Choice of postoperative radiation for stage IIIA pathologic N2 non-small cell lung cancer: impact of metastatic lymph node number

Postoperative radiation (PORT) is an option for non-small cell lung cancer (NSCLC) patients with resectable stage IIIA pathological N2 status (pN2). For patients with PORT, this study aims to investigate the impact of the exact number of positive lymph nodes (LNs) on overall survival (OS) and lung cancer-specific survival (LCSS). Within the Surveillance, Epidemiology, and End Results database, we identified 3373 patients with stage IIIA pathological N2 status (pN2) NSCLC who underwent a lobectomy or pneumonectomy from 2004 to 2013. OS and LCSS were compared among patients coded as receiving PORT or observation. The proportional hazards model was applied for investigation. OS and LCSS favored PORT for patients with stage IIIA (pN2) NSCLC. Multivariable analyses showed that PORT and the exact number of positive LNs (n ≤ 3) were independently associated with better OS and LCSS. Both better OS and LCSS emerged for positive LNs (n > 3) after the use of PORT in survival analyses, whereas the benefits of OS and LCSS were not observed anymore for positive LNs (n ≤ 3) group. More importantly, multivariable analyses showed that the use of PORT is an independent risk factor of survival for positive LNs (n > 3) but not for positive LNs (n ≤ 3). In Stage IIIA (pN2) NSCLC, the use of PORT demonstrated better survival results than no PORT for patients with positive LNs (n > 3), but not for patients with positive LNs (n ≤ 3).


Background
The presence of histologically confirmed lymph node metastases is an important prognostic factor for many malignancies. In non-small cell lung cancer (NSCLC), the nodal status with metastases has been suggested to be of significance. For patients with pathologic N2 NSCLC that is considered resectable, complete surgical resection is a favor choice of the management of localized non-small-cell cancer and the use of adjuvant platinum-based chemotherapy is also considered the standard of care presently [1,2].
However, even after complete surgical resection and adjuvant chemotherapy, node-positive patients still have a 20% to 40% risk of localregional recurrence (LRR), and LRR correlates independently with worse OS for patients with NSCLC [2,3]. Thus, postoperative radiation therapy (PORT) is often recommended to improve local tumor control and survival in IIIA (N2) NSCLC patients with good performance status. Additionally, for resectable stage IIIA N2 patients, National Comprehensive Cancer Network (NCCN) guidelines (Version4. 2016) also support the use of PORT (sequential or concurrent chemoradiation) for N2 nodal status patients regardless of whether the surgical margins are positive. Two singlecenter retrospective studies and one previous SEER based study suggested that the use of PORT improved survival for patients with N2 nodal disease [4][5][6]. Another postoperative trial also demonstrated a benefit to PORT in N2 disease [7]. However, a previous metaanalysis of randomized trials demonstrated no benefit with PORT, and the use of PORT could even result in a decrease in OS due to the cardiac and pulmonary toxicity from the radiotherapy itself [8]. In addition, a recent randomized trial refer to whether PORT is benefit or not for N2 NSCLC patients (NCT00410683, Lung ART in Paris) is still recruiting patients, which indicates that PORT or not was still a controversial issue.
In NSCLC, the number of nodal stations with metastases has been previously demonstrated to have significances on the survival of N2 diseases with PORT [9,10]; however, few studies focused on the numbers of positive LNs, and so far only one singlecenter study demonstrated the total number of positive LNs seems to be an independent prognostic indicator in patients with pN2 NSCLC [11]. Therefore, it is valuable to further analyze whether the number of positive LNs could impact the prognosis in pN2 patients with PORT.

