AIM: To compare the prognostic assessment of lymph node ratio and absolute number based staging system for gastric cancer after D2 resection.

METHODS: The clinical, pathologic, and long-term follow-up data of 427 patients with gastric cancer that underwent D2 curative gastrectomy were retrospectively analyzed. The relationships between the metastatic lymph node ratio (MLR), log odds of positive lymph nodes (LODDS), and positive lymph nodes (pN) staging methods and the long-term prognoses of the patients were compared. In addition, the survival curves, accuracy, and homogeneity were compared with stratification to evaluate the prognostic assessment of the 3 methods when the number of tested lymph nodes was insufficient (< 10 and 10-15).

RESULTS: MLR [hazard ratio (HR) = 1.401, P = 0.012], LODDS (HR = 1.012, P = 0.034), and pN (HR = 1.376, P = 0.005) were independent risk factors for gastric cancer patients. The receiver operating characteristic (ROC) curves showed that the prognostic accuracy of the 3 methods was comparable (P > 0.05). Spearman correlation analysis confirmed that MLR, LODDS, and pN were all positively correlated with the total number of tested lymph nodes. When the number of tested lymph node was < 10, the value of survival curves staged by MLR and LODDS was superior to those of pN staging. However, the difference in survival curves between adjacent stages was not significant. In addition, the survival rate of stage 4 patients using the MLR and LODDS staging methods was 26.7% and 27.3% with < 10 lymph node, respectively which were significantly higher than the survival rate of patients with > 15 tested lymph nodes (< 4%). The ROC curve showed that the accuracy of the prognostic assessment of MLR, LODDS, and pN staging methods was comparable (P > 0.05), and the area under the ROC curve of all 3 methods were increased progressively with the enhanced levels of examined lymph nodes. In addition, the homogeneity of the 3 methods in patients with ≤ 15 tested lymph nodes also showed no significant difference.

CONCLUSION: Neither MLR or LODDS could reduce the staging bias. A sufficient number of tested lymph nodes is key to ensure an accurate prognosis for patients underwent D2 radical gastrectomy.

Gastric carcinoma is one of the most common cancers in many Asian countries including South Korea and Japan, and the second most common cause of cancer-related death worldwide[1]. There are nearly 470 000 newly diagnosed cases every year in China. Of these cases, approximately 75% of the patients will die, making gastric cancer the third leading cause of cancer deaths[2]. Because of its long-term efficacy, D2 radical gastrectomy has been accepted in most countries, including those in the Europe and the United States, as the standard surgery for gastric cancer[3-5]. The pathological staging of gastric cancer after D2 radical gastrectomy is not only closely related to the long-term survival of patients but is also the main basis to guide subsequent adjuvant therapy. In the currently accepted criteria of postoperative tumor-node-metastasis (TNM) staging of gastric cancer, the staging of regional lymph node metastasis (N) is of great significance. This staging is currently controversial and changes frequently. In both the latest 7^th edition of the American Joint Cancer Committee (AJCC)[6] and the 14^th edition of the Statute of Gastric Cancer Treatment in Japan[7] in 2010, the absolute number of positive lymph nodes (pN) in the perigastric region was used as the staging basis, and the staging criteria for each stage were unified. Meanwhile, many studies have supported the N staging by computing the metastatic lymph node ratio. Currently, there are 2 main methods in the staging of relative number of positive lymph nodes, the metastatic lymph node ratio (MLR)[8] and the log odds of positive lymph nodes (LODDS)[9]. The former calculates the ratio of the number of pN over the total number of the tested lymph nodes, while the latter calculates the log value, log[(pnod + 0.5)/(tnod - pnod + 0.5)], of the ratio between positive and negative lymph nodes. Previous studies have shown that, especially when the number of the tested lymph nodes was insufficient, the staging of MLR and LODDS could more accurately assess the prognosis of patients with gastric cancer than staging using the absolute value (pN)[9-13]. However, a unified standard of specific staging for relative number of positive lymph nodes is not currently available, and whether this ratio is superior to the pN staging is also unknown[14,15]. Therefore, the clinical data and long-term follow-up results of the gastric cancer patients that received D2 radical gastrectomy were retrospectively analyzed in this study, and the values of the above staging methods for regional lymph node metastasis in assessing patient prognosis were compared.

