Abstract
Objective: To assess the therapeutic effect of the extent of lymph node dissection performed in patients with a stage pI non-small-cell lung cancer (NSCLC). Methods: We analysed data on 465 patients with stage I NSCLC who were treated with surgical resection and some form of lymph node sampling. The median number of lymph node sampled was 10 and the median number of ipsilateral mediastinal lymph node stations sampled was two. We chose to define a procedure that harvested 10 or more lymph nodes and sampled two or more ipsilateral mediastinal stations as a lymphadenectomy, by contrast with sampling when one or both criteria were not satisfied. The effect of the surgical techniques: lymph node sampling (LS; n=207) vs. lymphadenectomy (LA; n=258) on 30-day mortality and overall survival were investigated. Results: A total of 6244 lymph nodes was examined, including 4306 mediastinal lymph nodes. The mean (±SD) numbers of removed lymph nodes were 7±6.1 per patient following LS vs.18.6±9.3 following LA (P=0.001). An average mean of 1±0.90 mediastinal lymph node station per patient was sampled following LS vs. 2.7±0.8 following LA (P≪10−6). Overall 30-day mortality rates were 2.4 and 3.1%, respectively. LA was disclosed as a favourable prognosticator at multivariate analysis (Hazard Risk: 1.43; 95% Confidence Interval: 1.00–2.04; P=0.048), together with younger patient age, absence of blood vessels invasion, and smaller tumour size. Conclusions: Importance of lymph node dissection affects patients outcome, while it does not enhance the operative mortality. A minimum of 10 lymph nodes assessed, and two mediastinal stations sampled are suggested as possible pragmatic markers of the quality of lymphadenectomy.
1 Introduction
Long term results of the surgical treatment of stage I non-small-cell lung cancer (NSCLC) are disappointing. Although surgical resection is considered as the standard of care because of the presumed local nature of the disease at that stage, as high as 30% of patients having undergone a complete resection experience recurrence, generally in distant organs. One of the explanation may be due to the technique of the bronchopulmonary and mediastinal lymph node dissection performed at surgery. Indeed, the extent of lymphadenectomy continues to be debated in the literature. The ‘adequate’ lymph node dissection is not yet established. Current surgical practice varies from mere visual inspection of the unopened mediastinum to radical lymphadenectomy.
Recently, major publications dealing with this topic were reported, based on prospective or large retrospective analysis [1–7]. However, the debate remains still open, and it appeared to us important to add our experience by conducting a retrospective analysis of a group of 465 consecutive patients who underwent resection of stage I NSCLC during the last 12-year period in one single institution. We investigated on the impact of the extent of the lymph node evaluation at surgery, lymph node sampling (LS) vs. lymphadenectomy (LA), on 30-day mortality and overall survival.
2 Materials and methods
2.1 Patients
The medical records of 636 patients who underwent lung resection and some form of lymph node evaluation for a pathological stage I non-small-cell lung cancer, as defined by the new International Classification System [8], were reviewed for the 12-year period from January 1992 to December 2003. Patient charts were identified by screening of a database into which data were entered prospectively for any patient undergoing surgery for lung malignancy at our department. Patients who had a history of another primary extrapulmonary malignancy (n=48), who presented with a metachronous lung carcinoma (n=8), who received neoadjuvant therapy (n=9), or who had undergone an atypical resection or a segmentectomy (n=24) were excluded from analysis. Finally, further 82 patients were excluded by reason of incomplete information about surgery and/or pathology.
There were 465 patients, 398 males and 67 females, whose age ranged from 31 to 85 years (mean: 62±10 years). Local-regional lung cancer staging was made with chest X-ray, computed tomography (CT) scan and fiberoptic bronchoscopy. Abdominal ultrasound and CT scan of adrenal glands was obtained routinely, whereas brain CT scan and nuclear bone scan were performed selectively in patients presenting with symptoms. Functional evaluation included spirometry, blood gas analysis at rest, and carbon monoxide diffusion capacity as a routine. Split perfusion nuclear scan was done in patients with an impaired pulmonary reserve (estimated post-resection FEV1 of less than 1l). All patients were operated on by or on the supervision of the same four staff surgeons.
