Impact of Different Types of Lymphadenectomy Combined With Different Extents of Tumor Resection on Survival Outcomes of Stage I Non-small-cell Lung Cancer: A Large-Cohort Real-World Study

Background: To investigate the prognostic impact of different types of lymphadenectomy with different extents of tumor resection on the outcomes of stage I non-small-cell lung cancer (NSCLC). Methods: Patients were classified into lobectomy and sublobectomy groups, and then each group was subdivided according to the types of lymphadenectomy. The end points of the study were overall survival (OS) and disease-free survival (DFS). Propensity score matched (PSM) comparative analysis and univariate and multivariate Cox regression analyses were performed. Result: A total of 1,336 patients were included in the current study. Lobectomy was associated with better OS and DFS. In the lobectomy group, lobectomy with bilateral mediastinal lymphadenectomy (BML) was associated with better OS than lobectomy with systematic nodal dissection (SND) or lobe-specific systematic node dissection (L-SND). Lobectomy with SND or L-SND was associated with better OS than lobectomy with systematic nodal sampling (SNS) or selected lymph node biopsy (SLNB). Additionally, lobectomy with BML or SND was associated with better DFS than lobectomy with L-SND or SNS or SLNB. After PSM, compared with lobectomy with SNS or SLNB, lobectomy with SND resulted in more favorable OS and DFS. There was no survival difference between different types of lymphadenectomy for patients who underwent sublobectomy. A multivariable analysis revealed independent associations of lobectomy with BML or SND with better OS and DFS compared with those of lobectomy with SNS or SLNB. Conclusion: This study reveals an association of lobectomy with more systematic and complete lymph node dissection, such as BML or SND, with better prognosis in stage I NSCLC patients.

Background: To investigate the prognostic impact of different types of lymphadenectomy with different extents of tumor resection on the outcomes of stage I non-small-cell lung cancer (NSCLC).
Methods: Patients were classified into lobectomy and sublobectomy groups, and then each group was subdivided according to the types of lymphadenectomy. The end points of the study were overall survival (OS) and disease-free survival (DFS). Propensity score matched (PSM) comparative analysis and univariate and multivariate Cox regression analyses were performed.
Result: A total of 1,336 patients were included in the current study. Lobectomy was associated with better OS and DFS. In the lobectomy group, lobectomy with bilateral mediastinal lymphadenectomy (BML) was associated with better OS than lobectomy with systematic nodal dissection (SND) or lobe-specific systematic node dissection (L-SND). Lobectomy with SND or L-SND was associated with better OS than lobectomy with systematic nodal sampling (SNS) or selected lymph node biopsy (SLNB). Additionally, lobectomy with BML or SND was associated with better DFS than lobectomy with L-SND or SNS or SLNB. After PSM, compared with lobectomy with SNS or SLNB, lobectomy with SND resulted in more favorable OS and DFS. There was no survival difference between different types of lymphadenectomy for patients who underwent sublobectomy. A multivariable analysis revealed independent associations of lobectomy with BML or SND with better OS and DFS compared with those of lobectomy with SNS or SLNB.

