Morbidity, mortality and survival after 110 consecutive bilobectomies over 12 years

Abstract

OBJECTIVES

To analyse statistical aspects of mortality, morbidity and survival after bilobectomy (BT), an operation rarely studied in the literature.

METHODS

One hundred and ten cases were studied, comprising 58 upper-middle bilobectomies and 52 lower-middle bilobectomies performed between 1999 and 2010. Indications were of 9 benign diseases, 12 carcinoid tumours, 5 metastases and 84 non-small cell lung cancers (2 stage 0; 34 stage I; 22 stage II; 25 stage III and 1 stage IV).

RESULTS

Mortality was nil. Twenty-six percent of patients experienced significant morbidity, influenced in multivariate analysis by the presence of three or more comorbidities (P = 0.03) and by a forced expiratory volume in 1 s of <60% (P = 0.01). Lower-middle BT was associated with more postoperative complications than upper-middle BT (P = 0.012). The 5-year survival rate of patients with non-small cell lung carcinoma was 82% in stage I, 59% in stage II and 20% in stage IIIA. Survival was significantly influenced by stage (P = 0.0018) and tobacco weaning (P = 0.0012).

CONCLUSIONS

BT can be achieved with low mortality, and survival results that are comparable with those unregistered after standard lobectomy. However, almost one quarter of patients experienced significant postoperative complications. Surgical techniques aiming to reduce residual pleural space should be especially considered after lower-middle BT, due to the highest morbidity being associated with this procedure.

INTRODUCTION

Since the first bilobectomy (BT) performed by Churchill in 1933 and the first series published in 1988, the postoperative course of this procedure has been generally considered as intermediate between that of lobectomy and pneumonectomy [1-6]. Owing to the small number of studies and their age—all but two having been carried out over 15 years ago—the postoperative course and the risk factors relating to this operation are not well established. The objectives of our study were to analyse the mortality, morbidity and 5-year survival rate associated with BT in our institution in the previous 12 years.

MATERIALS AND METHODS

Collection of data

This was a retrospective case review. Clinical records of patients who underwent BT at the University Hospital of Caen between 1st January 1999 and 31st December 2010 were reviewed. No case was excluded and, at the completion of the study (1st December 2011), no patient had been lost during follow-up. Operative reports were reviewed to determine the indication(s) for BT, type of BT (i.e. upper and middle bilobectomy vs lower and middle bilobectomy), type of bronchial stump closure (stapling or handle) and number of drains. Hospital charts were also studied to establish age, gender, histological pathologic diagnosis and stage (according to the current 7th edition of the American Joint Commission for Cancer TNM classification [7]), perioperative mortality and postoperative complications, duration of drainage, prolonged air leakage (defined by bubbling exceeding 7 days) and post-discharge complications. Tobacco intoxication (in pack-years of smoking) and date of quitting were also recorded. One hundred and ten patients (26 women and 74 men) with a mean age of 60 years (range 19–82 years) were recorded. Fifty-eight upper-middle bilobectomies and 52 lower-middle bilobectomies were performed. Nineteen patients (17%) were older than 70 years, 18 of them presenting cancer. A smoking history was found in 91 patients (83%) with a median tobacco exposure of 36 pack-years (range 8–150).

Preoperative assessment

The extent of tumour was evaluated by chest X-Ray, fibreoptic bronchoscopy, thoracoabdominal and brain computed tomography scans. Since 2004, positron emission tomography (PET) scan was also performed as a routine examination. Respiratory function was evaluated by spirometry and blood gas analysis. Mediastinoscopy was not routinely performed, except to rule out contralateral N3 disease or extensive N2 disease, in case of which neoadjuvant chemotherapy was undertaken and indications for surgery were reconsidered after evaluation of the chemotherapy response. Patients with significant chronic obstructive pulmonary disease (COPD) underwent complementary investigations to estimate postoperative breathing capacity (by lung ventilation perfusion scan) and/or maximal consumption of O₂ during exercise (VO₂ max). Moreover these patients were carefully checked by echocardiography to exclude pulmonary hypertension or chronic heart failure and coronarography was carried out if there was any suspicion of coronaropathy.

