Performance of prolonged air leak scoring systems in patients undergoing video-assisted thoracoscopic surgery segmentectomy

Abstract

Open in new tabDownload slide

OBJECTIVES

We assessed the accuracy of 3 validated lobectomy scoring systems to predict prolonged air leak (PAL) in patients undergoing video-assisted thoracoscopic surgery (VATS) segmentectomy.

METHODS

We reviewed all consecutive patients who had a VATS segmentectomy between January 2016 and October 2020. We determined PALs on postoperative day 5. These findings were correlated with the calculated Brunelli (gender, age, body mass index [BMI], forced expiratory volume in 1 s < 80 and pleural adhesion), Epithor (gender, location, dyspnoea score, BMI, type of resection and pleural adhesion) and European Society of Thoracic Surgeons (ESTS) (gender, BMI and forced expiratory volume in 1 s) scores of each patient.

RESULTS

A total of 453 patients (mean age: 66.5 years, female/male sex ratio: 226/227) underwent a VATS segmentectomy for malignant (n = 400) and non-malignant (n = 53) disease. Postoperative cardiopulmonary complications and in-hospital mortality rates were 19.6% and 0.4%, respectively. Median chest tube drainage duration and hospital stay were 2 (interquartile range: 1-4) and 4 (interquartile range: 3-7) days, respectively. On day 5, the prevalence of PAL was 14.1%. The ESTS, Brunelli and Epithor scores for the treated population were, respectively, class A (6.8%), class B (3.2%), class C (10.8%) and class D (28.2%); very low and low (0%), moderate (5%), high (6.3%) and very high (21%); and class A (7%), class B (13.2%), class C (24%) and class D (27.8%). All scores correlated with PAL (p ≤ 0.001). The areas under the receiver operating characteristic (ROC) curve were 0.686, 0.680 and 0.644, respectively.

CONCLUSIONS

All 3 scoring systems were correlated with PAL > 5 days following the VATS segmentectomies. ESTS scores seem easier to introduce in clinical practice, but validation by a multicentre cohort is mandatory.

INTRODUCTION

Prolonged air leak (PAL) is a common complication of lung resection. It is defined by ESTS (European Society of Thoracic Surgeons)/Society of Thoracic Surgeons consensus as an air leak lasting for more than 5 days [1]. A PAL may occur in 6-26% of cases after anatomical lung resections and is associated with extended hospital length of stay and increased costs and morbidity [2–5]. Several preoperative predictive factors for PAL have been identified such as the presence of emphysema or male gender. Different scoring systems have been elaborated and validated, albeit originally for patients undergoing lobectomies (by open or VATS approach) [6–10]. Pulmonary VATS segmentectomy procedures have increased in practice over the past years to manage early-stage lung cancer in patients with poor pulmonary lung functions (unfit for lobectomy) or to manage/diagnose centrally located nodules [11–13]. Pulmonary segmentectomy is more complex technically than standard lobectomy because it requires a deeper dissection of the hilum and the division of single/multiple intersegmental planes within a thick lung parenchyma. Thus, combining the technical difficulties of a VATS segmentectomy with poor pulmonary function puts the patients at higher risk to develop PAL. Therefore, the preoperative identification of patients with an increased risk of PAL would be helpful to implement additional intraoperative (sealant, buttressing or pleural tent) or postoperative measures (suction device) and to anticipate extended hospital stays due to longer drainage times. In the context of a lobectomy, various scoring systems have been validated to predict postoperative PAL: the Brunelli score, which was developed for open lobectomies to predict PAL of >5 days [7]; the Epithor score for anatomical resection to predict PAL >7 days[8] and the ESTS score for VATS lobectomies to predict PAL > 5 days [6]. How these scoring systems predict PAL in a segmentectomy is unknown. The goal of this study was to determine the accuracy of these 3 validated scores to predict PAL in a cohort of patients undergoing a VATS segmentectomy.

METHODS

Ethics statement

This retrospective study was approved by the local ethics committee (CER-VD in Lausanne) (referral number: 2021–00620), which waived the need to obtain informed patient consent.

