MRI Biomarkers And The Predictive Models For The PD-1/PD-L1 Expression And Prognosis In ICC

Objective:To explore the effect of programmed cell death protein 1(PD-1)and programmed cell death protein ligand 1(PD-L1)expression on the prognosis of intrahepatic cholangiocarcinoma(ICC).To preliminarily explore the correlation between the magnetic resonance imaging(MRI)biomarkers related to the PD-1/PDL1 expression and prognosis and the immune microenvironment,and fibrotic stroma in the tumors.To predict the PD-1/PD-L1 expression and prognosis in ICC based on MRI biomarkers.Materials and Methods:The expression of PD-1 and PD-L1,the counts of CD3,CD4,and CD8 T lymphocytes,and the morphology and abundance of fibrotic stroma in 98 patients with ICC were evaluated by histopathological methods.The effect of PD-1 and PDL1 expression on the prognosis of ICC patients were analysed based on Kaplan-Meier and Cox regression.Radiomic features were extracted from MRI in the arterial phase and portal vein phase,respectively,and Radiomics features related to the PD-1/PDL1 expression and prognosis in ICC patients were selected by Spearman correlation analysis and minimum absolute contraction and selection operator.The MRI features and clinicopathological factors related to the PD-1/PD-L1 expression and prognosis in ICC patients were selected by univariate and multivariate analysis.Pearson or Spearman was used to analyze the correlation between MRI features,MRI Radiomics features,and the counts of CD3,CD4,CD8 T lymphocytes,and the morphology and abundance of fibrotic stroma in the tumors.The MRI Radiomics features with the best performance in predicting the PD-1/PD-L1 expression and prognosis were selected and the corresponding Radiomics score(Radscore)was calculated as the biomarkers.Radscore,MRI features,and clinicopathological factors were used to predict the PD-1 and PD-L1 expression,individually and in combination.A clinical overall survival(OS)model was developed with independent features based on Cox regression,to stratify ICC patients into high-risk and low-risk groups using the median of the model.A calibration curve was plotted to evaluate its performance.Results:The 5-year survival rates of PD-1 positive and PD-1 negative cases were 12.5%and 48.3%,respectively(p<0.05),whereas 5-year survival rates were 21.9% and 39.4%for PD-L1 positive and PD-L1 negative cases,respectively(p<0.05).The best model for the prediction of PD-1 expression was the combined model including Radscore,MRI features,and clinicopathological factors,with an area under the curve(AUC)of0.897.The best model for the prediction of PD-L1 expression was based on the MRI Radiomics extracted arterial phase,with an AUC of 0.890.The MRI features and MRI Radiomics features related to PD-1/PD-L1 expression and prognosis were partly correlated with CD3,CD4,and CD8 T lymphocyte counts,and the morphology and abundance of fibrotic stroma in the tumors,either weak positively(r=0.21-0.41,P<0.05)or negatively(r=-0.21--0.25,P<0.05).PD-1 expression,PD-L1 expression,lymph node metastasis,carcinoembryonic antigen,imaging classification,and the Radscore were independent predictors for OS.The clinical OS model(C-index,0.786;95% CI: 0.740,0.832)incorporating these independent predictors was developed to divide ICC patients into high-risk(median,2.69;95% CI: 2.16,3.19)and low-risk groups(median,0.97;95% CI: 0.3,1.4)based on a median of 1.8.The 1-,3-,and 5-year survival probability of the two groups were55.1%,10.2%,and 6.1% and 89.8%,67.3%,and 61.2%,respectively(p<0.001).The calibration curve of the clinical OS model demonstrated excellent agreement between predictions and observations.Conclusion:ICC patients with PD-1 or PD-L1 positive expression have a poor clinical outcome,and the PD-1/PD-L1 pathway can be used as a therapeutic target.MRI features and MRI Radiomics features can be used as non-invasive biomarkers to predict the PD-1/PD-L1 expression and prognosis,some of which are related to the immune microenvironment and fibrotic stroma in the tumors.The predictive models of PD-1/PD-L1 expression and prognosis based on MRI biomarkers are effective,and the results need to be validated by an external multi-center.