Comorbidity factors

The following variables, considered as comorbidity factors, were especially studied: body mass index in kg/m², alcoholism, persistent tobacco intoxication, high blood pressure, hypertension, diabetes mellitus, peripheral artery disease, coronary heart disease, arrhythmia, presence of COPD (according to the GOLD criteria definition i.e. forced expiratory volume in one second/forced expiratory vital capacity (FEV1/FVC) ratio<0.7 [8]), American Society of Anaesthesiologists (ASA) score >2 (patient with severe systemic disease) and preoperative chemotherapy. Performance status was assessed using the Eastern Cooperative Oncology Group (ECOG) score running from 0 to 5, with 0 denoting perfect health and 5 indicating death. Six patients suffered minimal COPD (FEV1 > 80%), whereas 19 patients had moderate COPD (80% < FEV1 < 50%): 3 patients with 80% < FEV1 < 70%, 9 with FEV1 < 70% and 7 with FEV1 < 60%. Among these 19 patients, pneumonectomy was contraindicated in 7 cases and considered as an ‘at risk’ procedure in 9 cases because predictive postoperative FEV1 was less than 40% with maximal oxygen uptake (VO₂ max) less than 20 ml/kg.

Cardiopulmonary morbidity was defined by the following criteria, according to the ESTS Database (Annual Report 2011, produced by the ESTS Database committee; printed by Tecnoprint srl AN, April 2011; available at : www.ests.org):

• respiratory failure and mechanical ventilation for more than 24 h

• reintubation at any time of the postoperative period

• pneumonia

• atelectasis requiring bronchoscopy

• pulmonary embolism or oedema

• supraventricular arrhythmia requiring treatment

• ventricular arrhythmia

• acute myocardial ischaemia

• stroke

• renal insufficiency

• cardiac failure.

Forty-three percent of our patients had comorbidity factors and 8% had coronary artery disease. As indicated in Table 1, the most frequent comorbidity factors were high blood pressure (35%) and COPD (23%). When considering that persistence of tobacco intoxication until operation was a comorbidity factor, 89% of patients (98/110) presented at least one comorbidity factor: 31 patients had one, 21 had two, 16 had three and 30 had four or more of these factors (Table 1).

Table 1:

Preoperative comorbidity factors experienced by patients

Comorbidity	n	%
Age (years)
Mean	59.6
<70 years	91	82.5
>70 years	19	17.5
Sex ratio	74 men/26 women
Body mass index (kg/m²)
Mean	25.6
<20 kg/m²	9	8
20–30 kg/m²	82	74.5
>30 kg/m²	19	17.5
Performance status
0	55	50
1	49	44.5
2	6	5.5
COPD
None	85	77.5
Minimal	6	5
Moderate	19	17.5
Alcoholism
Yes	10	91
No	100	9
Tobacco (pack-years) [min–max]	30.3 [0–150]
Quit smoking date
Yes	89	81
No	21	19
High blood pressure
Yes	39	35.5
No	71	64.5
Diabetes mellitus
Yes	9	8
No	101	92
Peripheral vascular disease
Yes	20	19
No	90	82
Coronary artery disease
Yes	9	8
No	101	92
Cardiac arrhythmia
Yes	8	7.5
No	102	92.5
FEV1 (%)
Mean	87.2
50–60%	7	6.5
60–70%	9	8
70–80%	22	20
80–90%	28	25.5
>90%	44	40
ASA score
1	74	67
2	16	15
3	20	18
Preoperative chemotherapy
Yes	22	20
No	88	80

COPD: chronic obstructive pulmonary disease, FEV1: forced expiratory vital capacity; ASA: American Society of Anaesthesiologists.

Patient management, indication and type of operation

Preoperative respiratory physiotherapy was required for 2 weeks. Despite robust persuasion to quit smoking, 21 patients were still smokers at the time of bilobectomy. Prophylactic antibiotics using a first generation cephalosporin (cefazolin 2 g intravenous) were administered before the induction of anaesthesia according to French national guidelines (sfar.org). A double-lumen tube was inserted and bilobectomy with systematic radical lymph node resection was performed through a posterolateral thoracotomy, sparing the serratus anterior muscle.

