Differentiation Of Benign And Malignant Ascites And Establishment Of Mathematical Model

Background and purpose:Ascites is a common clinical symptom,ascites can be divided into benign ascites and malignant ascites,benign ascites can be divided into tuberculous ascites and non tuberculous benign ascites.The differentiation of benign and malignant ascites is closely related to the choice of clinical treatment plan and prognosis,but it is still a difficult problem to distinguish the part of benign and malignant ascites in clinical. At present,the most commonly used method is exfoliative cytology,although it has a high the specificity of diagnosis,but its sensitivity and accuracy are low,it is limited in clinical application.lt is urgent to find a simple and highly accurate method for the diagnosis of ascites now.In clinic,the accuracy of the diagnosis of ascites is still low by detecting the single laboratory index, the detection of multiple indicators can improve the accuracy of diagnosis,but this is a complicated processd.This paper is a retrospective study of clinical cases.The single factor analysis is used to screen out the meaningful laboratory indexes for the differential diagnosis of ascites,and then set up the mathematical diagnostic model by multivariate logistic regression analysis.The selected laboratory indicators as a whole,with the quantitative data to reflect the possibility of the disease in order to improve the sensitivity,specificity and accuracy of diagnosis.Methods:In this paper, retrospective analysis was conducted to analyze the clinical data of 172 patients with ascites who were hospitalized from Department of Gastroenterology,the First Hospital of Lanzhou University from January 2010 to December 2016.The serum levels of AFP,CEA, CA125,CA199, CA724, LDH, GLU, ALP, GGT,TP, ALB, ascites AFP, CEA, CA125,CA199, CA724, LDH, GLU, ALP, GGT, TP, ADA and TP difference of 23 laboratory indexes werecollected.First,these indexes were analyzed by single factor analysis,the statistical significance of the indicators were used as independent variables.Benign and malignant ascites,tuberculosis and non tuberculous ascites were used as dependent variables, respectively.Then the multivariate logistic regression analysis was used to establish the mathematical model for the diagnosis of malignant ascites and tuberculous ascites,and the diagnostic performance was analyzed by ROC curve.Results:1.The three major causes of ascites were decompensated cirrhosis,malignant tumors and tuberculous peritonitis.2.The age of malignant ascites patients were higher than the benign ascites patients,the morbidity rate of female were higher than the male;Tuberculous ascites patients were younger than the non-tuberculous ascites patients,there was no significant difference between male and female.3.Differential diagnosis of benign and malignant ascites:we had gotten an equation for the diagnosis of malignant ascites,the regression equation:P1 = 1/[1 + e^{-（- 3.859 + 0.082X1+0.001X2+ 0.003X3）}].Among them,X1=ascites CEA,X2= ascites CA125,X3=ascites LDH,P1 was the predicted probability. e was natural logarithm.The area under ROC curve of the mathematical diagnosis model was 0.944,the best of cutoff value is 0.515,the sensitivity was 77.78%,the specificity was 98.31%,the accuracy was 91.86%, the missed diagnosis rate was 22.22%,the misdiagnosis rate was 1.69%,the Youden index was 76.09%,the positive predictive value was 95.45%,the negative predictive value was 90.63%,the positive likelihood ratio was 46.023,and the negative likelihood ratio was 0.23.4.Differential diagnosis of tuberculous and non tuberculous ascites:we had gotten an equation for the diagnosis of tuberculous ascites,the regression equation:P2 = 1/[1 + e^{-（-7.466-0.005X1+0.104X2 + 0.010X3 + 0.130X4）}] Among them, X1=ascites LDH, X1= ascites TP,X3, ascites GGT,X4= ascites ADA,P2 was the predicted probability,e was natural logarithm.The area under ROC curve of the mathematical diagnosis model was 0.978,the best of cutoff value was 0.302,the sensitivity was 93.18%, the specificity was 94.53%,the accuracy was 94.19%, the missed diagnosis rate was 6.82%,the misdiagnosis rate was 5.47%, the Youden index was 87.81%,the positive predictive value was 85.42%, the negative predictive value was 97.58%,the positive likelihood ratio was 17.03, and the negative likelihood ratio was 0.07.5.Comprehensive identification of malignant ascites and tuberculous ascites: if Pre-1>0.515,Pre-2<0.302;highly suspected malignant ascites;Pre-1<0.515, Pre-2>0.302,highly suspected tuberculous ascites,Pre-1 <0.515,Pre-2<0.302,diagnosis of benign non-tuberculous ascites;Pre-1>0.515,Pre-2>0.302,comparing which was closer to 1 that was more consistent with the diagnosis of one of the diseases.Conclusion:Logistic regression analysis was used to establish a mathematical model.It linked the selected laboratory index into an organic whole and used the quantitative data to identifiy the properties of ascites.It was improve the diagnostic sensitivity,specificity and accuracy.