| Background and purpose:Gastric cancer is one of the most common malignant tumors and one of the leading causes of tumor death worldwide.Due to the concealment of early symptoms and lack of early screening,many patients with gastric cancer are at an advanced stage or even have distant metastases,which greatly affects the survival time and quality of life of patients,and also creates a heavy family burden on society and individuals.To date,To date,there are no effective treatment guidelines and clinical studies for distant metastatic gastric cancer,and clinicians often base their treatment strategies on past experience and lack scientific evidence.This study retrospectively analyzes the clinicopathological characteristics of patients with distant metastatic gastric cancer and establishes a survival prognosis model for such patients,which is intended to provide a reference for clinicians in formulating individualized treatment plans.Method:This study retrospectively analyzed the demographic characteristics,clinicopathologic features,and survival information of patients with metastatic gastric cancer(mGC)diagnosed by histology from 2010-2016 using the National Institute of Surveillance,Epidemiology,and End Results(SEER)database,with a total of 3742 eligible patients were screened according to inclusion and exclusion criteria.In this study,prognostic models predicting overall survival(OS)and gastric cancer specific survival(CSS)in patients with mGC were established,and then each predictive model was visualized and the predictive power of the models was evaluated by using nomogram.Firstly,3742 patients were randomly divided into a development set(n=2620)and validation set(n=1122)at a ratio of 7:3,and OS prognosis model was constructed using modeling the development set.The variables with P<0.05 were screened and excluded using stepwise regression analysis(backward method),and AIC(Akaike information criterion)as a criterion.Incorporation into the Cox proportional risk model after multivariate analysis: the OS nomogram survival prediction model was constructed on this basis;the validation set tested effectiveness of the prediction model: the Brier score and calibration curve were used to assess consistency between the predicted and actual observed survival rate;the discrimination was measured by the concordance index(Cindex)and the area receiver operating characteristic curve(AUC).Patient grouping and screening for variables was done in the same way as in the first model,with the same development set of patients used to construct the CSS nomogram survival prognosis model.The obtained variables are subjected to multivariate analysis and incorporated into Fine-Gray’s competition risk model: a CSS nomogram survival prediction model is constructed;the validation set tests the model’s performance in the same way as the first model.Finally,the relationship between prediction scores and survival prognosis was described: the k-means clustering algorithm was used to artificially divide the prediction scores of OS nomogram into three layers,and the Kaplan-Meier survival curve was used to fit the relationship between survival and prediction scores during follow-up time to show the differences and associations between the risks.P< 0.05 considers the difference to be statistically significant.Result:1.Stepwise multivariate Cox regression showed that primary tumor site,T stage,N stage,tumor diameter,bone metastasis,surgery,chemotherapy were independent prognostic factors for OS in patients with mGC.In the validation group,brier score and calibrate curve reflect a good consistency between predicted survival and actual observed survival rate;C-index and AUC indicate that the model have a good prediction effect,with OS nomogram model predicting that the accuracy of 6,12 and 18 months are different.The Brier score is 0.19(95%CI: 0.17-0.21),0.14(95%CI: 0.13-0.16)and 0.10(95%CI: 0.09-0.11)and AUCs of 0.77(95% CI: 0.74-0.80),0.75(95%CI: 0.72-0.78)and 0.75(95CI: 0.71-0.79),respectively.2.After excluding confounding factors of non-tumor cause death,stepwise multivariate competitive risk model analysis suggested that primary tumor site,grade,tumor diameter,brain metastasis,surgery and chemotherapy were independent prognostic factors for CSS in metastatic gastric cancer patients.In the validation group,brier score and calibrate curve reflect a good consistency between predicted survival and actual observed survival rate;C-index and AUC indicate that the model have a good prediction effect,and the accuracy of the CSS nomogram model predictions at 6,12 and 18 months varied.The Brier scores are 0.20(95% CI: 0.18-0.21),0.16(95% CI: 0.15-0.17)and 0.12(95% CI: 0.10-0.13),AUCs of 0.76(95% CI: 0.73-0.79),0.73(95% CI: 0.68-0.75)and 0.70(95% CI: 0.65-0.75),respectively.3.The k-means clustering algorithm classified the prediction scores of the validation cohort into three risk strata: low,medium and high,and the difference in survival among patients in these three layers showed statistically significant(P<0.001).Conclusions:1.Age,primary tumor location,histological grade,depth of tumor invasion,bone metastasis,lung metastasis,distant lymph node metastasis are risk factors for mGC patients,and surgery and chemotherapy are protective factors for mGC patients.2.The independent predictors of CSS in mGC patients are histological grade,T stage,tumor diameter,lung metastasis,distant lymph node metastasis,insurance status,and whether to undergo surgery or chemotherapy.3.According to the prediction scores of OS nomogram,patients in the low-risk group had a better prognosis and the high-risk group had the worst long-term prognosis.4.Both the OS and CSS nomograms provide a good assessment of survival prognosis for mGC patients,providing them with a personalized survival scoring scale. |
PDF Full Text Request