Establishment of Lymph Node Metastasis Prediction Model For T1a Gastric Cancer

Background and Aims: Endoscopic resection has been widely used in the treatment of early gastric cancer recent tears. For patients undergoing endoscopic resection, lymph node metastasis is an important prognostic factor. Thus, we built a model to predict the likelihood of lymph node metastasis in early gastric cancer. Methods: Data of 789 gastric cancer patients were obtained from the Surveillance, Epidemiology, and End Results (SEER) database. We picked several variables and chi-square analysis was used to come up with statistically signicant factors. Then, we used these factors to build a nomogram predicting the lymph node metastasis risk in T1a gastric cancer. Finally, we evaluated this nomogram using C-index, ROC curve and DCA. Results: We built a nomogram based on four factors. The C-index of this nomogram was 0.750. The AUC value of ROC curve for this nomogram was 0.750. DCA showed that it would be benecial to use this nomogram if the threshold probability was between 0.01 and 0.41. Conclusion: This study built a nomogram to predicted the lymph node metastasis risk in T1a gastric cancer based on four factors. Upon evaluation, this nomogram has a good ability of predicting the risk of lymph node metastasis.


Introduction
Gastric cancer is a malignant tumor originate from gastric epithelium. It is the sixth most common cancer and its mortality ranks second among all cancers globally (1). East Asia, including China, Japan and Korea, has a high incidence of gastric cancer (1).
Murakami rstly de ned early gastric cancer (EGC) as the tumor limited to mucosa and submucosa of the stomach, regardless of tumor size or lymph node metastasis (2)(3). However, this de nition has been controversial (4, 5). Saragoni suggested that gastric cancer with lesions that are con ned to the mucosa or invade minimally the submucosa and without lymph node metastasis was de ned as EGC (5). Such staging allows for more appropriate treatment, especially endoscopic resection in gastric cancer.
Generally speaking, a standard gastrectomy, including the resection of more than two thirds of stomach and a D2 lymph node dissection, is the major treatment method for EGC and advanced gastric cancer (AGC) (2). However, gastrectomy is always related with long hospital stay and high complication rate (6).
This can also lead to poor quality of life for patients (7). Compared with gastrectomy, endoscopic resection, including endoscopic mucosal resection (EMR) and endoscopic submucosal dissection (ESD), had no signi cant difference in overall survival and fewer complications (8,9). Among them, ESD has better e cacy, but it is di cult to operate (10)(11)(12). Endoscopic resection has been widely used in the treatment of EGC in many areas (13,14). Japanese gastric cancer treatment guidelines suggested that, for those gastric cancers with a low possibility of lymph node metastasis, EMR or ESD would be a better choice (15). In this guideline, the absolute indication for EMR/ESD is "A differentiated-type adenocarcinoma without ulcerative ndings [UL (−)], of which the depth of invasion is clinically diagnosed as T1a and the diameter is ≤ 2 cm". Besides, there are three expanded indications for ESD: Tumors clinically diagnosed as T1a and (1) Of differentiated-type, UL (−), but > 2 cm in diameter. (2) Of differentiated-type, UL (+), and ≤ 3 cm in diameter. (3) Of undifferentiated-type, UL (−), and ≤ 2 cm in diameter. So far, endoscopic resection has been widely performed on the patients with EGC worldwide.
For patients with EGC who undergo EMR/ESD, lymph node metastasis is an important prognostic factor (16)(17)(18). So, before performing EMR/ESD, we should make sure that the patient has a low possibility of lymph node metastasis. According to Japanese gastric cancer treatment guidelines, the object of our study was con ned in patients with T1a gastric cancer. In this study, using the Surveillance, Epidemiology, and End Results (SEER) database, we screened several important factors (Sex, Race, Age, Primary site, Tumor size, Grade and Histologic type) and used these factors to build a model to predict the likelihood of lymph node metastasis in EGC.

