Development And Validation Of A Preoperative Diagnostic Model For Microvascular Invasion And A Predictive Model For Postoperative Recurrence In Hepatocellular Carcinoma

Objective:Hepatocellular carcinoma(HCC)is a deadly disease because of high recurrence rate and microvascular invasion(MVI)is an important risk factor for early recurrence of HCC.However,the diagnosis of MVI can only be obtained from pathological examination of surgical specimens,which is lack of guidance in perioperative management of HCC patients.In this study,we will develop a preoperative diagnostic model for MVI and explore the implication of this model in surgical plans preoperatively.Material and methods:The clinical and imaging features of patients who underwent surgery in Cancer Hospital,Chinese Academy of Medical Sciences during January,2015 and December,2017 were retrospectively analyzed.A total of 357 patients were assigned into training cohort(n=257)and validation cohort(n=100)according to surgery time.Independent-sample T-test,Chi-square test and Fisher exact test were applied to assess the difference between the two cohorts,univariate and multivariate Logistic regression analysis were conducted to reveal the risk factors for MVI.And then a diagnostic nomogram was developed and validated.The discrimination and calibration were analyzed using area under receiver operating characteristic(AUROC)curve and calibration curve with Hosmer-Lemeshow test.Decision curve analysis was performed to evaluate the prospect of the nomogram in clinical application.Kaplan-Meier survival analysis with log-rank test was conducted and the implication of the nomogram in surgical plans were revealed.Results:The clinical and imaging features were balanced between training cohort and validation cohort,and postoperative pathology revealed MVI in 140(39.2%)patients.Preoperative imaging features including:tumor diameter,intratumoral arteries,tumor type,and serum alpha fetoprotein were independent risk factors for MVI.A nomogram incorporating these factors was developed and the AUROC in training cohort and validation cohort were 0.803(95%CI,0.746-0.860)and 0.814(95%CI,0.720-0.908),respectively.Calibration curve with Hosmer-Lemeshow test indicated favorable calibration in the two cohorts(P=0.84,and P=0.79).Decision curve analysis revealed promising clinical application of the diagnostic nomogram.In addition,patients with nomogram predicted MVI suffered higher recurrence rate(P<0.001).Anatomic resection improved recurrence free survival in small HCC patients with nomogram predicted MVI.Conclusion:The nomogram achieved favorable performance in preoperative MVI diagnosis and was promising in clinical application.Patients with nomogram predicted MVI suffered from higher recurrence risk.And anatomic resection was recommended for patients with nomogram predicted MVI.Objective: Recurrence is the leading cause of failure in the treatment of hepatocellular carcinoma(HCC).The existing staging systems and prognosis models are less effective in risk stratification of recurrence in patients with HCC.The aim of this study was to explore biomarkers associated with recurrence and to establish a prognosis model to stratify the recurrence risk of HCC patients.Material and methods: The clinical and pathological variables of 371 HCC patients in TCGA were analyzed to explore the risk factors for early recurrence in HCC patients.Differentially expressed genes in patients with early relapse or not were analyzed using DESeq2,and the GO enrichment analysis of differential genes was performed.Univariate COX regression analysis,LASSO regression combined with multivariate COX regression analysis were conducted to screen genes associated with recurrence-free survival in HCC patients.Univariate and multivariate COX regression analysis revealed the associated factors of disease-free survival in HCC patients.Furthermore,we established a comprehensive prognosis model combing gene prognosis model and TNM staging system,time-dependent ROC curve and decision curve analysis were used to evaluate the application of the three models in the TCGA dataset and GSE14520 dataset.Results: The ECOG score(P=0.025),Child-Pugh grade(P=0.040),and TNM stage(P<0.001)were significantly associated with early recurrence of liver cancer.One hundred and seventy-nine differentially expressed genes(73 up-regulated genes and 106 downregulated genes)were screened with a fold change > 4 and adjust P value <0.01.GO enrichment analysis indicated that the differentially expressed genes were closely related to immune function.CD79 A and PDZRN4 were candidate independent prognostic genes for early recurrence of HCC and validated in TCGA and GSE14520 datasets.Univariate and multivariate COX regression analysis revealed that TNM staging system and gene prognosis model,were independent prognostic factors for recurrence-free survival(P=0.001,P =0.043);time-dependent ROC curves demonstrated similar predictive efficacy between the gene prognosis model and TNM staging system at 1 st 2nd,and 3rd year after surgery(TCGA dataset: 0.68 vs.0.76,P=0.106;0.74 vs.0.71,P =0.675 and 0.71 vs.0.68,P= 0.554;GSE14520 dataset: 0.63 vs.0.65,P=0.812;0.69 vs.0.69,P=0.923 and 0.63 vs.0.68,P=0.258);comprehensive prognosis model can improve the predictive efficacy of TNM staging system(TCGA dataset: 0.81 vs.0.76,P=0.033;0.82 vs.0.71,P<0.001 and0.77 vs.0.68,P=0.003;GSE14520 dataset: 0.67 vs.0.65,P=0.080;0.74 vs.0.69,P=0.001 and 0.70 vs.0.68,P=0.081).Decision curve analysis suggested that the gene prognosis model and the comprehensive prognosis model were promising in clinical application.Conclusion: We explored the clinicopathological features of early recurrence in HCC patients,screened biomarkers associated with early recurrence,and established a two-gene prognosis model for early recurrence in HCC patients.Further analysis showed that the model were similar to TNM staging system in predicting postoperative recurrence-free survival,and can improve the predictive performance of the TNM staging system.