Coverage or space reduction techniques such as pleural tent, sealants, buttressed staple lines or pneumoperitoneum were not routinely used. Manual sutures were generally used when the bronchus had been opened to examine the proximal extent of the tumour and ensure clear margins before closure. A total of 76 bronchial stumps were closed automatically with staples and 34 manually by interrupted 4/0 nylon sutures.

Since 2005, most of the patients were drained using a single Blake 24-French drain rather than the two silicone rubber 30-French drains previously used, in accordance with a study we made that showed the safety and benefit of a single drainage in thoracic surgery [9]. As a result, 47 patients had double drainage and 63 had single drainage chest tubes, all of which were removed when persistent drainage was less than 200 ml/24 h and without air leak for over one day.

All patients who were supported for a neoplastic disease [non-small cell lung cancer (NSCLC) i.e. carcinoid tumours and metastases] underwent a complete lymph node dissection.

After surgery, patients were monitored for at least 24 h in the intensive care unit. Multimodal analgesia was provided associating morphine-based patient controlled analgesia with nefopam administration. Chest X-rays were performed daily until chest drain removal and up to hospital discharge. Serial C-reactive protein measurements were routinely performed at 3-day intervals for early detection of infection such as empyema.

In order to reduce the size of the residual pleural cavity, patients received active respiratory physiotherapy twice per day. This treatment was associated with optimizing bronchodilator therapy, to promote adherences between the lung and the chest wall. Patients were generally discharged a few days after the removal of drainage, usually either at home or at a nursing convalescent home.

There was no change support both indications, surgical technique and protocols for anaesthesia and analgesia during the 12 years of this study

Indications for operation are depicted in Table 2.

Table 2:

Surgical anatomical indications for bilobectomy

Type of bilobectomy	n	Indication for bilobectomy
Upper-middle bilobectomy (n = 58)	20	Involvement of the minor fissure
	20	Ventral tumour close to incomplete or absent fissure
	8	Disease into two lobes (5 NSCLC, 2 metastasis, 1 benign disease)
Lower-middle bilobectomy (n = 52)	22	Internal or external involvement of the bronchus intermedius
	26	Invasion of the anterior part of the major fissure
	4	Disease into two lobes (2 NSCLC, 1 metastasis, 1 benign disease)

Pathological examination

There were 9 benign lesions (1 bronchiectasis, 1 botryomycosis, 1 Wegener's granulomatosis, 1 tuberculosis, 1 plasmocytary granuloma, 2 aspergilosis and 2 hamartochondromas), 12 carcinoid tumours, 5 metastases of various origins and 84 NSCLC. Among the NSCLC, there were 41 adenocarcinomas, 36 squamous cell carcinomas and 7 other cell carcinomas (5 large cells, 1 sarcomatoid and 1 bronchiolo-alveolar). Among the 84 NSCLC, 2 were stage 0 (complete histological response after chemotherapy), 34 stage I (19 stage IA and 15 stage IB), 22 stage II (12 stage IIA and 10 stage IIB), 25 stage III (24 stage IIIA and 1 stage IIIB) and 1 stage IV. Among the 24 stage IIIA, there were 7 synchronous double cancers.

Statistical methods

Data are expressed as mean and standard deviation and median (interquartile range) for non-normally distributed variables. Normality was assessed with the d'Agostino–Pearson test. Comparison of two means was performed using the unpaired student's t-test; comparison of two medians was performed using the Mann–Whitney test. Categorical variables were each expressed as number and percentage. Comparison of proportion was performed using Pearson’s chi-squared test and Fisher's exact method, as appropriate, was used for comparison of qualitative data.

We performed a logistic regression to assess variables associated with the occurrence of postoperative complications. The binary dependent variable was defined as patients without complications or with minor complications (group 0 and 1) and patients suffering important or severe complications (groups 2 and 3). Since we observed 28 events, 3 independent predictors were included in the model, in order to limit over-fitting. The odds ratios and their 95% confidence interval of variables selected by the logistic model were calculated. Overall performance of the model was evaluated using Nagelkerke's R². Calibration and discrimination were evaluated using the Hosmer–Lemeshow test and the c-statistic, respectively. The overall performance optimism was evaluated using a bootstrapping procedure with 1000 bootstrap samples.

