Clinical data pertaining to 3,484 patients who underwent gastrectomy for gastric cancer between January 2010 and December 2014 were retrospectively collected from the medical records of 9 institutions [7]. Written informed consent for surgery and the use of clinical data was obtained from the patients, as required by the Institutional Review Board of Nagoya University (approval number 2017-0104) and all the participating institutions. As we had aimed to conduct clinical research using only retrospective clinical data without intervention, we applied the opt-out recruitment strategy, according to the policy adopted by the Japanese Government. The purpose, design, and objectives of the study have been published on our homepage (https://www.med.nagoya-u.ac.jp/medical_J/ethics/rinsyoukansatsu.html) to provide an opportunity to the patients for declining to participate in our study. Out of the 3,484 patients, data pertaining to 401 patients were considered for analysis based on the following inclusion criteria: R0 resection with systematic lymphadenectomy performed in accordance with the Japanese Gastric Cancer Treatment Guidelines, pathological examination showing pT1N+ or pT2–3N0 gastric cancer based on the 3rd English edition of the Japanese Classification of Gastric Carcinoma [8] that corresponds to the tumor-node-metastasis (TNM) system of classification of the International Union Against Cancer Classification of Malignant Tumors, 8th edition [9], and the availability of sufficient data for analysis (Fig. 1) [89]. We excluded patients with gastric stump cancer, with cancers having histologies other than the common histological classification types, such as carcinoid tumor and carcinoma with lymphoid stroma, who underwent extended surgery, such as pancreaticoduodenectomy and Appleby's procedure, who received neoadjuvant or adjuvant chemotherapy, and those with a history of other primary cancers.

The following data were collected from the medical records of the 9 institutions: demographic and clinical data, including sex, age, tumor location, tumor diameter, and macroscopic type; surgical details, such as type of procedure, grade of lymphadenectomy, concomitant resection of other organs, duration of operation, intraoperative blood loss, and postoperative complications; and histological tumor findings, including tumor location, size, macroscopic type, histological type, lymphatic invasion, and venous invasion (classified according to 3rd English edition of the Japanese Classification of Gastric Carcinoma) [8]. The tumors that were classified as differentiated included well-differentiated tubular, moderately differentiated tubular, and papillary adenocarcinomas. The tumors that were classified as undifferentiated included poorly differentiated tubular adenocarcinoma, signet ring cell carcinoma, and mucinous adenocarcinoma [10]. Recurrence-free survival (RFS) was defined as the date of surgery to the date when recurrence was first detected. Only gastric cancer-related deaths were considered for calculating DSS, and subjects who died of other causes were not considered.

The patients underwent postoperative follow-up for 5 years or until recurrence. The follow-up included physical examinations and laboratory tests, including tests for serum tumor markers every 3 months, enhanced computed tomography of the chest and abdominal cavity every 6 months, and upper gastrointestinal endoscopy at 1, 3, and 5 years, as described in the Japanese Gastric Cancer Treatment Guidelines [4]. Treatment after recurrence was based on the patient's consent and condition, and the evidence available at that point of time.

Survival curves were calculated using the Kaplan-Meier method and compared using the log-rank test. Cox proportional hazard models were used to perform univariate and multivariate analyses. A χ² test was performed for analyzing the categorical variables. All the statistical analyses were performed using JMP 13 software (SAS Institute, Cary, NC, USA). A P-value of <0.05 was considered to be statistically significant.

The patient characteristics are enlisted in Table 1. The median age of the 401 patients was 70 years, and the male:female ratio was approximately 2.2:1. The most frequently performed surgical procedure was distal gastrectomy (73%) followed by total gastrectomy (22%). More than half (54%) of the patients underwent D2 lymphadenectomy. Approximately 20%–30% of the patients underwent non-D2 lymphadenectomy, and included those with clinical T2–4 gastric cancer (Supplementary Fig. 1). There were 78 patients who underwent non-D2 lymphadenectomy despite having clinical T2–4 gastric cancer (74 patients) and clinical T1N+ (4 patients). Of these 78 patients, 64 (82%) were aged 70 years or older. Of the total 401 patients studied, 85 (21%) experienced postoperative complications of grade II or higher, according to the Clavien-Dindo system of classification, and 40 patients (10%) had postoperative complications of grade III or higher. There were 2 operative deaths (0.5%), and both the patients died of pneumonia. Lymphatic and venous invasions were detected in 279 (69.6%) and 169 patients (42.1%), respectively. The patients were categorized into the lymphovascular invasion (LVI)-positive group (n=300) and the LVI-negative group (n=101) based on the presence and absence of LVI, respectively. Most patients with pT1N+ disease (86%) were at the pN1 stage (1 or 2 lymph node metastases). It was observed that the number of patients with pT2N0 disease was greater than the number of patients with pT3N0 disease. The median follow-up time for the 401-patient cohort was 48.9 months.

The 5-year overall survival, DSS, and RFS rates for the 401 patients were 84.3%, 95.2%, and 93.1%, respectively (Supplementary Fig. 2). Univariate analysis revealed that age ≥70 years and the presence of LVI were significantly associated with poor RFS (hazard ratio [HR], 2.81; 95% confidence intervals [CIs], 1.18–7.75; P=0.019 and HR, 7.82; 95% CI, 1.65–139.9; P=0.005, respectively). In the multivariate analysis, age ≥70 years and the presence of LVI remained the independent prognostic factors for poor RFS (HR, 2.62; 95% CI, 1.09–7.23; P=0.030 and HR, 7.88; 95% CI, 1.66–140.9; P=0.005, respectively) (Table 2).

The LVI-positive group had significantly worse RFS in comparison to that of the LVI-negative group (P=0.017) (Fig. 2A). In total, 24 (6.0%) patients experienced recurrence following curative gastrectomy, of which 23 patients had LVI. The most frequently observed sites of initial recurrence were the liver and lymph nodes (8 patients each), followed by the lung (n=5), peritoneum (n=4), and local recurrence (n=2). Multiple sites of recurrence were observed for 2 patients at the time of initial detection. The distribution of the sites of recurrence according to the LVI status is depicted in Fig. 2B. There were no significant associations between lymph node recurrence and lymphatic invasion (P=0.700) or between hematogenous recurrence (liver and lung metastasis) and venous invasion (P=0.110) (Supplementary Table 1).