Lymph Node Ratio Is an Independent Prognostic Factor for Patients with Siewert Type II Adenocarcinoma of Esophagogastric Junction: Results from a 10-Year Follow-up Study

Emerging evidences suggest that lymph node ratio (LNR), the number of metastatic lymph node (LN) to the total number of dissected lymph nodes (NDLN), may predict survival in multiple types of solid tumor. However, the prognostic role of LNR in adenocarcinoma of the esophagogastric junction (AEG) remains uninvestigated. The present study is intended to determine the prognostic value of LNR in the patients with Siewert type II AEG. A total of 342 patients with Siewert type II AEG who underwent R0 resection were enrolled in this study. The optimal cutoff of LNR was stratified into tertiles using X-tile software. The log-rank test was used to evaluate the survival differences, and multivariate Cox regression analyses were performed to determine the independent prognostic variables. The optimal cutoff of LNR were classified as LNR = 0, LNR between 0.01 and 0.40, and LNR > 0.41. Patients with high LNR had a shorter 5- and 10-year disease-specific survival (DSS) rate (8.5%, 1.4%) compared with those with moderate LNR (20.4%, 4.9%) and low LNR (58.0%, 27.5%) (P < 0.001). Multivariate Cox regression analysis indicated that LNR was an independent factor for DSS after adjusting for confounding variables (P < 0.05). Furthermore, after stratification by NDLN between NDLN < 15 group and NDLN ≥ 15 group, the LNR remained a significant predictor for DSS (P < 0.05). LNR is an independent predictor for DSS in patients with Siewert type II AEG regardless of NDLN. Patients with higher LNR have significantly shorter DSS.


Introduction
The morbidities of esophagogastric junction carcinoma (EGJ) have dramatically increased in recent decades [1][2][3]. EGJ is characterized by pathologically heterogeneous tumors developing in the border between the esophageal squamous epithelium and gastric adenomatous epithelium, and includes cardiac adenocarcinoma, adenocarcinoma of the distal esophagus, and squamous cell carcinoma of the distal esophagus [4]. Among the EGJ, adenocarcinoma of esophagogastric junction (AEG) is the most common pathological type [3]. Due to its unique anatomic location, special gene expression profile, and aggressive behavior, AEG is gaining more and more attention as an independent disease [5,6]. At present, surgery remains the mainstay for resectable AEG [7], but the overall efficacy is not satisfactory in patients with advanced stage diseases [8]. Currently, lymph node (LN) status is a stronger prognostic factor for survival in AEG, and the absolute number of positive LNs is used for determining nodal staging. However, a number of investigators have argued that the current nodal staging based on the positive LNs could be influenced by total number of dissected lymph nodes (NDLN), and probably leading to stage migration [9,10]. Therefore, it is imperative to identify a more reliable prognostic marker to tailor subsequent chemotherapy or radiation therapy.
Increasing evidences have suggested that lymph node ratio (LNR), the count of positive LN to the NDLN, may affect the prognosis in a variety of solid carcinomas [11,12], including esophageal carcinoma [9,13] and gastric cancer [10,14]. Up to now, the LNR has not been evaluated specifically for Siewert type II AEG, and it is crucial to identify optimal indicator with reliable prognostic information. The purpose of this study was to evaluate the prognostic value of LNR in patients with Siewert type II AEG.

Patients
According to the Siewert classification system, AEG can be categorized into three types as follows: AEG type I, the tumor center located 1-5 cm above the gastric cardia; AEG type II, the tumor center between 1 cm above and 2 cm below the gastric cardia; and AEG type III, the tumor center located 2-5 cm below the gastric cardia [15,16]. In our hospital, Siewert type I AEG and Siewert type III AEG were classified as esophageal cancer and gastric cancer, respectively. Therefore, all patients enrolled in this study were diagnosed with Siewert type II AEG. The exclusion criteria are as follows: (1) patients rejected further therapy after diagnosis; (2) patients with distant metastasis; (3) patients received neoadjuvant chemotherapy or radiation therapy before surgery; (4) patients without R0 resection. From January 1, 2000, to December 31, 2010, 342 of patients with Siewert type II AEG were eligible for inclusion. The primary tumor size (T), regional lymph node (N), distant metastasis (M), and the pTNM staging for each patient were according to 8 edition of American Joint Committee on Cancer (AJCC) and Union for International Cancer Control (UICC) [17]. Tumor histologic type and differentiation were also evaluated according to the AJCC/UICC classification [17]. Routine laboratory tests for liver and kidney function, chest electrocardiography, esophageal barium meal examination, and chest and abdomen computed tomography scans were performed in order to exclude surgery contraindications. This study was conducted with the approval of the Ethics Committee and the Institutional Review Board of the Cancer Hospital of Shantou University Medical College (IRB serial number: # 04-070).