Data collection
The Surveillance, Epidemiology, and End Results (SEER) database is a national cancer surveillance program that collects information on all incident cancer cases from 18 areas of United States and covers approximately 26% of the population. In this study, identified data for patients with stage IIIA pathological N2 NSCLC were obtained from the SEER database for patients treated from January 2004 to December 2013. Pathologic IIIA stage patients derived from AJCC stage group (6th and 7th edition). N2 LNs status were defined according to CS LNs codes manual. The specific histologic types selected were those coded as non-small cell carcinoma, large cell carcinoma NOS, adenocarcinoma NOS and squamous cell carcinoma NOS. We then only chose patients with positive LNs according to regional nodes positive codes. Patients were finally included if they underwent a radical surgery of either a lobectomy or pneumonectomy. Subsequently, only those patients coded as receiving no radiation and/or cancer-directed surgery were considered not treated by postoperative radiotherapy, and those who coded as radiation after surgery were defined treated by postoperative radiotherapy. Radiation method was then restricted to beam radiation and radiation NOS according to radiation codes. Overall survival (OS) and lung cancer-specific survival (LCSS) were determined from SEER cause-specific death classification and SEER other cause of death classification codes. OS was defined as the time from surgery until death as a result of any cause, and LCSS was defined as the interval from surgery until death as a result of lung cancer. To reduce the immortal time bias, we excluded patients who survived less than 4 months. Fig. 1 details the selection process for inclusion of patients.

Statistical analysis
The Pearson χ2 test was used to analyze categoric variables. We used Kaplan-Meier method to determine OS and LCSS for patients underwent PORT or not. The log-rank test was used to compare the survival curves between Port and No PORT groups. Multivariable Cox proportional hazards models were used to calculate adjusted hazard ratios (HRs) and their 95% CIs relating to the variables as described.
Results were considered to be statistically significant when P < 0.05. All data were analyzed using the SPSS 22.0 (SPSS, Chicago, IL), and the survival curve was drawn with GraphPad Prism 5.0 (GraphPad Software, San Diego, CA).

Baseline characteristics and outcomes
A total of 3377 stage IIIA (N2) NSCLC patients with positive LNs were included in overall survival analysis, and 3077 patients were included in lung cancer-specific survival analysis. In the OS analysis, comparative treatment strategy was PORT in 1198 patients (35·5%) and no PORT in 2179 (64·5%). In the LCSS analysis, 1094 patients (35·6%) and 1983 (64·4%) underwent PORT and no PORT separately. Table 1 details the baseline characteristics. According to the results illustrated above, PORT was less performed in stage IIIA (N2) NSCLC patients, especially in elderly patients. And patients underwent lobectomy constitute the vast majority of included patients. Additionally, no significance differences emerged in sex, race location, tumor size, T stages, laterality and histology between PORT and No PORT patients.
The survival analysis by Kaplan-Meier plots showed that PORT was significantly associated with better OS (log-rank test, p = 0·0013) and LCSS (log-rank test, p = 0·0094) for all NSCLC patients ( Fig. 2a and b). These results were similar to E, Lally's study [4].

Comparison of positive lymph nodes (n ≤ 3 and n > 3)
Subset characteristic analyses were then performed for patients classified by the number of positive LNs (n ≤ 3 and n > 3). Patients with unspecified number of positive LNs were excluded, and so 194 and 180 patients were excluded from the following OS and LCSS multivariable analyses. Additional file 1: Table S3 and Table S4 detailed the baseline characteristics of OS and LCSS. No significance differences emerged in sex, race location, tumor size, T stages, laterality and histology between two groups, except T stage categories in the No. ≤3 group of LCSS patients' characteristic.
The survival analyses were also investigated based on positive LNs categories (n ≤ 3 and n > 3). For patients with positive LNs (n ≤ 3), no significant differences was observed in OS (p = 0·1435) and LCSS (p = 0·1227) ( Fig. 3a and b). However, for patients with positive LNs (n > 3), there was a significant difference in survival between PORT and No PORT both in OS (p = 0·0015) and LCSS (p = 0·0087) (Fig. 3 c and d).
The Cox proportional hazards regression model ( Table 2) was then applied to study the superiority of PORT in subgroups of positive LNs categories (n ≤ 3 and n > 3). In the positive LNs (n ≤ 3) subgroup, the use of PORT did not have a significant impact on survival both in OS and LCSS (OS with No PORT vs. PORT: HR,

