Operative versus Nonoperative Treatment in Patients with Advanced Non-Small-Cell Lung Cancer: Recommended for Surgery

Background There is currently limited evidence for a correlation between the recommended operation and overall survival (OS) in patients with advanced non-small-cell lung cancer (NSCLC). Methods NSCLC patients with stages III and IV, recommended for operation, were identified in the US National Cancer Institute Surveillance, Epidemiology, and End Results database (SEER).We used propensity score matching (PSM) and multivariable Cox proportional hazards regression to ensure the robustness of our findings. The cumulative rates of death were compared between patients with and without recommended operations using the Kaplan−Meier curves. Results Operation was recommended for 3331 patients but was not performed in 912 (27.4%) patients (then on-operative group). After PSM, 553 pairs matched. Compared to the nonoperative group, the hazard ratios (HRs) in the operative group were 0.46 (95% CI 0.23–0.95 and p=0.037) in stage IIIA and 0.54 (95% CI 0.42–0.68 and p < 0.001) in stage IVA. However, in stages IIIB, IIIC, and IVB, the recommended operative group was not associated with better OS. The OS was not different in stage IIIA-N2, stage IVA-N1, and stage IVA-N3 patients between groups (p=0.28, p=0.14, and p=0.79, respectively). Moreover, the recommended operative group had better OS than the nonoperative group in stage IIIA-N0 (p=0.00085), stage IIIA-N1 (p=0.009), stage IVA-N0 (p < 0.001), and stage IVA-N2 (p=0.034). Conclusion Compared to the nonoperative group, recommended operation improved survival in NSCLC patients with stage IIIA-N0, stage IIIA-N1, stage IVA-N0, and stage IVA-N2. However, in stages IIIA-N2, IIIB, IIIC, IVA-N1, IVA-N3, and IVB, recommended operation did not lead to significantly improved survival time.


Introduction
Lung cancer is the world's leading cause of cancer death [1][2][3]. Nearly 80% of all lung cancer patients are diagnosed with non-small-cell lung cancer (NSCLC) [3,4]. In clinical practice, approximately 75% of patients already have an advanced stage of NSCLC at the time of diagnosis [5,6]. Despite signifcant improvements in the treatment of advanced NSCLC in recent years, survival remains poor, with a fve-year survival rate below 6% [3,7]. Te role of operation as one part of multimodality management for advanced-stage patients is persistent but controversial [8,9].
As the primary local therapy approach, the oncologist performed operative resection of the primary tumor in selected patients with advanced NSCLC [10]. A series of small retrospective research studies provided contradicting results on the benefts of operations [11][12][13]. Randomized controlled trials suggested that operative resection may not enhance overall or progression-free survival in NSCLC patients with stage IIIA-N2 [14,15], noting that one of the trials only included patients with unresectable tumors. Previous studies have mainly focused on operative treatment for stage I-IIIA NSCLC and provided less operative information on more advanced NSCLC [16,17]. Moreover, according to the NCCN guidelines, doctors should consider aggressive local therapy for patients with limited metastases in the context of multimodality treatment [18][19][20][21]. However, there is little basis for physicians to make robust judgments about the appropriate treatment strategies and protocols for NSCLC patients [22,23]. Hence, retrospective cohort research based on real-world populations may be valuable for clinicians to identify operative candidates who are likely to have improved survival outcomes and thus further support better treatment decisions.
Terefore, we conducted a large-scale retrospective cohort study through the US National Cancer Institute Surveillance, Epidemiology, and End Results (SEER) database. Te primary purpose was to investigate the relationship between operative treatment and overall survival (OS) in advanced NSCLC patients. Te second purpose was to determine the clinical characteristics of patients associated with overall survival benefts from the operation, consisting of patient information at the time of the initial diagnosis of NSCLC.

Data.
Tis cohort research adopted the SEER database (the November 2021 submission). We obtained the patient database through SEER * Stat software (SEER Stat 8.4.0).

