Positive lymph node ratio is an index in predicting prognosis for remnant gastric cancer with insufficient retrieved lymph node in R0 resection

The staging system of remnant gastric cancer (RGC) has not yet been established, with the current staging being based on the guidelines for primary gastric cancer. Often, surgeries for RGC fail to achieve the > 15 lymph nodes needed for TNM staging. Compared with the pN staging system, lymph node ratio (NR) may be more accurate for RGC staging and prognosis prediction. We retrospectively analyzed the data of 208 patients who underwent R0 gastrectomy with curative intent and who have ≤ 15 retrieved lymph nodes (RLNs) for RGC between 2000 and 2014. The patients were divided into four groups on the basis of the NR cutoffs: rN0: 0; rN1: > 0 and ≤ 1/6; rN2: > 1/6 and ≤ 1/2; and rN3: > 1/2. The 5-year overall survival (OS) rates for rN0, rN1, rN2, and rN3 were 84.3%, 64.7%, 31.5%, and 12.7%, respectively. Multivariable analyses revealed that tumor size (p = 0.005), lymphovascular invasion (p = 0.023), and NR (p < 0.001), but not pN stage (p = 0.682), were independent factors for OS. When the RLN count is ≤ 15, the NR is superior to pN as an important and independent prognostic index of RGC, thus predicting the prognosis of RGC patients more accurately.

www.nature.com/scientificreports/ gery being total gastrectomy with Roux-en-Y anastomosis. None of the patients received neoadjuvant chemotherapy. All of the study participants were followed up through telephone or written communications until death or end of the study period (December 31, 2015). We declared that the informed consent from patients has been obtained prior to our study. As some patients invovled in our study died during the follow up, we obtained the informed consent from their guardians. The median and mean follow-up periods were 60 and 83.15 months (range 1-346 months), respectively. In total, 11 patients were lost to follow-up, and 17 died during the postoperative period. Finally, 208 patients were included in the analysis. Postoperative chemotherapy regimens included 5-fluorouracil (5-FU) plus cisplatin, 5-FU plus mitomycin and epirubicin, 5-FU plus leucovorin and cisplatin, 5-FU plus cisplatin and epirubicin, and cisplatin plus oxaliplatin.
Statistical analyses. All statistical analyses were performed using SPSS software version 20.0. We used the Kaplan-Meier method to determine overall survival (OS) and the log-rank test for comparison. Cox's proportional hazard model (forward-stepwise method) was used for the multivariate analysis to determine the independent factors for prognosis prediction. We divided the multivariate analysis into two steps. In step 1, we analyzed all significantly important prognostic factors using multivariate analysis, with the exception of the revised NR stage. In step 2, we performed multivariate analysis with consideration of all significantly important prognostic factors, including the revised NR stage. Two-tailed p-values and differences were considered significant at p < 0.05.

Multivariate analyses.
To determine which is more accurate between pN stage and NR stage as a prognostic index for the OS of RGC patients, we conducted a two-step multivariate analysis of prognosis. In step 1, we performed a multivariate analysis of all the important prognostic factors (tumor size, Borrmann type, histology, Lauren classification, lymphovascular invasion, pT stage, pN stage, postoperative adjuvant therapy), except for NR staging. We found that tumor size (p = 0.007), lymphovascular invasion (p = 0.002), and pN stage (p < 0.001) were independent factors predicting prognosis (Table 2). However, NR stage was superior to pN stage (UICC/ AJCC) as an independent prognostic indicator ( Table 2), as determined in step 2 of the analysis. Moreover, pT stage was remarkably not an independent prognostic index for RGC patients, according to our analysis.

Migration.
We further calculated the migration among the different stages according to the new NR stages.
We found that 25 and 5 cases of pN1 were re-divided into rN2 and rN3, and 33 cases of pN2 were re-divided into rN3, respectively (Table 3A). Migration of the corresponding modified TNM stage using NR (mTNM-NR) staging was also observed, as shown in Table 3B.
As shown in Fig. 1, there was a significant difference in prognosis between pN and NR stages among the RGC patients. According to the analysis, there was no significant difference between pN2 and pN3 (p = 0.511), but there was a significant difference between rN2 and rN3 (p = 0.034) in their predictive capability for the prognosis of RGC patients. In addition, we presented the Kaplan-Meier curves based on the TNM staging (UICC/AJCC 8th edition) and mTNM-NR staging system as shown in Fig. 2 www.nature.com/scientificreports/ Correlations analysis. We further analyzed the correlation between NR and other clinicopathologic features. As shown in Table 4, there was significant positive correlation between NR and the following parameters: tumor size (p = 0.000), Borrmann type (p = 0.000), macroscopic serosal classification (p = 0.002), lauren's Table 1. Descriptive data of the patients involved in this cohort (n = 208). UICC union for international cancer control, AJCC American joint committee on cancer, NR lymph node ratio. Characteristic of lymph nodes metastasis. The total number of lymph nodes retrieved was 1152, of which 426 were metastases. We present the distribution and basic statistics of the total retrieved LN numbers with metastatic LN numbers in Table 5.