INTRODUCTION
Non-small-cell lung cancer (NSCLC) is the malignancy with the highest morbidity and mortality rates worldwide (1). However, improved radiologic imaging and widespread low-dose computed tomography screening have led to increased detection of early-stage (stage I) NSCLC (2). For surgically resectable lung cancer, surgery is the best therapeutic option. Additionally, lymph node (LN) dissection during surgery is essential. For these reasons, determining the appropriate model of resection of original tumors and LN dissection has attracted more attention.
Sublobar resection has been indicated to result in a similar survival rate for early-stage NSCLC patients compared with that of lobectomy, and it can retain more pulmonary function (5). However, some researchers hold the opposite view (6), and they believe that sublobectomy results in an inferior survival rate. Several previous studies have indicated that among patients who underwent sublobectomy, whether LN dissection was performed did not influence the survival outcomes (7). However, there are also some studies reporting that LN dissection can result in a better prognosis even in patients who have undergone sublobectomy (8). The need for SND has also been questioned. Whether SND compared with SNS can provide more accurate staging and survival benefit or lead to more complications in stage I NSCLC patients has not yet been confirmed (9,10). Some studies have also shown that L-SND results in the same survival rate as that of SND, although it decreases the duration of the surgery and incidence of postoperative complications. However, some studies have found that L-SND may be ineffective in some Abbreviations: ACCP, American College of Chest Physicians; AJCC, American Joint Committee on Cancer; ALK, anaplastic lymphoma kinase; BML, bilateral mediastinal lymphadenectomy; CIs, confidence intervals; DFS, disease-free survival; EGFR, epidermal growth factor receptor; ELND, extended lymph node dissection; ESTS, The European Society of Thoracic Surgeons; HRs, hazard ratios; KM curves, Kaplan-Meier curves; LN, lymph node; L-SND, lobe-specific systematic node dissection; NCCN, National Comprehensive Cancer Network; NSCLC, non-small-cell lung cancer; OS, overall survival; PSM, propensity score matched; SLNB, selected lymph node biopsy; SND, systematic nodal dissection; SNS, systematic nodal sampling. N2-positive patients, influencing the pathological grading and adjunctive therapy (11,12).
Therefore, in our study, we used a large cohort of patients to compare the outcomes of different LN dissection models combined with different extents of resection in stage I NSCLC patients to formulate guidelines regarding the extent of tumor resection and lymphadenectomy.

Patients
This study initially included 2,102 consecutive cases consisting of clinical and pathological stage I NSCLC patients who underwent surgical treatment between 1999 and 2014 in the Department of Thoracic Surgery in Sun Yat-sen University Cancer Center. This study was approved by the Institutional Review Board of Sun Yat-sen University Cancer Center.
All patients were confirmed to be pathological stage I according to the 8th edition of the American Joint Committee on Cancer (AJCC) lung cancer staging classification (13) and met the following criteria: (1) primary NSCLC; (2) preoperatively considered node negative; and (3) pathologic stage was T1a-2aN0M0. The exclusion criteria were as follows: (1) a history of other primary cancer; (2) double primary lung cancer; (3) the patient received neoadjuvant therapy; (4) positive surgical margins; (5) the clinicopathologic and follow-up data were not complete; and (6) patients chose sublobectomy because they could not tolerate lobectomy (Figure 1). Finally, 1,336 patients were included in further analysis.

Study End Points
The outcomes of this study included overall survival (OS) and disease-free survival (DFS). The latest follow-up of the current study was performed on October 15, 2018.

Patients Grouping
The location and station of mediastinal LNs were based on the International Association for the Study of Lung Cancer LN map (14). For analytical purposes, we divided the patients into two major groups (Group Lobe: for those patients who underwent lobectomy, biolobectomy, or pneumonectomy; and Group Sublobe: for those patients who underwent wedge resection or segmentectomy) and then into five or three smaller groups, respectively, according to the LN dissection as follows: Group Lobe: A, patients with SLNB or SNS; B, patients with SND; C, patients with L-SND; D, patients without any LN dissection; and E, patients with bilateral mediastinal lymphadenectomy (BML); and Group Sublobe: F, patients with SND; G, patients with SLNB or SNS; and H, patients without any LN dissection ( Table 1). The definitions of all options in the intraoperative A propensity score matched (PSM) comparative analysis was performed to control the non-random variables among groups. We adjusted for potential differences between Group A (patients with SLNB or SNS) and Group B (patients with SND) (1:1 match), and finally 468 patients (234 in each group) were included in the PSM analysis. We generated a propensity score for the matched groups using logistic regression based on the patients' potential confounding baseline characteristics, including age, tumor size, tumor location, and surgical approach. We next created a balanced cohort using an optimized performance-matching algorithm with a caliper setting of 0.02. The procedure was conducted using SPSS 21.0.

