Fax: (717) 531-0882
Extent of lymphadenectomy and outcome for patients with stage I nonsmall cell lung cancer
Article first published online: 12 JAN 2009
Copyright © 2009 American Cancer Society
Volume 115, Issue 4, pages 851–858, 15 February 2009
How to Cite
Varlotto, J. M., Recht, A., Nikolov, M., Flickinger, J. C. and DeCamp, M. M. (2009), Extent of lymphadenectomy and outcome for patients with stage I nonsmall cell lung cancer. Cancer, 115: 851–858. doi: 10.1002/cncr.23985
- Issue published online: 2 FEB 2009
- Article first published online: 12 JAN 2009
- Manuscript Accepted: 17 JUN 2008
- Manuscript Revised: 11 JUN 2008
- Manuscript Received: 14 FEB 2008
- nonsmall cell carcinoma;
- mediastinal dissection;
- lymph nodes
It is uncertain whether lymphadenectomy (LA) affects overall survival (OS) or disease-free survival (DFS) rates for patients with stage I nonsmall cell lung cancer (NSCLC), as is the optimal number of lymph nodes that should be recovered.
There were 24,273 patients with stage I NSCLC diagnosed from 1992 to 2002 who were included in the Surveillance, Epidemiology, and End Results database and who underwent a definitive surgical procedure. Median follow-up was 35 months.
For the entire population, having LA was associated with an increase in the 5-year OS rate from 41.6% to 58.4% (P<.0001) and in DFS from 58.0% to 73.09%, compared with not having LA. Outcome improved with increasing number of recovered lymph nodes, with a plateau at 11 or more lymph nodes. For patients diagnosed from 1998 to 2002 undergoing only N1 or only N2 dissections, LA was also associated with statistically significant improvements in OS in both groups and a significant difference and trend for improved DFS in the 2 groups, respectively. The maximum differences in both OS and DFS between those with no LA and those with LA occurred when 11 to 16 lymph nodes were removed for the former group or 7 to 10 lymph nodes for the latter group, respectively.
LA was associated with increased rates of OS and DFS, compared with no LA. Our results also suggest the minimum number of recovered lymph nodes needed to see the maximum staging accuracy conferred by LA. Cancer 2009. © 2009 American Cancer Society.
There is no current consensus on whether to perform lymphadenectomy (LA) or on its optimal extent for patients with nonsmall cell lung cancer (NSCLC).1 For example, the Eastern Cooperative Oncology Group has defined lymph node sampling (MLNS) as removal of at least 1 lymph node and complete dissection (MLND) as removal of all lymph nodes from levels 4, 7, and 10 during right-sided thoracotomy or levels 5, 6, and 7 during left-sided thoracotomy.2 However, surgeons from the University Hospital of Zurich defined MLND as removal of all lymph nodes from levels 2 to 4 and 7 to 9 for right-sided lung cancers or levels 4 to 9 for left-sided lung cancers.3 Recently, the International Association for the Study of Lung Cancer (IASLC) Staging Committee proposed standards for adequate lymph node removal based on tumor location: for right upper and middle lobe cancers, subcarinal lymph nodes and at least 2 of 3 stations from the superior paratracheal, inferior paratracheal, and pretracheal regions; for right lower lobe tumors, subcarinal and right inferior paratracheal lymph nodes and either the paraesophageal or pulmonary ligament lymph nodes; for left upper lobe tumors, subcarinal, paraesophageal, and anterior mediastinal lymph nodes; and for left lower lobe cancers, subcarinal, paraesophageal, and pulmonary ligament lymph nodes.4
Similarly, there is continued controversy concerning the number of lymph nodes that need to be removed for proper staging or therapeutic purposes. The IASLC recommends removal of at least 6 lymph nodes, 3 from intrapulmonary and/or hilar locations and 3 from mediastinal locations, 1 of which must be the subcarinal station.4 Two retrospective reviews (containing 442 and 465 patients) found that removal of at least 6 lymph nodes and dissection of at least 3 mediastinal stations,5 or removal of 10 lymph nodes and dissection of at least 2 mediastinal stations,1 were associated with improved survival, respectively. To properly assess the value of MLND, the American College of Surgeons Oncology Group has completed a randomized trial (ACOSOG Z0300) of MLND or MLNS in patients with clinical N0 or N1 (less than hilar) lymph nodes. Results of this trial are currently pending.
There is also conflicting information about whether the extent of LA is associated with a survival benefit. Two retrospective studies revealed a survival benefit with more extensive LA,1, 5 but several studies demonstrated no difference in long-term survival rates, whether patients underwent MLNS or MLND.6–8 One trial that randomized 532 patients to either MLND or MLNS found that MLND improved survival,9 but 2 smaller trials (with 115 and 169 patients) did not.10, 11
Therefore, we decided to evaluate whether LA affects overall survival (OS) or disease-free survival (DFS) rates for patients with stage I NSCLC and how the number of lymph nodes recovered and their location (hilar vs mediastinal) are correlated with outcome.