Discussion
While RGC has been extensively studied, its staging system is yet to be established. Therefore, RGC staging generally follows the TNM staging of primary GC 1,7 . In the 8th UICC/AJCC TNM staging system, the amount of metastatic LNs is used to determine the LN stage, and the number of RLNs must be > 15 to arrive at a more accurate LN staging. However, this appears to be inappropriate for RGC where RLNs can be < 16, particularly in the case of radical lymphadenectomy for initial malignant disease 8 . Previous studies have shown that LN metastasis is an independent risk factor in RGC local recurrence. Moreover, complete resection and LN dissection are key to successful radical gastrectomy for RGC. Thus, appropriate LN staging can accurately indicate patient prognosis, guiding further treatment. By contrast, inadequate LN dissection can lead to residual cancer cells, which results in a higher recurrence rate after surgery. Ultimately, it remains unclear whether the 8th UICC/ AJCC pN staging can accurately assess the prognosis of RGC patients.    Table 5. Characteristic of lymph nodes metastasis.

Regional lymph node Retrieved LNs Metastatic LNs Rate of LN metastasis (%)
No. www.nature.com/scientificreports/ NR is the ratio of the amount of positive LN to that of the RLN. Despite the same number of metastases, cases with higher RLN and thus lower NR have better prognosis 9 . Due to the common occurrence of low RLN cases during RGC resection, we hypothesized that NR was more accurate than the present pN stage, which was confirmed by our results. The NR classification system stratifies each stage better; in our analysis, we used a multiple Cox regression model to select the NR stage as a substitute for the pN stage. In addition, lymphovascular invasion and tumor size were also proven to be independent risk indexes for prognosis.
The NR system has obvious advantages and attractive prospects for clinical application in RGC. Dapeng et al. showed that NR plays a role as an independent prognostic factor after radical operation for GC 10 . Lee et al. also suggested that among patients with stage IIIA GC, the prognosis differed between those with RLN ≤ 15 and > 15 11 . Ichikura's study found that the amount of positive LNs is closely related to the RLN 12 . Both a low and exceedingly high RLN may lead to inaccurate staging, leading to inaccurate assessments of patient prognosis. Therefore, some researchers suggested using the LN metastasis rate for LN staging as it can reflect both the amount of dissected LN and positive LN 13 . On the other hand, there is no consensus on the prognostic value of NR in RGC patients.
Lu et al. reported that the tumor size may be more hierarchical than the pN (7th UICC) stage system in RGC 14 . Similarly, we found that tumor size is an independent risk factor of prognosis. RGC is complicated to manage because of its diversity, and the initial operation significantly influences the second operation. Most researchers stipulate that initial gastrectomy alters the lymph flow from the remnant stomach and LN metastasis due to reconstruction and adhesion, leading to the occurrence of RGC 15,16 . Several authors have reported that the risk of RGC was higher after Billroth II reconstruction. Gandolfi et al. found that Billroth II was the most commonly used approach (95%) for reconstruction during the initial surgery of RGC patients 17 . With respect to the type of surgical anastomosis, it was reported that compared with Billroth I, Billroth II surgery increased the risk of RGC by four times 18 . Xiang et al. found that RGC mostly occurred at the anastomotic sites, and that LN metastasis of the jejunal mesentery and mesenteric root was universal, particularly in patients who underwent Billroth II anastomoses 16 . In addition, our results demonstrate that T stage was not an independent prognostic factor for patients with RGC. We speculated that RCG is characterized by alterations in the normal anatomic lymphatic flow of the stomach and its association with the surrounding tissue. Adhesion after primary operation may influence the progression of cancer and the infiltration of surrounding organs, especially for RGC at the anastomosis site. In our study, we enrolled 208 RGC cases, there were mainly cases with the depth of invasion at T2 (n = 35, 16.8%), T3 (n = 103, 49.5%) and T4 (n = 62, 29.8%). Only 8 cases were T1 stage (3.9%). Alberto et al. 19 reported that RGC with pT1 stage had lower rate of lymph node metastasis. Once the depth of