| Purpose: To explore reliable prognostic risk factors and establish a prognostic prediction model for accurate survival stratification of patients with Glioblastoma(GBM).In order to improve the prognosis of GBM patients,we quantitatively evaluated the Tumor Purity(TP)and Tumor Associated Macrophages(TAMs)of the microenvironmental characteristics of GBM,explore the relationship between TP,TAMs and survival time,and analyze the correlation between TP,TAMs and MR imaging findings.Finally,a comprehensive model was constructed to predict the prognosis of GBM patients by integrating clinical risk factors,Tumor Microenvironment(TME)and radiomics features.Materials and Methods:1.A total of 156 patients with GBM,IDH wild-type diagnosed by histopathology and molecules from January 2017 to April 2022 in the Lanzhou University Second Hospital were collected retrospectively.Magnetic Resonance Imaging(MRI)images were analyzed blindly by radiologists with experience in CNS disease diagnosis,tumor location,tumor diameter,peritumoral edema,cystic degeneration/necrosis,hemorrhage,contrast enhancement T1-Weighted Imaging(CE-T1WI)signal intensity and apparent diffusion coefficient(ADC)value were recorded or measured.Neuropathologists performed image segmentation of multiple Region of Interest(ROI)based on Hematoxylin-Eosin staining(H&E)and CD163,CD68 immunohistochemical staining whole-slide imaging(WSI).TP and TAMs were quantitatively evaluated by Image J software.Kaplan-Meier method was used to draw survival curve,and Log-rank test was used to analyze the difference of survival curve between high expression group and low expression group of TP and TAMs in GBM patients.Spearman correlation analysis was used to test the correlation between continuous variables and TP,TAMs.T-test,one-way ANOVA or MannWhitney U test and Kruskal-Wallis test were used to compare the differences of TP and TAMs between groups.2.A total of 149 patients diagnosed as GBM,IDH wild-type by histopathology and molecular diagnosis from January 2017 to April 2022 in the Lanzhou University Second Hospital were collected retrospectively,they were randomly divided into training set(n=104)and test set(n=45).The prognostic factors related to the overall survival(OS)of GBM were selected by univariate Cox in clinical factors,tumor microenvironment and radiomics features,and the best combination of features related to OS was selected by Lasso-Cox regression to established clinical models,radiomics models,clinical-radiomics models,clinical-tumor microenvironmentmodels and clinical-radiomics-tumor microenvironment comprehensive prediction models,respectively.Calculate the model consistency index(C-index)to evaluate the prediction ability of the model.Evaluate the accuracy of the model in predicting the survival probability at different times,calculate the area under ROC Curve(AUC),accuracy,sensitivity,specificity of the prediction model.The nomogram of the model was established on the basis of the constructed model,and the clinical utility of the model was determined by decision curve analysis(DCA).Kaplan-Meier plotting survival curve,Log-rank test to evaluate survival rate difference between high-risk group and low-risk group,test model the prognostic stratification ability.The net reclassification improvement index(NRI)and integrated discrimination improvement index(IDI)were used to compare the improvement of the integrated model compared with other models.Results: 1.156 GBM,IDH wild-type patients,93 males and 63 females,with an average age of 52.24±11.79 years and a median OS of 12 months.The average TP of GBM was 66.13%,the median CD163 positive(CD163+)TAMs was 14.94%,and the median CD68 positive(CD68+)TAMs was 7.89%.Survival analysis showed that high/low TP and high/low TAMS expression in GBM were related to survival rate.TP was divided into high TP and low TP with 64.00% as the critical value(P=0.038);CD163+TAMs and CD68+TAMs were divided into high expression and low expression with the critical values of 10.00% and 14.80%,respectively(P=0.033,P=0.011).The survival rate of patients with low TP and high TAMs expression decreased rapidly and the survival time was relatively short.TP was negatively correlated with CD163+TAMs and CD68+TAMs(r=-0.327,P=0.000;r=-0.235,P=0.003).Analysis of the correlation between the expression of TP,TAMs and clinical imaging signs,TP was positively correlated with tumor cell proliferation activity Ki-67(r=0.263,P=0.001),TP was negative correlations with peritumoral edema,ADCmean and ADCmin(r=-0.177,P=0.027;r=-0.377,P<0.001;r=-0.323,P<0.001,respectively).The expression of CD163+TAMs and CD68+TAMs was positively correlated with the age of GBM patients(r=0.213,P=0.027;r=0.167,P=0.038).CD163+TAMs expression was positively correlated with peritumoral edema and ADCmean value(r=0.231,P=0.004;r=0.189,P=0.018).The difference of CD68+TAMs in GBM with different enhancement degree was statistically significant(H = 10.189,P = 0.006).2.149 GBM,IDH wild-type patients,there were 104 cases in the training set,including 59 males and 45 females,with an average age of 52.44 ±11.89 years and a median survival time of 11 months;there were 45 patients in the test set,including 27 males and 18 females,with an average age of 51.80 ± 12.11 years and a median survival time of 15 months.By univariate Cox and Lasso-Cox analysis,the age(HR=1.021,95% CI 1.004-1.038,P=0.018),peritumoral edema(HR=1.015,95% CI 1.003-1.027,P=0.012),TP(HR=0.978,95% CI 0.963-0.993,P=0.003),CD163 + TAMS(HR=1.027,95% CI 1.003-1.052,P=0.024)and 21 radiomics features based on CET1 WI and T2 WI were screened associated with OS.The above factors form the best features combination for establishing prognosis prediction model of GBM patients.The established clinical model,radiomics model,clinical-radiomics model,clinicalTME model,clinical-radiomics-TME comprehensive model,in the training set,the Cindex is 0.646,0.693,0.718,0.656,0.727,respectively.Clinical-radiomics-TME model has best survival prediction,the AUCs of the 12-month,24-month,and 36-month survival predictions of the training set reached 0.842,0.802,and 0.826,respectively,and the AUCs of the test set were 0.704,0.772,and 0.713,respectively.Compared with clinical-TME model and clinical-radiomics model,the ability of clinical-radiomics-TME comprehensive model to predict survival probability at 12 months,24 months and 36 months was improved.The training set NRI were 0.213,0.233,0.202 and 0.040,0.034,0.091,respectively,and the training set IDI were 0.148,0.155,0.184 and 0.030,0.017,0.051,respectively.Conclusion: TP and TAMS in GBM are related to the prognosis of patients,GBM patients with low TP and high TAMS expression have short survival time and poor prognosis.The TP and TAMs are correlated with age at diagnosis,peritumoral edema degree,ADC value and CE-T1 WI intensity of GBM patients.Preoperative monitoring of MRI images features and quantitative parameters can reflect GBM microfeatures to a certain extent.The five prognostic prediction models established by combining different prognostic risk factors of GBM patients have good predictive efficiency.Among them,the clinical-radiomics-TME comprehensive model integrating clinical,radiomics features,TP and TAMs has the best efficiency in predicting the prognosis of GBM patients.The prediction ability of the comprehensive model has been positively improved compared with other models.Radiomics features,TP and TAMs play an important role in prognostic models,reflecting the clinical value as prognostic risk factors.The prognosis model which integrates multi-factor information promotes the accurate clinical management of prognosis stratification in patients with GBM. |
PDF Full Text Request