MATERIALS AND METHODS
Data were obtained from the Surveillance, Epidemiology, and End Results (SEER) program of the United States National Cancer Institute (NCI), using the SEER-13 Registries 1992-2002 data set. The SEER-13 database is derived from a set of geographically defined, population-based cancer registries operated under direct contract with NCI by local nonprofit organizations in Connecticut, Iowa, Hawaii, New Mexico, Utah, Atlanta, Detroit, Los Angeles, San Francisco-Oakland, San Jose-Monterey, Seattle-Puget Sound, rural Georgia, and among Alaskan Native populations.12 The case ascertainment rate of the SEER registries has been reported to be 97.5% and is felt to accurately represent the entire American population.13 Data were accessed on May 3, 2006. Because we used existing data without individual subject identification, informed consent by the study participants was not necessary.
We included patients in this database diagnosed with American Joint Committee on Cancer stage I NSCLC,14 who underwent a definitive surgical procedure, including wedge or segmental resection (called “less than a lobectomy” for the remainder of this analysis), lobectomy or bilobectomy, and pneumonectomy or extended pneumonectomy. The following histologic types were defined as NSCLC: nonsmall cell carcinoma not otherwise specified (NOS) (which included undifferentiated carcinoma NOS and anaplastic carcinoma NOS); large cell carcinoma (large cell neuroendocrine carcinoma, large cell carcinoma with rhabdoid phenotype, lymphoepithelioma-like carcinoma, basaloid carcinoma, and clear cell carcinoma); adenocarcinoma (including acinar cell carcinoma, mixed subtype adenocarcinoma, alveolar adenocarcinoma, bronchioloalveolar carcinoma, nonmucinous bronchioloalveolar carcinoma, mucinous bronchioloalveolar carcinoma, mixed mucinous and nonmucinous bronchioloalveolar carcinoma, clear cell adenocarcinoma, mucinous cystadenocarcinoma, mucinous or colloid adenocarcinoma, mucin-producing adenocarcinoma, papillary adenocarcinoma, signet ring adenocarcinoma, solid adenocarcinoma, and well-differentiated fetal adenocarcinoma); adenosquamous carcinoma; and squamous cell carcinoma (basaloid squamous cell carcinoma, papillary squamous cell carcinoma, clear cell squamous cell carcinoma, keratinizing squamous cell carcinoma, large cell nonkeratinizing squamous cell carcinoma, and small cell nonkeratinizing squamous cell carcinoma).
The extent of LA and its effect on survival was analyzed for the entire group of stage I patients who underwent definitive surgery from 1992 to 2002. For patients diagnosed from 1998 to 2002, outcome was examined separately for patients having only N1 (2683 patients) or only N2 nodal dissections (1019 patients).
A total of 24,273 patients met the eligibility criteria. Their median follow-up time was 35 months. The median follow-up of patient subsets having only N1 or N2 dissections was found to be 23 months and 19 months, respectively.
Demographic, histopathologic, and treatment characteristics of the entire cohort, those having only N1 dissections after 1997, and those having only N2 dissections after 1997, are listed in Table 1. The total population, as well as the N1 and N2 dissection groups, were found to be similar. However, the patient group that had no lymph nodes removed was found to have a smaller median tumor size of 2 cm (compared with 2.5 cm in the other groups) and was more likely to have a resection less than a lobectomy (70% vs 6%-24%). Table 2 demonstrates the number of lymph nodes recovered for each group.
|Variable||All Patients (N=24,273)||Only N1 Dissection, 1998-2002 (n=2683)||Only N2 Dissection, 1998-2002 (n=1019)||No Nodal Dissection 1998-2002 (n=1414)|
|Age, mean, y (range)||67 (17-85)||67 (17-67)||72 (32-85)||72 (32-87)|
|Tumor size, median, mm (range)||25 (1-800)||25 (1-475)||25 (1-175)||20 (1-160)|
|Nodes examined, median (range)||5 (0-90)||4 (0-33)||5 (0-71)|
|Less than lobectomy||16%||6%||24%||70%|
|Prior to surgery||1%||1%||1%||0.5%|
|Left upper lobe/lingula||27%||26%||27%||26%|
|Left lower lobe||14%||16%||15%||16%|
|Left lung NOS||0%||0%||0%||0%|
|Left 2+ lobes||0%||0%||0%||0%|
|Left main stem bronchus||0%||0%||0%||0%|
|Right upper lobe||35%||35%||33%||34%|
|Right middle lobe||5%||6%||6%||6%|
|Right lower lobe||16%||16%||17%||17%|
|Right lung NOS||1%||0%||0%||0%|
|Right 2+ lobes||1%||1%||1%||1%|
|Right main stem bronchus||0%||0%||0%||0%|
|N level dissected|
|N1 and N2||36%||0%||0%||0%|
|N1, N2, and N3||2%||0%||0%||0%|
|Large cell cancer||6%||6%||7%||5%|
|Nonsmall cell cancer||4%||4%||5%||3%|
|Number of Nodes Examined||All (N=24,273)||Only N1 Dissection (n=2683)||Only N2 Dissection (n=1019)|
|0||3426 (17%)||6 (0.0%)||6 (1%)|
|1-3||4774 (24%)||877 (39%)|
|4-6||4540 (23%)||696 (31%)||181 (22%)|
|7-10||3483 (18%)||431 (19%)||168 (20%)|
Having an LA was associated with an increase in the 5-year OS rate from 41.6% to 58.4% for the entire study population (P<.0001) and in DFS from 58.04% to 73.09% (P<.0001), compared with not having an LA. Outcome improved with an increasing number of recovered lymph nodes, until a plateau was reached at 11 or more lymph nodes (Fig. 1).