Follow-up
The primary outcome of this study was disease-specific survival (DSS), which was defined as the duration from the date of cancer diagnosis to cancer-associated death. The postoperative follow-up was carried out by regular out-patient visit and telephone interview. The endpoint time was December 31, 2018, with a median follow-up of 26 months (ranging from 2 to 201 months) and a total of 279 (81.6%) deaths. Patients that died of non-cancer-related causes, as well as those that were lost to follow-up, were also included in the survival analysis, but the information regarding the mortality of these patients was considered censored data.

Statistical Analysis
The distribution of continuous variables was examined using the Mann-Whitney U test. Chi-square or Fisher's exact test was used to compare nominal data. X-tile software (version 3.6.1) was applied to determine the optimal cutoff for LNR based on the minimal P value method [18]. Survival analysis was examined by Kaplan-Meier plots and the log-rank test. Univariate analysis was performed to identify potential prognostic factors. Then, multivariate Cox proportional hazard analysis was used to explore independent prognostic factors that were statistically significant (P < 0.10) and clinically meaningful in the univariate analysis. Forward stepwise selection with a likelihood-ratio test was performed to determine variables for the Cox regression analysis. SPSS statistical software (version 18.0, SPSS Inc., Chicago, IL, USA) was used for all the analyses. The P value, based on two-sided test, less than 0.05 was considered statistically significant.
As shown in Table 1, Kaplan-Meier analysis showed that the patient age, tumor histopathological type, tumor differentiation, tumor Bormann's type, tumor lesion length, and T stage, N stage, pTNM stage, and surgery procedure were significantly associated with DSS (P < 0.05). No significant difference was observed in the association between DSS and patient's gender, whether having received chemotherapy or radiotherapy.

Multivariate Cox Regression Analysis
Based on the above univariate analysis, the impact of patient age, tumor histopathological type, tumor differentiation, tumor Bormann's type, tumor lesion length, T stage, N stage, pTNM stage, LNR, and surgery procedure on DSS was further tested by multivariable analysis (Table 2). Finally, patient's age, tumor lesion length, pTNM stage, and LNR were significantly associated with DSS in multivariate Cox regression analysis ( Table 2). Patients in the group of LNR = 0 had a longer DSS. Compared with the group of LNR = 0, patient with moderate LNR (0 < LNR ≤ 0.4) had a 1.64-fold increased risk of cancer-related death (HR = 1.64; 95% CI, 1.04-2.58; P = 0.034), and patient with high LNR (LNR > 0.4) had a 2.61-fold increased risk of cancer-related death (HR = 2.61; 95% CI, 1.62-4.20; P = 0.001).

Multivariate Survival Analysis on DSS by Stratification of the Number of Dissected Lymph Nodes
The NCCN guidelines recommend that at least 15 LNs should be removed for adequate nodal staging [19]. The NDLN is a most important factor associated with LNR. Therefore, we investigated whether the LNR was still associated with DSS by stratifying patients into two groups, NDLN < 15 and NDLN ≥ 15. As shown in Suppl Table 2, there were significantly difference on tumor differentiation, tumor length, T stage, N stage, LNR, number of positive LN, number of resected LNs, pTNM stage, chemoradiation, and surgery procedure between the subgroup of NDLN < 15 and NDLN ≥ 15. More patients had advanced tumor stage, poor tumor differentiation, and having received chemotherapy or radiation in the subgroup of NDLN ≥ 15. In addition, more patients underwent total gastrectomy with partial esophagectomy in the subgroup of NDLN ≥ 15 (Suppl Table 1). Furthermore, the DSS in the subgroup of NDLN < 15 was longer than the subgroup of NDLN ≥ 15 (Suppl Fig. 1). However, the results of multivariate Cox regression analysis demonstrated that LNR was significantly associated with DSS both on the subgroup of NDLN < 15 and NDLN ≥ 15 (Table 3). In the subgroup of NDLN < 15, the 5-and 10-year DSS were 58.1%, 15.4%, 11.1% and 31.2%, 5.9%, 2.2% for patients with LNR = 0, 0.01 < LNR ≤ 0.40, and LNR > 0.4, respectively ( Fig. 2a; P < 0.001). In the subgroup of NDLN ≥ 15, the 5-year DSS were 56.7%, 25.6%, and 3.6% for patients with LNR = 0, 0.01 < LNR ≤ 0.40, and LNR > 0.4, respectively ( Fig. 2b; P < 0.001).

