Building Of Mode Diagnosing Liver Malignancy And Mode Predicting Posthepatectomy Liver Failure

Object: to explore the joint application of a number of tumor markers in thediagnosis of hepatic malignant lesions and establish the diagnostic formula; toevaluate liver failure incidence after hepatectomy by using the commonly liverfunction examination, and estabilish a prediction formula. Methods:99patients withhepatectomy admited in the Affiliated Hospital of LuZhou Medical College betweenApril12009to December312011were enrolled in this study.(1)99patients wasdivided into benign disease group and malignancy disease group according topathological diagnosis, alpha-fetoprotein, carcinoembryonic antigen, carbohydrateantigen19-9, alpha-fucosidase, serum ferritin were compared by using the Wilcoxonrank sum test between two groups and statistical difference variables was entered intothe logistic regression model after natural logarithmic transformation and establisheda formula predicting malignant hepatic lesions, the model would be evaluated byusing ROC curve.(2)99patients was divided into liver failure group and thenon-liver failure group according posthepatectomy recovery. Demographicscharacteristic, liver function, renal function, blood lipids, blood routine examination,electrolytes, coagulation, surgical situation, and other projects was compared by t test,χ2test, rank sum test between the two groups before hepatectomy, statisticallydifferent projects were entered in the logistic regression model and established aformula predicting liver failure after hepatectomy, the model was be tested by ROCcurve to evaluate the diagnostic performance. Results:(1)66patients with malignant lesions in99cases, including:50patients with hepatocellular carcinoma,4patientswith cholangiocarcinoma,2patients with hepatic mixed carcinoma,1patient withhepatic sarcomatoid carcinoma,1patient with liver adenocarcinoma,3patients withmetastasis liver cancer,1patient with gallbladder carcinoma permeating liver.37patients of benign lesions, including:11patients with hepatic bile duct stones,25patients with hepatic hemangioma,1patient with hepatic inflammatory pseudotumor.statistical Variable difference between the two groups is alpha-fetoprotein,carcinoembryonic antigen, alpha-L-fucosidase, serum ferritin (P <0.10), they enteredin the logistic regression model, carcinoembryonic antigen is removed. Finally,alpha-fetoprotein, alpha-L-fucosidase, serum ferritin went into the regressionequation, the regression coefficients are:0.691,2.397,1.337,respectively. the modeldiagnosing malignant liver lesions: P=1/[1+e-(-13.152+0.691×ln(alpha-fetoprotein,ng/ml)+1.337×ln(serumferritin,ng/ml)+2.397×ln(alpha-L-fucosidase,U/L)]. the best cutoff point of the model is0.60, In whichsensitivity and specificity of the model, the total coincidence rate, the rate ofmisdiagnosis and missed diagnosis rate, Youden index, positive predictive value,negative predictive values are:0.897,0.952,0.920,0.048,0.103,0.8490.963,0.870.(2)15patients with posthepatectomy liver failure is in the99patients after hepatectomyliver failure in15cases, of which1died,1patient with prothrombin time activity lessthan50%and elevated bilirubin more than50μmol/L after5th days postoperatively,8patients with postoperative refractory ascites,4patients with refractory ascites andelevated bilirubin more than50μmol/L after5th days postoperatively.1patient withelevated bilirubin more than50μmol/L after5th days postoperatively. Thevariables of statistic difference between liver failure group and the non-liver failure group is: the nature of the lesion, cirrhosis, hepatitis B surface antigen,alpha-fetoprotein, aspartate aminotransferase, direct bilirubin, total bile acids,alkaline phosphatase, lipoprotein (a), white blood cells, platelets, serum albumin,cholinesterase, alpha-L-fucosidase, serum calcium, prothrombin time, internationalnormalized ratio of prothrombin time, prothrombin time activity, activated partialthromboplastin time, respectively(P<0.05), which will be entered in the logisticregression model, and ultimately total bile acids, cholinesterase, prothrombin time,hepatitis B surface antigen entered into the regression equation, the regressioncoefficients were:2.095,-2.684,12.936,1.985,respectively. The model predicting liverfailure after hepatectomy is: P=1/[1+e-(2.095×ln(TBA,)-2.684×ln(CHE)+12.936×ln(PT)+1.985×HBsAg(thepositive is1, the negative is0)-18.33]. the best cutoff point of the model is0.20, in whichsensitivity and specificity of the model, the total coincidence rate, the rate ofmisdiagnosis and missed diagnosis rate, Youden index, positive predictive value,negative predictive value are:0.923、0.867、0.875、0.133、0.077、0.790、0.545、0.985. Conclusion: The joint application of alpha-fetoprotein, serum ferritin,alpha-L-fucosidase can improve accuracy in diagnosing malignant hepatic lesions;high level total bile acid, low level cholinesterase, prothrombin time prolonged,hepatitis B surface antigen-positive was risk factors for posthepatectomy liver failure,which can be used to predict the incidence of postoperative hepatic failure.