Inclusion/Exclusion Criteria.
Patients diagnosed with NSCLC (malignant neoplasm of the lung and bronchus, NSCLC histology, and one primary) between 2010 and 2019 were recruited from the database.
Inclusion criteria were (1) patients with pathologically confrmed NSCLC and recommended for surgery, (2) those with stage III and IV following the 8th edition of the American Joint Committee on Cancer (AJCC) TNM classifcation, and (3) those who were diagnosed as the frst primary malignancy.
Exclusion criteria included (1) patients diagnosed with other histological types (e.g., small cell lung cancer, ICD-0-3 8041-8045) and (2) patients with incomplete data (e.g., incomplete survival months, unknown primary tumor location, or unspecifed diagnostic confrmation). Te data and codes were documented by the North American Association of Central Cancer Registries. Te site and histology of primary cancer were coded using the International Classifcation of Diseases (ICD-O-3). Te Research Ethics Board of the Chinese PLA General Hospital exempted the study from ethical approval because the author could not get the patient's identity information. We obtained data agreement according to the requirements of the SEER database.

Variables.
Operation treatment was defned as a record of the following: (1) surgery performed or (2) recommended but not performed in "Reason no cancer-directed surgery" in the SEER database.

Covariates.
According to the published guidelines and research, we obtained the following variables: (1) demographic information, (2) variables that could afect cancer-directed operation for NSCLC or OS reported by previous literature, and (3) other relevant information on account of clinical experience. Te following variables were adopted to construct the adjusted models: age, sex, race, marital status, primary site, grade, laterality, histology, AJCC stage, radiation, chemotherapy, and type of surgery.
2.5. Outcomes. Te outcome was OS. Since the date is a confdential variable in some US registries, a process was established so that a SAS code could be downloaded from the data site. A registry can run the SAS code locally and provide the length of survival in the month for analysis.
2.6. Statistical Analysis. All skewed or normally distributed continuous data were presented as median (Mdn) and interquartile range (IQR) or mean ± standard deviation (SD) as appropriate. Categorical data were expressed in frequency or as percentages. We compared the characteristics of the operative group with the nonoperative group using t-tests (normal distribution), Mann−Whitney tests (skewed distribution), or χ 2 tests (categorical variables), where appropriate.
Tis article followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement. No imputation was performed because the percentage of missing data was small (0-6%). We used propensity scorematched analysis (PSM) to minimize baseline diferences. Baseline matching variables consisted of age, sex, race, marital status, primary site, laterality, histology, AJCC stage, radiation, and chemotherapy. We paired the nonoperative group and the operative group using exact matching with a caliper size of 0.2 based on the propensity scores. Te cumulative rates of death were compared using the Kaplan−Meier curves.
We established multivariate Cox proportional hazard models to investigate the factors associated with overall survival. We assessed the associations between operative treatment and OS using hazard ratios (HRs) and 95% confdence intervals (CIs). Due to the diferences in distribution between the groups, the adjusted Cox proportional hazards models included age, sex, race, marital status, primary site, grade, laterality, histology, AJCC stage, radiation, chemotherapy, and type of surgery.
Te statistical software package R version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria) and free statistics software version 1.3 performed all statistical analyses. A p value < 0.05 (two-sided) was statistically signifcant.

Baseline Characteristics of NSCLC Patients.
A total of 3331 patients with NSCLC who were recommended for surgery were enrolled in this research. Of these patients, 912 (27.4%) were recommended for surgery but not performed (the nonoperative group), and the remainder underwent surgery (the operative group). After propensity scorematched analysis, patient characteristics were balanced across groups, and 1106 patients with NSCLC who were recommended for the operation were enrolled in this research ( Figure 1). Te demographic and clinical characteristics of NSCLC patients before and after PSM are summarized in Table 1. Additionally, 3331 advanced NSCLC patients who were recommended for surgery were alive or dead due to cancer in this study.

Cox Proportional Hazard Regression Analysis.
We constructed multivariate models to evaluate the associations between the operation and OS ( Table 2). Te operation was independently associated with improved OS in the entire advanced cohort (Model 3: HR � 0.57, 95% CI 0.47-0.69, and p < 0.001). Compared to the nonoperative group, the hazard ratios (HRs) in the operative group were 0.46 (95% CI 0.23-0.95 and p � 0.037) in stage IIIA and 0.54 (95% CI 0.42-0.68 and p < 0.001) in stage IVA. However, the recommended operation was not independently associated with improved OS in stages IIIB, IIIC, and IVB. Moreover, HRs were similar between the unadjusted model and the adjusted models with imputed covariate data.

