The value of lymph nodes ratios in the prognosis of resectable remnant gastric cancer through the retrospective propensity score matching analysis

From January 2010 to December 2015, the clinicopathological data of 195 patients with RGC who underwent curative gastric resection (including total resection of the gastric stump with lymphadenectomy) at Changhai Hospital were retrospectively collected. In this study, RGC was defined as a carcinoma arising in the gastric remnant after gastrectomy, regardless of the histology of the previous lesion (benign or malignant), its risk of recurrence, the extent of initial resection, or methods of reconstruction [23]. The TNM classification (7th edition) was reported as a practical staging system for RGC, so the RGC stage in this study was conducted according to the TNM classification (7th edition) by AJCC [24, 25]. All the patients enrolled received standard therapy, which meant curative operation and postoperative adjuvant chemotherapy if necessary according to the NCCN Clinical Practice Guidelines in Oncology Gastric Cancer. This study was approved by the Medical Ethics Committee of Changhai Hospital of the Second Military Medical University. The informed consent requirement was exempted.

Inclusion criteria: 1) age 18 to 85 years, life expectancy > 3 months; 2) Histopathological examination confirmed as stomach adenocarcinoma; 3) Benign gastric lesions above 5 years after the initial partial gastrectomy, new cancers on the remnant gastric; 4) gastric malignant lesions undergo partial gastric resection and more than 5 years after R0 resection, new cancers of the remnant gastric; 5) complete clinical-pathological data; 6) adequate organ functions (leukocyte count > 3,500/μl, platelet count > 100,000/μl, hemoglobin > 10.0 g/dl, serum creatinine < 1.25 times the upper limit of normal (ULN), transaminases and alkaline phosphatases < 2.5 times ULN or < 5 times ULN in patients with liver metastasis, bilirubin < 1.5 times ULN, and prothrombin time < 12.0 s). Exclusion criteria: 1) concomitant other primary cancers; 2) incomplete or unavailable medical record data, incomplete preoperative examination, or refusal of surgical treatment; 3) patients whose initial lesions are malignant and have not been removed by R0; 4) patients with incomplete medical records.

Between January 2010 and December 2015, a total of 195 patients with RGC and 2001 patients with GC were admitted. Under case selection criteria, a total of 77 patients with RGC and 624 patients with GC were excluded. Reasons for ineligibility included patients incompatible with the operation (8/71 cases), positive margin (3/41 cases), inadequate organ function (4/155 cases), incomplete clinical data (6/162 cases), palliative operation or M1 stage (51/195 cases), and wedge gastrectomy (5/0 cases). Finally, this study included 118 resectable RGC patients and 1377 patients with GC who performed the curative operation.

Follow-ups were conducted once every 28-day for half a year, once every 3-month for 2 years, once every 6-month for 3 years, and yearly thereafter. Follow-up terms consisted of physical examination, complete blood test, chest radiography, and ultrasound of the abdomen as clinically indicated. Computed tomography (CT) scanning or magnetic resonance imaging (MRI) would be performed if necessary. The follow-up data of patients were collected either through telephone or outpatient service.

Medical records of patients enrolled were collected retrospectively by two assistants from Medbanks Network Technology CO. Collected data included patients’ background, time intervals between primary and secondary operation, pathologic diagnosis of RGC, surgical data, and tumor characteristics. The data errors were corrected by the Department of General Surgery Changhai Hospital.

To minimize the influence of other confounders on the outcome, we used a propensity score analysis to match RGC patients with GC patients. RGC patients were matched in a 1:4 ratio with GC patients using the nearest neighbor matching and based on gender (male or female), age (≤ 60 or > 60 years), tumor Grade (well-differentiated, moderately differentiated, poorly differentiated), tumor size (≤ 3 cm or > 3 cm), number of lymph nodes (LN) harvested, TNM stage (7th edition, AJCC) and pathological lymph nodes positive ratio (pLNR, number of positive LN/number of resected LN).