2.2 Operative technique of lymph node evaluation
The technique of the lymph node evaluation was performed at surgeon discretion. The terms used by the surgeon in the operative report to qualify the technique of lymph node evaluation were not taken into consideration. The median number of lymph node sampled was 10 and the median number of ipsilateral mediastinal lymph node stations sampled, as described in the AJCC cancer staging manual [9], was two. Levels 10 and higher were considered N1 lymph nodes. Thus, we chose to define a procedure that harvested 10 or more lymph nodes and sampled two or more ipsilateral mediastinal stations as a lymphadenectomy, by contrast with sampling when one or both criteria were not satisfied. The choice of this cut off for these two variables (number of lymph node and number of mediastinal lymph node stations sampled) was reinforced by the results of ROC curves for the state of death (median of lymph node number=10, sensibility 63%, specificity 48.2% and median of mediastinal lymph node stations number=2, sensibility 66.5%, specificity 43.2%).
Patients were assigned to each study group by an independent observer after reviewing of the operative and pathology reports, and according to the following quantitative criteria. These two techniques were studied for comparison: lymph node sampling (LS; n=207) vs. lymphadenectomy (LA; n=258).
2.3 Statistical analysis
Continuous variables were expressed by means and standards deviations, whereas categorical data were presented using frequencies and percentage. Unpaired t-test were performed when appropriate to compare mean scores across subgroups whereas chi-square or Fisher exact test were used to compare frequencies. Operative mortality included all deaths occurring within 30 days after pulmonary resection or during the same hospital stay.
Survival data were updated for December 2003, by contact with the patient or his referring physician. For patients lost to medical follow-up, missing data were obtained by consulting the City Hall registry.
The following variables were considered as potential prognostic factors of survival: age, gender, surgical procedure (pneumonectomy vs. lesser resections), tumour size (cm), pathology (adenocarcinoma, squamous cell carcinoma vs. others), visceral pleura invasion (absent vs. present), lymphatic or/and blood vessel invasion (absent vs. present), technique of lymph node sampling (LS vs. LA). Survival was estimated using the product limit method of Kaplan–Meier from the date of the operation and included the operative mortality as well as any cancer-related and unrelated death. Comparisons were made using the log-rank test. P-values of 0.05 or less were considered statistically significant. Univariate and multivariate analysis were performed using Cox proportional hazards regression model to determine factors potentially predicting postoperative survival. The Cox model was used to incorporate in the same model any explanatory variables with a P-value less than 0.20. This statistical analysis was performed by using the SPSS V10.0 software package (SPSS Inc., Chicago, IL).
3 Results
3.1 Characteristics of patients
The characteristics of these 465 patients are shown in Table 1. No significant differences were found in terms of main clinical characteristics between the two groups of patients defined by surgical techniques (LA vs. LS) using chi-square or Fisher exact test. There were 381 lobectomies including bronchial sleeve lobectomy with (n=3) or without reconstruction of the pulmonary artery (n=6), and 23 bilobectomies including one bronchial sleeve resection.
A total of 6244 lymph nodes was examined, including 4306 mediastinal lymph nodes. The mean (±SD) numbers of removed lymph nodes were 7.0±6.1 per patient following LS vs.18.6±9.3 following LA (unpaired t-test: P≪0.0001). An average mean of 1.0±0.90 mediastinal lymph node station per patient was sampled following LS vs. 2.7±0.8 following LA (unpaired t-test: P≪0.0001).
3.2 Postoperative morbidity and mortality
Fifty major complications (10.8%) occurred (Table 2) and resulted in 13 postoperative fatalities (2.8%). The postoperative mortality rates were 2.4 and 3.1% for LS (n=5) and LA (n=8), respectively (P=0.66). Causes of death were all related to adult respiratory distress syndrome (ARDS).
3.3 Survival
At completion of the study, 325 patients were alive and 140 were dead. The median follow-up for the whole series was 31 months (mean: 39±31 months). The clinical status of survivors was as follow: alive and well (199/325, 61%), alive with disease (44/325, 14%), alive with another cancer (11/325, 3.3%), and unknown (71/325, 22%). The cause of death was related to the operation in 9.3% (13/140) of the cases, to the lung cancer in 38% (53/140), and to another cancer in 4.3% (6/140). Cancer-unrelated deaths accounted for 11% (15/140) of all deaths. The cause of death was not documented in 38% (53/140) of cases.
