Accuracy Of Preoperative Grading In Brain Gliomas By Arterial Spin Labeling Based On Meta-analysis

Purpose: To carry out a Meta-analysis on studies that used arterial spin labeling(ASL)for preoperative grading of brain gliomas,so as to assess the value of ASL for predicting high/low-grade brain gliomas.Materials and Methods: 1.Relevant Chinese and English studies published between January 1992 and November 2017 were searched in Pub Med,Cochrane Library,Web of Science,China Biomedical Literature(CBM),CNKI,and Wanfang Database.The studies that met the criteria were included and their data and parameters were extracted.2.For dichotomous data,Meta-Disc 1.4 software was used to perform heterogeneity test and quantitative syntheticanalysis,and the source of heterogeneity was identified through meta-regression;Rev Man 5.3software was used to calculate the four-fold table through indirect data,as Quality Assessment of Diagnostic Accuracy Studies tool(QUADAS-2)was employed.Begg and Egger tests were performed with Stata 12 software to determine the presence or absence of publication bias.3.For continuous data,Rev Man 5.3 was used for QUADAS-2.Stata 12 software was used to perform heterogeneity test and quantitative synthetic analysis.Meta-regression was carried out on the heterogeneous research groups to identify the source of heterogeneity,and Begg and Egger tests were performed to determine the presence or absence of publication bias.Results: 1.There were a total of 14 articles that met the inclusion criteria of dichotomous data.The diagnostic data were the maximum tumor blood flow(TBFmax)and three normalized perfusion parameters:maximum tumor blood flow/contralateral normalized white matter blood flow [n TBF(WM)],maximum tumor blood flow/contralateral normalized gray matter blood flow [n TBF(GM)],and maximum tumor blood flow/contralateral normalized hemisphere blood flow [n TBF(H)].The TBFmax、n TBF(WM)and n TBF(GM)group showed no heterogeneity,but the n TBF(H)group had heterogeneity.Meta-regression failed to identify the source of heterogeneity.The diagnostic sensitivities were summarized as 0.90(95%CI 0.84~0.95),0.87(95% CI 0.82~0.91),0.89(95% CI 0.82~0.94),and 0.93(95% CI 0.87~0.96),while the specificity were summarized as 0.78(95% CI 0.70~0.85),0.88(95% CI 0.82~0.93),0.89(95% CI0.79~0.95),and 0.87(95% CI 0.78~0.93).The areas under the SROC curve corresponding to each group were 0.9371,0.9378,0.9453,and 0.9142,respectively.No publication bias was detected among researches of TBFmax,n TBF(WM)and n TBF(GM)group,while it was detected among researches of n TBF(H)group.2.There were a total of 22 articles that met the inclusion criteria of continuous data.The diagnosticparameters were same asthe dichotomous data.Heterogeneity test of quantitative data showed heterogeneity between the TBFmax and n TBF(WM)groupsand good homogeneity between the n TBF(GM)and n TBF(H)groups.Meta-regression were conducted on the data of TBFmax and n TBF(WM)groups,suggesting that the major source of heterogeneity was the diverse ASL techniques.In TBFmax group,SMD=1.559(1.235~1.882);in n TBF(WM)group,SMD=1.611(1.287~1.936);in n TBF(GM)group,SMD=1.846,(1.564~2.127);in n TBF(H)group,SMD=1.691(1.443~1.938).Subgroup analysis was performed on the TBFmax group based on different ASL techniques,SMD of CASL and 3D p CASL subgroup was 1.405(1.117~1.693),SMD of PASL subgroup was 2.738(1.997~3.479).The maximum perfusion values and the three normalized parameters of the high-grade gliomas were significantly higher than those of the low-grade ones,all the results were significant statistically.Conclusion: For the dichotomous data groups,all four parameters could provide accurate prediction of glioma grades preoperatively,but n TBF(H)group has publication bias,and their diagnostic value needs to be further verified in the future.For the continuous data groups,all four parameters could accurately grade the gliomas,the TBF perfusion values of the high-grade gliomas were significantly higher than those of the low-grade ones.Both parts of the data showed that the diagnostic efficacy of normalized parameters were better,especially the n TBF(GM)group.