Analysis Of Prognostic Factors And Establishment Of The Mortality Prediction Model For Patients With Gastric Cancer Patients Based On Tumor Immune Microenvironment

Background and AimGrowing evidences are showing that the immune infiltration status is closely associated with prognosis of patients with gastric cancer.The aim of the present study is to:(1)build and assess the immunoscore model for prognostic prediction;(2)build and assess the nomogram model for the individualized prediction of mortality risk in gastric cancer.MethodsWe downloaded the gene microarray data,the pathological and prognostic profiles of gastric cancer patients from the public database Gene Expression Omnibus,and made use of the CIBERSORT algorithm platform to analyse the fractions of 22 types of immune cells;we then randomly sorted the above patients into the training group and the verifying group as 7:3 and built the immune risk score,IRS model,using the LASSO COX regression model.We then made assessment on the predictive value of IRS model in the verifying cohort and the entire patients cohort.Finally,based on the pathological parameters of patients,we built the nomogram model to predict the mortality risk of gastric cancer patients,and similarly,assessed the accuracy and conformity of its predictive value in the verifying cohort and the entire patients cohort.Results1.The construction and assessment of the IRS model:by using the LASSO COX regression model,we screened 11 immune cells to build the immunoscore model,IRS,the formula was:IRS=(0.322 X the fraction of M2 macrophages)-(0.162 × the fraction of M1 macrophages)-(0.179 × the fraction of CD8+T cells)-(0.469 × the fraction of activated CD4+ memory T cells)-(0.556 × the fraction of regulatory T cells)-(0.443 × the fraction of activated dendritic cells)-(0.220 × the fraction of activated NK cells)-(0.459 × the fraction of plasmocytes)+(0.369 × the fraction of??T cells)+(0.318 × the fraction of resting dendritic cells)+(0.713 × the fraction of naive CD4+T cells).The univariate survival analysis indicated that the IRS model could differentiate the gastric cancer patients into high-and low-risk groups,in the case of total survival,and the multivariate analysis showed that IRS is an independent prognostic factor.the receiver operating characteristic curve,ROC analysis showed that the IRS model has promising value in predicting the 2-,3-and 5-year survival rates.Results of the verifying and the entire patients cohorts also supported our findings.2.The construction and assessment of the nomogram model:based on the results of the multivariate analysis in the training group,we constructed the nomogram model by the R software to make individualized prediction of the mortality risk in gastric cancer patients.The nomogram model was internally verified using the Bootstrap self-sampling method,and the C index was 0.77,significantly better than the 6th TNM staging(p<0.001).The calibration curve confirmed the high consistency of the nomogram and the actual observation.The decision-making curve showed that the nomogram model performed better in clinical application than the TNM staging system.Besides,results in the verifying and the entire patients cohorts also supported our findings.Conclusion:1.The IRS immunoscore model we proposed is an independent prognostic factor for the total survival of patients with gastric cancer.2.Compared with the traditional TNM grading system,the nomogram,based on clinical pathological variables and IRS,could make individualized prognostic prediction of gastric cancer patients,more precisely and of higher clinical value.