Patients

We reviewed the records of all consecutive patients who underwent a lung VATS segmentectomy from January 2016 to October 2020 at the Thoracic Center of Western Switzerland (Centre Universitaire Romand de Chirurgie Thoracique), which includes 4 affiliated centres (University Hospital of Lausanne, University Hospital of Geneva, Hospital of Valais and Hospital of Neuchâtel). All VATS segmentectomies (for malignant and non-malignant diseases) were included. All patients with non-small-cell lung cancer (NSCLC) or other lung malignancies (including metastasis) were discussed by a multidisciplinary tumour board. In patients with adequate pulmonary functions, VATS segmentectomies were performed for NSCLCs that were than 2 cm in diameter and were well located in a segment with no nodal involvement. Segmentectomy was also proposed for larger tumours if the patients could not undergo lobectomy due to poor pulmonary functions. Patient staging was assessed by fluorodeoxyglucose positron emission tomography scans and magnetic resonance imaging of the brain. Patients who required a conversion thoracotomy (n = 16, 3.4%) or a completion lobectomy were excluded from this study, as were patients undergoing associated wedge resection, lobectomy or pneumonectomy. This study was reported according to the STROBE criteria for observational studies.

Surgical technique

All segmentectomies were performed by 5 surgeons. Each surgeon had performed > 100 VATS lobectomies before performing segmentectomies. Anatomical segmental resections were accomplished by removing 1 or more pulmonary segments to achieve complete resection of the lesion. The VATS segmentectomies were performed with the patients under general anaesthesia with lung isolation by double lumen intubation. Surgical resection was undertaken using a standardized three-port approach (utility incision in the fourth intercostal space, 1 for a10-mm 30° thoracoscope in the seventh intercostal space anteriorly and 1 posteriorly), or a single port approach from 2018. Segmentectomies were performed with individual dissection of the segmental bronchus, artery(ies) and vein(s). All bronchovascular structures were transected using endoscopic staplers or energy devices. The intersegmental plane was defined using indocyanine green when necessary and divided using mechanical staplers. In patients who had NSCLC, systemic hilar and mediastinal nodal dissections were performed. In situations where nodal disease was suspected, the latter was confirmed on fresh frozen sections, and a completion lobectomy was generally undertaken. No techniques for preventing air leaks, such as pleural tents, buttressing material or pneumoperitoneum were used with these patients. Sealants were rarely applied and when they were, it was at the discretion of the surgeon. At the end of the surgical procedure, no test for air leak was performed. One chest tube was placed at the end of the operation and left on suction with a traditional or electronic device (Thopaz, Medela, Baar, Switzerland) at -15 to -20 cm H₂O. Patients were extubated in the operating room and admitted to a specialized dedicated thoracic ward. A chest x-ray was taken immediately after recovery to check the full expansion of the lung and the positioning of the chest tube. Chest tubes were removed when there was no air leak and when the amount of pleural fluid was <400 ml over 24 h. Postoperative pain was controlled by opioids (subcutaneous morphine), non-steroidal anti-inflammatory drugs and paracetamol on postoperative day 0. The subcutaneous opioids (e.g. morphine) were replaced by oral opioids (e.g. tramadol) following removal of the chest tube.

Data were collected in the authors’ electronic database. Data included demographics, comorbidities, body mass index (BMI), cardiopulmonary functional testing, tumour stage and histological diagnosis, operative characteristics and clinical outcomes up to 30 days after discharge, including length of stay from the day of surgery, readmissions, reoperations and cardiopulmonary complications. Air leak was assessed twice daily. We checked to determine if PAL was present on postoperative day 5. Simple segmentectomies were defined as segmentectomies requiring the division of a single intersegmental plane, such as S6, left upper division and lingular segmentectomies. Complex segmentectomies included other segmentectomies that required the division of >2 intersegmental planes, including cases with 2 segmental resections.