For patients undergoing only N1 dissection, LA was also associated with statistically significant improvements in OS and DFS, compared with no LA (P = .056 and .043, respectively). The maximum difference between those with no LA and those with only N1 dissection was seen when 11 to 16 lymph nodes were removed (4-year OS, 40.0% vs 68.2%; DFS, 40.0% vs 78.6%) (Fig. 2). Of note, the 61 patients who had 17 or more lymph nodes removed had nonsignificant trends for worse DFS and OS (P = .096) than the 158 patients with 11 to 16 lymph nodes removed. The 90-day crude mortality rates in these 2 groups were 1.6% and 1.9%, respectively.
For patients undergoing only N2 dissection, LA was associated with a statistically significant improvement in OS (P = .045) and a trend for a DFS benefit (P = .086), with the maximum difference between those with no LA and those with 7 to 10 lymph nodes removed (4-year OS, 75.0% vs 77.7%; DFS, 75.0% vs 85.0%) (Fig. 3). Patients with 11 or more removed N2 lymph nodes had significantly worse OS than those with 7 to 10 lymph nodes removed (P = .023). The 90-day crude mortality rates in these 2 groups were 5.6% and 1.8%, respectively.
To the best of our knowledge, our study contains the largest number of patients of any examining the relationship between OS and DFS and whether LA is performed or not and, if performed, the number of removed lymph nodes for patients with stage I NSCLC. Both performing LA and having a greater number of recovered lymph nodes were associated with a statistically significant improvement in outcome. These data cannot, however, distinguish between 2 possible mechanisms that could cause such an effect. The first (and probably most important in our view) is that performing LA and recovering larger numbers of lymph nodes result in more accurate identification of patients with mediastinal or hilar nodal involvement, who are then no longer characterized as having stage I disease, thus improving the prognosis of the remaining patients called stage I. The second is the possibility that LA serves to remove unrecognized nodal metastatic disease, thus exerting a direct therapeutic effect.17
As shown in Table 1, the patient group having no lymphadenectomy differed greatly from the patient groups having N1 only and N2 only lymph node dissection in terms of the percentage of limited resections (70% vs 6%-24%) and tumor size (2.0 cm vs 2.5 cm in both groups). In our experience, there are generally 2 patient groups who receive limited lung resections. Those groups include patients with small primary tumors or patients with poor prognostic characteristics who probably could not tolerate a lobectomy or pneumonectomy. It must be remembered that our patient population includes some surgical patients who were treated before the Lung Cancer Study Group reported the local control benefits of lobectomy versus limited resection in 1995.18 Unfortunately, the SEER database does not include information regarding morbidity score, performance status, and/or preoperative forced expiratory volume in 1 second. We would assume this is because the no lymphadenectomy group contains both patients with a poor prognostic characteristics and those with smaller tumors, but we only have proof of the latter. Nevertheless, despite higher local recurrences with limited resection as compared with lobectomy, no significantly higher death rate or death from cancer rate was noted in the only prospective, randomized trial comparing these 2 surgical procedures.18
Interestingly, there appeared to be a survival detriment to having a very large number of recovered lymph nodes in those patients having only N1 or only N2 dissections. The small number of patients with only N1 dissections and removal of 17 or more lymph nodes (61 patients or 3% of the N1 population) must be taken into consideration. However, the subgroup who underwent N2 dissections with the removal of 11 or more lymph nodes constituted a substantial portion (180 patients, or 22%) of the patient group having only N2 dissections.
