Background

Lung cancer is the first leading cause of cancer death in China [1] and around the world [2]. Despite the incorporation of standard multimodality therapies tailored by the staging system, non–small cell lung cancer (NSCLC) remains an aggressive disease with a dismal prognosis [3–5], highlighting the importance of seeking other prognostic factors and applying adjuvant or biological therapy for patients with poor prognosis.

Although tumor size and invasiveness have been used for many years in determining the T category of various cancers, tumor location is undervalued. Although this factor is not incorporated in the seventh edition of the International Association for the Study of Lung Cancer tumor-node-metastasis (TNM) staging system [6], tumor location has long been suggested as a potential, albeit controversial, prognostic factor [7–11]. Previous investigators have found that non-upper lobe tumor location was an adverse prognostic factor for patients with stage I or III NSCLC [8–11]. However, a recent study by Puri et al. [12] indicated that tumor location does not predict survival of patients with pathologic stages I and II NSCLC. Kudo et al. [13] also reported that tumor location influences only the long-term survival of NSCLC patients with lymph node metastasis. In addition, Whitson et al. [14] found that, although tumor location is associated with lymph node yield, it is not an independent predictor of survival.

The conflicting results from previous studies may be partially due to the variation in the method by which the tumor locations, such as non-upper and upper, right- and left-sided, and left lower lobe (LLL) and non-LLL tumor, were compared. Moreover, various surgical approaches, including pneumonectomy, wedge resection, and segmental resection, were used in their studies, possibly influencing the survival of patients with resectable NSCLC [15, 16]. In the present study, we therefore selected NSCLC patients who had undergone lobectomy followed by systematic lymph node dissection and explored whether tumor lobe location was an independent prognostic factor for cancer-specific survival (CSS) based on nodal status.

Patients and methods

Patient selection

This study was approved by the Ethics Committee of the Sun Yat-sen University Cancer Center and was conducted using a database of 1578 consecutive patients who had undergone surgery for NSCLC at the Department of Thoracic Oncology between January 2000 and December 2008. Eligible patients included those with histologically proven NSCLC and had undergone complete resection, such as lobectomy (including bilobectomy) and systematic lymph node dissection (R0 resection). Patients who had undergone wedge resection, sleeve resection, segmentectomy, and pneumectomy and those with mediastinal lymph nodes (MLNs) being dissected at less than three stations were not included in the study. Patients who met any of the following criteria were also excluded: tumor located at a fissure and incorporating more than one lobe, the presence of distant metastasis before operation, neoadjuvant therapy before operation, previous or concurrent malignancy, or postoperative death within 1 month. All patients provided written informed consent before surgery.

Therapeutic process

Thoracotomy or video-assisted thoracoscopy were performed. The tumor location was marked during surgery as the right upper lobe (RUL), right non-upper lobe (RMLL), left upper lobe (LUL), or LLL. Following fixation in 10% neutral buffered formalin for approximately 24 h at room temperature, the tumor size was measured and therefore the maximum diameter of the tumor was recorded. At least 6 stations of lymph nodes and no less than three stations of MLNs were dissected with the guidance of the Mountain/Dresler lymph node map [17]. The mediastinal lymphadenectomy procedure was performed in accordance with the approach described in Allen’s study [18]. Each patient was staged according to the seventh edition of TNM classification [6].

Four to six cycles of cisplatin-based adjuvant chemotherapy were recommended to high-risk N0 and node-positive patients undergoing surgery after 2004.

Follow-up

Follow-up was generally carried out 1 month after surgery, every 6 months for the first 3 years, and annually thereafter. The routine examinations included a physical examination, chest radiography, and ultrasound of the abdomen (including the liver, pancreas, spleen, kidney, and adrenal glands). Chest and upper abdominal computed tomography and brain magnetic resonance imaging (MRI) were conducted every year. Bone scan, bronchoscopy, and vertebral MRI were performed only in symptomatic patients. Patients who were lost to follow-up were censored at the last time of contact. The median length of follow-up for surviving patients was 84 months (range 1–163 months).

Statistical analysis

All statistical analyses were performed with SPSS version 16.0 (SPSS, Chicago, IL, USA). Categorical variables were compared by using χ2 test or Fisher’s exact test, when applicable. The differences in the numerical variables between the two groups were analyzed using the Mann-Whitney U test. Univariate and multivariate survival analyses with calculation of the hazard ratio (HR) were performed by using Cox proportional hazards regression model. Potential prognostic factors with a P value of less than 0.10 were entered into multivariate analyses. A stepwise forward procedure was used to derive a final model of the variables. The CSS was defined as the duration from the date of the operation to either the date of death from NSCLC or the last follow-up (August 2, 2013). A significant difference was declared if the P value from a two-tailed test was less than 0.05.