Multivariate analysis was performed by a Cox proportional hazard model, entering the three most significant factors due to the low number of deaths in our study. Statistical analysis was performed with Statistics, Epidemiology, Medicine (SEM) software and R version 2.14.2 (www.r-project.org) with specific packages.

All P values are two-tailed and values <0.05 were considered as significant.

RESULTS

Mortality and morbidity

The operative mortality was nil. All patients survived at least 90 days after their operation or until the end of their hospitalization.

Thirty-five patients (32%; group 0) did not show any complication and had a mean hospitalization time of 8.7 days (5–15 days). Seventy-five patients (68%) experienced complications, listed in Table 3. For better comprehension, these complications were classified into three groups of increasing severity: minor, severe and major. This severity was corroborated by a significant increase of the mean hospitalization time (P < 0.00001) (Table 3), but not by the mean durations of drainage, which were, respectively, 5.6, 10.2, 12.7 and 12.8 days in groups 0, 1, 2 and 3. In this study of 110 patients, the total duration of drainage was 9.5 days and the overall mean hospitalization time was 15.5 days.

Table 3:

Postoperative complications and mean hospitalization time.

Group	n	Type of complications	MHT in days [min–max]
No complications	35	None	8.7 [5–15]
Minor complications n = 46 Group I	22	Prolonged air leak	13 [9–22]
	16	Retained secretion requiring fibroscopy
	7	Atrial fibrillation
	1	Recurrent laryngeal nerve injury
Important complications n = 19 Group II	15	Pneumonia	19 [11–29]
	1	Chylothorax
	1	Haemothorax
	1	Acute pancreatitis
	1	Contralateral pneumothorax requiring VATS
Severe complications n = 10 Group III	5	ARDS	43 [5–111]
	2	Empyema without bronchopleural fistula
	2	Empyema with bronchopleural fistula
	1	Massive pulmonary embolism

MHT: mean hospitalization, VATS: video-assisted thorascopic surgery, ARDS: acute respiratory distress syndrome.

Concerning postoperative complications, 74% (n = 81/110) of patients (groups 0 and 1) experienced no complications or only minor ones, with a mean hospitalization time of 11 days. In contrast, 26% of patients (29/110; groups 2 and 3) experienced significant complications, with a mean hospitalization time of 27.3 days. In group 3, two bronchopulmonary fistulas were recorded: one occurred in a high-risk patient who had previously received chemotherapy and high-dose radiation for treating an oesophageal cancer; the other occurred in a non-observant smoker patient with diabetes mellitus. These two fistulas were successfully treated by apposition of an intercostal muscle flap. In total, seven patients of the entire series required a second operation: one for evacuating a haemothorax, one for treating an incidental contralateral pneumothorax, one for thrombectomy of a pulmonary artery embolism using cardiopulmonary bypass, one for treating an empyema by an open window thoracostomy after failure of pleural lavage and two for closure of a bronchopleural fistula, as mentioned above.

Factors influencing morbidity

Factors associated with postoperative complications are presented in Table 4.

Table 4:

Factors influencing postoperative complications in univariate analysis

Factors	Groups 0 & 1 (n)	Groups 2 & 3 (n)	P values
Sex
Male	56	28	0.0029
Female	25	1
Pathological examination
NSCLC	57	27	0.013
Other	24	2
Number of comorbidity factor
0 to 3	64	16	0.013
4 or more	17	13
FEV1 (%)	88.76	82.97	0.24
<60%	2	5	0.019
>60%	79	24
COPD
Yes	14	11	0.023
No	67	18
Tobacco (pack-years)	33	43	0.038
Tobacco intoxication
Stopped or never smoked	67	19	0.048
Active intoxication	14	10
ASA score
1	14	2	0.068
2	56	18
3	11	9
Age (years)
58.9	61.4	0.46	0.087
<70 years	70	21
>70 years	11	8
Performance status
0	45	10	0.15
1	32	17
2	4	2
Preoperative chemotherapy
Yes	14	8	0.23
No	67	21
Coronary artery disease
Yes	5	4	0.37
No	76	25
Body mass index (kg/m²)	25.5	25.9	0.43
Peripheral vascular disease
Yes	16	4	0.48
No	65	25
Alcoholism
Yes	6	4	0.52
No	75	25
Cardiac arrhythmia
Yes	6	2	0.92
No	75	27
Type of bilobectomy
Upper-middle	42	16	0.76
Lower-middle	39	13
High blood pressure
Yes	29	10	0.9
No	52	19
Diabetes mellitus
Yes	6	3	0.92
No	75	26