Discussion
The involved LN is recognized to be a vital prognostic factor for predicting patient's survival after the operation of gastrointestinal cancer. However, the accuracy of predicting patient's prognosis that harvested sufficient number of suspected metastatic LNs is markedly higher than those without adequate resection of potential involved LNs. A number of investigators have dedicated to identify and develop alternative reliable survival predicting factors or models for patients with gastrointestinal cancer, particularly after removal of the tumor, which could outperform the traditional pTNM staging method [20][21][22]. In this study, we defined the optimal cutoff of LNR as LNR = 0, LNR between 0.01 and 0.40, and LNR > 0.41 using X-tile software [18]. The 5 and 10 years of DSS in the patients with high, moderate, and low LNR were 8.5%, 20.4%, 58.0% and 1.4%, 4.9%, 27.5%, respectively. We also demonstrated that patient's age, tumor lesion length, pTNM The TNM staging system predicts survival according to the extent of local involvement of primary tumor, as well as the presence or absence of regional or distant metastasis. However, the adequacy of the current TNM staging system for AEG is increasingly being questioned [23]. The significance of the LNR for predicting AEG prognosis has been  [24,25]. Sisic and colleagues also demonstrated that the N stage does not prognosticate survival for preoperatively treated patients, while the prognostic role of the LNR remains significant after neoadjuvant treatment [24]. The parameter of LNR combines the number of involved LNs with the absolute NDLN is in fact particularly appealing.
The key merit of LNR is that it can summarize the information reflecting both tumor biology (the number of involved LNs) and treatment modalities (the DNLN) [26], with features of becoming a better indictor to predict the prognosis of patients with AEG. A recent study comprised of 735 patients from the SEER (Surveillance, Epidemiology, and End Results) CI confidence interval, HR hazard risk, LNR lymph node ratio, NDLN number of dissected lymph nodes database supported that LNR, incorporating with a continuous variable, was better than N staging for prognostic prediction of AEG [25]. In line with their results, our study showed that LNR and pTNM stage, but not N stage, were independent prognostic factor for DSS. Therefore, the application of LNR might reduce the stage migration and provide more prognostic information to overcome limitations of current staging systems. The best cutoff for LNR in AEG remains undetermined presently. Most published studies on optimal cutoff of LNR values used an arbitrary calculation, mean values, or quartiles to discriminate patient groups based on their prognosis [22,24,25,27]. In this study, we determined the optimal cutoff using X-tile program. Log-rank statistics was used to calculate the minimum P values in this program, which can minimize the loss of information from multiple testing through crossvalidation and control the inflated type I error [18].
Obviously, it is more appropriate and reliable from a methodological and statistical point of view. In this study, the continuous variate of LNR was classified trichotomy as LNR = 0, LNR between 0.01 and 0.40, and > 0.41. Our results showed that LNR was associated with DSS regardless the number of harvesting LNs. A recent study classified LNR into LNR = 0, 0 < LNR2 ≤ 0.125, 0.125 < LNR3 ≤ 0.425, and LNR4 > 0.425 using X-tile software [21], and suggested that LNR was a unique prognostic factor for survival. But subgroup analysis was not presented in their results. Another study using previous published cutoff value stratified the LNR into LNR = 0, LNR less than 0.30, LNR between 0.31 to 0.60, and LNR > 0.60 [26], by which, they identified LNR as an essential survival predictor. However, further analysis showed that the LNR had no significance for prognosis in the subgroup of NDLN < 15. Therefore, the optimal cutoff value of LNR in this study still needs further validation.
