Survival After Lobectomy Versus Sublobar Resection for Stage IA Large-Cell Neuroendocrine Carcinoma of the Lung: A Population-Based Study

Background: Due to the low incidence of pulmonary large cell neuroendocrine carcinoma (LCNEC), the survival analysis for comparing lobectomy and sublobar resection (SLR) for stage IA LCNEC remains scarce. Methods: Patients diagnosed with pathological stage IA LCNEC between 1998 and 2016 were extracted from the Survenillance, Epidemiology, and End Results (SEER) database. The oncological outcomes were cancer-specic survival (CSS) and overall survival (OS). Kaplan-Meier analysis and Cox multivariate analysis were used to identify the independent prognostic factors for OS and CSS. Furthermore, propensity score matching (PSM) was performed between SLR and lobectomy to adjust the confounding factors. Results: A total of 308 patients with stage IA LCNEC met the inclusion criteria: 229 patients (74.4%) received lobectomy and 79 patients (25.6%) received SLR. Patients who underwent SLR were older (P<0.001), had smaller tumor size (P=0.010) and fewer lymph nodes dissection (P<0.001). The 5-year CSS and OS rates were 56.5% and 42.9% for SLR, and 67.8% and 55.7% for lobectomy, respectively (P=0.037 and 0.019, respectively). However, multivariate analysis did not identify any differences between SLR group and lobectomy group in CSS (P=0.135) and OS (P=0.285); and the PSM also supported these results. In addition, age at diagnosis and laterality of tumor were identied as signicant predictors for CSS and OS, whereas the number of lymph nodes dissection was a signicant predictor for CSS. Conclusions: Although SLR not inferior to lobectomy in terms of oncological outcomes for patients with stage IA LCNEC, more lymph sampled during be considered early-stage NSCLC: non-small cell lung cancer; DNA: deoxyribonucleic acid; EGFR: epidermal growth factor receptor; ALK: anaplastic lymphoma kinase; SEER: the Survenillance, Epidemiology, and End Results; ICD-O-3: International Classication of Disease for Oncology, 3rd edition; PSM: propensity score matching.

However, multivariate analysis did not identify any differences between SLR group and lobectomy group in CSS (P=0.135) and OS (P=0.285); and the PSM also supported these results. In addition, age at diagnosis and laterality of tumor were identi ed as signi cant predictors for CSS and OS, whereas the number of lymph nodes dissection was a signi cant predictor for CSS.
Conclusions: Although SLR is not inferior to lobectomy in terms of oncological outcomes for patients with stage IA LCNEC, more lymph nodes can be dissected or sampled during lobectomy. Lobectomy should be considered as a standard procedure for patients with early-stage LCNEC who are able to withstand lobectomy.

