Prognostic Impact of the Signet Ring Cell Type in Node-Negative Gastric Cancer

Little is known regarding the prognostic impact of the signet ring cell (SRC) histotype on negative lymph nodes (LNs) in gastric cancer (GC). In this study, we aimed to investigate the differences between SRC and non-SRC GC patients without LN metastasis. The medical records of patients with GC who underwent gastrectomy at Sun Yat-Sen University Cancer Centre from 1996 to 2012 were reviewed to analyse the clinicopathologic characteristics associated with survival. A total of 480 cases of GC patients without LN metastasis were identified, which included 90 SRC GC patients and 390 non-SRC GC patients. Between the two groups, there were a host of significant differences in the American Joint Committee on Cancer, 7th edition (AJCC) stage. We found that SRC histology was correlated with a poor prognosis in terms of recurrence in node-negative GC patients and that SRC histologic analysis combined with AJCC staging maybe an effectual method for prediction of the recurrence rate. Additionally, we found that SRC GC presents a more dismal overall prognosis in patients with perineural or vascular invasion.

All characteristics in which differences were observed were included in our multivariable models. We also compared the characteristics of individuals comprising those excluded from this sample using a multivariable model (Supplemental Table 2). The outcomes indicated that there were no differences between the two groups based on gender distribution, age, or tumour location. Thus, we concluded that our analytic sample broadly represented the demographic and clinical presentation characteristics of all identified LN-negative GC cases, with the above caveats.
Survival (overall survival and relapse-free survival). Kaplan (Fig. 2). To determine whether the SRC type has other prognostic relevance, all patients positive and negative for perineural or vascular invasion with known clinical records and long-term follow-up were analysed. In all of these analyses, the results suggest that the SRC type had an obvious independent influence on prognosis in perineural (10.4 vs. 14.2 years; P = 0.005) and vascular (5.6 vs. 11.3 years; P = 0.037) invasion-positive disease (Fig. 3). Kaplan-Meier curves for relapse-free survival are shown in Fig. 4a. Of all 480 GC patients, 148 had a GC relapse. The prognosis of SRC GC, with median relapse-free survival of 5.3 years, was significantly worse compared with non-SRC adenocarcinomas, with median relapse-free survival of 7.0 years (P < 0.001). Moreover, the patterns of recurrence were different between the two groups. Patients with SRC GC had less frequent late recurrence but more frequent early recurrence (P = 0.032) (Supplemental Tables 3 and 4).  The results of the univariate analyses for factors associated with relapse-free survival in node-negative GC are summarized in Table 3. SRC histology (univariate Cox HR, 1.83; 95% CI, 1.23 to 2.70; P = 0.003), age (univariate Cox HR, 1.03; 95% CI, 1.01 to 1.04; P < 0.001), perineural invasion (univariate Cox HR, 1.85; 95% CI, 1.32 to 2.60; P < 0.001), vascular invasion (univariate Cox HR, 2.84; 95% CI, 1.91 to 4.22; P < 0.001), advanced tumour grade, and increasing AJCC and tumour stages were significantly associated with relapse-free survival. Moreover, the number of LNs (univariate Cox HR, 2.23; 95% CI, 0.98 to 5.07; P = 0.055) tended to be associated with relapse-free survival. In the current study, gender and tumour location were not associated with relapse-free survival.

Predictors of mortality and
To determine the independent prognostic factors associated with overall and relapse-free survival in patients with node-negative GC, probable factors in the multivariate analysis were input into a Cox proportional hazards model (Supplemental Tables 5 and 6). The age at diagnosis (multivariable Cox HR, 1.03; 95% CI, 1.01 to 1.04; P = 0.006), female sex (multivariable Cox HR, 1.65; 95% CI, 1.04 to 3.09; P = 0.022), perineural invasion (multivariable Cox HR, 1.64; 95% CI, 1.14 to 2.36; P = 0.008), and increasing AJCC and tumour stages (Supplemental Table 5) were independently associated with mortality. SRC histology (multivariable Cox HR, 1.13; 95% CI, 0.70 to 1.85; P = 0.612) and vascular invasion (univariate Cox HR, 1.32; 95% CI, 0.84 to 2.08; P = 0.230) were not found to be independent predictors of mortality. Interestingly, SRC histology (multivariable Cox HR, 2.37; 95% CI, 1.42 to 4.26; P = 0.033), age (multivariable Cox HR, 1.03; 95% CI, 1.02 to 1.05; P < 0.001), vascular invasion (multivariable Cox HR, 2.16; 95% CI, 1.41 to 3.30; P < 0.001) and increasing AJCC and tumour stages (Supplemental Table 6) were significantly associated with relapse-free survival. Therefore, we determined the receiver operating characteristic curves for the prediction of relapse-free survival in node-negative GC patients using the AJCC stage, SRC histology, or a combination of both ( Fig. 4b-d). The area under the curve (AUC) for the AJCC staging-based prediction (0.729) was higher than that for the SRC histology-based model (0.681), and the combination of both factors achieved the highest AUC value (0.784). Moreover, multivariate Cox regression survival analysis adjusting for SRC histology, age, AJCC stage 3, T 4 , and vascular invasion consistently validated strong correlations with shorter relapse-free survival in patients, as is also clearly shown in (Fig. 5).

