Indications of Lymph Node Metastasis and Survival Analysis in T1N+M0 Gastric Cancer: a Population-Based Study


 Background

Since the definition of early gastric cancer (EGC) was first proposed in 1971, the treatment of gastric cancer with or without lymph node metastasis (LNM) has changed a lot. The present study aims to identify risk factors for LNM and prognosis, and to further evaluate the indications for adjuvant chemotherapy (AC) in T1N + M0 gastric cancer.
Methods

A total of 1291 patients with T1N + M0 gastric cancer were retrieved from the Surveillance, Epidemiology, and End Results (SEER) database. Univariate and multivariate analyses were performed to identify risk factors for LNM. The effect of LNM on overall survival (OS) and cancer-specific survival (CSS) was compared with patients grouped into T1N0-1 and T1N2-3, as the indications for AC.
Results

The rate of LNM was 19.52%. Multivariate analyses showed age, tumor size, invasion depth, and type of differentiation and retrieved LNs were associated with LNM (p < 0.05). Cox multivariate analyses indicated age, sex, tumor size, N stage were independent predictors of OS and CSS (p < 0.05), while race was indicator for OS (HR 0.866; 95%CI 0.750–0.999, p = 0.049), but not for CSS (HR 0.878; 95% CI 0.723–1.065, p = 0.187). In addition, survival analysis showed the proportion of patients in N+/N0 was better distributed than N0-1/N2-3b. There were statistically significant differences in OS and CSS between patients with and without chemotherapy in pT1N1M0 patients (p༜0.05).
Conclusions

Both tumor size and invasion depth are associated with LNM and prognosis. LNM is an important predictor of prognosis. pT1N + M0 may be appropriate candidates for AC. Currently, the treatment and prognosis of T1N0M0/T1N + M0 are completely different. An updated definition of EGC, taking into tumor size, invasion depth and LNM, may be more appropriate in an era of precision medicine.

adopted in many clinical studies [4]. Since the indications were rst introduced for endoscopic mucosal resection (EMR) in 1987, therapeutic EMR and endoscopic submucosal dissection (ESD) have been accepted as routine option for EGC [5]. However, the presence of LNM is as an absolute contraindication to ESD.
Over the past years, many studies indicated the depth of invasion is the most important factor which is associated with LNM. The incidence of LNM in EGC with submucosal and mucosal invasion is 8-25% and 2-5%, respectively [6][7][8][9]. LNM is also considered as one of the most important prognostic factor.
The 5-year survival rate for patients with and without LNM is 87.3% and 94.2%, respectively [10]. In addtion, the indication for AC in T1N + M0 gastric cancer is T1N2-3bM0, according to the latest Japanese gastric cancer treatment guidelines 2018 (5th edition) [11]. Whether AC is really unnecessary for T1N1M0 gastric cancer remains controversial. Moreover, the majority of gastric cancer occurs in Asia, data regarding LN metastasis for T1 cancer from non-asian patients is rare. Therefore, our study aims to identify risk factors for LNM and to further evaluate the indications for AC in T1N + M0 gastric cancer using Surveillance, Epidemiology, and End Results (SEER) database from USA.

Patients and Methods
All analyzed data were collected from SEER database between January 1988 and December 2012. The inclusion criteria included: (1) patients aged at least 20 years with gastric adenocarcinoma con rmed by histology (2) primary gastric cancer, (3) underwent radical surgery (4) The tumor pathological stage was T1M0 according to the 7th AJCC stage, (5) at least 1 lymph node examined; Patients were excluded due to: (1) unknown T and/or N, M category; (2) patients with previous malignancies; (3) patients received radiotherapy or chemotherapy prior to surgery; (4) patients without exact examined lymph node.
All cases have been recoded using the International Classi cation of Disease for Oncology third edition (ICD-O-3). The histological types were categorized into intestinal type and diffuse type. According to the ICD-O-3, diffuse type EGC includes signet ring cell carcinoma (M8490), diffuse carcinoma (M8145), and linitis plastica (M8142). Intestinal type EGC includes carcinoma (not otherwise speci ed; M8010), adenocarcinoma (not otherwise speci ed; M8140), tubular (M8211), and intestinal type (M8144). The patients were divided into two groups according to age (≤ 84 and > 85 years old) and tumor size (≤ 3 cm and > 3 cm). Tumor sites were divided into eight groups, as follows: cardiac, fundus, body, antrum, pylorus, lesser curve, large curve and overlapping/NOS.

Statistical Analyses
Statistical analyses were performed with SPSS 21.0. Fisher's exact or chi-square tests was used for categorical variables. Logistic regression analyses were performed to identify risk factors for LNM. Cumulative survival rates of OS and CSS were analyzed using the Kaplan-Meier method. Multivariate Cox regression was performed to explore the potential risk factors for poor OS and CSS. A P value < 0.05 was considered statistically signi cant.

Risk factors for LNM
The characteristics were compared between LN negative group and LN positive group ( Table 1). The results showed that there were no signi cant differences between the groups in sex, race and Lauren classi cation (p > 0.05). The diffuse histologic type showed no more frequent LNM than the intestinal type (P = 0.061). But the two groups signi cantly differed in age, primary site, tumor size, invasion depth (T stage), tumor grade and retrieved LNs (p < 0.05). The proportion of patients with LNM was lower in elder patients (≥ 85) than younger patients (p = 0.025). A greater tumor size (> 3 cm), deeper invasion depth (T1b) were related with more frequent LNM (p < 0.001). In the patients with poorly differentiated type cancer, the rate of LNM was much higher than the differentiated type (p < 0.001). Multivariate analysis showed that age, tumor size, invasion depth, type of differentiation, and retrieved LNs were associated with LNM in T1N + M0 gastric cancer patients (Table 2).