Estimated overall survival rate was 61.9% [56.3–67.5] at 5 years. The survival rates of stage I lung cancer patients with reference to various discrete categorical variables are shown in Table 3. Univariate analysis identified five of them as significant for a poor prognosis: older age, male gender, pathology, bigger tumour size and blood or/and lymphatic vessels invasion. LA was disclosed as a favourable prognosticator at multivariate analysis (Hazard Risk: 1.43; 95% Confidence Interval: 1.00–2.04; P=0.048, Fig. 1), together with younger patient age, absence of blood and lymphatic vessels invasion, and smaller tumour size (Table 4).
4 Discussion
We conducted this study to assess the effect of the extent of lymph node dissection in patients with a stage pI non-small-cell lung cancer on 30-day mortality and overall survival. We found that lymphadenectomy, defined as the removal of at least 10 lymph nodes and a minimum of two mediastinal stations explored, affected patients outcome, while it did not enhance the operative morbidity and mortality. Despite its retrospective design, our study presents with two characteristics reinforcing the meaning of its results: the prospective compilation of the data, and the comparison of two groups of patients treated concurrently that proved to be relatively homogeneous regarding to the main clinical variables. The most important pathological difference between the two groups relied to the pT status of the tumour, with significantly more T2 tumours in the LA group than in the LS group. Nevertheless, this difference, which would have had logically a detrimental impact on survival, did not prevent the manifestation of a survival improvement in the LA group. Furthermore, lymphadenectomy was found as an independent prognosticator of improved survival at multivariable analysis, suggesting a beneficial effect of the technique.
Controversies still exists as to the need for lymphadenectomy vs. lymph node sampling for the cure of early stage NSCLC. The optimal lymph node dissection during lung cancer surgery remains controversial among thoracic surgeons, even in the same team. Proponents of lymphadenectomy point out that the procedure improves the accuracy of disease staging and improves overall results by allowing a complete R0 resection and thereby decreasing local-regional recurrence, and directing more accurate adjuvant therapy. Opponents feel that lymphadenectomy is futile when facing a potentially systemic disease, and postulate higher morbidity and mortality.
The first issue to be addressed is the absence of standardized surgical methods for both sampling and lymphadenectomy, thus hampering any attempt of a comprehensive analysis of internationally available data. Surgeon's inability to intraoperatively determine the presence or absence of tumour within the mediastinal lymph nodes without biopsy was demonstrated by Gaer and Goldstraw [10], who compared the intraoperative visual and tactile evaluation of resected mediastinal lymph nodes with the ultimate histologic diagnosis. Thus, it may be reasonably presumed that the nodal status evaluation through an unopened mediastinal pleura would produce even less accurate results. These data explain why the technique of systematic node sampling should be preferred to random sampling. Systematic node sampling consists of lymph node biopsy at multiple predetermined levels within the mediastinum and bronchopulmonary areas. Complete lymphadenectomy is a concept even less well defined. Indeed, detailed data on the number of removed lymph nodes, as well as the description of nodal stations sampled are seldom reported in the surgical literature, thus precluding any comparison between series. It has been recently suggested that the quality of lymphadenectomy may be reflected quantitatively by the number of totally removed lymph nodes during thoracotomy and pathologically evaluated [6,7,11]. However, reported cut-off values vary greatly from 6 to 15. From a pragmatic point of view, the number of lymph nodes may be increased artificially when pieces of lymph nodes had been fragmented during dissection. Furthermore, as demonstrated by Riquet et al., there are wide variations of the number of lymph nodes at a given station between individuals [12]. All these considerations led us to define in a mixed way a procedure that harvested 10 or more lymph nodes and sampled two or more mediastinal stations as a lymphadenectomy, by contrast with sampling when one or both criteria were not satisfied. Of note, these definitions were in accordance with those used by the Eastern Cooperative Oncology Group in the ECOG 3590 trial (Intergroup, INT 0115) [3].
While it is commonly accepted that for an appropriate staging a minimum of hilar and mediastinal lymph nodes has to be removed and examined [2,3], the therapeutic efficacy of mediastinal lymphadenectomy is still under debate. The first randomised studies published in the English-written literature included a few number of patients, and failed to demonstrate any significant difference of outcome in patients having received radical systematic lymphadenectomy (LA) when compared to those having undergone lymph node sampling (LS) [1,2]. However, analysis of subgroups of patients according to the nodal status suggested that LA had a tendency to prolong relapse-free survival significantly with a borderline effect on overall survival in patients with minimal N disease (pN1 disease or pN2 disease with involvement of only one lymph node level) [2]. A Chinese group demonstrated recently on the basis of a large randomised clinical trial a significant survival advantage with LA over LS, and the subgroup analysis disclosed a significant benefit in stages I and IIIA NSCLC patients, and a marginally significance in stage II patients [5].