Statistical analysis was performed using SPSS software (18th version, SPSS, Chicago, IL, USA). Enumeration data were presented in percentage, measurement data was presented in average number and standard deviation, and abnormal distribution data was displayed by median and quartile. The associations between categorical variables were analyzed by the Pearson chi-square test. Continuous variables were assessed using t test or Cochran-Mantel–Haenszel test as appropriate. Survival curves and univariate analysis were performed in accordance with the Kaplan–Meier method, and the log-rank test was used to evaluate statistical significance. Cox regression analysis was used in multivariate analysis to assess independent prognostic factors. A probability (P) value < 0.05 was considered to indicate statistical significance. All tests were two-sided.

The flowchart of the program for participants in this study is shown in Fig. 1. The baseline characteristics of patients with GC in the pre-match and post-match samples are presented in Table 1. The two groups did not differ significantly in terms of age, sex, tumor grade, and TNM stage. However, compared with the control group, the RGC group had significantly larger tumors, higher pLNR, and fewer LNs harvested (all p < 0.05).

The overall survival rate was 46.61% for RGC patients compared to 55.08% for control groups (P < 0.01), and the mean overall survival time of RGC patients was 40.23 ± 32.27 months, compared with 55.06 ± 34.29 months in the control group (P = 0.023 after matching) (Fig. 1). As shown in the OS curves, the OS rates of RGC patients with TNM stage I (92.31%) and stage III (14.81%) were almost similar to GC patients (91.33% and 20.51%) (Fig. 2A-E). But the OS rate of RGC patients with TNM stage II was 45.00%, which was much worse than the control group (75.45%) (Fig. 2F, G). Another interesting finding showed that there was no significant difference in OS rates among RGC patients with TNM stage IIb, GC patients with TNM stage IIIa (P = 0.736), and IIIb (P = 0.105) (Fig. 3).

To find the relationship between primary disease and clinicopathological characteristics in RGC, all RGC patients were divided into two sub-groups (RGC-B for primary benign disease vs. RGC-M for primary malignant disease). Primary benign disease meant benign ulcer, while primary malignant disease referred to gastric adenocarcinoma. In all 118 cases, 61 (51.69%) RGC patients had primary GC in the first operation. The time interval, number of LN harvested, and pN stages were statistically different between the two groups (Table 2). Theoretically speaking, the primary disease was an important personal factor so it may influence the prognosis, however, our study showed there were no obvious survival differences attributed to it (40.99 ± 33.50 months vs. 39.52 ± 31.34 months; P = 0.26).

There was an interesting finding that the OS of RGC patients with TNM stages II varied much from each other (Fig. 2A, B). The OS of patients with stage IIb was much worse than IIa (P < 0.001) and similar to IIIa (P = 0.463, Fig. 3A) and IIIb (P = 0.014, Fig. 3B), which indicated that the current TNM classification (7th edition) might not be suitable to the patients with RGC. Further, we deemed it might be caused by the number of lymph nodes harvested in RGC patients being much less than in GC patients (P < 0.001), which might affect the results of TNM classification evaluated by the number of lymph nodes status (Fig. 4).

Univariate log-rank test analysis revealed that tumor size and pLNR were significant prognostic factors for RGC. Multivariate Cox proportional hazards model analysis revealed that pLNR (HR: 2.405, P = 0.028; Table 3) was an independent prognostic factor to OS in RGC. Further, the correlation analysis also showed there was a significant correlation between the number of metastatic lymph nodes and the number of lymph nodes harvested in RGC patients (r = 0.79, P < 0.001). So we deemed the TNM classification evaluated by LNR status might be more suitable for RGC patients.

