| Objective:To build and evaluate a predictive model for early relapse after the radical surgery in hepatocellular carcinoma(HCC)patients with microvascular invasion(MVI).Material and methods:The patients undergoing surgical resection for tumor and confirmed as HCC with MVI by the postoperative pathological examination in Cancer Hospital of Chinese Academy of Medical Science from Jan 2014 to June 2019 were enrolled continuously.Univariate and multivariate Cox analysis were first performed in all subjects to determine the independent risk factors of relapse-and then logistic regression analysis was performed to determine the independent risk factors of early relapse.Based on the independent risk factors of early relapse,a predictive model was established,which was presented in a form of nomogram and webpage calculator.Subsequently,the value of this predictive model for discriminability,accuracy and clinical practicability was compared with that of AJCC staging system(Version 8.0)and BCLC staging system-a calibration curve was used to evaluate the accuracy of predictive model-clinical decision curve was plotted to evaluate the clinical practicability of predictive model.Finally,based on the ROC curve,the best threshold value of early prediction was calculated for predictive model,AJCC staging system and BCLC staging system separately-the sensitivity and specificity of these three methods were compared to further evaluate the predictive model.Results:A total of 406 patients were finally enrolled.These patients were allocated into early relapse group(160 patients)and non-early-relapse group(246 patients).All included patients were randomly divided into training group and validation group at a ratio of 7 to 3.The independent risk factors of relapse were determined using the univariate and multivariate Cox analysis.Subsequently,with early relapse(Yes v.s No)as dependent variable and the screened independent risk factors of relapse as independent variable,the independent risk factors of early relapse were screened.According to the relevant literatures,the following aspects were finally determined as independent risk factors of postoperative early relapse:preoperative blood AFP level,preoperative HBeAg status,type of MVI,diameter of tumor,number of tumors,invasion to hepatic envelope and satellite node.By applying above factors,a predictive model was established,which was displayed in nomogram and webpage calculator.In either modeling group or validation group,C index of discriminability in predictive model was superior to that in AJCC staging system(Version 8.0)and BCLC staging system:in modeling group:C index was 0.737,0.60 and 0.57 for predictive model,AJCC staging system and BCLC staging system,respectively-in validation group:C index was 0.737,0.63 and 0.60 for predictive model,AJCC staging system and BCLC staging system,respectively.In both modeling group and validation group,the resampling was done for 1000 times using Bootstrap method and the calibration curve was plotted to evaluate the level of calibration-and the results showed,in either modeling group or validation group,the predictive probability better concord with actual probability-and the early relapse was accurately predicted with the predictive model.As shown by the decision curve analysis(DCA),the predictive model was of greater value for clinical practicability.Based on the optimum cut-off value of 120 scores calculated using Youden index,the patients were randomly allocated into high risk group and low risk group to discriminate whether the patients had an early relapse or not.In both modeling group and validation group,the sensitivity and specificity of predictive model were better(sensitivity:74%Vs 76%-specificity:61%Vs 64%).Conclusion:The HCC patients with MVI was at higher risk of postoperative early relapse after the resection of liver cancer.In such patients,the following aspects are independent risk factors of postoperative early relapse:type of MVI,HBeAg status,preoperative blood AFP level,number of tumors,diameter of tumor,satellite node and invasion to hepatic envelope.The predictive model established by applying above factors can accurately predict the early relapse after the surgical resection in HCC patients with MVI.Background and objective:Comprehensive treatment based on surgery is the main treatment method to obtain no evidence of disease status for colorectal cancer patients with liver metastases.However,the patients still had a high recurrence rate after surgery.The traditional staging system based on clinical and pathological characteristics cannot accurately assess the prognosis of patients.With the development of high-throughput sequencing technology and bioinformatics,clinicians can quickly,accurately,and efficiently obtain patient genome and transcriptome information.The new staging system based on genomic data has been applied in practice and showed a good predictive value of prognosis and drug effect.However,few studies have focused on the transcriptomic characteristics,prognostic molecular