Background
Pulmonary large cell neuroendocrine carcinoma (LCNEC), rst reported in 1991, is a rare (around 3% of all lung cancer) but aggressive subtype of lung tumors (5-year overall survival [OS] rate is only 16.2%) [1][2][3]. On basis of the latest World Health Organization (WHO) classi cation of lung tumors, LCNEC is eliminated from the cluster of large cell carcinoma, and is regrouped into the high-grade neuroendocrine tumor (NET) along with small cell lung cancer (SCLC) [3]. LCNEC and SCLC share some histological features (including rosette, trabeculae, molding of nuclei and palisading, etc.), same immunohistochemical neuroendocrine markers (including cluster of differentiation 56, neural cell adhesion molecules, etc.), and similar oncological outcomes (including metastatic behavior and poor prognosis) [2][3][4][5][6]. However, of note, several recent studies with aid of the next-generation sequencing and cell-free DNA detection found that LCNEC also comprised non-SCLC (NSCLC)-like subset, which may cause the individual response to different chemotherapy regimen; and the common driver mutations, including epidermal growth factor receptor (EGFR) mutation, anaplastic lymphoma kinase (ALK) rearrangements etc., were rarely detected [4,7]. Therefore, systemic treatment for pulmonary LCNEC remains under debate in clinical practice [4,5,7]. Based on this fact, surgical resection is still recognized as the standard scheme for early-staged pulmonary LCNEC patients to obtain long-term survival [8,9].
In 1995, a pivotal randomized trial administered by the Lung Cancer Study Group (LCSG) revealed higher localrecurrence rate and inferior OS rate in patients who received sublobar resection (SLR, including non-anatomical wedge resection [WR] and anatomical segmentectomy) rather than lobectomy for stage I NSCLC [10]. Since then, lobectomy with mediastinal lymph nodes dissection as the criterial surgical procedure even in stage IA NSCLC has been upheld in the guidelines until recently [8,9]. However, several retrospective studies have disputed this dogma by demonstrating non-inferior oncological outcomes of SLR and lobectomy for select patientswith stage IA (with or without less than 2 cm) NSCLC [11][12][13][14]. Also, for stage IA SCLC patients in the Survenillance, Epidemiology, and End Results (SEER) database, there is no statistically difference of cancerspeci c survival (CSS) or OS between SLR group and lobectomy group [15]. Potential bene ts of SLR include preserving cardiopulmonary function, reducing perioperative morbidity and mortality, and providing chance for repetitive resections [13,16,17]. On the other hand, a debate remains about the lower rate of local recurrence after lobectomy that might bring survival bene t, particularly in patients with a good physical status. Speci cally, due to the low incidence rate of pulmonary LCNEC and the majority of rst diagnosed patients in the advanced stages, the investigations on the oncological clearance of SLR and the equivalency between SLR and lobectomy among pulmonary LCNEC patients remain scarce [2,18]. Therefore, in this present study, the data retrieved from the SEER registry was used to compare the oncological outcomes following SLR and lobectomy in stage IA pulmonary LCNEC.

Study cohort
The public-access SEER database with additional treatment elds (1975-2016 varying) used in this study was released on April 2019, which covered approximately 34.6% of the U.S. population [19].

Variable De nitions
The continuous variable of age at diagnosis was grouped into three groups according to the WHO recommendations on physical activity for health: less than 60 years old, 61-74 years old, and 75 years old and above. Year of diagnosis was divided into 4-year intervals as follows: 2001-2004, 2005-2008, 2009-2012, and 2012-2016. The insurance status of insured, any medicaid, and insured but no speci cs were all de ned as insured; similarly, the marital status of single included never married, widowed, divorced, and single (never married). In addition, all patients were restaged to IA1, IA2, and IA3 on basis of the American Joint Committee on Cancer 8th Tumor-Nodes-Metastasis (TNM) staging manual for lung cancer.

Outcomes
Survival time, measured in months, was calculated from the day of surgical resection to the day of death or the day of last follow-up if patient was documented as alive. Only patients occurred LCNEC-related death were recorded as uncensored in CSS, but OS included any cause of death. The operative mortality was de ned as any death within 30 days of surgical resection.

Statistical analysis
The best cutoff values for tumor size (≤ 20 mm/>20, ≤ 30 mm) and the number of lymph nodes dissection (≤ 5/≥6) were determined by the X-tile software (version 3.6.1, copyright Yale University 2003). The continuous variables were presented as mean ± standard deviation (SD) and were compared by the Student's t test between the SLR set and lobectomy set. For categorical variables, the chi-squared test or Fisher's exact test was applied to compare the proportions between the two sets. Survival distributions between variables were analyzed by the Kaplan-Meier method. Thereafter, potential predictors (P < 0.05, by the log-rank test) identi ed by univariate analysis for CSS and OS were enrolled into to the Cox proportional hazards regression models to calculate the hazard ratios (HRs).
All above statistical analyses were calculated by using SPSS software (version 24.0, IBM SPSS, Armonk, NY, USA), and the threshold value for signi cance was less than 0.05 at two sides. With aid of the R software (version 3.6.1), propensity score matching (PSM) was implemented by the "matchit" package to reduce potential bias between the SLR group and the lobectomy group, and the 1-to-1 nearest neighbor matching (ratio = 1) was set. In addition, the survival curves before and after PSM were plotted by the "survival" and "survminer" packages.

