Among 122793 patients with gastric malignancies, 74303 were male (60.51%), and 48490 were female (39.49%); 89157 patients (72.6%) were over 60 years of age; 31754 patients (36.2%) had a diagnosis and treatment interval of ≤ 1 month; 56061 patients (63.8%) had a diagnosis and treatment interval of > 1 month; and 765363 patients (62.5%) were mainly from metropolitan counties with a population of more than 1 million (Table 1). For the AJCC staging of tumors, stage IV was the most common stage, with 12419 cases(38.0%); Most lesions were ≤ 5 cm in size, with 25964 cases(61.4%); The tumors were mainly located in the cardia: 34019(34.0%); The main pathological type was adenocarcinoma, with 96259 cases(82.7%)(Table 1). The main treatment methods were surgical treatment in 31494 patients (35.9%), chemotherapy in 19220 patients (21.9%), surgery, chemotherapy and radiotherapy in 12740 patients (14.5%), surgery plus chemotherapy in 11012 patients (12.6%), and chemotherapy plus radiotherapy in 8698 patients (9.9%). By 2019, 91869 (74.82%) of the 122793 patients had died. The median follow-up time was 11.7 months (0-239 mo). The MST of the patients with gastric malignancies was 13.7 months, and the OS rates at 1 year, 3 years and 5 years were 38%, 28% and 23%, respectively.
Univariate analysis using the Kaplan‒Meier method showed that sex, age, tumor size, tumor site, AJCC stage, rural/urban setting, histopathological type, and treatment mode were prognostic factors for gastric malignancy patients (log rank test P < 0.05), but there was no significant difference between diagnosis and treatment (log rank test P = 0.726, Table 1 and Fig. 1).
A multivariate Cox regression model was used to analyze gender[female vs. male, HR(95% CI),0.899(0.682–0.938)],age[60–74 vs. 0–44,HR(95% CI),1.189(1.088-1.300),P < 0.001; \(\ge\)75 vs. 0–44,HR(95% CI),1.728(1.577–1.894),P\(<\)0.001], tumor size[> 5 cm vs. 0–5 cm, HR(95% CI),1.172(1.076–1.275)], tumor site[Fundus vs. Cardia, HR(95% CI),0.903(0.818–0.997),P\(=\)0.043;Gastric body vs. Cardia, HR(95% CI),0.840(0.782–0.903),P\(<\)0.001;Gastric antrum vs. Cardia, HR(95% CI),0.856(0.808–0.907),P\(<\)0.001;Gastric lesser curvature vs. Cardia, HR(95% CI),0.817(0.760–0.878),P\(<\)0.001;Gastric greater curvature vs. Cardia, HR(95% CI),0.859(0.781–0.945),P\(=\)0.002],AJCC stage[IB vs. IA,HR(95% CI),1.505(1.370–1.655),P\(<\)0.001;IIA vs. IA,HR(95% CI),2.092(1.919–2.282),P\(<\)0.001;IIB vs. IA,HR(95% CI),2.815(2.581–3.072),P\(<\)0.001;IIIA vs. IA,HR(95% CI),3.989(3.633–4.381),P\(<\)0.001;IIIB vs. IA,HR(95% CI),4.862(4.432–5.334),P\(<\)0.001;IIIC vs. IA,HR(95% CI),6.673(6.121–7.275),P\(<\)0.001;IV vs. IA,HR(95% CI),7.040(6.363–7.790),P\(<\)0.001],pathology[Neuroendocrine tumors vs. Adenocarcinoma, HR(95% CI),0.578(0.513–0.651),P\(<\)0.001;Mesenchymal tumor vs. Adenocarcinoma, HR(95% CI),0.297(0.267–0.329),P\(<\)0.001;Other vs. Adenocarcinoma, HR(95% CI),0.685(0.545–0.861),P\(=\)0.001], rural/urban setting[Counties in metropolitan areas of 250,000 to 1 million pop vs. Counties in metropolitan areas ge 1 million pop, HR(95% CI),1.088(1.036–1.142),P\(=\)0.001;Counties in metropolitan areas of lt 250000 pop vs. Counties in metropolitan areas ge 1 million pop, HR(95% CI),1.261(1.166–1.362),P\(<\)0.001;Nonmetropolitan counties adjacent to a metropolitan area vs. Counties in metropolitan areas ge 1 million pop, HR(95% CI),1.273(1.171–1.385),P\(<\)0.001;Nonmetropolitan counties not adjacent to a metropolitan area vs. Counties in metropolitan areas ge 1 million pop, HR(95% CI),1.292(1.174–1.423),P\(<\)0.001] and treatment[Surgery vs. Chemotherapy, HR(95% CI),0.673(0.624–0.726),P\(<\)0.001;Radiotherapy vs. Chemotherapy, HR(95% CI),2.246(2.000-2.524),P\(<\)0.001;Surgery+Chemotherapy vs. Chemotherapy, HR(95% CI),0.448(0.414–0.484),P\(<\)0.001;Surgery+Radiotherapy vs. Chemotherapy, HR(95% CI),0.518(0.437–0.613),P\(<\)0.001;Surgery+Chemotherapy+Radiotherapy vs. Chemotherapy, HR(95% CI),0.421(0.389–0.455) P\(<\)0.001] as prognostic factors and independent risk factors(P < 0.001)(Fig. 2). The 1-year, 3-year and 5-year OS of each variable of gastric malignancy patients was obtained by the life table method (Table 2).
Based on the results of the multivariate Cox regression analysis, the nomogram included 8 variables with statistical significance in the construction of the nomogram prediction model for patients with gastric malignancies. According to the nomogram, the scores corresponding to different clinicopathological characteristics of the patient can be obtained, and the total scores obtained can be used to predict the OS of the patient at 1 year, 3 years, and 5 years (Fig. 3). The C-index of the model was 0.763 ± 0.002 after 1000 self-service resampling verifications using the bootstrap method. ROC curve analysis was used to further evaluate the differentiation of the model, and the AUCs of the patients with OS at 1 year, 3 years, and 5 years were predicted to be 0.76, 0.79, and 0.79, respectively. All calibration curves and ROC curves indicated that the predicted OS of the nomogram had a good fit with the actual OS, indicating that the prognostic evaluation model had good calibration (Fig. 4).