Results of univariate analysis

When patients without complications or with only minor complications (groups 0 and 1) were compared to patients suffering significant or severe complications (groups 2 and 3), seven variables were found to be important, in univariate analysis, for predicting the occurrence of a postoperative complication. These were male gender, type of tumour (NSCLC vs other histological disease), presence of COPD, FEV1 < 60%, presence of four or more comorbidity factors, total amount of tobacco intoxication in pack-years and active tobacco intoxication before surgery.

It is noteworthy that all six patients with minimal COPD presented complications (three pneumonias, two prolonged air leakages exceeding 7 days, and one retained secretion requiring fibroscopy), whereas 84% of patients (n = 16/19) suffering more severe COPD presented postoperative complications, often severe: five retained secretion requiring bronchoscopy, four pneumonias, two prolonged air leaks, two ARDS, one atrial fibrillation, one haemothorax and one empyema with bronchopleural fistula requiring surgery.

Significantly more patients demonstrated any form of morbidity after lower-middle bilobectomy than after upper-middle bilobectomy (when comparing groups 1 + 2 + 3 vs group 0) (P = 0.012). This difference was corroborated by a trend of longer duration of drainage in the lower-middle bilobectomy group (10.2 days vs 8.8 days; P = 0.12) and prolonged hospitalization time (17.8 days vs 13.5 days; P = 0.10). The presence of one drain vs two drains did not influence morbidity and no significant differences were observed in hospitalization time between these two groups.

Results of multivariate analysis

The presence of three or more comorbidity factors (odds ratio (OR) 2.75 [1.06–7.13]; P = 0.03) and preoperative FEV1 < 60% (OR 9.27 [1.44–59.46]; P = 0.01) were independent predictors of postoperative complications. A good calibration (Hosmer–Lemeshow test: P = 0.73) and discrimination (c-statistic = 0.72) were observed. The overall optimism evaluated on 1000 bootstrap samples was 0.06.

Survival

The overall survival of patients presenting benign and carcinoid tumours was 100% at the completion of the study, whereas 50% of patients who underwent resection of lung metastasis were still alive two years after the operation.

Factors influencing survival in the NSCLC sub-group

Survival of patients with NSCLC was very significantly influenced by tumour stage and lymph node status (Figs 1 and 2). In univariate analysis, no significant survival difference was observed, considering age (under or over 70 years), sex, type of bilobectomy and histology (adenocarcinomas vs squamous cell carcinomas). Seven factors were significantly associated with improved 5-year survival: these were cancer stage (P < 10⁻⁷), tobacco weaning (P = 0.00019), FEV1 > 65% (P = 0.026), absence of any kind of postoperative complication (P = 0.017), body mass index < 23 kg/m² (P = 0.012), absence of cardiac arrhythmia (P = 0.012) and complete resection (R0) (P = 0.05). In multivariate analysis, survival was significantly influenced by two factors: cancer stage (P = 0.0018) and tobacco weaning (P = 0.0012).

Figure 1:

Five-year survival of the 84 cases of NSCLC by stages. Stages were recorded according to the seventh edition of the TNM Classification revised in 2007 [7]. The 5-year survival was 100% for stage 0 (n = 2; curve 1), 82% for stage I (n = 34; curve 2), 59% for stage II (n = 22; curve 3), 26% for stage IIIA (n = 25; curve 4) and 0% for stage IV (n = 1; curve 5) (P < 0.0000001).

Figure 2:

Five-year survival by lymph node status: 78% for N0 (n = 52; curve 1), 33% for N1 (n = 18; curve 2) and 28% for N2 disease (n = 14; curve 3) (P = 0.0048).