The reasons for these findings in these subgroups are not clear. This phenomenon was not seen for the entire study population, and hence they may be due to chance. However, comprehensive nodal dissection may contribute to serious complications, such as bronchopleural fistula due to interruption of the blood supply to the bronchial stump, phrenic and recurrent laryngeal nerve injury, thromboembolic events, and an increased risk of chylothorax and hemothorax.19, 20 In addition, MLND can be associated with an increased risk of impaired lymphatic drainage, resulting in pulmonary edema, acute respiratory distress syndrome, and pneumonia.1 Although only 1 randomized trial noted a higher incidence of morbidity in patients undergoing MLND compared with MLNS; there was no surgically related mortality in either group, and the grade of these complications were not mentioned.10 Our study suggested an increased risk of postoperative (90-day) mortality in patients undergoing aggressive N2 only mediastinal dissection, but not in patients with more extensive N1 only dissections.
Three prospective, randomized trials have been performed to assess the possible survival advantage of MLND over that of MLNS. The largest, from China, included 532 patients and found a survival benefit in MLND,9 but 1 containing 169 patients conducted in Munich11 and 1 with 115 patients performed at Yamaguchi University did not find a survival benefit in MLND.10 However, these trials differed from each other in terms of entry criteria, preoperative workup, details of surgery, when radiotherapy was given postoperatively, and how it was done. Recently Manser and colleagues used a fixed-effects model to assess the possible survival benefit of LA over the first 4 years after randomization for these 3 trials and found that MLND significantly reduced the risk of death (pooled hazard ratio, 0.78).21 The American College of Surgeons Oncology Group (ACOSOG) Z0300 trial, which randomized patients with clinical N0 or N1 (less than hilar) disease to undergo either MLND or MLNS, has completed accrual, and the results are currently pending. However, by using SEER-13, we looked at a database that provides a cross-section of American centers (approximately 14% of US cancer patients are in this database) including both academic and community hospitals that prevents 1 or 2 large institutions from dominating the results. Furthermore, our study reveals the importance of not only a thorough N2 dissection, but of a complete N1 dissection as well. We do not believe that the ACOSOG trial is designed to answer whether or not a thorough N1 dissection is necessary.
The SEER database has substantial limitations. It does not provide information on performance status, weight loss, smoking, lymphatic and/or vascular invasion, type of resection (R0, R1, R2), lymph node level dissected, type of preoperative workup, or recurrence pattern. Furthermore, the pattern of failure (local, regional, or distant, or combinations of these) is also not recorded. Moreover, the overwhelming majority of patients included in this study were treated and staged in an era when positron emission tomography (PET) scans were not routinely performed. However, about 10% to 20% of patients with a negative PET scan and mediastinoscopy will still have pathologic mediastinal lymph node involvement at the time MLND, even those with T1 tumors.22–26 Finally, the SEER database fields regarding number of lymph nodes dissected does not discriminate between fragments of the same lymph node versus discreetly separate lymph nodes and/or nodal stations. This ambiguity regarding the definition of a sampled lymph node may limit the inferences regarding a precise or optimal number of lymph nodes for an adequate LA. Hence, we included appropriate lymph node ranges in this study.
However, knowing the limitations of the SEER database, we decided to limit our analysis to stage I disease, which by definition does not invade the mediastinum, chest wall, and/or visceral pleural. Therefore, we felt that the benefits of lymphadenectomy would be more clearly appreciated because of the rarity of positive resection margins in this patient group. In addition, in our N2 dissection only group, we assume that there were a minimum number of N1 lymph nodes that were removed in the definitive surgical treatment of primary lung cancer and missed, not examined, or not reported by the pathologist. Nevertheless, this patient group with “only N2 dissection” allowed us to see the beneficial and detrimental effects of mediastinal lymphadenectomy quite clearly. We believe that this retrospective study adds to the growing literature suggesting the beneficial effects of lymphadenectomy in stage I NSCLC.27
Lymphadenectomy was associated with increased rates of OS and DFS, compared with no lymphadenectomy in our study. Our results also suggest the optimal number of recovered lymph nodes needed to see the maximum staging accuracy of LA is 11 to 16 lymph nodes when dissection included only N1 lymph nodes and 7 to 10 lymph nodes when only N2 lymph nodes are removed. Common sense would dictate that a thorough and systematic evaluation of both N1 and N2 nodal stations for each clinical stage I NSCLC patient would provide the best outcome.
Conflict of Interest Disclosures
The authors made no disclosures.
- 12Surveillance, Epidemiology, and End Results (SEER) Program Public-Use Data (1973-2003), National Cancer Institute, DCCPS, Surveillance Research Statistics Branch. Available at:http://seer.cancer.gov/publicdata/ Accessed May 3, 2006.
- 13Surveillance, Epidemiology and End Results (SEER) Program. Data quality. Available at: http://seer.cancer.gov/about/quality.html Accessed May 3, 2006.
- 21Surgery for early stage non-small cell lung cancer. Cochrane Libr. 2006; 2: 1–39., , , et al.