Patient demographics
From January 1st, 2009 to December 31st, 2016, altogether 2649 pathologically confirmed incident gastric adenocarcinoma cases were enrolled into our cohort. Detailed information of patient clinicopathological characteristics are presented in Table 1. The short-term endpoints and survival outcome were depicted in Table 2. The median age at the time of diagnosis was 59 (range, 21–86) years.
There were 2568 patients in NTB group and 81 in TB group, respectively. As discovered from Table 1 and Table 2, there were some heterogeneities in baseline characteristics between the two groups. To be specific, TB group was associated with a higher proportion of male patients, prior pulmonary comorbidities, ex-smoker or current smoker, PPCs and a smaller tumor size and shorter PODs. In addition, significant differences were also observed in the distributions of pathological stages and Bormann types. Furthermore, PSM was performed to create the comparable groups. After PSM, all variables between 142 cases in NTB group and 73 in TB group revealed equal distributions of patient demographics and baseline characteristics (Table 1, Table 2, p>0.05 for all).
POCs and baseline factors
POCs between cohorts revealed some different distributions. In detail, significant differences were observed in the distribution of PPCs between cohorts, with the PPCs occurrence in 262 (10.2%) patients in the NTB group and 22 (27.2%) patients in the TB group. We further performed the univariate and multivariate logistic regression analysis to assess the risks for PPCs. The univariate analysis revealed that old TB, older age, higher ECOG score, ex-smoker or current smoker, comorbidities, perioperative blood transfusions, surgical procedure, lower third location and poorly differentiated histology (G3) were significantly associated with PPCs (p < 0.05 for all, Table 3). After the multivariate logistic regression analysis, TB history, older age, pulmonary dysfunction, lower third location remained to be risk factors for PPCs (p < 0.05 for all, Table 3).
Survival
Upon the last follow-up, 1003 out of the 2649 patients (39.1%) died, with a median OS time of 22.1 months. Meanwhile, 940 of the 1021 deaths (93.7%) were related to cancer. The median OS time since operation for all patients was 50.1 months, and the overall 1- and 3-year OS rates were 91.5% and 72.2%, respectively. There were significant differences in OS between the two cohorts (Fig.2a). The estimated 3-year OS rate in TB group was 84.7%, which was significantly higher than that of 69.2% in NTB group (p=0.005; Fig.2a). The median follow-up period in TB group was 57 months, about 7 months longer than that in NTB group (50 months, p = 0.011).
After one-to-two PSM, 73 pairs of patients were qualified for further analysis. Good balance was demonstrated for both of the matched patients, with no clinically significant differences in the covariates, as depicted in Table 1 and Table 2. Besides, significant differences in OS rates were also detected in the PSM cohorts (Fig.2b). The 3-year OS rate in TB group was 90.4%, which was higher than that of 69.2% in NTB group (p=0.001; Fig.2b). After adjusting for potential confounders by multivariate Cox regression analysis, the coexistence of old TB lesion remained significantly and strongly associated with the decreased tumor-related mortality (adjusted HR: 0.220, [95% CI: 0.107-0.454], p<0.001), as shown in Table 5.
Univariate and Multivariate Cox Regression
In univariate analysis, old TB, age, ECOG score, pulmonary dysfunctions, perioperative blood transfusions, tumor size, surgical procedure, surgical approach, lower third and diffused locations, histological grading, neural/vascular invasion, Bormann’s type, pathological T-N-M category, pathological TNM stage, POCs and adjuvant chemotherapy were associated with the patient’s OS (Table 4). Based on univariate analysis results, factors of TB history, age, ECOG score, comorbidities, blood transfusions, tumor size, surgical procedure, approach, location, histological grading, Bormann’s type, neural/vascular invasion, T-N-M stage, POCs and adjuvant chemotherapy were included for stepwise multivariate Cox regression analysis. It was discovered that old TB lesion and adjuvant chemotherapy were the independent factors to determine good survival, while older age, pulmonary dysfunctions, blood transfusion, bigger tumor size, diffused location, higher TNM classifications, PPCs predicted poor prognosis (Table 4).
After PSM, old TB, tumor size, blood transfusion, surgical procedure, location, Bormann type, T-N-M category, P-stage, PPCs and adjuvant chemotherapy were identified to be associated with OS in the univariate analysis. Stepwise analysis of the multivariate Cox regression included the same covariates as in the multivariate analysis before PSM. And finally, several independent positive prognosticators for OS-old TB, adjuvant chemotherapy, proximal gastrectomy and laparoscopic or robot-assisted gastrectomy, and several negative prognosticators including middle third or lower third locations and G3 and PPCs were identified after PSM. Detailed univariate and multivariate Cox regression analyses of OS after PSM are shown in Table 5.