Discussion
Through a large population-based cohort based on SEER database, we investigated whether the use of PORT will improve the prognosis of patients examined rare LNs metastasis. We detected age, race, sex, primary site, histology and so on when analyzing both OS and LCSS in pooled analysis. The outcomes of multivariable analyses demonstrated that the number of positive LNs and the use of PORT were independent risk factors. Patients with positive lymph more than 3 were found with poorer survival, and the use of PORT benefited patients significantly. These independent risk results were similar to a previous SEER based study and a retrospective study [4,11]. In order to detect how the number of positive LNs influences the OS and LCSS after the use of PORT, we assigned patients into positive LNs (n ≤ 3) group and positive LNs (n > 3) group according to the hazard ratios of each number category. Consequently, Kaplan-Meier curves and the cox proportional hazards regression models all demonstrated that the use of PORT significant improves survival for the patients with positive LNs (n > 3). And PORT was found not associated with the survival benefit in patients with positive LNs (n ≤ 3). The SEER data are retrospectively collected, so the potential for error or bias may exist. We recognize that confounding factors, such as margin status and performance status, may influence the treating physician's decision to recommend the use of PORT. This information is not available for analysis, and an estimated 1% to 17% of surgical resections could still result in positive surgical margins [12]. So we only selected patients underwent lobectomy or pneumonectomy to avoid the positive margin status bias as much as possible. In addition, patients underwent lobectomy or pneumonectomy tend to have a better performance status than those who taken partial, wedge or segmental resection. Details about the lymph resection are also not available in the database. According to NCCN guidelines, resection is considered  not appropriate for patients with multiple pathologically proven malignant LNs greater than 3 cm, so those patients with fusion or huge malignant LNs could have been be excluded after patient selection. Considering more than 3000 patients were included in this study, we hoped that the impact of the incomplete LNs resection could be minimized. To date, SEER database does not provide the data of adjuvant chemotherapy or target therapy, and SEER-Medicare database does not open to the users outside the United States. Although preoperative adjuvant chemotherapy is now considered the standard treatment for resectable pN2 status patients, more detail data is still need for the further research. For patients with PORT, details of the dose, range and the dose per fraction are not available as well. Presently, the standard PORT dose and dose per fraction were considered less than 54 Gy and 2 Gy respectively [6,13]. According to the latest NCCN guide lines, the total dose of PORT was recommended no more than 60 Gy and the dose per faction was recommended less than 2Gy. 50-54Gy was recommended for negative margins and 54-60Gy was recommended for extracapsular nodal extension or microscopic positive margins in guide lines. According to a previous SEER and National Cancer Data Base (NCDB) pooled analysis, NCDB contains data not available in SEER database, such as chemotherapy and RT dose. And the results demonstrated that the range of radiotherapy dose is from 45 to 82.8 Gy and the median dose is 54 Gy after screening patients [14]. Therefore, we only included patients underwent lobectomy or pneumonectomy to reduce the heterogeneity of radiotherapy stems from positive margins and different performance status.

Conclusions
In summary, results from our study demonstrated that the exact number of positive LNs in ipsilateral mediastinal nodal resection has an impact on survival for stage IIIA pN2 patients with PORT. Although there were biases in lacking the details of adjuvant chemotherapy and postoperative radiotherapy, standard chemotherapy regiments and modern radiation technology could minimize these biases. To our knowledge, PORT was deemed detrimental for patients with N0 or N1 nodal disease because of the increased rate of intercurrent deaths [4,15,16]. One explanation is undetected microscopic/residual is less in N0 and N1 disease, so the benefit gained by treatment with PORT is diminished from the radiation toxicity [4]. For pN2 nodal status, there is a larger lymphatic metastasis of disease compared with N0 or N1, so the use of PORT is often recommended. According to our results, different number of positive   Abbreviations: NOS not otherwise specified a : 4, 5, 6, ≥7, these four categoric variables were designed for hazard ratios in the right two columns