Statistical Analysis
Categorical variables were calculated using the χ 2 test, and continuous variables were analyzed using the t-test. All end points were estimated using the Kaplan-Meier method and compared using the log-rank test. Multivariable survival analyses were performed using the Cox proportional hazards model to identify prognostic factors for OS and DFS. For all analyses, a two-sided P < 0.05 was considered statistically significant. Hazard ratios (HRs), 95% confidence intervals (CIs), and Pvalues for each variable were determined using SPSS 21.0 software (IBM, Armonk, NY), and survival curves were drawn using Prism 7.0 (GraphPad software, La Jolla, CA).

Group Subgroup Description
Group Lobe A (n = 346) Lobectomy with selected lymph node biopsy or sampling B (n = 328) Lobectomy with systematic lymph node dissection C (n = 577) Lobectomy with lobar-selective lymph node dissection D (n = 13) Lobectomy without any lymph node dissection E (n = 28) Lobectomy with bilateral mediastinal lymphadenectomy Group Sublobe F (n = 3) Sublobectomy with systematic lymph node dissection G (n = 13) Sublobectomy with randomly lymph node dissection or sampling H (n = 28) Sublobectomy without any lymph node dissection

Patient Characteristics
In total, 44 patients underwent sublobectomy, and 1,292 patients underwent lobectomy, including 346 (25.9%) patients in Group A, 328 (24.6%) in Group B, 577 (43.2%) in Group C, 13 (1.0%) in Group D, 28 (2.1%) in Group E, 3 (0.2%) in Group F, 13 (1.0%) in Group G, and 28 (2.1%) in Group H. There were 822 males and 470 females in Group Lobe and 33 males and 11 females in Group Sublobe. The mean age of Group Lobe was 59.35 ± 10.1 years, and the median age was 60.0 years. In addition, in Group Sublobe, the mean age was 66.73 ± 11.9 years. Non-squamous cell carcinoma was the most common (1,002 and 40 patients, 77.6 and 90.9%, respectively) pathologic type in both groups. When the clinicopathologic characteristics were compared among groups, it was interesting to note that sex, histology, cell differentiation, smoking history, adjuvant therapy, and treatment after disease progression were well-balanced between the subgroups in both Group Lobe and Group Sublobe. Patients in Group Lobe were younger, had larger tumors, and had more LNs resected than those in Group Sublobe. Among the patients in Group Lobe, lobectomy was much more common than bilobectomy or pneumonectomy (1,234 vs. 33 and 25, respectively). Interestingly, in later procedures, there were more patients who underwent LN dissection rather than LN sampling in Group Lobe. Among Group Sublobe, most patients underwent wedge resection without any LN dissection. Notably, those who received SND in Group Sublobe underwent segmentectomy. The baseline characteristics of the patients in Group Lobe are summarized in Table 2, and those in Group Sublobe are summarized in Supplemental Table 1.

Comparison of survival between the lobectomy group and sublobectomy group
A Kaplan-Meier survival analysis and log-rank comparison revealed that compared with the Group Sublobe, the lobectomy group (Group Lobe) was significantly associated with better OS (log rank = 9.45, P = 0.002) and DFS (log rank = 3.97; P = 0.045) in patients with stage I NSCLC (Figures 2A,B).

Comparison of survival between different types of lymphadenectomy in Group Lobe
When the patients who underwent lobectomy were divided into Groups A to E, it was clear that the patients who underwent lobectomy without LN dissection (Group D) had the worst OS. Among those who underwent LN dissection, patients who underwent systematic LN dissection (Group B) or L-SND (Group C) had better outcomes than patients who underwent selective LN biopsy or sampling (Group A), although the difference between Group B and Group C was not significant. Patients who underwent BML (E) had the best postoperative survival among the subgroups in Group Lobe.  (Figures 3A,B).

Comparison of survival between different types of lymphadenectomy in Group Sublobe
A similar subgroup analysis was performed in Group Sublobe. However, there were no significant differences in either OS or DFS among Group F, Group G, and Group H (P = 0.621 and P = 0.954, respectively) (Figures 4A,B).