The clinicopathologic characteristics and prognosis of RGC had been investigated in a few studies, which had not reached a consensus yet [10, 27, 28]. According to most reports, RGC was frequently diagnosed at an advanced stage, which caused a relatively low rate of curative resection and poor prognosis, suggesting that RGC may have distinct biological features from primary GC [29]. But some authors have compared RGC with primary GC and discovered no significant difference in survival between RGC and primary GC [30]. In most studies, the lymph node N staging of RGCs still follows the grading criteria of the UICC. However, in patients with first-time GC, postoperative lymph node drainage changes, and the lymph nodes detected by RGC cannot fully define the N stage, especially since RGC occurs after GC. The total number of postoperative lymph node dissections during re-surgery usually does not exceed 10, which is significantly less than the number of lymph nodes dissected by RGC after surgery for benign lesions. This may result in inaccurate staging. Studies have shown that the total lymph node detection rate and peritoneal lymph node metastasis rate of RGC are lower than those of primary proximal gastric cancer (PPGC), and TNM grading may not be suitable for RGC. In our study, we found the OS of RGC was much worse than GC, although the TNM stages of patients in the two groups were similar. But we deemed that this phenomenon might be caused by an unsuitable TNM classification system for RGC, for there was nearly no difference in OS curve among stage IIb, IIIa, and IIIb RGC patients, which indicated the present TNM staging system could not evaluate the OS of advanced RGC patients with lymph nodes metastasis. So our study also lent further support that it is necessary to investigate a new indicator to evaluate the lymph node status in advanced RGC.

In this study, RGC patients were divided into the RGC-B group and RGC-M group according to the nature of the initial disease. The numbers of the two groups were 57 (48.31%) and 61 (51.69%), respectively. The number of cases in the RGC-M group was slightly higher than that in the RGC-B group. This may have a great relationship with the application of proton pump inhibitors and the increase in the rate of radical resection of primary GC in recent years. By comparing the clinicopathological features of patients in the RGC-B group and the RGC-M group, it was interesting to find that the clinicopathological features of RGC-M were similar to those of RGC-B, except for the time interval from the start of the first operation. It indicated the primary disease has no effects on the characteristics of RGC. And there was no difference in tumor location between RGC-M and RGC-B also lends further support to the previous conclusion that the conception of CRS was more reasonable in the definition of RGC [6, 31]. The survival analysis of this study showed that the prognosis of patients in the RGC-B group was better than that of the patients in the RGC-M group, but the difference was not statistically significant. This may be related to the small number of samples included in this study, and there may be some bias.

Lymph node status is an established prognostic factor in both GC and RGC [12, 32]. But the Japanese classification of gastric carcinoma nor the TNM classification raised a special lymph node classification system for RGC till now. At present, most staging system (UICC/AJCC) stratifies the N value simply on the basis of the count of metastatic lymph nodes. But the total number of lymph nodes harvested in RGC was much less than in GC, which might affect the results of the N value based on the count of metastatic lymph nodes. Accumulated evidence showed the lymph nodes ratio (LNR) could reflect both qualitative and quantitative lymph node spread as well as the extent of disease, which might serve as an additional prognostic indicator to better evaluate the outcome of patients with advanced RGC [33‐35]. In particular, we focused on the prognostic value and staging accuracy of the metastatic lymph nodes ratio (LNR) in patients with RGC undergoing curative resection, and we also found that LNR was an independent prognostic factor in RGC patients in multivariate analyses, for LNR was less influenced by the total number of lymph nodes harvested in operation. Since Okusa firstly demonstrated the impact of the “frequency of the metastases” on patients’ outcomes, there were few articles focusing on LNR in RGC to our knowledge [36]. In the present study, we reported that LNR might consider a simple, reproducible, and highly reliable prognostic factor in RGC patients. But the threshold of LNR in the TNM staging system of RGC needs a larger sample size to confirm.

There still existed some limitations to this study. First, the major limitation was its retrospective design in a single hospital. Secondly, we were unable to collect some information on the study patients, such as lifestyles, dietary preferences, and environmental features for statistical analysis, which might be important influencing factors. Further prospective studies were necessary to clarify the pathophysiology and development of RGC.