Scoring systems

We used the 3 existing scoring systems that had been published and validated to stratify the incidence of PAL in European populations: the Brunelli, the Epithor and the ESTS scores. The Brunelli score was validated in a cohort of 658 patients operated on with open lobectomies to predict PAL >5 days. It includes only 4 variables (age > 65 years, BMI< 25.5 m/kg², forced expiratory volume in 1 s (FEV1) < 80 and the presence of pleural adhesions) and predicts 4 risk classes (A, B, C and D) based on the number of items. Those patients in class A were at lower risk and those in class D were at higher risk of PAL. The Epithor score was validated with the French database and included 1233 patients with VATS anatomical resection and PAL > 7 days. The PAL (IPAL) index was gender (F = 0; M = 4)-(BMI-24) + 2 × dyspnoea score + pleural adhesion (no = 0; yes = 4) + pulmonary resection (wedge = 0; lobectomy or segmentectomy = 7; bilobectomy = 11; bulla resection = 2; volume reduction = 14) + location (lower or middle lobe = 0; upper = 4). Then, the IPAL score was determined using the formula 1/1 + exp [−(−4.213 + 0.1167 × IPAL)] and yielded 5 classes of PAL risk (very low, low, moderate, high, very high). In order to reach higher homogeneity and facilitate comparison between scoring systems, we tested the Epithor score with PAL measured at 5 and 7 days. Because we found no difference in the results, and because this procedure is clinically justifiable, we used this score with the PAL values at 5 days (and not at 7 days). Finally, the ESTS score was validated with the ESTS database and included 5069 patients operated on for VATS lobectomy who had PAL >5 days. Only 3 variables were used: gender (male), BMI < 18.5 kg/m² and FEV1 < 80% and predicted 4 risk classes of PAL > 5 days (A, B, C and D).

Statistical analyses

Continuous variables following a normal distribution are presented as means with standard deviation (SD). Nominal variables with large number categories or continuous variables not following a normal distribution are summarized as medians with an interquartile range. Binary variables are presented as numbers with percentages. Numerical variables with normal distribution were tested by the unpaired Student t-test whereas those without normal distribution were tested by the Mann–Whitney U test. Categorical variables were tested by the χ² test. Variables with a P-value <0.2 in univariable analyses were candidates for the multivariable models. Identification of the set of relevant factors associated with PAL was performed using a stepwise backward multivariable regression with the Akaike information criterion score. Odds ratios (OR) with 95% confidence intervals (CI) are reported.

The area under the receiver operating characteristic (ROC) curve estimates the discriminating value of the different score slopes described by the relation between the predicted and observed incidence values of PAL. An area of 0.5 suggests no discrimination (i.e. the ability to diagnose patients with and without the disease or condition based on the test), 0.7 to 0.8 is considered acceptable, 0.8 to 0.9 is considered excellent, and more than 0.9 is considered outstanding [14]. The AUC values were compared using a nonparametric method for paired data based on an empirical ROC curve estimation. The Stata version 16 software (StataCorp, College Station, TX, USA) was used for statistical analyses. All tests were two-tailed with a statistical significance threshold at 0.05.

RESULTS

Patient characteristics

From January 2016 to October 2020, a total of 453 patients with a mean age of 66.5 years (female/male: 226/227) underwent VATS pulmonary segmentectomies. Overall, the prevalence of PAL was 14.1% and 9.2% 5 and 7 days after the operation, respectively. Table 1 summarizes the characteristics of the patients with and without PAL.

When comparing patients without PAL with those with PAL, those with PAL presented significantly lower FEV1 (P < 0.001), lower diffusing capacity of the lung for carbon monoxide [DLCO] (P < 0.001), lower BMI (P < 0.001) and higher ASA scores (P = 0.034). We did not observe significant differences regarding PAL occurrence when comparing the surgical approaches (uni- or multiport) (P = 0.750), type of segment (unique/multiple, P = 0.665; simple/complex, P = 0.322), localization (upper/lower) (P = 0.258) or side (P = 0.186).

Patient 30-day mortality and morbidity were 0.4% and 26.9%, respectively (Table 2). PAL was associated with increased length of stay (median: 11 vs 5 days, P < 0.001) and increased rate of pneumonia (18.8% vs 7.8%; P = 0.006) and empyema (3.1 vs 0.3%; P = 0.036). Thirteen patients required replacement of a chest tube (2.9%) for postablative pneumothorax and were included in the group PAL for analysis.

On multivariable analysis, 3 factors were identified as risk factors for PAL: tobacco use, DLCO < 80 and the presence of pleural adhesions (Table 3). In contrast, BMI > 30 and diabetes mellitus were protective factors for PAL.