Material And Methods
Data on gastric cancer patients were collected from SEER database (http://seer.cancer.gov/), including the information of sex, age, race, primary site, tumor size, grade, 7th Edition T stage of The American Joint Committee on Cancer (AJCC), histologic type and regional nodes positive. Selection criteria: 1) All data have detailed 7th Edition T stage classi ed by AJCC and are clinically diagnosed as T1a; 2) All data have detailed ICD-O-3 histologic type; 3) All data have lymph node information. Exclusion criteria 1): unknown age; 2) unknown sex; 3) unknown race; 4) unknown histologic type; 5) unknown tumor size; 6) unknown grade; 7) unknown detailed primary site; 8) no lymph node resection.
Race was classi ed into 3 groups: black, white and others. Primary site was classi ed into 4 groups: upper 1/3, middle 1/3, lower 1/3 and others; Tumor size was classi ed into 4 groups: 1, 1-2, 2-3 and 3; Grade was classi ed into 3 groups: Well differentiated, moderately differentiated and poorly differentiated. Histologic type was classi ed into 4 groups: adenocarcinoma, intestinal type adenocarcinoma, signet ring cell carcinoma and others.
Chi-square test was used to evaluate the relationship between lymph node metastasis and the above variables (sex, race, age, primary site, tumor size, grade and histologic type) and P-value < 0.05 was viewed to have statistical signi cance. Multivariable logistic regression analysis was performed to construct a lymph node metastasis prediction model including the variables with a P-value < 0.05. Then, a nomogram based on the result of logistic regression analysis was constructed. After that, Harrell's concordance index (C-index), Calibration curve and Receiver Operating Characteristic (ROC) curve was used to evaluate the accuracy of the nomogram for lymph node metastasis prediction. Finally, Decision Curve Analysis (DCA) was performed to assess the clinical utility of this nomogram.
Besides, aimed at the expanded indications mentioned in the Japanese gastric cancer treatment guidelines, chi-square analysis was performed to explore whether diameter affects the rate of lymph node metastasis in differentiated gastric cancers and whether differentiation affects the rate of lymph node metastasis in gastric cancers no more than 2 in diameter. [ Figure 1]

Patients Characteristics
A total of 164269 cases of gastric cancer were obtained from SEER database. After screening according to the selection criteria and removing the non-conforming cases according to the exclusion criteria, 789 cases were remained [ Table 1]. There were 80 positive cases of lymph node metastasis and 709 negative cases of lymph node metastasis.

Establishment of lymph node metastasis prediction model
The result of chi-square analysis showed race, tumor size, grade and histologic type were signi cantly related to the risk of lymph node metastasis [ Table 1]. These four factors were selected to perform multivariable logistic regression analysis [ Table 2] and draw Nomogram [ Figure 2].

Accuracy test of Nomogram
The calibration curve of lymph node metastasis rate demonstrated a good agreement between our nomogram and the actual situation [ Figure 3]. The C-index of our nomogram was 0.750(95CI, 0.695 to 0.805). The ROC curve of nomogram was showed in Figure 4 and the Area Under Curve (AUC) values was 0.750. The result of DCA was showed in Figure 5, showing that it would be bene cial to use this nomogram if the threshold probability was between 0.01 and 0.41.

Evaluation of expanded indications
for differentiated intramucosal gastric cancer, the difference in lymph node metastasis rate between diameter 2cm group and diameter≤2cm group had statistical signi cance. For intramucosal gastric cancer with diameters no more than 2cm, the difference in lymph node metastasis rate between differentiation group and undifferentiation group didn't have statistical signi cance [ Table 3].