DISCUSSION

As shown in previous studies [2–6], tumour extension across a fissure, internal or external involvement of the bronchus intermedius, absence of fissure, and synchronous double cancers were the main indications for bilobectomy (42, 20, 18 and 11%, respectively, in our series). As indicated in Table 5, three databases have recently reported a mortality rate between 4.3 and 6.9%. There were the 2011 ESTS Database Annual Report 2011, available at: http://www.ests.org (accessed 10 June, 2012) and the Base de données EPITHOR de la Société Française de Chirurgie Thoracique et Cardio-Vasculaire, huitième (Juin 2011) et neuvième rapport (Juin 2012), available at: http://www.sfctcv.net (accessed 1st June, 2012) [10]. This rate lies midway between that of standard lobectomy and pneumonectomy and ranges, respectively, from 1.9–3.6% and 6.2–8.0% in the same databases. Unfortunately, because precise reports concerning bilobectomy are still lacking in these databases, we could only hypothesize to explain the lower mortality that we registered.

Table 5:

Thirty-day mortality after bilobectomy in the literature

Series	Period	Number of death/ number of patients	% of death
Keller et al. [2]	1974–1985	7/166	4.2
Deneuville et al. [3]	1978–1988	9/148	6.1
Massard et al. [4]	1980–1992	4/112	3.5
Allen et al. [18]	1999–2004	2/42	4.8
STS database [10]	1999–2006	12/353	3.4
Kim et al. [5]	1984–2007	4/92	4.3
Galetta et al. [6]	1998–2009	2/146	1.4
French database	2005–2010	77/1115	6.9
ESTS database	2007–2011	42/930	4.5
Current series	1999–2009	0/110	0
All series	1974–2011	159/3214	4.94

In our experience, our lower mortality rate (around 1%) after bilobectomy was not significantly different from that of lobectomy in recent years and could be due to several hypothetical causes:

• Firstly, an appropriate preoperative preparation of our patients selected for bilobectomy, through 2–3 weeks respiratory physiotherapy and tobacco weaning. Even if controversial, we still consider that persistent tobacco intoxication, up to the operation, is a major risk factor for postoperative complications [11]. We observed, at least in univariate analysis, that tobacco weaning significantly reduced morbidity in bilobectomy. Therefore, whenever the tumour is not too large, delay of the operation to allow complete cessation of smoking should be discussed.

• Secondly, the better medical condition of our patients selected for bilobectomy: mortality after pulmonary resection has been often considered to be correlated to the severity and numbers of comorbidity factors or to an increased ASA score [12]. Regarding the number of comorbidity factors, our 42% rate of patients with three or more comorbidity factors was higher than the 41% rate of patients with two or more comorbidity factors, who undergone lobectomy in the French Epithor database (French Society of Thoracic and Cardiovascular Surgery; Epithor database; available at: http://www.epithor.net/). In contrast, our 18% rate of ASA 3 score patients seemed lower than the 26% and 27% rates after lobectomy, respectively, recorded on the Epithor and ESTS databases, although ASA scores and number of comorbidity factors (more than 3 or 4) were significant or close to being significant factors in our study for predicting morbidity in univariate analysis, which considered male gender and FEV1 < 60% as significant. Male gender has previously been implicated as a risk factor for mortality and postoperative complication after lung surgery [13]. We may emphasize the importance of COPD as previously described by Licker et al. [14], because more than 80% of our patients with COPD demonstrated at least one complication, often pulmonary and severe. Our absence of mortality could be related to a lower rate of severe COPD although, in 7 cases, pneumonectomy was functionally contraindicated. Cardiopulmonary morbidity, a well-known risk factor for morbidity and mortality following lung surgery, reached 43% in our study. This was not significantly different to the 35% rate recorded in the ESTS database (n = 326/930 vs 47/110; P = 0.11). Our 8.2% rate of patients with coronary artery disease was similar to the 9% recorded by the same databases for patients undergoing lobectomy or pneumonectomy.

• Thirdly, to an appropriate postoperative care and/or a ‘happy hazard effect’: none of our four patients with acute respiratory distress syndrome (ARDS) died, in comparison to the 40% and 50% mortality rates, respectively, of this complication after lobectomy and pneumonectomy [15].

• Finally, our trained and experienced surgical team, considering hospital volume and experience and human medical factors, has been shown to be essential to explain the low mortality in several reports [10, 16–18], such as that of Galetta et al. [6] who also recently reported a 1.4% mortality rate after bilobectomy (Table 5).