Scoring systems

We evaluated the performances of the 3 published scoring systems (Table 4): The Brunelli score was associated with class A in 6.8% of cases, class B in 3.2%, class C in 10.8% and class D in 28.2%. The ROC area was 0.686 (CI 95%: 0.642–0.729) and correlated with PAL (P < 0.001). The Epithor score used for PAL > 5 days was very low and low (0%), moderate (5%), high (6.3%) and very high (21%). The ROC area was 0.680 (CI 95%: 0.635 to 0.723) and correlated with PAL (P = 0.003). The ESTS risk score was associated with class A (7%), class B (13.2%), class C (24%) and class D (27.8%) and correlated with PAL (P = 0.001) with an ROC area of 0.644 (CI 95%: 0.599–0.689). The ROC areas were not significantly different among the 3 scoring systems (P = 0.47) (Fig. 1).

DISCUSSION

A PAL is defined by the presence of an air leak for more than 5 days [1]. It is a common complication, occurring in up to 26% of cases after pulmonary resection [4, 15–17]. The presence of PAL has been reported to increase morbidity, length of hospitalization and hospital costs [3]. Considerable efforts have been made to develop surgical adjuncts to prevent PAL. Several surgical techniques have been described to decrease PAL, including pleural tenting, pneumoperitoneum, mechanical pleurodesis and staple line buttressing, but these procedures either have their own limitations or are prohibitively expensive if used routinely.

The advantage of segmentectomy over lobectomy is preservation of lung function, but pulmonary segmentectomy is a technically more challenging operation [18]. Initially reserved for unfit patients with early-stage cancer, the operation recently gained visibility as the JCOG0802 trial demonstrated better overall survival for segmentectomy than lobectomy in patients with early-stage lung cancer but an association with increased local recurrence [19]. The results of this study will most likely influence the surgical approach for cancers smaller than 2 cm in diameter. As a consequence, the number of segmentectomies will increase. However, this procedure requires deeper dissection in the hilum and separation of several intersegmental planes, thus potentially exposing the patient to a higher likelihood of PAL. In addition, the application of VATS and the growing expansion of segmentectomies in high-risk patients with early-stage lung cancer will theoretically increase the operability rate. Indirectly, it will also increase morbidity in fragile patients with PAL. We observed that the presence of PAL was associated with a hospital stay on average 7 days longer than those in patients without PAL.

Various scoring systems have been developed with the goal of preoperatively identifying patients with an elevated risk of PAL [6–10]. Achieving this goal would allow the application of intraoperative measures to prevent air leak in high-risk patients with PAL, which in turn would facilitate PAL prevention and allow surgeons to provide better preoperative counselling to the patients at risk of developing PAL, of having a chest tube in place for a longer time or of being discharged with a chest drain.

These scoring systems were developed to predict the risk of PAL after anatomical resection, performed either by VATS or by open approaches, and we wanted to validate this concept in patients having segmentectomies. We decided to use only those scoring systems that represented the European population [6–8]. Hence, only 3 scoring systems were eligible for our analysis, yielding PAL scores initially published for open lobectomies.

We report an incidence of PAL of 14.1% and 9.2%, respectively, after 5 and 7 days post-segmentectomy. This incidence of PAL is reported to range between 6.5% and 8.8% [12, 17, 19]. The demographics and characteristics of the population presented in this study might explain the discrepancy we observed (mean age of 66.5 years; 41.5% of patients older than 70 years; 37.5% of patients with a predicted FEV1 < 80%; 58.1% with a predicted DLCO < 80%). On the other hand, the ESTS database reported 7.7% of PAL in the segmentectomy group [17]. Interestingly this study did not report differences compared with the “lobectomy” group, although patients having lobectomies were younger and had better pulmonary function (FEV1: 88% vs 80%), forced vital capacity (97% vs 91.4%) and diffusion capacity of CO (74% vs 68%). Nevertheless, the pulmonary function values of the segmentectomy group in the ESTS database were better than those of our cohort. No mention of simple/complex segments was included.