Patient characteristics
A total of 308 patients diagnosed with stage IA LCNEC of the lung met inclusion criteria were nally retained in this study (Fig. 1). In demographic characteristics (Tables 1 and 2), the average and median age at diagnosis of whole cohort were all 67 years old (range, 39-89 years old), and most of the patients were white race (88.6%), female (55.8%), married (54.2%), and insured (68.8%). In addition, the pulmonary LCNEC was mostly located in the right lower lobe (37.0%), followed by right upper lobe (30.8%), left lower lobe (15.6%), right middle lobe (10.7%) and left upper lobe (5.8%).   Fig. 2b). Unexpectedly, after Cox multivariate regression analysis, there were no signi cant difference between the SLR group and the lobectomy group in CSS or OS (all P > 0.05, Table 2).

Page 11/19
The imbalanced variables, including age at diagnosis, tumor size and lymph nodes dissection (yes/no), were included into the 1-to-1 PSM. After that, a total of 158 patients were selected, with 79 patients in both the SLR group and lobectomy group. No signi cant difference was observed between the two groups in 13 of the 15 variables after PSM, except for whether to dissect lymph nodes and the number of lymph nodes dissection (Table S1). On Kaplan-Meier survival analysis, there was no signi cant difference of CSS or OS between the two matched groups as well (all P > 0.05, Fig. 2c and 2d).

Prognostic Factors For Css And Os
The median follow-up time of the 308 patients was 33.5 months (range from 1 month to 191 months). The 3- year and 5-year CSS and OS rates were 73.5% and 63.6%, 64.9% and 52.4%, respectively. Table 2 showed the univariate and multivariate analyses. Age at diagnosis (Fig. 3a and 3b) and laterality of tumor ( Fig. 3c and 3d) were identi ed as independent prognostic factors for CSS and OS. In addition, the number of lymph nodes dissection less than 5 was a risk factor for CSS (Fig. 3e), but not for OS.

