The Methodology Of SROC Technologies And Its Medical Application In Diagnostic Test

Meta analysis is a statistical method which can synthesize the research results of the same research, the certain conditions. Meta analysis is one of the most important methods of reliability and accuracy of the evaluation.However, SROC technology is a Meta analysis method which is not generally applied in the diagnostic test, which according to their diagnostic odds ratio, using a comprehensive curve to express the same test indexes from different tests, this curve is called the SROC curve. SROC curve organic integrate the multiple independent diagnostic information of ROC curve, certainly will get better quality of the performance evaluation of the diagnostic test. Recently, the technology of fitting SROC curve has made certain progress on the model. Based on this, Meta analysis of diagnostic test about the different threshold value(that is, the classification results) can be divided into three types: single variable, bi-variable and multi-variable. We explore the key is the SROC based on bivariate random effect model in this study.At present, because of the single variable SROC technology( namely, fixed effect model)without making full use of original datainformation, so it ignores the differences of the effect of different diagnostic threshold and hides the real diagnosis performance of diagnostic test. In order to solve the single variable SROC technical shortcomings, this study propose the SROC based on bivariate random effect model, in fact,contains 5 kinds of technology:: η on ξ、ξ on η、D on S、R&G(Rutter and Gatsonis), major axis method to solve these shortcomings.In this paper, we randomly selected two literature of Meta-analysis on diagnostic tests, used nlmixed and mixed statements in the proc process step of SAS to do parameter estimation and model fitting. The results were evaluated by the area under the curve(AUC) and the Youden index. In addition, in order to better explore the different SROC technology, the diagnostic test of the total bile acid in intrahepatic cholestasis of pregnancy was studied for the first time. Stata12.0 software was used to analyze the heterogeneity.Research results show: 1) The performance of Rutter and Gatsonis(R&G) was better than other methods through two examples analysis data(example 1:AUC=0.875,Youden index:0.75; example 2:AUC=0.780,Youden index:0.66;). However,ξ on η has the worst effect(example 1:AUC=0. 107; example 2:0.448;). 2) Based on the evidence of evidence-based medicine, collected a total of 6 references, and involved a total of 1394 objects(834 healthy pregnant women and 560 patients with ICP), which total bile acids in pregnant women range from 6.02-19.2umol/L. The data was divided into groups based on the country(China or foreign), gestation period(second or third trimester) and study design(prospective or retrospective) using subgroup analysis. The results found that the diagnostic odds ratio for each of the subgroups was statistically significant regardless of country, study design, or gestation periods, we could detect that there were obvious heterogeneity among studies( I2 >50%, p < 0.05). Egger results showed that the Meta-analysis results that there was no publication bias(p = 0.190). 3) In the evaluation of diagnostic value, AUC of SROC 5 techniques of R & G still was high(η on ξ:0.711;ξ on η:0.448;D on S :0.730;R&G:0.801;Major axis method :0.703, Youden index:0.56). Therefore, the accuracy value of TBA was higher in second or third trimester, especially in the third trimester pregnant women. 4) While Youden index was between 0.50 and 0.8, R&G method the accuracy is perfect.SROC5 technology methods are based on the linear model of bivariate random effect model. We can separate the effect of degree of sensitivity and specific estimates. Therefore, the advantage of five techniques can solve both the inter-study heterogeneity and between- study heterogeneity,which is the development of traditional SROC technology. Simultaneously,our study implement the results combined with the excact estimation of optimal variance, to make the model results more reliable. It helps to improve the accuracy of diagnostic test. At the same time, this studycomprehensive evaluate the diagnosis performance for the development of total bile acid, based on conventional means of evidence-based and five SROC curve technolog. This study not only scientifically evaluate the diagnostic value of total bile acid in ICP, but also discusses the five characteristics of SROC technology, which provides a reference to subsequent development of SROC technology.