Results

Demographic characteristics

A total of 627 patients were included in this study. Their clinicopathologic characteristics are summarized in Table 1. Histologically, 363 tumors were classified as adenocarcinoma, and 191 tumors were classified as squamous cell carcinoma; the remaining 73 tumors were miscellaneous types, including adenosquamous carcinoma, large cell carcinoma, and carcinoid carcinoma. A total of 587 patients underwent a single lobectomy; the remaining 40 patients underwent bilobectomy, including 8 lesions in the RUL, 6 lesions in the right median lobe, and 26 lesions in the right lower lobe.

Table 1 Clinicopathologic characteristics of the entire non–small cell lung cancer (NSCLC) patient cohort (n = 627)
Full size table

Tumor location and CSS in the entire cohort

During the follow-up period, 275 patients died, including 12 from non-cancer-related causes, such as myocardial infarction, cerebral infarction, pneumonia, and vehicular accidents. The 5-year CSS rate of the entire cohort was 61.6%. The relationship between clinicopathologic characteristics and CSS is shown in Table 2. Univariate survival analysis indicated that tumors located in the non-upper lobe (P = 0.013), LLL tumors (P = 0.007), tumors with advanced T category (P = 0.001), lymph node metastasis (P < 0.001), a tumor size of >4 cm (P < 0.001), the presence of pleural invasion (P = 0.015), a high histology grade (P = 0.006), and bilobectomy (P = 0.044) were risk factors significantly associated with CSS. Multivariate analysis demonstrated that LLL (HR: 1.465, 95% confidence interval CI 1.090–1.969, P = 0.011), lymph node metastasis (HR: 2.742, 95% CI 2.145–3.507, P < 0.001), and a tumor size of >4 cm (HR: 1.474, 95% CI 1.151–1.888, P = 0.002) were independently associated with dismal CSS in patients with resected NSCLC.

Table 2 Univariate and multivariate analysis of prognostic variables for patients with resected NSCLC (n = 627)
Full size table

Tumor location and CSS based on nodal status

We performed a sub-analysis of NSCLC patients stratified in terms of nodal status. The relationship between the clinicopathologic characteristics and CSS in patients with or without lymph node metastasis was examined (Table 3). The 5-year CSS rate of node-negative patients was 73.8%. Univariate analysis indicated that gender (P = 0.022), tumor location (upper vs. non-upper, P = 0.024; LLL vs. non-LLL, P = 0.038), T category (P = 0.001), pleural invasion (P = 0.029), tumor size (P < 0.001), and pathology (P = 0.013) were statistically associated with CSS. However, multivariate analysis demonstrated that the only independent risk predictor for CSS was a tumor size of >4 cm (HR: 1.889, 95% CI 1.324–2.696, P < 0.001, Fig. 1a). In node-positive patients, the 5-year CSS rate was 40.8%, and the presence of LLL tumors was the only factor significantly associated with unfavorable CSS (HR: 1.528, 95% CI 1.015–2.301, P = 0.042, Fig. 1b).

Table 3 Univariate analysis for NSCLC patients with or without lymph node metastasis
Full size table
Fig. 1
figure 1

Kaplan–Meier survival analysis in patients with resectable non–small cell lung cancer (NSCLC). a The relationship between tumor size (≤4 vs. >4 cm) and cancer-specific survival in patients with resectable NSCLC without nodal involvement (P < 0.001); b the relationship between tumor location [left lower lobe (LLL) vs. non-LLL] and cancer-specific survival in patients with resectable NSCLC and nodal involvement (P = 0.042)

Full size image

The relationship between tumor location and clinicopathological factors

As shown in Table 4, the clinicopathologic characteristics were compared between LLL and non-LLL tumors. After the patients were dichotomized into node-positive and node-negative groups, we found that LLL tumors associated with large tumor size (P < 0.001) and high histological grade (P = 0.023) in node-negative but not node-positive patients. Irrespective of nodal status, the association between LLL tumors and a smaller number of dissected MLNs was significant (P < 0.001). There were also significantly fewer dissected MLNs in left-sided than in right-sided tumors [node-negative: 5 (range 3–21) vs. 9 (range 3–44), P < 0.001; node-positive: 5 (range 3–17) vs. 9 (range 3–38), P < 0.001]. However, the difference between the number of MLNs dissected in upper and non-upper tumors was not significant [node-negative: 7 (range 3–44) vs. 7 (range 3–29), P = 0.085; node-positive: 7 (range 3–38) vs. 7 (range 3–36), P = 0.772]. No significant differences in sex, age, T category, TNM stage, pathology, or pleural invasion were observed in any subgroups. However, in node-positive cases, more patients with LLL tumors than those with non-LLL tumors received adjuvant chemotherapy (P = 0.050), whereas no significant difference existed in node-negative patients (P = 0.063).

