Prognosis of patients with esophageal squamous cell carcinoma after esophagectomy using the log odds of positive lymph nodes.

To compare the log odds of positive lymph nodes (LODDS) with the number of positive lymph nodes (pN), lymph node ratio (LNR), removed lymph node (RLN) count, and negative lymph node (NLN) count in determining the prognosis of patients with esophageal squamous cell carcinoma (ESCC) after esophagectomy. The records of patients with ESCC who received esophagectomy were retrospectively reviewed. The log-rank test was used to compare curves for overall survival (OS), and Cox regression analysis was performed to identify prognostic factors. The prognostic performance of the different lymph node staging systems were compared using the linear trend chi-square test, likelihood ratio chi-square test, and Akaike information criterion. A total of 589 patients were enrolled. Univariate Cox analysis showed that pN stage, LNR, RLN count, NLN count, and the LODDS were significantly associated with OS (p < 0.05 for all). Multivariate Cox analysis adjusted for significant factors indicated that LODDS was independent risk factor on overall survival (OS), and a higher LODDS was associated with worse OS (hazard ratio = 3.297, 95% confidence interval: 2.684–4.050, p < 0.001). The modified Tumor-LODDS-Metastasis staging system had better discriminatory ability, monotonicity, and homogeneity, and better optimistic prognostic stratification than the Tumor-Node-Metastasis staging system in determining the prognosis of patients with ESCC. The LODDS staging system was superior to other lymph node classifications in determining the prognosis of patients with ESCC after esophagectomy. LODDS may be incorporated into esophageal staging system if these results are eventually confirmed by other studies.


INTRODUCTION
Esophageal cancer (EC) is a common malignancy of the gastrointestinal system, and esophageal squamous cell carcinoma (ESCC) is the major type of EC in China [1,2]. Currently, the pathologic nodal (pN) stage of EC is determined by the number of positive lymph nodes (PLNs) [3]. However, inaccurate lymph node dissection and pathological evaluation may cause understaging of pN stage and inappropriate treatment, the so-called phenomenon of stage migration [4,5]. There are several nodal staging systems which have been proposed and investigated to provide a more accurate nodal staging system, including the lymph node ratio (LNR), the number of removed lymph nodes (RLNs), and the number of negative lymph nodes www.impactjournals.com/oncotarget (NLNs) [6][7][8][9][10][11][12][13]. The number of RLNs is an important indicator of the accuracy of staging. However, the therapeutic and prognostic value of the number of RLNs remains controversial [7,[14][15][16]. In addition, for nodenegative EC patients, the LNR status is the same as pN0 classification, and thus LNR does not provide a more accurate prognosis than Tumor-Node-Metastasis (TNM) staging. Moreover, several studies have suggested that a higher number of NLNs is associated with better survival [9][10][11].
Log odds of positive lymph nodes (LODDS), defined as the log of the ratio between the probability of the number of PLNs and the number of NLNs [16][17][18][19][20]. The LODDS has been determined as novel prognostic factor in patients with gastric cancer, colorectal cancer, and pancreatic cancer [17][18][19][20][21][22][23]. Until now, no study accessing the prognostic value of the LODDS in patients with ESCC has been reported. Thus, the aim of this study was to compare the prognostic value of the LODDS with the pN stage, LNR, NLN count, and RLN count in patients with ESCC after esophagectomy.
We determined whether the influence of LODDS staging system on OS was modified by the pT stage. In both pT1-2 and pT3 patients, LODDS staging system was significantly associated with OS (log rank p < 0.001 for pT1-2, and p < 0.001 for pT3 patients) (Figure 2A-2B).
We examined the prognostic effect of the LODDS staging system according to different pN stage. In both pN0 (log rank p < 0.001) and pN1 patients (log rank p = 0.018), LODDS staging system was significantly associated with OS ( Figure 3A-3B). There were trends for prognostic effect of the LODDS staging system in pN2-3 patients, but these were not statistically significant (log rank p = 0.092) (Table 3C). We then examined the influence of LODDS classifications on OS according to histologic grade. The effect of LODDS classifications significantly difference across any histologic grade group (log rank p < 0.001 for G1 patients, p < 0.001 for G2 patients, and p = 0.027 for G3 patients) ( Figure 4A-4C).
We finally compared the OS rates of patients in different LODDS classifications according to the number of RLNs. The prognostic effect of LODDS classifications was also found in patients with ≤ 14 RLNs (log rank p < 0.001) and more than 14 RLNs (log rank p < 0.001) ( Figure 5A-5B).

