Development And Evaluation Of Diagnostic Models Basedlin Clinical Data For The Differentiation Of Hepatocellular Carcinoma And Benign Focal Hepatic Lesions

Background:With the spread development of various imaging examinations,more and more focal hepatic lesions(FHL)have been detected clinically.However,it is critical to make accurate differential diagnosis due to the totally different of the treatment and prognosis for different FHL.It remains a challenge for differentiating malignant and benign FHL just through individual laboratory indexes or imaging features.Therefore,it is of great significance to establish a diagnostic model with high accuracy and efficiency for the clinical diagnosis and treatment.Objectives:To compare the diagnosis value of single,similar clinical indexes and multiindex model for hepatocellular carcinoma and benign FHL.To build an effective predicting model for differentiating hepatocellular carcinoma and benign FHL,and to verify the diagnostic efficiency of the model finally.Methods:1.Clinical data collection: Retrospectively reviewed the clinical data of patients with FHL who hospitalized in the First Affiliated Hospital of Nanchang University between June 2013 and June 2018,including sex,age,smoking and drinking etc.Laboratory findings(complete blood count,blood biochemistry and tumor markers etc.),surgical and pathological data,and other imaging results(CT,US or MRI)were collected.2.Analysis of univariate diagnostic value: The patients with FHL were divided into hepatocellular carcinoma group and benign FHL group.The receiver operating characteristic(ROC)curves of hepatocellular carcinoma and benign FHL were established by clinical indexes,calculated the area under their respective cures and compared the diagnostic efficiency.3.Analysis of multivariate diagnostic value: Logistic regression modeling and ROC curve analysis were performed with tumor markers,liver function examination and imaging characteristics as variables.The diagnostic value of various models of AUROC was used to identify hepatocellular carcinoma and benign FHL.4.Multivariate model developing and diagnostic value evaluation: All FHLs were randomly divided into training set and test set according to 2:1,and the clinical indicators of training set patients were used to establish a differential diagnosis model of FHL.ROC curve analyze the diagnostic value of the model,and use the test set to verify the diagnostic value of the model.Results:1.Clinical data: A total of 433 FHL patients were included in the study,with an average age of 53.9 ± 12.6 years,280 males and 153 females,with a male to female ratio of 1.8:1,including 243 cases of hepatocellular carcinoma,100 cases of hepatic hemangioma,47 cases of liver abscess and 43 cases of hepatic cyst.The single index of tumor markers,liver function examination and imaging characteristics were observed statistically significant(P<0.05-0.001).2.The differential diagnostic value of single,similar and multi-index model for the identification of hepatocellular carcinoma and benign FHL: 433 patients with FHL were divided into benign FHL group(n=190)and hepatocellular carcinoma group(n=243).The AUROC of the single clinical index in diagnosing hepatocellular carcinoma and benign FHL was calculated respectively,and we found that the AUROC of AFP(0.942),AST(0.854),PTINR(0.814),GGT(0.798)and unclear boundary(0.779)were larger than others and the diagnostic value of hepatocellular carcinoma was relatively high.A ROC curve were developed with the similar indexes,we found that the AUROC of serum tumor markers(0.964)was larger than that of liver function index(0.880)and imaging characteristics(0.849),and with a relatively better diagnostic value.A ROC curve model was built with the multi-indexes and we found that the AUROC in differentiating hepatocellular carcinoma and benign FHL was 0.988,the diagnostic efficiency of which was significantly superior to single and similar indexes.3.The diagnostic value of multivariate diagnostic Model for the identification of hepatocellular carcinoma and benign FHL: 433 patients with FHL were randomly divided into training set(n=289)and test set(n=144).Except for the age and smoking history(P<0.05),other clinical features between the two groups were observed no significant difference(P>0.05).Multivariate stepwise logistic regression analyses based on the training set indexes showed that history of HBV,AFP,CEA,GGT,unclear boundary and solitary lesion which entered into this model were the independent risk factors of hepatocellular carcinoma(P<0.05).A ROC curve was constructed basing on the result and the AUC of ROC cure for the model was 0.979.The diagnostic efficiency of the diagnosis model is 0.983 in the test set,and the fitting degree and the calibration curve between the two models was good,which indicates that a reasonable efficiency of the diagnostic model established in this study.Decision curve analysis showed that the model was clinically useful.Conclusions1.The diagnostic efficiency of the model for differentiating hepatocellular carcinoma and benign FHL which combined with serum tumor markers,liver function indexes and imaging features,was significantly superior to that of the single or the similar indexes.2.HBV history,solitary lesion,unclear boundary,AFP,CEA and GGT were independent diagnostic factors of hepatocellular carcinoma.3.A diagnostic model was constructed basing on history of HBV,solitary lesion,unclear boundary,AFP,CEA and GGT,and providing a good diagnostic value for differentiating hepatocellular carcinoma and benign FHL.