The number of positive LNs identified may rely on the NDLN, while the value of LNR is highly dependent on both the positive harvesting LNs and NDLN. Accurate evaluation of LN metastasis and NDLN not only leads to appropriate staging but also reliably predicts prognosis [23,28]. Incorrect assessment of the LNR could result in insufficient NDLN, therefore, causing an inaccurate prognostic assessment. The NCCN guidelines recommendation for openable patients with AEG demonstrated that at least 15 LNs should be removed for adequate nodal staging [19]. Categorization of LNR could probably be affected by the extent of LN dissection and the stage migration phenomenon, particularly in cases with fewer than 15 examined LNs [29]. Therefore, we performed a subgroup analysis to demonstrate whether the value of LNR was still associated with DSS in the subgroup of NDLN < 15 and NDLN ≥ 15. Interestingly, the results of multivariate Cox regression analysis demonstrated that LNR was still associated with DSS both in the subgroup of NDLN <15 and NDLN ≥15. This may be partly due to LNR incorporating not only the nodal burden and cancer spread but also the extent and quality of surgical staging. Therefore, LNR may be more meaningful for identifying subsets of patients with similar prognosis.
Currently, whether the number of LNs harvested affecting the prognosis remains controversial. It is reasonable to hypothesize that a higher NDLN is associated with a better clearance of occult LN metastasis. This hypothesis is corroborated by several studies that patients with pN0 Siewert type II EGJ had a survival benefit from harvesting more than 15 of LNs dissection [5,30]. The NDLN might improve the patient survival, but possibly lead to increased morbidity and mortality. In particular, a variety of comparative studies between Western and Eastern counties had demonstrated that no superior survival rates were observed in more extensive lymphadenectomy [31][32][33].
More recently, a study by Cao and colleagues indicated that the survival between patients with NDLN < 15 and NDLN ≥ 15 had no significant difference [34]. This perspective was supported by several retrospective studies, which proposed that the NDLN had no impact on overall survival, while a lower LNR was connected to favorable survival [9,24]. Inconsistent with their results, our study showed that the DSS in the subgroup of NDLN < 15 was better than the subgroup of NDLN ≥ 15. The reason for this phenomenon may due to more patients had higher LNR, advanced tumor stage, and poor tumor differentiation in the subgroup of NDLN ≥ 15, which contributed to the poor survival. It is reasonable to propose that tailoring lymphadenectomy according to T and N stage for reducing complications in individual is preferred. Future treatment tactics may include accurate pre-operative assessment of the TNM staging to design an optimal surgical approach. Besides, discovering new biomarkers to locate metastatic nodes or developing novel sentinel node tracing techniques might be considered in future research [35,36].
There are a couple of limitations in our study. Firstly, it was a retrospective study from a single institution, which may have biases such as absence of random assignment and patient selection. Therefore, validation of our findings in another cohort is warranted. Secondly, the extent of LN dissection was not uniform due to our study spanned over a decade during 2000 to 2010, which the surgical paradigm might have shifted. However, we incorporated the LNR information in cases with limited LN dissection, which may compensate the shortage of this situation [37]. Finally, adjuvant chemotherapy or radiation was difficult to finish due to more complications and higher recurrence rate appeared after invasive surgery [38,39], particular for elderly patients. This may compromise the survival result. But our result suggested that more attention should be paid on these patients with higher LNR, since they may benefit from treating with aggressive adjuvant chemotherapy [40].

Conclusion
In summary, our study suggests that LNR provides reliable prognostic information in patients with Siewert type II AEG, regardless of the NDLN. Patients with higher LNR had shorter DSS. LNR can reflect the tumor aggressiveness and is not significantly affected by the extent of LN dissection.
Author Contributions Conception and study: Yuling Zhang, Chunfa Chen, and De Zeng; acquiring data: Yuling Zhang and Ditian Liu; statistical analysis: Yuling Zhang, Ditian Liu, and Chunfa Chen; drafting article and critical revision of article: Chunfa Chen and De Zeng. All authors read and approved the final manuscript.

Compliance with Ethical Standards
Conflict of Interest The authors declare that they have no conflict of interest.