invasion was T2 or above, the rate of LN metastasis increased significantly. The lymph node dissection for RGC should be extended beyond the initial operation. A high incidence of LN metastasis in the No. 10-14, 20, 110, 111 was found in the patients who had previously undergone gastric reconstruction. All the regional lymph nodes around the stomach that were not cleared in initial operation should be dissected. The LN metastasis rate was 42.9%, 34.2%, 15.9%, 29.2% and 37.2% in No. 10, 11, 12, 14 and 20 in our study. Therefore, it is not at all surprised that N categories replaced T categories being an independent risk factor for prognosis of RGC. In addition, the small sample size may contribute this result. We also observed LN metastasis of the jejunal mesentery during RGC surgeries (n = 2). However, this needs to be further validated and analyzed in future research.
In contrast to our findings, some studies reported that only pT and pN are the prognostic factors of RGC. Nakagawa et al. performed a similar study but obtained a contrasting conclusion, stating that NR was not better to pN (7th UICC) 4 . The contrasting findings may be attributed to the different NR cut-off points and the eligibility criteria between the current study and that by Nakagawa et al. In Nakagawa's study, the NR cut-offs were 0; > 0, ≤ 0.1; > 0.1, ≤ 0.4; and < 0.4. When Marchet analyzed the clinical data of 1853 patients with gastric cancer, NR was divided into 0%, 1-9%, 10-25%, and > 25% 20 . Satio et al. divided NR stages into 0%, > 5%, 5-10%, 10-20%, 20-30%, and > 30%, and found that NR was an independent prognostic factor in patients with gastric cancer; however, the number of positive LNs was not an independent factor 21 . Bando et al. 22 studied 650 GC patients who had > 15 LN dissections after operation. Their NR staging was divided into 0%, < 10%, 10-25%, and > 25%, and they found that NR staging was more effective than the JGCA LN staging. Moreover, Sun et al. identified the following best-fit cut-off values: rN0: 0%; rN1: 1-20%; rN2: 21-50% and rN3: > 50%, proving the advantage of NR stage in minimizing stage migration for GC patients with insufficient number or level of retrieved LNs 23 . Furthermore, few scholars have elaborated on the method of dividing NR staging nodes. There is no international consensus; hence, further research is needed to determine the appropriate staging nodes. In our study, we identified the NR stages as rN0: 0; rN1: > 0 and ≤ 1/6; rN2: > 1/6 and ≤ 1/2; and rN3: > 1/2.
Moreover, a significant portion of patients included in our study underwent Billroth I and Roux-Y during the initial surgical reconstruction (n = 41, 21%). However, we only focused on those who underwent Billroth II. In addition, we included cases with RLN ≤ 15, and excluded the very few cases with RLN > 16. By contrast, Nakagawa et al. included both cases of RLN ≤ 15 and > 15. One study showed that if the number of RLN after radical gastrectomy was > 15, then the prognostic value of NR for patients with GC was worse than that of N stage 24 , supporting our findings that NR stage can predict the prognosis of RGC patients more accurately than pN stage.
Our study has some limitations. First, it spanned a relatively long period of 15 years. Continuous improvements in surgical instrumentation and chemotherapy regimens may have affected the outcomes. We also evaluated a specific sample population, and thus, our findings need to be further validated in other types of RGC patients and cohorts, particularly in the Western population. Because the number of cases with RLN > 15 is too small to be representative, our study included only samples with ≤ 15 RLN. This may limit the possibility of the NR staging system being developed as a standard staging system in clinical practice, and our findings should hence be verified by further big data studies. We suggest that for RGC with RLN > 15, the original pN stage (8th edition) can be used. www.nature.com/scientificreports/ In conclusion, the metastatic LN ratio is an independent prognostic index for patients with RGC. When the amount of RLN is ≤ 15, the NR stage can more accurately predict the prognosis of patients with RGC than the pN stage, facilitating the selection of better treatment options.

Data availability
At present, the datasets generated during and analyzed during the current study are being used in another research; hence, these are not publicly available but could be available from the corresponding author on reasonable request.