网状Meta分析图形结果解读

2015,15(1):103109•103•©2015*I430060MetaMetaBrooks-Gelman-RubinMetaMetaResultInterpretationofNetworkMeta-analysisYIYue-xiong,ZHANGWei*,LIUXiao-yuan,ZHANGJuan,ZHUDing-jun,LVQiong-yingFirstDepartmentofGynecology,RenminHospitalofWuhanUniversity,Wuhan430060,ChinaAbstractComparisonamongmultipleinterventionshasbeenrealizedduetothedevelopmentofnetworkmeta-analysisandsofarmanystudieshavereporteditsimplementationprocess.However,itsresultsarerarelyinterpretedindomesticstudiesatpresent.Thisarticleinterpretstheresultsoftraceplots,densityplots,Brooks-Gelman-Rubindiagnosisplots,rankogram,surfaceunderthecumulativeranking,andnetworkplots,toprovidereferencesandassistanceforfurtherresearchregardingnetworkmeta-analysis.KeywordsNetworkmeta-analysis;Graphicresult;InterpretationDOI:10.7507/1672-2531.201402631984Email:[e-mail]*Email:[e-mail][1]Meta-MarkovChainMonteCarloMCMC[2-4]Rgemtc[5,6]Brooks-Gelman-Rubin[7]Winbugsmodelmtc.runmodel-mtc.model(network,type=consistency,factor=2.5,n.chain=3,linearModel=random)results-mtc.run(model,sampler=R2WinBUGS,n.adapt=5000,n.iter=20000,thin=1)GIsuraplotsuraplot2Meta2.1traceplotMCMCcon-vergence-MarkovChainMonteCarlochainMCMCchain[8,9]MCMCChinJEvid-basedMed2015,15(1):103-109•104•CJEBM©2015EditorialBoardofChinJEvid-basedMed[10][11]11a050MCMC1b02 0001a1c1 000501b1 0201 0401d5 000MCMC1e5 00020 000MCMC2.2densityplot[12][13]BandwidthBandwidth0[14]Bandwidth[15]Bandwidth[16]2catmodel$codeGId.G.Idnorm0,prior.precNBandwidth2a2b11n.adaptn.intera050b02 000c1 00050d5 0005 000e5 00020 0002015,15(1):103109•105•©2015 70.113 495%CImeana=0.451 095%CIa=–0.122 71.067 5meanb=0.3741 195%CIb=–0.175 30.914 52c5 0002bBandwidth0.038 742bmeanc=0.334 795%CIc=–0.154 810.826 02d50002cBand-widthBandwidthd=0.038 852cmean95%CImeand=0.343 595%CId=–0.151 810.840 6n.adapt=5 000n.iter=5 0002e20 000Bandwidthe=0.029 552dmeane=0.344395%CIe=–0.152 200.840 640 0002f2eBandwidthBand-widthe=0.025 7meanf=0.342 095%CIf=–0.156 700.838 0n.adapt=5 000n.inter=20 0002.3Brooks-Gelman-RubinBrooks-Gel-man-RubindiagnosisplotBrooks-Gelman-Rubinpotentialscalere-ductionfactorPSRF[17]BrooksGelman[18]shrinkfactor97.5%n[19]22n.adaptn.interbandwidthmean2.5%97.5%a0500.109 70.451 0–0.122 71.067 5b1 000500.113 40.374 11–0.175 30.914 5c1 0005 0000.038 740.334 7–0.154 810.826 0d5 0005 0000.038 850.343 5–0.151 810.840 6e5 00020 0000.029 550.344 3–0.152 200.840 6f5 00040 0000.025 70.342 0–0.156 700.838 0g10 00020 0000.029 830.337 7–0.161 600.839 4ChinJEvid-basedMed2015,15(1):103-109•106•CJEBM©2015EditorialBoardofChinJEvid-basedMed[18]gemtcValkenhoef[17]PSRF11.051PSRF1.05BrooksGelman[19]3n197.5%n1PSRF13Brooks-Gelman-Rubin50n.iter=500PSRFgelman.diag(results)3a100500shrinkfactor97.5%1PSRF1.183b10n.adapt=1 000n.iter=5 00097.5%3000PSRF1PSRF=1.0233a3cn.adapt=5000n.iter=20000500097.5%1PSRF1Brooks-Gelman-RubinPSRF2b3000PSRF1PSRF=1.0297.5%3 000PSRF=1.02Brooks-Gelman-Rubin32.4Rankprobabilitiesandrankogram4n.adapt=5 000n.iter=20 000ranks-rank.probabilityresults,preferreddirection=1rankprobabilities14n[20]ABA1110120.3%136.6%1493.2%B10.5%21.2%32.5%44.1%55.8%67.2%78.7%810.2%911.3%1012.3%1113.7%1214.0%138.6%140.0%Arankogram[21]42.5cumulativerankingplotandsurfaceunderthecu-mulativeranking[22][18]53Brooks-Gelman-Rubin3Brooks-Gelman-RubinPSRFn.adaptn.interPSRFa1005001.18b1 0005 0001.02c5 0002 00012015,15(1):103109•107•©2015=cumj,b∑a-1b=1a-1jbacumj,bjb[23]SUCRA010SUCRA1SUCRA10[24]SUCRA[25]5SUCRACSUCRA81.42%F78.88%CFHIDLJKBEGA2.6networkplotMetaFruchtermanReingold[26]6gemtcMeta[27]3gemtcsucraplotMetaBrooks-Gelman-RubinBrooks-Gelman-RubinBrooks-Gelman-RubinPSRFABCDEFGHIJKLMN0.00.20.40.60.8414414[,1][,2][,3][,4][,5][,6][,7][,8][,9][,10][,11][,12][,13][,14]A0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0030.0660.932B0.0050.0120.0250.0410.0580.0720.0870.1020.1130.1230.1370.1400.0860.000C0.3570.2630.0920.0540.0390.0290.0250.0230.0210.0220.0230.0290.0200.002D0.0230.0440.0750.1000.1140.1120.1060.0980.0890.0790.0690.0570.0370.000E0.0000.0020.0070.0190.0400.0640.0950.1260.1470.1580.1530.1240.0640.000F0.0890.1690.2730.1990.1140.0650.0400.0230.0130.0080.0040.0020.0000.000G0.0000.0000.0000.0010.0040.0120.0270.0540.1000.1630.2280.2510.1590.000H0.2810.2120.0860.0550.0420.0350.0300.0270.0280.0280.0330.0490.0720.021I0.1640.1350.1590.1170.0870.0660.0560.0470.0420.0380.0360.0370.0170.001J0.0410.0560.0750.0830.0850.0820.0770.0740.0740.0770.0820.0960.0970.002K0.0020.0110.0270.0570.0930.1240.1390.1440.1380.1180.0820.0470.0170.000L0.0040.0140.0350.0710.1180.1550.1670.1550.1250.0850.0470.0200.0040.000M0.0060.0160.0230.0320.0360.0400.0420.0510.0560.0680.0880.1400.3580.042N0.0290.0660.1220.1700.1700.1430.1080.0760.0550.0340.0180.0070.0020.000ChinJEvid-basedMed2015,15(1):103-109•108•CJEBM©20