Subgroups analysis on OS and CSS
We divided the patients into different groups according to the number of LNMs ( Fig. 2 and Fig. 3). Kaplan-Meier survival analysis showed that there were statistically signi cant differences in OS and CSS between pN0/N + or pN0-1/N2-3b categories (p 0.001), but an overlap in survival curves was found between categories pN0-1/N2-3b. The proportion of patients in pN0/N + was better distributed and pN0/N + category showed improved prognostic performance in predicting OS and CSS. A statistical assessment of the predictive performance of the two category methods revealed that the pN0/N + category had a higher χ2 (26.06 vs. 18.39) for OS, and the difference is much higher (45.47 vs. 24.37) for CSS. The pN0/N + classi cations for T1N + M0 seem to have an optimal prognostic strati cation.
In addition, Kaplan-Meier survival analysis showed that there were statistically signi cant differences in OS and CSS between patients with and without chemotherapy in pT1N1M0 patients (p 0.05). Therefore, pT1N1M0 may be an indication for AC in gastric cancer patients.

Discussion
Early gastric cancer (EGC) is de ned as tumor con ned to the mucosa or submucosa, regardless of LNM in 1971 by Murakami [3]. Even though, other classi cations, such as Kodama's classi cation [12] and Paris's classi cation [13] were also proposed to de ne EGC, the Murakami de nition is the still the most widely adopted one in recent studies. In an era of precision medicine, the diagnosis, treatment and prognosis of EGC with or without LNM are completely different. Currently, EMR and ESD are rst alternative choice for patients without LNM, while radical surgery is necessary for patients with LNM.
Several reports have focused on the risk factors of LNM and indicated that the lymph node status was an important prognostic factor [6,9,[14][15][16]. In addition, the indications for AC in T1N + M0 gastric cancer is T1N2-3bM0, according to the latest Japanese gastric cancer treatment guidelines 2018 (5th edition) [17].
Whether AC is really unnecessary for T1N1M0 gastric cancer remains to be decided, but few studies have focused on this topic. Therefore, we investigated the incidence and risk factors of LNM, and evaluate the survival of T1N + M0 gastric cancer using the SEER database.
In the present study, we found the LNM rate for patients with T1N + M0 gastric caner is 19.5%, which is comparable to previously reported 4%-24% [14,[18][19][20]. Our study also indicated that age, tumor size, invasion depth, type of differentiation and number of retrieved LNs were independent risk factors for LNM in EGC. Depth of invasion is the most important risk factors for LNM. Previous studies have reported the relationship of age and LNM. Higher risk for LNM was more indenti ed in young patients [21], and a lower risk in old patients [22]. However, other studies reported that LNM was not associated with age [23]. In the present study, old age (≥ 85) had a lower risk for LNM. Many previous reports have also con rmed that undifferentiated type is more aggressive [24], which was con rmed in our study.
In our study, the tumor size and depth of invasion are independent risk factors for LNM, which was consistent with previous studies [25]. Another important risk factor for LNM is the presence of lymphovascular invasion. However, among all these risk factors, depth of invasion might have the greatest impact.
In Oncology [17,26]. However, AC are not recommended for patients with pT1N0M0 and pT1N1M0 according to the latest Japanese gastric cancer treatment guidelines 2018 (5th edition) [11]. Our study revealed that there were statistically signi cant differences in OS and CSS between pN0/N + or pN0-1/N2-3b categories (p 0.001), but an overlap in survival curves was found between categories pN0-1/N2-3b. The proportion of patients in pN0/N + was better distributed and pN0/N + category showed improved prognostic performance in predicting OS and CSS. Moreover, multivariate analyses showed AC was associated with CSS. Therefore, pT1N1M0 may be appropriate candidate for AC, other studied have also showed these patients [27][28][29].
There are some limitations to our study. First, the retrospective study is based on SEER database, which represents only 28% of the U.S. population and lacks data on many medical details. Second, we just investigated the major risk factors in terms of LNM and prognosis, suggesting modi cation of the de nition with tumor size, invasion depth and LNM. Further clinical studies are warranted to investigate more de nitive parameters. The biological behavior and molecular mechanism of LNM in gastric cancer needs to be clari ed. An updated de nition, which combines macroscopic types, pathological morphology and molecular classi cation, may be useful to make appropriate treatment decisions and follow-up plans. Furthermore, a uni ed and standardized de nition makes different studies comparable.

Conclusions
In Authors' contributions PJ and YL contributed equally to this work and they were involved in study concept, data acquisition, analysis, and interpretation, and production of tables, wrote the rst draft, and revised it critically in light of comments from other authors; YTT was involved in study conception and design, data interpretation, manuscript revision, and discussion; SM, WZK, HL, and FHM were involved in data acquisition and literature review; HTH and WKL were involved in the manuscript revision and discussion.

Figure 1
Kaplan-Meier survival analysis showed that there were statistically signi cant differences in OS and CSS between patients with and without chemotherapy in pT1N1M0 patients (p 0.05).

Figure 2
CSS analysis of patients with T1N+M0 gastric cancer according to different lymph node status There were statistically signi cant differences in CSS between pN0/N+ or pN0-1/N2-3b categories (p 0.001), but an overlap in survival curves was found between categories pN0-1/N2-3b.

Figure 4
Flowchart of T1M0 gastric cancer patients included process