Due to the decent long-term survival rate, surgical resection, represented by standard D2 radical gastrectomy, is currently the preferred treatment for gastric cancer. However, in recent years, with the rise of the concept of individualized treatment and the application of new adjunct treatment in clinical practice, an accurate prognostic assessment of patients with gastric cancer after surgery is essential to the development of relevant follow-up treatment strategies[17,18]. Currently, the postoperative pathological TNM staging is accepted and widely applied as the prognostic evaluation indicator in clinical practice. With regard to the N portion of the TNM staging, there has been considerable controversy ranging from earlier staging based on anatomical sites of metastatic lymph nodes[19] to the specific staging criteria based on the number of regional metastatic lymph nodes[20,21]. The N staging criteria were not unified until the 7^th edition of the AJCC[4] and the 14^th edition of the Statute of Gastric Cancer Treatment in Japan[5] unified the criteria for the first time in 2010. However, many researchers still believe that when the staging is based on the absolute number of metastatic lymph nodes, the number of pN is easily influenced by the numbers of removed and tested lymph nodes. When the number of tested lymph nodes is insufficient, staging bias may occur, affecting the accuracy of the prognostic assessment[22,23]. The N staging based on MLR can overcome the above shortcomings[24,25]. Therefore, when comparing the prognostic assessment of different lymph node metastasis staging methods in gastric cancer patients after D2 radical gastrectomy, this study focused on the impact of the 3 staging methods on long-term survival rate when the number of pathologically tested lymph nodes after surgery was insufficient.

To date, neither MLR nor LODDS staging has accurate and widely accepted criteria; therefore, the log-rank survival test was first conducted to verify the staging criteria of MLR and LODDS (Table 2). The 5-year survival rates of various stages according to the above criteria were similar to those of the corresponding pN stages (TNM staging criteria in the 7^th edition of AJCC/UICC). The correlation analysis of the 3 staging methods also showed that MLR and LODDS were significantly positively correlated with pN. The ROC curves also showed that the accuracy of prognosis assessment of the 3 staging methods in gastric cancer patients was not significantly different. The subsequent univariate and multivariate analyses both showed that the MLR, LODDS, and pN staging methods were all closely related to patient prognosis-they were all independent risk factors for the prognoses of gastric cancer patients. The above results suggest that MLR, LODDS, and pN staging methods can all be used for the prognostic assessment of gastric cancer, and the assessment efficacies of the 3 methods were similar.