Although mortality was nil, almost 25% (n = 29/110) of our patients experienced a severe complication that, in ten cases, was major and life threatening and that, in 7 cases, required reoperation. This ‘cloudy’ postoperative outcome, reflected by an overall mean hospitalization time of 15.5 days, was much closer to those of pneumonectomy. Than that generally observed after pneumonectomy, the typical lengths of stay for lobectomy being approximately 9 days in our experience and similarly reported by others [10, 13, 16]. Although mean hospitalization time is influenced by other factors unrelated to the patient's recovery from surgery, it is most often a good reflection of postoperative outcome. The longer mean hospitalization time we observed after bilobectomy was largely influenced by prolonged air leakage—the most frequent complication—that we recorded in 20% of our cases, a rate similar to the 15% and 19% reported by others [6, 18]. These rates were higher than those recorded in the ESTS database (12%) and the French Epithor database (11%) and higher than the 8% usually recorded after lobectomy in our experience and as seen by others [18]. To reduce prolonged air leakage, measures such as sealants, pleural tent or buttressed staple lines should be more routinely considered than the sparing use that we have made of the techniques.

Our study confirms the safety of the use of a single Blake drain, as we have previously reported [9]. This single drain may advantageously replace the third drain advocated by Massard et al. because its flexible quality allows safe drainage of the residual inferior pleural space [4]. Our incidence of bronchopulmonary fistula after bilobectomy was 1.8%: close to the 1.1% and 0.6% recently reported by others and lower than those previously reported in older series [3, 5, 6]. This result might be due to the use of modern stapling devices for bronchus closure and to the better general medical condition of patients currently selected for operation. In our two cases of bronchopleural fistula, closure of the fistula was successfully achieved by an intercostal muscle flap apposition.

As previously mentioned by several authors—but contrary to others—we confirm that lower-middle bilobectomy, a higher-risk procedure than upper-middle bilobectomy, was associated with higher morbidity (P = 0.012) [2, 3, 6]. In fact, after lower-middle resection, a very large inferior and posterior residual pleural space is left in place, due to the disparity of size and lack of congruence between the remaining upper lobe and the size of the thoracic cavity. This ‘dead-space’ cavity, which fills with pleural fluid, may be considered most often as an ‘after effect’ of this operation that should only be tolerated when it is not too large, knowing that stabilization or resolution generally occurs within one year [2, 19]. To reduce this cavity as much as possible, procedures such as resection of the 6th rib, insertion of a supplemental drain with a high suction level of 150 mmHg, creation of a pneumoperitoneum, or phrenic nerve scratch should be considered [3, 4, 6]. We thought that early removal of the drain would increase the risk of creating too large a residual pleural cavity. Therefore, when bubbling was minimal or absent, we applied simple water-seal drainage and the patient was encouraged to walk in the surgical unit. However this attitude should be appraised in prospective studies to determine the best method of drainage and its optimal duration.

As shown in the figures, our 5-year survival rates for NSCLC are similar to those reported by IASLC studies [7]. We confirm that bilobectomy with radical lymphadenectomy is a valuable option for treating lung cancer, that can be planned as a routine operation, even for patients who might have tolerated pneumonectomy on functional results [2–6]. Unlike others, we found no difference in survival between adenocarcinomas vs squamous cell carcinomas [5], nor between upper-middle bilobectomy and lower-middle bilobectomy [6]. In our series, tobacco weaning positively influenced long-term survival after resection of lung cancer.

Limits of the study

Our study has some limitations because of its retrospective and monocentric design. Our results in terms of morbidity are perhaps not exportable to other centres, although our population seems to be representative of patients undergoing lung surgery (NSCLC, metastasis, carcinoid tumours and benign diseases). Data were collected retrospectively, and not prospectively, from a database service, which increases the risk of loss of information. However, none of the cases was rejected and the data collection was found complete for all patients. Further prospective studies should be performed to evaluate the results of this operation and to clarify the best option (bilobectomy vs lobectomy and wedge) for treating cancer with minimal invasion of the adjacent lobe, a problem which is currently debated [20].

Conflict of interest: none declared