Whether or not removing healthy lymph nodes improves survival in lung cancer is still a moot point. Considering our data and these recent publications [5–7], it may be suggested that there is a place for mediastinal lymphadenectomy for pathological stage I disease to improve survival, even if the ‘adequate’ definition of mediastinal lymphadenetomy will still remain an open issue. Obviously, the value of mediastinal lymphadenectomy in early stage NSCLC needs to be confirmed by at least another phase III study, now in progress (American College of Surgeons Oncology Group, ACOSOG protocol Z0030). Even if stage migration, the so-called ‘Will Rogers’ phenomenon, interferes clearly with these findings between LA and LS, a therapeutic efficiency of lymphadenectomy may be hypothesized. This therapeutic efficiency of LA may be grounded through the presence of occult micrometastatic disease in the lymph nodes. The concept of occult micrometastatic disease originates from the detection of metastatic cells which are not detected by current clinical staging examinations and conventional histopathologic methods, using immunohistochemical or molecular methods. The presence of micrometastatic cells in the lymph nodes, the bone marrow, and the peripheral blood, can be considered as a marker for primary tumors with a high metastatic potential [13–15]. Osaki et al. demonstrated recently that micrometastasis in the bone marrow and the lymph nodes of patients with completely resected stage I NSCLC were detected with the same frequency (27.8%), without evidence of any relationship between these two sites, suggesting different regulations or mechanisms [13]. LA allows to remove micrometastatic lymph nodes and, in turn, increases the proportion of complete R0 resections, leaving no residual disease. A German group investigated on the therapeutic effectiveness of LA vs. LS in N0 patients as determined by conventional histopathological analysis, but with nodal micrometastases as detected by using immunohistochemical methods. Unfortunately, they failed to demonstrate any influence on prognosis of the type of lymphadenectomy in such patients with early disseminated tumour cells in lymph nodes [4]. Paradoxically, a survival benefit of LA was suggested in patients with a N0 disease as defined with both routine histology and immunohistochemistry.
The operative mortality observed in this series was in line with current standards [16]. The risk from pulmonary resection is dependent on the extent of resection. In that way, it is important to point out that the proportion of patients having undergone a pneumonectomy was about 13% in the present series, a rate that is in accordance with what is usually reported in this early stage, ranging from 5 to 20% [6,7,17–20]. Of concern is the fact that the technique of lymph node sampling did not influence postoperative mortality or major morbidity accordingly to previous claims [5,7], but in contrast with others [1]. Potential complications of lymphadenectomy may arise from injuries to the bronchial arteries and nerves, the tracheobronchial tree itself, and recurrent nerves. Mediastinal lymphadenectomy also impairs the lymphatic backflow and may add to pulmonary oedema. The occurrence of respiratory complications, i.e. pneumonia and ARDS, was very similar in the two groups. The incidence of laryngeal nerve paralysis was higher in the LA group although the difference did not reach statistical significance. An impaired laryngeal mobility can be dramatic after lung resection, and it should be tracked and treated as soon as possible. Its consequences include swallowing disorder and ineffective cough, both concurring to life-threatening events [21]. Our policy included immediate laryngeal examination and delayed oral feeding, and sometimes early thyroplasty as advocated by some authors [22]. This approach maintained related respiratory events at a relatively low level.
To conclude, importance of lymphadenectomy affects patients outcome, while it does not enhance the operative mortality. A minimum of 10 lymph nodes assessed, and two mediastinal stations sampled are suggested as possible pragmatic markers of the quality of lymphadenectomy.
Survival curves according to the type of lymph node dissection (LS vs. LA) in surgically resected pathological stage I non-small-cell lung cancer.
Characteristics of the two groups of patients (LS vs. LA)
Major postoperative complications
Relationship between categorical variables and survival for patients with surgically resected pathological stage I non-small-cell lung cancer
Univariate and multivariate analyses of prognostic factors in surgically resected pathological stage I non-small-cell lung cancer
Presented at the joint 18th Annual Meeting of the European Association for Cardio-thoracic Surgery and the 12th Annual Meeting of the European Society of Thoracic Surgeons, Leipzig, Germany, September 12–15, 2004.