Discussion
In this study, we examined the effect of SRC histology in LN-negative GC patients on survival and tumour presentation. Our study first demonstrated that SRC patients have worse recurrence and survival rates, yet SRC histology was not an independent prognostic factor for overall survival in LN-negative GC patients. Additionally, another important finding in this study strongly suggests that SRC histology is an independent prognostic factor associated with overall survival in patients with perineural or vascular invasion.
In recent decades, the subset of LN-negative patients who still present a poor prognosis has been the subject of considerable research and controversy [12][13][14] . Previously, we found that the number of LNs is an independent prognostic factor in LN-negative GC patients. Moreover, we demonstrated that patients with negative LNs who undergo D2 dissection should have at least 16 LNs retrieved for accurate staging 15 . A result consistent with this was also reported by Baiocchi et al. in a multicentre study 16 .
Currently, although understanding of the prognostic implications of SRC histology in GC is limited and somewhat conflicting, the majority of studies still show that SRC variants are frequently related to a more aggressive clinical course, especially in Eastern patients 17,18 . Therefore, a history of SRC histology might be another factor that contributes to the minority of LN-negative GC patients having poor overall survival.
The primary findings of our study indicate that SRC histology has a correlation with poor prognosis due to recurrence in node-negative GC patients and that SRC histologic analysis combined with AJCC staging maybe an effective method for predicting the recurrence rate of node-negative GC patients. To our knowledge, SRC tumours are a variant of diffuse carcinoma according to Lauren's recent classification 19,20 . To exclude the impact of diffuse carcinoma, we first compared the diffuse type with the non-diffuse type in our enrolled patients, and we subsequently found that the diffuse carcinoma type was not an independent prognostic variable in GC patients with negative LNs (Supplemental Table 7, Supplemental Figs 1 and 2). Furthermore, we found that SRC histology was associated with a poorer overall prognosis in patients with perineural and vascular invasion, which is consistent with the results of other studies on oesophageal cancer and advanced GC patients 21,22 . There were several potential reasons for our novel findings. First, SRC GC presents with a more aggressive microscopic appearance that is characterized by abundant intracytoplasmic mucin that pushes the nucleus to the periphery, similar to a signet ring 23 . Second, SRC tumours are considered to be biologically distinct from well-differentiated tumours, which themselves differ according to tumour location and the degree of differentiation 24 . Third, patients with perineural and vascular invasion are usually diagnosed at advanced stages due to late clinical manifestations. The poorer prognosis of GC with SRC histology might be more related to the stage at diagnosis than to histology, but further evidence is needed. For these reasons, SRC GC patients may need more aggressive treatments, such as neoadjuvant or adjuvant chemotherapy, rather than surgery alone. A recent study proposed that preoperative chemotherapy followed by surgery might be a more appropriate treatment for SRC-containing adenocarcinomas of the oesophagogastric junction and stomach 25 .
The secondary finding of our study is that SRC histology was not independently associated with mortality compared with non-SRC histology in LN-negative GC patients. Overall survival did not differ between the two tumour types, though there was a distinct divergence in the overall survival Kaplan-Meier curves presenting before 10 years of follow-up. This finding suggested that overall survival favoured the non-SRC group and that a difference in survival might become evident if more patients are included in future studies. When adjusted for the AJCC stage, the analysis showed that patients with SRC GC had better survival than those with non-SRC GC (12.3 vs. 15.8 years) in AJCC stage 1. However, this increase in survival by 3.5 years was not statistically significant. In addition, the difference in survival between the two groups was not statistically significant in stage 2 or stage 3 GC patients. Several previous studies have reported improved survival with early stages of SRC GC compared with non-SRC GC and relatively worse survival in later stages of the disease 21,22 . Among all patients in the present study, 275 patients (70.5%) were stage 1 in the non-SRC group, and 48 patients (53.3%) were stage 1 in the SRC group; the submucosal invasion rates were 52.6% and 83.3%, respectively. Thus, most patients were not in an advanced GC stage, which might explain the rise in the unconformity trend. We did not completely analyse the later stages of GC because we excluded LN-positive patients. In addition, the different results in our study might indicate that the SRC subtype behaves differently in patients in China. Such differences in outcomes could reflect the aggressive screening and resection strategies employed in our country.
Interestingly, we also confirmed that SRC GC has a distinct presentation compared with non-SRC GC, presenting in patients at a more advanced stage. This is in accordance with previous studies conducted in Asian countries and other areas 25,26 . According to earlier publications, the two subtypes appear to present at different anatomic locations, with the SRC subtype being more likely to present in the body or the lower stomach 27 . This trend was also observed in the present study, but the difference was not statistically significant. It has been reported that SRC GC patients are younger in age and more likely to be women and present higher tumour grades compared with other types of GC 27,28 , but we did not observe these characteristics in the current study. This discrepancy might have been due to the exclusion of patients with LN metastasis, who often present with deeper tumour invasion and a worse prognosis 29 .
These differences in the presentation of SRC GC completely support the emerging concept that SRC GC might actually be a disease distinct from other types of GC, and numerous studies have clearly described this distinction [30][31][32] . Beyond the results reported here, the histology of SRC can be distinguished based on genetic expression data. Certain researchers have suggested that SRC GC may even be a completely different entity 33 . Furthermore, Sheng et al. reported that different molecular subtypes of GC had different responses to PI3-kinase inhibitors and 5-fluorouracil 34 .
There are certain limitations to our study. Notably, an important bias is that SRC features were present in 19.2% of patients in our database, which was significantly higher than the published rate (approximately 10% in all GC patients) 35 . In addition, we must consider that we did not compare different histologic types but rather the SRC type and the non-SRC type. For this reason, we might have overlooked different behaviours depending on the histologic type. Future research on this topic could explore other clearly defined gastric subtypes.
In conclusion, our study's results are currently the best evidence showing that SRC histology is an independent predictor of poor prognosis due to recurrence in LN-negative GC patients and that SRC histology is also an independent predictor of overall prognosis in patients with perineural and vascular invasion.