Disscusion
In the past, a standard gastrectomy, including the resection of more than two thirds of stomach and a D2 lymph node dissection, is the major treatment method for EGC and AGC. But, with the development of endoscopic technique, endoscopic resection, including EMR and ESD, has been widely used in the patients of EGC. Recently, researches have showed that in EGC, there was no signi cant difference in overall survival between the patients underwent radical gastrectomy and endoscopic resection, although the latter had a higher recurrence rate (8) (9). And the complication rate of endoscopic resection is lower than that of radical gastrectomy (8).
For those EGC patients who underwent endoscopic resection, lymph node metastasis has a signi cant adverse effect on the prognosis. So, endoscopic resection was suggested to be used in the EGC with a low possibility of lymph node metastasis. In our screened data, the lymph node metastasis rate of T1a gastric cancer is 10.1%, while in all T1a gastric cancer patients in SEER database, this rate is 9.7%. Choi (19) reported that the lymph node metastasis rate of T1a gastric cancer in western population is 7.8%, which is relatively close to our data. But Gotoda (20) suggested the lymph node metastasis rate of intramucosal cancers in the Japanese people is 2.2%. This could be because of ethnic differences, but it could also be because of other factors and all residents over 40 years old in Japan will be screened for gastric cancer (21). Therefore, they can always detect gastric cancer at a very early stage, which may be one of the reasons for its low lymph node metastasis rate. However, even after careful examination, it is still possible to misjudge whether there is lymph node metastasis. A method to predict the rate of lymph node metastasis is needed clinically.
Previous studies have shown that for EGC, age, tumor size, tumor location, ulceration, histological type, grade, macroscopic appearance (Such as ulcer, border, color) and depth of invasion are risk factors of lymph node metastasis (22)(23)(24)(25). Especially the depth of invasion is the most critical risk factors. Only patients with T1a gastric cancer, whose depth of invasion is limited to the mucosa are suitable for ESD/EMR treatment in Japanese gastric cancer treatment guidelines. Thus, the object of our study were the patients with T1a gastric cancer. In this study, we selected seven factors (Sex, Race, Age, Primary site, Tumor size, Grade and Histologic type) to study their relationship with lymph node metastasis in EGC.
After chi-square analysis, we found that for EGC, four factors (Race, Tumor size, Grade and Histologic type) were signi cantly related to the risk of lymph node metastasis. By logistic regression analysis, Race, Tumor size and Grade were considered to be the independent factors of lymph node metastasis. EGC with diameter 3cm or poor differentiation is prone to lymph node metastasis. In addition, blacks and whites are more likely than others to develop lymph node metastasis. Many studies had focused on the lymph node metastasis rate in EGC and its in uencing factors, but there is a lack of a suitable method to predict lymph node metastasis of EGC clinically.
Nomogram has been widely used in the prediction of various clinical events. This is the rst time a nomogram was used to predict lymph node metastasis in EGC. In this study, we construct a nomogram to directly predict the possibility of lymph node metastasis of EGC in T1a stage based on these four variables (Race, Tumor size, Grade and Histologic type) and the results of C-index, Calibration curve and ROC curve had revealed its great ability of prediction on lymph node metastasis. With this nomogram, we can calculate the score based on the level of the four variables, and then get the corresponding lymphatic metastasis rate according to the score in clinical practice. For patients with a low risk of lymph node metastasis, endoscopic resection might be a better treatment. In contrast, the patients with a high risk of lymph node metastasis are better candidates for standard gastrectomy or endoscopic resection with laparoscopic lymph node dissection (26).

Conclusion
Through the screening of SEER database, we obtained a large sample of patients with T1a gastric cancer.
After analyzing the data, we found that 10% of T1a gastric cancer patients had lymph node metastasis. If ESD/EMR was performed on these patients, the prognosis might may not be ideal. We designed a nomogram to predict the probability of lymph node metastasis to help patients choose surgical methods and decide whether to expand the operation after surgery. We believe that patients with high probability of lymph node metastasis still need standard gastrectomy or endoscopic resection with laparoscopic lymph node dissection.
Besides, we veri ed two of the expanded indications mentioned in the Japanese guidelines for gastric cancer treatment using the data obtained from SEER database. Through chi-square analysis, we found that for differentiated intramucosal gastric cancer, there is signi cant difference in lymph node metastasis rate between diameter 2cm group and diameter≤2cm group. But for intramucosal gastric cancer with diameters no more than 2cm, differentiation did not signi cantly affect the rate of lymphatic metastasis.
There are some limitations in our study. Due to the limited clinical information contained in the SEER database, some risk factors were not included in our study. However, the results of accuracy test show that nomogram based on the existing 4 factors has a good accuracy.
In conclusion, this study constructs a nomogram to help clinicians predict lymph node metastasis rates in EGC patients. Based on the calculated risk of lymph node metastasis, the clinician can choose the appropriate treatment.

Consent for publication
Not applicable.

Availability of supporting data
All data supporting the ndings in our research are available in SEER database (http://seer.cancer.gov/).

Competing interests
All authors declare no competing interests.     Nomogram of lymph node metastasis risk. Decision curve analysis on nomogram of lymph node metastasis risk