*3.1. Demographic and Clinicopathologic Characteristics*

Of the 77 patients with pathologic stage IIIA-N2 NSCLC, 35 were male, and 42 were female, with a mean age of 59 years (SD, 12.2 years; range, 34 to 82 years). Thirty-five (45.5%) patients were diagnosed as N2 disease before surgery, and 42 (54.5%) were diagnosed unexpectedly during surgery. Forty-one (53.2%) patients underwent VATS, and 36 (46.8%) underwent open thoracotomy. The most common histology was adenocarcinoma (62, 80.5%), followed by squamous cell carcinoma (9, 11.7%). The mean size of tumor was 2.9 cm (SD, 1.0 cm). Forty-five (58.4%) patients had tumors of 3 cm or less in diameter, and 32 (41.6%) patients had tumors greater than 3 cm. With respect to lymph node involvement, multiple N2 station was seen in 21 (27.3%) patients and median N2 ratio was 33.3% (IQR, 13.8–50%). Sixty-five (84.4%) patients received adjuvant chemotherapy, of which 23 patients received postoperative radiotherapy. Demographic and clinicopathologic characteristics of the patients are shown in Table 1.


**Table 1.** Demographic and clinicopathologic characteristics of 77 patients with resected stage IIIA-N2 non-small-cell lung cancer.

SD, standard deviation; IQR, interquartile range; ECOG, Eastern Cooperative Oncology Group; VATS, video-assisted thoracoscopic surgery; CEA, carcinoembryonic antigen; y, years.

#### *3.2. Overall Survival*

Figure 1 depicts that the one-, three-, and five-year OS rates were 91.9%, 61.3%, and 33.5%, respectively. The mean length of follow-up was 38.1 months.

**Figure 1.** Overall-survival curves of 77 patients with completely resected stage IIIA-N2 non-small-cell lung cancer.

In univariate analysis, the median OS was significantly influenced by tumor size. The median OS was 52.0 months (95% CI: 45.3–66.1) in patients with tumors of 3 cm or less, worsening to 32.6 months (95% CI: 23.2–43.6) in patients with tumors greater than 3 cm (log-rank *p* = 0.002) and corresponding to a five-year OS rate of 43.3% and 21.7%, respectively (Figure 2). Moreover, patients with VATS approach had significantly better OS compared with those who received open thoracotomy (five-year OS: 63.5% vs. 18.3%; log-rank *p* = 0.009). On the other hand, OS rates were significantly worse in patients with elder age (versus those with age under 65 years, five-year OS: 24.2% vs. 39.0%; log-rank *p* = 0.031) and those with ECOG 1 (versus those with ECOG 0, 5-year OS: 19.3% vs. 49.4%; log-rank *p* = 0.016).

**Figure 2.** Overall survival curves of patients stratified by tumor size.

Multivariate analysis showed that tumor size <3 cm (HR: 0.373, 95% CI: 0.194–0.714, *p* = 0.003) and VATS approach (HR: 0.383, 95% CI: 0.178–0.824, *p* = 0.014) were significant predictors for OS. Univariate and multivariate data are shown in Tables 2 and 3.


**Table 2.** Univariate analysis of factors associated with overall survival.

**Table 3.** Multivariate analysis of factors associated with overall survival.


CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; VATS, video-assisted thoracoscopic surgery.

#### *3.3. Disease-Free Survival*

The one-, three-, and five-year DFS rates were 53.4%, 24.5%, and 12.5%, respectively.

In univariate analysis, the median DFS was significantly influenced by tumor size. The median DFS was 18.4 months (95% CI: 11.9–33.6) in patients with tumors of 3 cm or less, worsening to 11.0 months (95% CI: 7.1–15.6) in patients with tumors greater than 3 cm (log-rank *p* = 0.016) and corresponding to a three-year DFS rate of 33.4% and 12.5%, respectively (Supplementary Materials Figure S1). There was a non-significant trend between poor prognosis and both clinical N2 disease (versus unsuspected N2 disease, three-year DFS: 16.2% vs. 31.1%; log-rank *p* = 0.077) and elevated CEA level (versus CEA level less than 3 ng/mL, three-year DFS: 18.2% vs. 33.3%; log-rank *p* = 0.053).

Multivariate analysis showed that tumor size <3 cm (HR: 0.451, 95% CI: 0.235–0.865, *p* = 0.017) and clinical N2 versus unsuspected N2 disease (HR: 2.525, 95% CI: 1.340–4.757, *p* = 0.004) were significant

Variables with *p*-values of less than 0.2 were tested in multivariate analysis. CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; VATS, video-assisted thoracoscopic surgery; CEA, carcinoembryonic antigen.

predictors for DFS. Both univariate and multivariate data are shown in Supplementary Materials Tables S1 and S2.

### **4. Discussion**

The AJCC TNM staging system is the global standard for lung cancer staging [8]. Compared with the seventh edition, the eighth edition has been validated in several cohorts [9,10], demonstrating better survival stratification and prognosis prediction. With regard to the major changes in the T classification, former stage IIIA-N2 disease is further separated into stage IIIA and IIIB, based on tumor size, which is suggestive of distinct prognosis between the two subgroups. Sui et al. [9] retrospectively analyzed a Chinese cohort including 3599 patients with pathological stage IA to IIIA between 2005 and 2012. Of 772 former stage IIIA patients, stage migration to IIIB was found in 180 (23.3%) patients, and associated with lower five-year survival rate (26.1% vs. 41.7%, *p* < 0.001). Therefore, we focused on updated stage IIIA-N2 NSCLC, which represents a heterogenic group of patients and complex treatment modalities, including surgical resection.