Comparison of survival via Cox regression
In  Table 3). A similar Cox regression analysis was also performed in the Group Sublobe, and the result is shown in Supplemental Table 2.

Comparison of survival among specific groups after PSM
Frontiers in Oncology | www.frontiersin.org   After PSM, pairs were formed between Group A and Group B. A total of 468 patients (234 pairs) were included in the PSM analysis, and the baseline characteristics of these patients are summarized in Supplemental Table 3. A Kaplan-Meier survival analysis and log-rank comparison revealed that Group B had better OS (log rank = 3.97, P = 0.046) and DFS (log rank = 7.00, P = 0.008) than Group A (Figures 5A,B).
Multivariable Cox regression analysis also revealed that lobectomy plus selective LN biopsy or sampling was independently associated with worse OS (HR, 1.305; 95% CI, 1.008-1.877; P = 0.050) and DFS (HR, 1.417; 95% CI, 1.020-1.968; P = 0.038) than other combinations. Additionally, advanced age and poor differentiation without adjuvant therapy were identified as being negatively correlated with OS.
Only lobectomy plus systematic LN dissection was positively correlated with DFS ( Table 4).

DISCUSSION
The extent of tumor resection and LN dissection are important elements in the treatment of resectable early-stage NSCLC (15,16). Some previous studies demonstrated that sublobectomy results in similar survival outcomes to lobectomy in earlystage NSCLC patients (5,17). However, other studies have reached different conclusions. In one study, despite adjustment for patient-and tumor-related characteristics, sublobectomy was associated with worse survival, even for stage IA patients (6). Furthermore, in another study, patients  undergoing sublobectomy were more likely to have inadequate lymphadenectomies and positive margins than those undergoing lobectomies (18). The results of our study suggest that for surgically resectable stage I NSCLC, most patients undergo lobectomy and that patients who undergo sublobectomy are older and have more comorbidities, smaller tumors, and worse survival.
For LN dissection, the standard model for early-stage NSCLC remains controversial even among different guidelines. The National Comprehensive Cancer Network (NCCN) guidelines recommend that for clinical node-negative stage I NSCLC and patients who underwent sublobectomies, SNS or selected LN biopsy is also acceptable (4). The American College of Chest Physicians (ACCP) guidelines note that for preoperative stage I patients, if they are intraoperatively determined to be node negative, selective LN sampling and dissection are both recommended (19). The European Society of Thoracic Surgeons (ESTS) guidelines recommend SND or sampling for all lung cancer patients. For peripheral T1 SCC, L-SND is also acceptable (3).
Some studies suggested that SND was associated with more accurate staging and better survival, although it results in longer operative times, greater intraoperative blood loss, and recurrent nerve injury; however, the results are inconsistent among studies (20). The results of the American College of Surgery Oncology Group Z0030 trial demonstrated that if systematic sampling of the mediastinal and hilar LNs returns a negative result, mediastinal LN dissection does not improve survival, although it can provide a more accurate staging (9,10). However, in contrast to the Z0030 trial, which included some pN1 or pN2 and other pIIIA or pIIIB patients, we focused only on pN0 and stage I patients. Moreover, we not only compared SND with SNS by PSM but also more accurately included L-SND and BML in our analysis. All of these may cause a different result between our result and the Z0030 trial result. We found that SND and BML were associated with better OS and DFS than other modalities. Regarding the controversy of LN dissection in stage I patients, the guidelines concentrate mainly on the survival outcome difference between SND and SNS. In our study, a PSM comparison was performed between these two groups, which had not been performed in the Z0030 trial. The results revealed that even after