Tumor-LODDS-Metastasis (TLM) staging
We used the likelihood ratio chi-square test, the linear trend chi-square test, and the AIC analysis to compare the prognostic performance of the different lymph node staging systems. Of note, the LODDS staging system performed superiorly to the other lymph node staging systems (Table 4).   As expected, analysis of patients by cancer stage from the 2009 UICC/AJCC classification system indicated poorer OS in patients with more advanced cancer (p < 0.001) ( Figure 6A). We developed a Tumor-LODDS-Metastasis (TLM) staging system by replacing the pN stage with the 4 different LODDS classifications. In this new staging system, there was no patient with stage IA. Survival analysis showed that the TLM staging system was also effective in predicting the survival of patients with ESCC (p < 0.001) ( Figure 6B).  We used the likelihood ratio chi-square test, the linear trend chi-square test, and the AIC analysis to compare the prognostic performance of the TNM and TLM staging systems. The results showed that the TLM staging system had a better prognostic value than the 2009 UICC/AJCC staging system for all 3 tests (Table 5). These results indicated that the new TLM staging system considering the LODDS classification rather than pN stage, is effective in determining the prognosis of patients with ESCC.     AIC, Akaike information criterion; LNR, lymph node ratio; RLN, removed lymph node; NLN, negative lymph node; LODDS, log odds of positive lymph nodes.

DISCUSSION
We examined 589 patients who received radical surgery of ESCC, and compared the prognostic performance of different lymph node staging systems in determination of the OS. The results showed that LODDS was independently associated with OS and that it had better prognostic value than other lymph node staging systems, including pN stage, LNR, NLN count, and RLN count.
Lymph node status is one of the most important factors used to predict survival in ESCC patients. In the original TNM staging system, lymph nodes were simply categorized as absent (N0) or present (N1) [29]. However, there is increasing interest in using the detailed status of lymph nodes for determining the prognosis of patients with EC. The staging system for EC from the 7 th edition of the AJCC/UICC reclassified the pN stage according to the number of PLNs [3]. However, the number of RLNs affects the number of PLNs, so staging by PLN count alone has limited use because it may lead to stage migration [4,5]. In recent years, several nodal staging systems including LNR, RLN count, and NLN count indicated that these parameters had prognostic value for EC patients [6][7][8][9][10][11][12][13]. The LODDS value is a novel indicator of lymph node status that has the potential to further improve the accuracy of pN classification and thereby provide a more reliable and accurate prognosis.
Recent studies that considered the LODDS focused on colorectal cancer, gastric cancer, and pancreatic cancer. These studies showed that use of the LODDS may help to provide a more accurate prognosis of such patients, and that it was superior to other systems used for the staging of lymph nodes [17][18][19][20][21][22][23]. To the best of our knowledge, the present study is the first to confirm the prognostic performance of the LODDS in ESCC.
The univariate analysis of the present study indicated that all five lymph node classifications had significant prognostic value. We used multivariate analysis to compare the performances of these different classifications. In particular, we used 6 multivariate models, including all of the five lymph node classifications (pN, LNR, RLNS, NLNs, and LODDS). In the first 4 classifications, pN, LNR, RLNs, and NLNs were significantly and independently associated with OS. In model 5, the LODDS classification was superior to pN classification. In Model 6, we used all five lymph node classifications together in the multivariate analysis, and the results showed that the LODDS classification was superior to these other lymph node staging systems. With that, we believe that the LODDS classification is more likely to minimize the effects of stage migration. In particular, the prognostic value of LODDS of ESCC patients was not significantly different by the extent of lymph node dissection.
In this study, by using the TNM staging system for esophageal cancer as a foundation, we replaced the N stage with the LODDS classification to establish a novel TLM staging system. The results indicated that this new staging system had a better performance than the TNM staging system for the likelihood ratio chi-square test, the linear trend chi-square test, and the AIC. Thus, we suggest that investigators consider incorporation of the LODDS into the lymph node staging used for ESCC in future.
The current study had several limitations that should be considered. First, this was a single-center retrospective analysis. Thus, prospective studies are needed to confirm our results. Second, we excluded patients with stage IV EC, so the relevance of our findings to these patients is unclear. In addition, we used cut-off points of LODDS classification determined according to the interquartile range. To date, no study has investigated the use of LODDS classification in EC, and so the optimal cut-off points have not yet been established.
In conclusion, our findings indicate that the LODDS is independently and significantly associated with survival of patients with ESCC, and that its prognostic value is superior to the other lymph node staging systems including pN stage, LNR, RLN count, and NLN count. LODDS may be incorporated into esophageal staging system if these results are eventually confirmed by other studies.