Although the total number of tested lymph nodes in the Cox proportional risk regression model was not a significant independent risk factor for patient prognosis, univariate analysis showed that as the number of tested lymph nodes increased, the 5-year survival rate of patients exhibited a downward trend (P = 0.039); moreover, a correlation analysis showed that MLR, LODDS, and pN were all positively correlated with the number of tested lymph nodes. When only the correlation coefficient of the number of tested lymph nodes was considered (pN > MLR > LODDS), the impact of the number of tested lymph nodes on the MLR and LODDS was smaller than that of the absolute number of pN, which suggests that compared with pN staging system, the MLR and LODDS were less affected by the total mumber of tested lymph nodes. The subsequent results of the survival curve of patients with insufficient tested lymph nodes also showed that, when the number of tested lymph node was < 10, the MLR and LODDS staging methods appeared to better assess prognosis than pN staging. However, as shown in Figure 2, although the MLR and LODDS staging methods could more accurately assess the 5-year postoperative survival rate of gastric cancer patients at the early-middle stages (stage 0-2), the difference in the prognostic assessment of the patients at middle-late stages (stage 3 and 4) was not significant. The main reason was that, although the ratio could reduce the impact of sampling error compared with the absolute number, as the number of pN increased, the impact of the sampling error due to the insufficient number of tested lymph nodes also increased. Therefore, the difference in the prognoses of patients in the middle-late stages was not significant. In addition, for the patients in the middle-late stages of MLR and LODDS, especially those in stage 4, the 5-year survival rates were all ≥ 20%, which was significantly higher than that of the patients with > 15 (< 4%) tested lymph nodes. The reason for this result may be that when the total number of tested lymph nodes was insufficient, the sampling sites were too concentrated near the lesion; therefore, the ratio of pN was higher, resulting in an overestimation of the actual pathological staging of patients. Moreover, the comparison of the 5-year survival rate of the 3 methods in patients with ≤ 15 tested lymph nodes also confirmed that the accuracy and homogeneity of the staging methods based on the MLR, LODDS or the absolute number of pN were similar, with no significant difference. At the same time, the comparison of the survival curve and ROC curve of the 3 staging methods in the < 10, 10-15 and > 15 group showed that the difference in the 5-year survival rate between stages and the assessment accuracy of survival rate were all increased progressively with the enhanced levels of examined lymph nodes, and the 3 staging methods exhibited no significant difference. The above results all confirmed that, regardless of the staging method, a sufficient number of tested lymph nodes was the key factor. When the number of tested lymph nodes was ≤ 15, the staging based on MLR or LODDS could not compensate for the inadequacy of pN staging, and thus could not accurately assess patient prognosis.

Although the number of cases did not affect the results of the statistical analysis significantly, it could be observed from the survival curve that the LODDS staging method appeared to better assess prognosis for patients at MLR and pN stage 0 with an insufficient number of tested lymph nodes. However, the advantage of the LODDS staging method was only apparent when the number of tested lymph nodes was < 10. Additionally, because the 5-year survival rate for patients in stage pN0 was relatively high, the survival rates of patients in various LODDS stages were not significantly different after re-staging. Finally, the calculation method of LODDS was complicated. All of the above factors limited the practical application of LODDS staging method, and the practical value was low.

Some studies have found that the staging methods based on the MLR and LODDS could more accurately predict the prognosis of gastric cancer patients than the staging method based on the absolute number (pN), especially when the number of tested lymph nodes was insufficient[9-13]. The above conclusions in this study appeared to be inconsistent with those previous findings. Different surgical methods may be the main cause of the contradictory findings[26]. In a study recently published in Annals of Surgery in 2012[13], the postoperative clinical, pathologic and follow-up data of 18 043 gastric cancer patients retrieved from the Surveillance, Epidemiology, and End Results database of United States were retrospectively analyzed, and the results showed that when the number of tested lymph nodes was insufficient, the MLR staging method could more accurately assess the patient prognosis than the pN staging method. However, only 10% of the patients in this study underwent D2 radical gastrectomy, and the scope of lymph node removal in the remaining patients was D1 or below. The insufficient number of tested lymph nodes was mainly limited by the scope of lymph node removal. Some studies have confirmed that the average number of removed lymph nodes during D2 radical gastrectomy could reach 32[27]. The smaller the number of tested lymph nodes, the greater the sampling error. A sufficient number of tested lymph nodes is key to reducing sampling error. Therefore, to accurately assess the prognosis of patients after D2 radical gastrectomy, no staging method can replace a sufficient number of tested lymph nodes.

In summary, the MLR, LODDS and pN are all independent risk factors for the long-term postoperative survival of gastric cancer patients. The accuracy of the prognostic assessment of the MLR and LODDS staging methods is comparable to that of the pN staging method in gastric cancer patients. However, for the patients that undergo a D2 radical gastrectomy, when the number of tested lymph nodes is insufficient (≤ 15), neither the staging method based on metastatic lymph node ratio nor the pN staging method can avoid staging bias. Therefore, as D2 radical gastrectomy is increasingly accepted, a sufficient number of tested lymph nodes is the only key to ensure an accurate prediction of gastric cancer patient prognosis.