model,and immune microenvironment for colorectal cancer patients with liver metastases.To comprehensively evaluate the expression profile,immune microenvironment characteristics of primary colorectal cancer and liver metastases and look for factors affecting patients’ prognosis.We performed transcriptome sequencing of the primary tumor and liver metastases and analyzed the immune microenvironment.Material and methods:In this study,specimens and related clinicopathological data of 20 patients of colorectal cancer patients with liver metastases were collected.We performed RNA-seq on the primary colorectal cancer,liver metastases and matched normal colorectal and liver tissues of each patient.After upstream data processing,we obtain the expression matrix of each sample.We subsequently got the differentially expressed genes of primary colorectal cancer VS liver metastases,liver metastases VS normal liver tissue,and primary colorectal cancer VS liver metastases.We performed an enrichment analysis of differentially expressed genes.Considering that the tissue-specific expression profile may affect the results,we deleted the tissue-specific gene set of liver or intestine tissue in the comparative analysis of colorectal cancer VS liver metastases.In terms of the microenvironment,we used six algorithms to calculate the proportion or abundance of different cells in the microenvironment of primary colorectal cancer and liver metastases.We also calculated the ESTIMATE score,immune type,and CMS type to evaluate the patient’s immune microenvironment status and expression profile characteristics.Finally,combining the patients’ clinical-pathological data and bioinformatics results,we conducted a comprehensive return to school to screen the factors that affect the prognosis of the patients.Finally,we conducted a thorough analysis and screened the possible prognosis factors combining the clinical-pathological data and bioinformatics results.Results:1.When analyzing the differential genes between different tissues,we found that the PCA results between different tissues,the results of the differential genes of liver metastases VS normal liver tissues,and the results of differential genes of primary colorectal cancer VS liver metastases showed tissue specificity.2.In order to compare the differential genes of liver metastasis VS colorectal cancer primary,We used the method of TSGS to correct tissue-specific gene set.It was found that the EMT ability of liver metastases was decreased and the MYC-related pathways may be enhanced.3.When using the ESTIMATE algorithm to evaluate the stromal score,immune score and tumor purity of each tumor sample,we found that the tumor purity of some samples was low.We found that the immune score and stromal score of CMS4 specimens were significantly higher than other subtypes.So the tumor purity calculated by ESTIMATE algorithm was lower.Therefore,we believe that the low tumor purity of CMS4 type is the characteristics of its mesenchymal infiltrating type expression profile.Survival analysis found that high ESTIMATE score and low tumor purity were risk factors for the prognosis of patients.4.Type CMS2 occupies a dominant position in the primary and metastatic foci of colorectal cancer.There is no CMS3 type in liver metastases.We believe that it may be caused by the higher metabolic expression profile in the liver that masks the expression profile of CMS3 metabolic disorder.Further research is needed.The CMS classification of liver metastases is related to the prognosis of patients.The prognosis of CMS type 2 and 4 is better than that of type 1 and 3.The immunophenotype of the primary colorectal cancer is related to the patient’s prognosis,and the immunophenotype 1 is better than type 2.5.Whether in the primary colorectal cancer or liver metastases,a variety of algorithms have shown that the increase in the proportion of infiltrating neutrophils in the tumor microenvironment is related to the poor prognosis of patients.Conclusion:1.Liver metastases of colorectal cancer retain the characteristics of the expression profile of normal colorectal tissue.It need to be corrected in the analysis.2.Compared with primary colorectal cancer,liver metastases of colorectal cancer have decreased EMT ability,and MYC-related pathways may be enhanced3.Regardless of the primary tumor or metastasis of colorectal cancer,the expression profile of CMS4 type has apparent characteristics of interstitial infiltration.Its immune score and stromal score evaluated by the ESTIMATE method are significantly higher than other subtypes.It may not be accurately assessed based on the ESTIMATE method—the purity of this subtype of tumor.For colorectal cancer,the ESTIMATE method score and its calculated tumor purity are only a feature of its expression profile.4.The immune classification of primary colorectal cancer and the CMS classification of liver metastasis is related to patients’ prognosis.5.The increase of neutrophils in the tumor microenvironment is related to patients’ poor prognosis. |
PDF Full Text Request