Survival Analysis.
We conducted Kaplan−Meier analyses to compare the survival rates of the operative and nonoperative groups. Te operative patients had better OS than the nonoperative patients in stages IIIA and IVA (p < 0.001, Figure 2). However, the improved OS was not signifcantly diferent between the operative and nonoperative groups in stage IIIB, IIIC, and IVB patients (p � 0.21, p � 0.16, and p � 0.16, respectively, Figure 2).
We evaluated several secondary outcomes to investigate potential factors that might have contributed to the benefts of the operation in stages IIIA and IVA. Te operative patients had better OS than the nonoperative patients in stages IIIA-N0 and N1 (p � 0.00085 and p � 0.009, respectively); however, there was no diference in stage IIIA-N2 patients between the groups (p � 0.28) (Figure 3). In stage IVA-N0 and N2, the operative group had better OS than the nonoperative group (p < 0.001 and p � 0.034, respectively); however, there was no diference in stage IVA-N1 and N3 patients between the groups (p � 0.14 and p � 0.79, respectively) ( Figure 4).

Discussion
Recently, for patients with advanced-stage NSCLC, chemotherapy and radiation therapy have been the primary management modalities [24], and the role of recommended operative treatment remains ambiguous. Our study found that recommended operative treatment was not signifcantly associated with improved survival outcomes in NSCLC patients with stages IIIB, IIIC, and IVB. In addition, our results showed that recommended operative treatment appeared to result in improved survival for NSCLC patients with stages IIIA-N0, IIIA-N1, IVA-N0, and IVA-N2. Our fndings will provide the basis for clinicians to select which kinds of advanced NSCLC patients would beneft from recommended operative treatment.
Among patients with stages IIIA-N2, IIIB, and IIIC, we found that recommended surgery did not signifcantly enhance overall survival, but patients with stages IIIA-N0 and N1 did. As mentioned previously, randomized controlled trials suggested that in the selected stage IIIA-N2 patients with responses to induction chemotherapy, the efect of operative treatment on overall or progression-free survival was not signifcantly better than radiotherapy [14]. In view of this, clinicians might adopt radiotherapy as the preferred choice of local treatment for these advanced NSCLC patients [14,15]. Teir conclusions are consistent with our fndings, and our research further strengthens the previous results. However, the abundance of this patient cohort allowed us to further investigate the potential benefts of having recommended the surgery in appropriately selected NSCLC patients with stage III.
Te latest treatment guidelines have recommended operative resection in selected NSCLC patients with stage IV, such as those cases with early-stage lung cancer and limited extrathoracic metastatic tumors, but there was minimal evidence [19]. Our result proved that stage IVB NSCLC patients who undergo the recommended surgery are unlikely to have better OS than the nonoperative group, but those with stages IVA-N0 and IVA-N2 might have a better OS. Previous studies have shown that surgery for cT1-2, N0-1, M1 or cT3, N0, and M1 disease did not seem to afect prognosis when compared with nonoperative therapy; however, because surgery does not provide an obvious beneft, they should not be recommended to stage IV patients with mediastinal nodal cases or more locally advanced tumors [13]. Tese conclusions are inconsistent with our fndings. However, some other studies were consistent with our results. Yamaguchi et al. [12] reported that by utilizing the local treatment for distant metastases and therapeutic Canadian Respiratory Journal pneumonectomy, some M1b-cStage IV NSCLC patients had more prolonged survival than others. Kawano et al. [11] reported that operative treatment could prolong the survival of NSCLC patients with stage IV on the premise that patients can tolerate surgery. In the clinical setting, stage IV NSCLC patients rarely undergo curative-intent resections. However, a strength of our study was that we were able to use a robust statistical method and an extensive comprehensive population-based SEER database to determine the OS gain among NSCLC patients. After controlling for age, sex, race, marital status, primary site, grade, laterality, histology, AJCC stage, radiation, chemotherapy, and type of surgery. We found that stage IVA patients with chemotherapy might have a signifcantly better impact on the survival benefts. Tis fnding implied that chemotherapy might have a more positive efect in determining the survival benefts of stage IV patients.