Patients
We retrospectively reviewed the records of patients with ESCC who received radical esophagectomy from  surgical margin (R0); (iv) no preoperative radiotherapy or chemotherapy. Patients died within 3 months after esophagectomy were excluded. This study was approved by the ethics committee of Sun Yat-Sen University Cancer Center. All patients provided written consent for storage of their information in the hospital database and for use of this information in our research.

Patient characteristics and lymph node status
Patient clinicopathologic factors including sex, age, pathologic tumor (pT) stage, pN stage, histologic grade, tumor location, LNR, RLN count, NLN count, and LODDS were used to assess the risks of overall mortality. The pT and pN stages were determined according to the 7 th edition of the AJCC/UICC staging system. The total number of RLNs was the number of dissected lymph nodes from the cervical, intrathoracic, and abdominal regions. The number of NLNs was obtained by subtracting the number of PLNs from the total number of RLNs. The LNR was defined as the ratio of the number of PLNs to the number of RLNs. The LODDS was defined as log 10 ([PLNs + 0.5]/[NLNs + 0.5]), in which 0.5 was added to both the numerator and the denominator to avoid an infinite number [24].

Surgical procedures
Surgery was performed according to our previously reported [6,11]. Briefly, all patients received radical esophagectomy and lymphadenectomy. The most common surgical approaches included the left thoracotomy, the Ivor-Lewis approach, and the cervicothoraco-abdominal procedure. Extensive lymph node dissection in the posterior mediastinum and abdomen was routinely performed. The lymph nodes in the middle and lower mediastinum, including the periesophageal, parahiatal, subcarinal, and aortopulmonary nodes were systematically dissected. In the abdomen, an upper abdominal and retroperitoneal lymph nodes dissection was performed including the celiac, splenic, common hepatic, left gastric, lesser curvature, and parahiatal nodes. Cervical lymphadenectomy was not routinely undertaken. Three-field lymphadenectomy was performed in those with suspected supraclavicular lymph node metastasis.

Examination of lymph nodes
All resected lymph nodes were submitted for pathologic assessment. The pathologists evaluated the resected lymph nodes, which were separately labeled by the surgeons in a routine manner. Lymph node metastasis was evaluated by hematoxylin and eosin (HE) staining. The number of RLNs was the sum of the cervical, intrathoracic, and abdominal lymph nodes.

Follow-up and survival endpoints
Follow-up after surgery consisted of a hospital visit, telephone call, or mail. In general, patients were scheduled for clinical follow-up every 3 to 6 months, including a physical examination, laboratory tests of complete blood counts, liver function test, chest radiography, ultrasonography, esophagography, and endoscopy. In addition, a fluorine-18-fluorodeoxyglucose positron emission tomography scan was performed if necessary. The endpoint of this study was overall survival (OS). OS was calculated as the time from the date of diagnosis to the date of death from any cause or the date of the last follow-up.

Statistical analysis
All data were analyzed using the SPSS statistical software package (version 16.0; IBM Corporation, Armonk, NY, USA). The χ 2 and Fisher's exact probability tests were used to assess differences in qualitative data. Recognizing that RLN count, NLN count, and LODDS may have been incompletely counted or affected by natural inter-individual variation in nodal distribution, there were examined as a categorical variable based on quartiles. The optimal cutoff point of RLN count and NLN count was 14 and 13, respectively, according to our previously reported [9]. The survival curves were plotted by the Kaplan-Meier method and p-values were calculated using the log rank test. Cox regression analysis was performed to identify significant prognostic factors. A p-value less than 0.05 was considered statistically significant.
The variance inflation factor (VIF) was used to verify the multicollinearity of pN stage, LNR, RLN count, NLN count, and LODDS, a VIF of 10 and above indicates a multicollinearity problem. The prognostic performances of the different lymph node staging systems were compared in terms of homogeneity, discriminatory ability, and monotonicity. We used the likelihood ratio chisquare test, the linear trend chi-square test, and the Akaike information criterion (AIC) within the Cox regression model to compare performances among different lymph node staging systems; the likelihood ratio chi-square test to measure homogeneity, the linear trend chi-square test to measure the discriminatory powers and monotonicities across categories, and the AIC to measure the optimistic prognostic stratifications [25][26][27][28]. A higher likelihood ratio chi-square score means better homogeneity, higher linear trend chi-square scores show better discriminatory ability and monotonicity, and smaller AIC values represent better optimistic prognostic stratification.

ACKNOWLEDGMENTS AND FUNDING
This work was supported by grants from the National Natural Science Foundation of China (No. 81402527), the Sci-Tech Office of Guangdong