Methods and Patients
Data source and study sample. The Gastric Cancer database of the Sun Yat-sen University Cancer Centre (SYSUCC), which contains clinicopathologic data from more than 4,000 GC patients, was searched, and 556 patients and 107 patients were identified as having non-metastatic LNs and SRC gastric carcinoma, respectively, as confirmed by postoperative pathologic assessment. These patients all underwent gastrectomy and at least D2 lymphadenectomies for GC, and most of the patients presented a retrieved number of nodes greater than 15 (Supplemental Fig. 3).
Further assessment was performed using the following exclusion criteria: unknown vital status and loss of follow-up data (n = 30), distant metastasis (n = 20), inadequate surgical data (n = 14), unspecified tumour location (n = 10) and unspecified staging (n = 2). As a result, the final group comprised 480 included patients.
All enrolled patients routinely underwent an endoscopic ultrasound or electronic gastroscopy and abdominal computed tomography (CT) examination before surgery to identify the invaded region. After pathologic examination, the diagnosis of GC and the pathologic type were confirmed. The patients underwent routine preoperative examination, including a chest x-ray, an abdominal ultrasound, and blood testing. All of the results    reconstruction after gastrectomy. Additionally, 23 (4.8%) and 112 (23.3%) patients received neoadjuvant and adjuvant chemotherapy, respectively. All specimens were examined by surgeons and pathologists shortly after surgery, and we reviewed all of the resected specimens at this time (Supplemental Fig. 4). According to the World Health Organization criteria for the histologic typing of GC, we defined SRC GC as an adenocarcinoma in which a predominant component (greater than 50% of the tumour) consisted of isolated or small groups of malignant cells containing intracytoplasmic mucin 36 . The status of the LNs was evaluated by pathologists using haematoxylin-eosin staining. Follow-up data were primarily obtained from the institutional databases, which were regularly updated based on clinical chart reviews, tumour registry information, physician records, patient correspondence, and telephone interviews. Moreover, this study was approved by the Sun Yat-sen University Cancer Centre review board in accordance with Chinese bioethical regulations, and all enrolled patients signed informed consent forms.

Statistical analysis.
To examine unadjusted (univariate) associations, t tests and χ 2 tests were used. Survival was determined based on cause-specific mortality. Survival curves were generated using the Kaplan-Meier method and were compared using the log-rank test. A Cox proportional hazards model was utilized for multivariate analysis, which included the age at diagnosis, gender, the type of surgical resection and the AJCC TNM staging. Stage was the stratified variable. The data were analysed using SPSS 19.0 statistical software. P < 0.05 was considered statistically significant.