The role of surgical resection for patients with stage IIIA-N2 NSCLC remains controversial, with different management preferences between Europe and America [11]. In Europe, surgeons tend to perform upfront resection, without induction therapy, for single-station, non-bulky N2 disease. The European Society for Medical Oncology (ESMO) guideline recommends that surgical resection, followed by adjuvant chemotherapy, is a reasonable treatment option for single-station N2 disease [8]. By contrast, in America, the standard treatment has been induction chemotherapy or chemoradiation, followed by surgical resection. A Cardiothoracic Surgery Network survey [12] demonstrated that more than 80% of thoracic surgeon preferred induction therapy for stage IIIA-N2 NSCLC, whereas only 12% preferred surgical resection followed by adjuvant therapy. For macroscopic single station N2 disease, 62% would consider surgical resection only if N2 clearance was achieved, whereas 18% considered this inoperable and offer definitive concurrent chemoradiation. Regarding the preference of induction therapy followed by surgical resection in America, considerations include better tolerance to full-dose chemotherapy preoperatively, better control of the systemic micro-metastases, assessment of treatment response before decision of surgery, and possible parenchymal sparing surgery [13]. Therefore, the approach of induction therapy is supported by the National Comprehensive Cancer Center Network (NCCN) guideline [7]. However, despite the high agreement and guideline recommendation, substantial variation in clinical practice existed in The Society of Thoracic Surgeons General Thoracic Surgery Database [14]. Of 3319 clinical stage III-N2 patients, 54% received direct surgical resection and 46% received induction therapy, with five-year survival rates of 36% and 35%, respectively. Considering the controversial role of surgical resection for patients with updated stage IIIA-N2 NSCLC, our study was aimed to investigate prognostic factors to guide therapeutic decisions.

For former stage IIIA-N2 NSCLC, previous studies have well demonstrated prognostic factors, including number of positive lymph nodes [15], microscopic N2 [16], single-station N2 [17–20], VATS approach [21], lobectomy approach [22], postoperative radiotherapy [23,24], and pathological response after induction therapy [25,26]. However, there is still some concern about changes of T classification and stage migration in the eighth edition. In the study, we presented a single-center retrospective study of 77 surgically resected IIIA-N2 NSCLC patients, staged according to the eighth edition of the AJCC staging system. Our first finding is that tumor size <3 cm was associated with better prognosis (HR: 0.373, *p* = 0.003). The possible reason is that tumor size was correlated with occult systemic micro-metastases. Yang et al. [27] reported that the proportions of cases with N0M0 status with tumor size <2 cm and >7 cm were 70.79% and 33.33%, respectively. Cho et al. [28] analyzed the data of 1821 patients with clinical N0-1 NSCLC, in which they found that tumor size >3 cm was a common predictor for unsuspected N2 and multiple-station N2 disease. Based on our finding and major changes of T classification, further large-scale studies are warranted to confirm the role of tumor size in patient selection and treatment strategy. Our second finding is that VATS approach was associated with better prognosis (HR: 0.383, *p* = 0.014). Previous studies showed similar results [21,26]. Despite the possible

selection bias of our study, the consistency of these findings suggests that the VATS approach can be employed safely, without compromised prognosis.

Our 33.5% five-year OS rate is slightly lower than that of the IASLC database [5], in which the five-year OS rates for clinical and pathological stage IIIA disease are 36% and 41%, respectively. The relatively poorer prognosis in our patients highlights the importance of patient selection and the multimodality treatment approach. First, in patients with stage IIIA-N2 NSCLC undergoing surgical resection, the prognostic value of degree of lymph node involvement has been well documented. The ESMO guideline [8] highlights that single-station N2 disease is the most important features while evaluating resectability. Several studies [17–20] have also demonstrated that multiple-station N2 involvement indicates a poorer prognosis, regardless of whether the induction therapy is given: five-year OS rate is usually below 25%. Given the poorer prognosis and higher risk of systemic micro-metastases, upfront surgical resection should be avoided in patients with multiple-station N2 disease. However, in our study, 21 (27.3%) patients with multiple-station N2 disease received surgical resection, whereas only 10 (13%) patients received induction therapy. Second, regardless of whether to offer surgical resection, the implementation of multimodality treatment is of most importance [7,8]. There is pooled evidence in a network meta-analysis [29] where patients with stage IIIA-N2 NSCLC treated with single modality treatment of either surgery or radiotherapy alone seemed to have the worst outcomes. Nevertheless, in our study, 12 (15.6%) patients received surgical resection, only without adjuvant therapy. The lack of multimodality treatment would also explain the poorer outcome of the study.

Our study has some limitations. First, given the nature of retrospective analysis, patients in our study were highly selected by multidisciplinary team screening and not representative of all patients with stage IIIA-N2 NSCLC. In addition, it is not possible to answer the question whether upfront surgical resection is superior to other multimodality approaches. Second, the number of cases in our study was small. The uneven distribution of clinicopathologic characteristics (e.g., single- or multiple-station N2) and treatment approaches (e.g., induction therapy) complicated the interpretation, and the statistical power could be limited.