Table 4 Relationship between tumor location and clinicopathologic factors of NSCLC patients
Full size table

Discussion

This study demonstrated that both non-upper and LLL tumors were unfavorable factors in CSS of patients with resectable NSCLC, as indicated by univariate analysis. However, only LLL tumors retained significance by multivariate analysis. The underlying reasons for this finding are not entirely clear, but we speculate that the lymphatic channel differs between lobes, likely resulting in different prognoses. Lower lobe tumors have the tendency to spread more frequently than upper lobe tumors to the subcarinal station, the passage of the hilar lymph nodes that connects with the contralateral nodes [19]. Hence, contralateral mediastinal drainage from the non-upper tumors may be more common, leading to unfavorable prognoses [20]. Moreover, left lung tumors spread to both the contralateral and ipsilateral lymph nodes with the same frequency [21], and the rate of occult metastasis to the right upper mediastinal nodes in left-sided NSCLC reached 72.7% [22]. Nohl-Oser et al. [23] also demonstrated that the highest metastatic rate to the contralateral mediastinum was 40% in LLL tumors, compared with 22% in LUL and 7% in RLL tumors.

Interestingly, after stratification based on lymph node status, the unfavorable prognosis of LLL tumors was found to be affected by the presence or absence of nodal involvement. LLL tumors were independently associated with CSS only in patients with lymph node metastasis, whereas a tumor size of >4 cm is the only independent risk predictor in the node-negative subgroup. This finding is consistent with that in a previous report by Kudo et al. [13], in which an influence of tumor location on the prognosis of patients with lymph node metastasis was also identified. It is tempting to attribute the dismal outcome of LLL to the fact that the mediastinal lymph nodes are prone to metastasis. However, after calculating the mediastinal metastatic rate of LLL and non-LLL tumors in node-positive patients, we found that tumor location presented no association with the metastasis of the mediastinal lymph nodes. Therefore, the LLL tumor was an intrinsic prognostic factor in this subgroup of patients. The observed disparity may be attributed to the different rates of occult metastasis in the contralateral nodes between node-positive and node-negative patients. In a recent study, Sakao et al. [24] revealed that >50% of patients with ipsilateral mediastinal nodal involvement suffered from occult pN3 disease. LLL tumors may be easier to be understaged to pN3 disease in node-positive patients than in node-negative patients, leading to a more unfavorable prognosis.

Contralateral MLN dissection was not performed in the present study; thus, the uncertainty of the status of the contralateral MLNs casts doubt on our results. However, we performed a systematic MLN dissection using the approach described in Allen’s study [18], which has been recognized as a standard procedure for staging resected NSCLC. Significantly fewer MLNs were dissected in LLL tumors than in non-LLL tumors, potentially because the left upper mediastinal lymph nodes were not routinely dissected in left-sided tumors in the mediastinal lymphadenectomy procedure used. The results revealed that there was only a significant difference in the number of dissected MLNs between the left- and right-sided tumors and not between the upper and non-upper tumors.

In this study, LLL tumors were positively associated with large tumor size and high histology grade in node-negative patients. This association may be because the left heart shadow makes it difficult to find LLL tumors. Tumors of larger size and poor differentiation are more invasive and progressive than smaller and well-differentiated ones. The close association between tumor location, size, and differentiation may explain why LLL tumors without lymph node metastasis are associated with poor survival in univariate analysis but not in multivariate analysis. However, in node-positive patients, tumor size and differentiation were not significantly different between LLL and non-LLL tumors, making tumor location the only significant prognosticator.

The limitations of this study include its retrospective nature, the small sample size, and the long, 9-year time-span of the evaluated cases. The treatment strategy changed numerous times over these years, possibly rendering the evaluation of the association between tumor location and prognosis more complicated. In this study, more node-positive patients with LLL tumors received postoperative chemotherapy than those with non-LLL tumors. Nevertheless, their CSS remains far from satisfactory. Routine and standard adjuvant chemotherapy for node-positive patients was not performed until 2004, and this limitation makes a prospective study on multimodality therapy versus tumor location necessary.

Conclusions

This study revealed that the presence of LLL tumors was a robust unfavorable prognostic factor in resected NSCLC patients. However, its role accounted for node-positive but not node-negative patients, potentially because tumors located in the LLL might be upstaged in node-positive patients. Further studies are needed to determine the optimal management strategy for this subgroup.