Consistent with the results of previous studies, we found that poor pulmonary function, active smoking, lower BMI and intraoperative pleural adhesions were all risk factors for PAL [6, 7, 9, 10, 20, 21]. Whereas patients with older age and male gender were more likely to have PAL according to the different scoring systems, we did not observe this correlation in our cohort. We did observe that lower BMI (<18) was a risk factor for PAL whereas obesity (BMI > 30) was protective against PAL. This finding might be explained by the variation in respiratory rates and tidal volumes in obese patients. Indeed, in this population, the abdominal pressure is elevated and the diaphragm is situated in a higher position. These factors can lead to a decrease in the residual space and facilitate pleural apposition and sealing of any possible air leak. Whereas complex segmentectomy has been associated with an increased risk of PAL [19], we did not observe a difference in terms of the incidence of PAL based on the type of segment (unique/multiple, P = 0.665; simple/complex, P = 0.322), localization (upper/lower) (P = 0.258) or side (P = 0.186). The presence of pleural adhesions has also been consistently described as a PAL risk factor [7, 8]. However, the goal of a scoring system is to determine the risk of PAL preoperatively. The presence of pleural adhesions is rarely determined preoperatively, and we think that this item should not be included in scoring systems.

When comparing the 3 different scoring systems, we found that the ROC areas under the curve (AUC) were similar, making them equivalent in their preoperative flagging power. We found that the observed PAL rate was higher than expected in the highest risk classes. We can assume that the addition of risk factors is probably more important than just a risk factor for prediction of PAL. Whereas all AUCs were > 0.5, which was consistent with good classification and better-than-random discriminating power, none of the scores presented an AUC > 0.7. This finding is probably due to the lack of correlation between gender and PAL in our cohort. When comparing the predicted incidence of PAL using the different scoring systems and the incidence of the PAL among those in our cohort who had VATS segmentectomies, we observed close concordance among the different risk classes.

Determination and validation by a large multicentre study seem necessary to better determine risk factors. When considering a scoring system for a VATS segmentectomy, we believe that 2 elements are important: 1) simplicity of implementation: In this sense, the ESTS scoring system includes readily available variables, thus making them user-friendly and easy to implement in daily practice; 2) the specificity of the VATS segmentectomy population: different elements should be included in the scoring system (FEV1, DLCO, tobacco, high BMI).

Our study has some limitations. Not only is it a retrospective study, but we use a scoring system initially designed for patients having lobectomies and apply it to patients having VATS segmentectomies, which comprises a different population regarding function and comorbidities. Furthermore, some data were missing from the records that we could study: We did not have information on the number and type of endostaplers used or on the use of energy devices to develop the intersegmental plane or on the use of sealants. Similarly, we had no data regarding the types of chest drainage systems used (digital vs traditional). The lack of these data makes our study harder to generalize but opens an interesting door for subsequent, multicentre studies.

In conclusion, we found that all 3 scoring systems were correlated with the occurrence of PAL following a VATS segmentectomy. However, the ESTS scoring system seems easier to introduce in clinical practice, and validation by a large multicentre cohort is mandatory.

Author contributions statement

Conceptualization: MG

Data curation: MG, JP

Formal analysis: MG

Investigation: MG

Methodology: MG, JP

Supervision: MG, JP

Validation: JP, CF, MC, MS, TK, WK, FT, MG

Visualization: MG

Writing—original draft: MG, JP

Writing—review & editing: JP, CF, MC, MS, TK, WK, FT, MG

Meeting presentation: The manuscript was presented at the annual European Society of Thoracic Surgeons meeting during the Best Mixed Abstract III Session VII on Monday 21 June 2021.

Clinical registration number: CER-VD (Switzerland). Project ID: 2021-00620

Acknowledgements

The author would like to thank data managers of the CURCT, Matthieu Zellweger Ph.D. and Gilles Kratzer, Ph.D., for their help.

Funding statement

None.

Conflict of interest: None.

Data availability statement

The data underlying this article will be shared on reasonable request to the corresponding author.

References

Abbreviations

 
• BMI

  body mass index

 
• ESTS

  European Society of Thoracic Surgeons

 
• FDG-PET

  fluorodeoxyglucose positron emission tomography

 
• FEV1

  forced expiratory volume in 1 second

 
• IQR

  interquartile range

 
• NSCLC

  non-small-cell lung cancer

 
• OR

  odds ratio

 
• PAL

  prolonged air leak

 
• ROC

  receiver operating characteristic

 
• VATS

  video-assisted thoracoscopic surgery