Discussion
The present study was conducted to investigate the CSS and OS associated with SLR and lobectomy in 308 patients with stage IA LCNEC retrieved from the National Cancer Institute's SEER database, by using Cox proportional hazards regression model and PSM to adjust the potential confounding factors. We found that SLR was not inferior to lobectomy for CSS and OS. However, multivariate analyses demonstrated that advancing age and right-sided lung resection were signi cant risk factors for CSS and OS; in addition, patients receiving lymph nodes dissection less than ve was associated with decreased CSS.
In line with the published literatures on patients with stage IA LECNEC of the lung who underwent lobectomy (27.0-67.0%), the 5-year OS rate was 67.8% in this population-based study, which was worse than other NSCLC [18,[20][21][22][23]. However, owing to the low incidence and a lack of evidence-based surgical treatment for LCNEC, there was limited study to evaluate the survival after SLR [6,18,22,24]. Reported by the NETs Working Group, 24 patients collected from eight high-volume European Thoracic Surgery Institutions received SLR, including 13 WR and 11 lobectomy; furthermore, compared to 81 patients received lobectomy, these patients received SLR showed a equivalent 5-year OS rate [6]. Similar survival outcome was verti ed by another single-institution study, which was the largest study cohort between 1991 and 2006 [22]. Conversely, in 2019, a prognostic analysis comparing SLR (133 cases) versus lobectomy or bilobectomy (425 cases) utilizing the SEER database found the ve-year OS rate was 22.5% and 42.5%, respectively (P < 0.001) [24]. However, of note, patients diagnosed with stage I through IV were all enrolled in above studies and most of the statistical analyses were performed on small size samples. fully consistent with our ndings, patients receiving SLR were more likely to have increased age, smaller size, and less likely to have lymph nodes dissection than those receiving lobectomy. However, SLR showed no inferior to lobectomy for CSS and OS of stage IA LCNEC by multivariate analysis and propensity score analysis in present study. We speculate that only PSM performed in the study of Waseem et al. may strengthen the power to identify statistical difference in survival. As known, if the ratio of the patients in control group (lobectomy) to the patients in study group (SLR) is less than 10-to-1, plenty of patients in study group could not match the nearest control cases during the process of PSM, which may increase the man-made selection bias [25]. Therefore, it is necessary to conduct the multivariate analysis to adjust potential confounding factors before PSM and/or after PSM [2,13,23,25]. Additionally, the reasons for SLR include intention-to-treat and compromise-to-treat in clinical practice, and the intentional SLR is appropriate for patients with poor pulmonary reserve or other major comorbidity; moreover, the su cient parenchymal resection margin is vital when surgeons perform the SLR [26]. Regrettably, these variables were not available within the SEER database. On the whole, the above weaknesses in the published and our studies may be preliminarily explain why the results were different between the two large size samples, and the debate on the oncological outcomes following SLR compared to lobectomy for patients with stage IA LCNEC remains [6,18,22,24].
We observed notable difference in the number of lymph nodes dissected by lobectomy and SLR, and the proportion of the patients without lymph nodes dissection in SLR group was signi cantly higher than that in lobectomy group, which were fully consistent with other studies for comparing lobectomy and SLR for stage IA NSCLC based on the NCDB [14,27]. It meant that patients receiving SLR for stage IA lung cancer in the NCDB and SEER database may not receive adequate lymph nodes dissection. Previous studies have found that adequate lymph nodes examination for lung cancer, especially in SLR, was associated with more accurate pathological staging and better survival [14,27,28]. Similarly, our result showed that the number of lymph nodes dissection greater than six yielded improved long-term bene t in CSS for stage IA LCNEC patients, and more lymph nodes dissection did not increase the operative mortality. In addition, line with other studies on early-stage LCNEC, Cox regression analysis revealed advancing age to be associated with worse survival [23].
Right-side radiation or pneumonectomy for lung cancer was regarded as independent risk factor for long-term survival [29,30]. The study on the right ventricular (RV) response to lung resection by using cardiovascular magnetic resonance imaging, reported by Philip et al., found that RV dysfunction occurred immediately following lung resection, especially right-sided resection, and persisted two months or more, which may associated with the dyspnea and reduced functional capacity [29]. In addition, Carolyn et al. reported that RV end-diastolic volume and center venous pressure would signi cantly increased after right-sided lung resection as well [30]. Therefore, the right-sided lung resection for early-stage LCNEC may also have negative impact on the cardiopulmonary function, postoperative complications, and long-term survival, which was similar to a prognostic analysis in patients with stage IA SCLC [15].
Undeniably, this population-based study had several limitations. First, the prospective study was di cult to carry out due to the rarity of early-stage LCNEC, therefore, this study was conducted from a retrospective viewpoint. However, multivariate analysis and PSM were performed to reduce the confounding factors and selection bias. Second, several important variables associated with the oncological outcomes after limited resection, such as reason for SLR, preoperative cardiopulmonary function, status of resection margin, local recurrence etc, were not documented in the SEER database. Third, limited by the number of patients who underwent SLR for stage IA LCNEC, we could not further compare the outcomes of segmentectomy and WR.

Conclusion
Although SLR is not inferior to lobectomy in terms of survival for patients with stage IA LCNEC of the lung, more lymph nodes can be dissected during lobectomy. Lobectomy should be condidered as a standard procedure for patients with early-stage LCNEC of the lung who are able to withstand lobectomy.  The ow diagram of the selection process for this population-based study cohort.

Figure 2
Kaplan-Meier survival curves of cancer-speci c survival (a, c) and overall survival (b, d) between patients with wedge resection, segmentectomy and lobectomy before and after propensity score matching.