submitted to anatomic lung resection, one-third of them through a minimally invasive approach.
The trend of the air leak measured at the ventilator was in line with the one assessed by the subjective evaluation of the air leak using the bubbling score.
Nevertheless, the bubbling score was not considered as a predictor of prolonged air leak by most of the studies that developed risk models for the postoperative occurrence of this complication [6, 10, 20, 21], perhaps for the subjective nature of this measurement. In fact, in several articles, the bubbling score was a factor considered for selecting the category of patients that would benefit from intraoperative additional measures for reducing the air leak, rather than for predicting the postoperative air leak [22, 23]. Moreover, in our study, for 19 patients treated by a VATS lobectomy procedure, we were not able to obtain a reliable evaluation of the air leak using the submersion test (poor visibility and exposure after reinflation of the lung). So we verified that the anticipation of an air leak in the postoperative course was possible for the entire cohort exclusively using the ventilator measurement.
We found that an IAL >500 ml/min measured at the ventilator was associated with longer air leak duration during the postoperative period. In particular those patients with an IAL >500 ml/min had a mean postoperative air leak of 10.1 days vs 1.5 days (P < 0.001) for those with an IAL <500 ml/min.
Moreover 55% of those patients with an IAL >500 ml/min experienced a prolonged (>5 days) air leak, while this complication occurred only in 6% of the cases with an IAL <500 ml/min (P < 0.001).
This represents novel information that, to the best of our knowledge, has never been reported before in the literature.
We verified that, independently from the IAL measurement, other operative factors were associated with the duration of the postoperative air leak. These were represented by the side (the left side is at higher risk for air leak than the right side) and the site of the resection (the upper side is at higher risk for air leak than the lower side).
Intraoperative parameters, such as the side and site of resection, were found as predictors of prolonged air leak by several articles [6, 24, 25]. In particular, we previously analysed a group of 588 patients submitted to open pulmonary lobectomy in order to generate a model for predicting the postoperative air leak in 2004. During that study, the objective evaluation of the intraoperative air leak was not considered as a possible predictor of prolonged air leak and we found, after logistic regression with bootstrap validation, that the risk factors for this complication were a reduced predicted postoperative forced expiratory volume in 1 s (regression coefficient: −0.033; P < 0.0001), the presence of pleural adhesions (regression coefficient: 0.74; P = 0.003) and an upper resection (regression coefficient: 0.75; P = 0.006).
In the present study, we verified that, in comparison with the 2 other significant risk factors (side and site of operation), the objective measurement of the IAL at the ventilator resulted the most reliable predictor of postoperative prolonged air leak at the bootstrap analysis.
Combining these 3 parameters, we were able to obtain an equation for estimating the duration of postoperative air leak. As a consequence, a hypothetical patient submitted to a left upper lobectomy and with an IAL >500 ml/min measured after resection would have an expected air leak duration of 15 days.
Limitations
This study may have the following potential limitations:
First, for the present series, we were not able to verify whether any postoperative factor could have influenced the association between IAL and duration of postoperative air leak. In fact, despite the patients were managed using standardized pathways of care and protocols, we did not collect data about pain control, mobilization, postoperative rehabilitation and the use of steroids.
Second, given the size of the study cohort and the low rate of the outcome tested (a PAL was found in 10% of the patients), we could not control the result of this analysis stratifying the population in different subgroups. Possibly, the accuracy of the IAL measurement in predicting the postoperative air leak duration may vary in relation to the age, chronic obstructive pulmonary disease grading and body mass index. These factors need to be tested in future analyses on larger populations.
Third, considering the fact that the present study was carried out within an institution with a dedicated research field and specific protocols for the management of the air leak, the reproducibility of these results in other settings needs to be verified by other independent investigations. In particular, the protocols for the chest tube removal and the use of digital system for assessing the postoperative air leak could have influenced the results. The same problem of generalizability applies for centres using regularly sealants, buttressing materials or energy devices to deal with lung parenchyma.
Clinical implications and conclusions
The measurement of the IAL at the ventilator after a lung resection is a reliable method for predicting a postoperative course complicated by a PAL.
This method of air leak assessment is objective and reproducible both in open and VATS patients.
Using the threshold of a measured air leak >500 ml/min, it was possible to identify the group of patients for which the adoption of additional preventative measures as well as tailored postoperative pathways of care would be of benefit in order to reduce the risk of PAL.
The advantage of objectively predicting a potential PAL intraoperatively offers the chance of minimizing the leakage during the same intraoperative phase using specific surgical procedures, such as pleural tent, pleurectomy or the use of sealants. Finally, this threshold can be used to select patients in future efficacy studies testing products or postoperative strategies aimed at preventing the occurrence of PAL (i.e. sealants and chest tube management policies).
Conflict of interest: Alessandro Brunelli received a speaker honoraria in the previous year from C.R. Bard Inc. for an educational activity not related to this study. Other authors have no conflicts of interest to disclose.
REFERENCES
Author notes
Both the authors are currently working at St. James's University Hospital, Leeds, UK but were working at Ospedali Riuniti Ancona, Italy at the time of data collection and study recruitment.