PSM, those who underwent SND had significantly better OS and DFS.
Previous studies also demonstrated that the lymphatic drainage of lung cancer might be lobe specific (21), which provides the basis for L-SND. Some studies suggested that L-SND proved to be as effective as SND (22), but the rate of detection of pN2 was significantly higher in the SND group (12). Our study showed that a significant difference was not found in OS between L-SND and SND. However, compared with SND, L-SND resulted in a worse DFS.
Our study also investigated the relationships between different types of lymphadenectomy in sublobectomy and survival outcomes. However, there were no significant differences between subgroups; this result was attributed to the small sample size. However, other studies found that LN removal appears to decrease locoregional recurrence and may be associated with a survival benefit, while it does not increase morbidity or the length of the hospital stay (23).
Several studies have also demonstrated an association between an increased number of dissected LNs following resection for NSCLC and better long-term survival rates. They indicated that examining a greater number of LNs in patients with stage I NSCLC treated with resection increases the likelihood of proper staging and affects patient outcomes (15). A more recent analysis identified a significant reduction in the mortality rate among patients with at least 16 dissected LNs and clinically certified node-negative NSCLC (24). In the current study, the numbers of resected LNs varied among groups. A greater number of resected LNs was associated with a better survival outcome. Patients who underwent lobectomy with BML may be associated with the greatest number of LNs resected in this subgroup (42.71 ± 21.3).
In our study, we specifically targeted stage I lung cancer and explored the associations between long-term survival and different extents of tumor resection and lymphadenectomy models. Meanwhile, we aimed to clarify which type of lymphadenectomy should be considered a recommended procedure for the surgical treatment of stage I NSCLC. We found that for patients who underwent lobectomy, long-term survival was associated with the type of lymphadenectomy. It is clear that patients who underwent only lobectomy had the worst OS and DFS, and BML resulted in the best OS. For DFS, BML, or SND was associated with better outcomes than that of other subgroups.
These results may be attributed to the following reasons; First, LN involvement can occur in patients with any tumor size (25), and metastasis and micrometastasis can be missed in some cases (26). Therefore, potential metastases may remain if the patient undergoes a sublobectomy or inadequate lymphadenectomy. Second, SND can resect all the possible metastatic tissue, and it can harvest more LNs, which could enhance staging accuracy (24). BML performed better in this respect; thus, it resulted in the best rates of OS and DFS. This may indicate that extended LN dissection may have survival benefits in not only advanced stage NSCLC but also stage I NSCLC. Third, it has been reported that more stringently defined mediastinal LNs are associated with better separation in the prediction of survival. BML can provide the most thorough nodal examination among the different types of lymphadenectomy, which may explain the better outcomes observed among patients who underwent BML and were declared node negative (16).
This study has several limitations of note. Given the retrospective nature of this single-institution study, selection bias is inherent in our study population. In addition, postoperative complications were not considered in our series.

CONCLUSION
In conclusion, our study shows that lobectomy with more systematic and complete LN dissection, such as BML or SND, is associated with better survival outcomes in stage I NSCLC patients. However, due to the small sample size, the efficacy of BML still needs to be further verified by a large-scale randomized clinical trial. We recommend lobectomy with SND as the recommended procedure even for stage I NSCLC patients to improve their prognosis. BML is also recommended if it is technically feasible. Despite the use of appropriate statistical methods, our study has some inevitable limitations; more evidence from largesample-size, multicenter prospective or real-world studies is warranted.

DATA AVAILABILITY
All datasets generated for this study are included in the manuscript and/or the Supplementary Files.

ETHICS STATEMENT
This retrospective study was approved by the Institute Research Medical Ethics Committee of Sun Yat-sen University Cancer Center. The reference number is B2018-011. All patients included in this research gave written informed consent according to the Research Ethics Committee of Sun Yat-Sen University Cancer Center.

AUTHOR CONTRIBUTIONS
LZ was responsible for conception and design of the study. WW, DC, and KX analyzed and interpreted the patient data and were major contributors to the writing of the manuscript. YC, XZ, and YW collected and assembled the data. ZH, XY, GW, and RZ performed the follow-up survey. All authors read and approved the final manuscript.

FUNDING
This study was supported by the grant 2016YFC0905400 from the National Key R&D Program of China. Conflict of Interest Statement: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Copyright © 2019 Wang, Chen, Xi, Chen, Zhang, Wen, Huang, Yu, Wang, Zhang and Zhang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.