A logistic univariate analysis revealed the following possible reasons for refusal to recommend surgery for patients with advanced NSCLC: stage at diagnosis, age, gender, race, marital status, and histology. Surgery is the preferred treatment for stage I lung cancer since it is curable. Tere are other powerful and healing therapies available. However, some patients only receive palliative treatment, while others get no treatment at all [25]. According to the search [25], comorbidities, patient preferences, and illness progression are the main causes of treatment refusal. Terefore, in addition to the previously mentioned reasons in this study, the impact of comorbidities, patient decisions, disease progression, patients' fnancial ability, and acceptance of doctors and techniques on whether advanced NSCLC patients refuse recommended surgery is thus poorly understood and requires further research.
Primary treatment strategies might be changing as systemic therapies develop, but some researchers have suggested that positive responses to systemic therapy would make more NSCLC patients eligible for operative management [18,26,27]. According to the Robinson classifcation, specifc cases of N2-patients had a higher possibility of survival as a part of multimodal therapy after an operative therapy [28,29]. Immunotherapies have revolutionized the treatment of advanced NSCLC. Evidence has emerged that it can be used for stage III disease [30,31]. However, how best to combine surgery with other new therapies needs more studies. In addition, for stage III and IV patients, further studies are needed to fnd better therapeutic options for NSCLC patients.
Nevertheless, this study has several limitations. Firstly, our analysis is based on the assumption that clinicians followed consistent criteria (e.g., NCCN guidelines) to recommend the most appropriate treatments to patients. Secondly, our study was powerless to assess the efect of other therapies (e.g., the multimodality approach) on survival when combined with resection surgery. Although we recruited 3331 advanced NSCLC patients who were recommended for surgery, all of whom were alive or dead due to cancer, we were unable to assess the impact of comorbidities on outcomes. Tirdly, although recommended primary tumor resection might not have an impact on the survival benefts in stage IVA-N1 and N3 patients,      the sample size was limited. Finally, this article relies heavily on statistical analysis, and we must acknowledge the limitations of statistics. In future studies, a larger sample size would be preferred to validate the fndings shown in this study. Besides, we would like to emphasize that correlation does not imply causation. Further study is required NA †, histology and other types included: transitional cell papillomas and carcinomas (n � 1), mucoepidermoid neoplasms (n � 5), cystic, mucinous, and serous neoplasms (n � 72), and complex epithelial neoplasms (n � 74). * , local tumor destruction included laser ablation or cryosurgery and electrocautery; fulguration (included the use of hot forceps for tumor destruction). Excision or resection of less than one lobe included excision, laser excision, bronchial sleeve resection only, wedge resection, and segmental resection (including lingulectomy). Te model adjusted for age, sex, race, marital status, primary site, grade, laterality, histology, AJCC stage, radiation, chemotherapy, and type of surgery. 8 Canadian Respiratory Journal to understand the direct cause of improved OS in the specifc advanced NSCLC patients after recommended resection surgery.
In conclusion, this NSCLC population-based study found that recommended operation was associated with prolonged survival in stage IIIA-N0, stage IIIA-N1, stage NA †, histology and other types included: transitional cell papillomas and carcinomas (n � 1), mucoepidermoid neoplasms (n � 5), cystic, mucinous, and serous neoplasms (n � 72), and complex epithelial neoplasms (n � 74). * , local tumor destruction included laser ablation or cryosurgery, electrocautery; fulguration (included the use of hot forceps for tumor destruction). Excision or resection of less than one lobe included excision, laser excision, bronchial sleeve resection only, wedge resection, and segmental resection (including lingulectomy). Te model adjusted for age, sex, race, marital status, primary site, grade, laterality, histology, AJCC stage, radiation, chemotherapy, and type of surgery.

Total
Non-operation operation    IVA-N0, and stage IVA-N2 patients compared with those who were recommended for operation but not performed. However, in NSCLC patients with stages IIIA-N2, IIIB, IIIC, IVA-N1, IVA-N3, and IVB, no evidence suggests that the recommended operation could signifcantly improve survival time.

Data Availability
Te data used to support the fndings of this study have been deposited in the SEER repository (https://seer.cancer.gov/).

Conflicts of Interest
Te authors declare that they have no conficts of interest.

Authors' Contributions
Hui Wang, Di Yang, and Yan Lv contributed equally to this work.

Supplementary Materials
Supplemental