循證醫(yī)學(xué)課件：循證醫(yī)學(xué)與Meta分析從理論到實(shí)踐

循證醫(yī)學(xué)與Meta分析從理論到實(shí)踐病例2007年2月中旬，長(zhǎng)海醫(yī)院消化內(nèi)科重癥胰腺炎監(jiān)護(hù)室35歲男性、暴飲暴食加大量飲酒后突發(fā)中上腹痛，伴惡心嘔吐體溫38.5度、心率120次/分、呼吸25次/分血常規(guī)：WBC13.0G/L；血淀粉酶1200U/mL(20-120U/mL)、肝腎功能基本正常、尿量正常。發(fā)病48h后胰腺CT診斷：重癥急性壞死性胰腺炎（酒精性）治療：臥床休息、禁食、補(bǔ)液、監(jiān)測(cè)重要臟器功能處理教授：患者體溫高、伴WBC升高，胰腺有壞死，要用抗生素預(yù)防后期的繼發(fā)感染。主治1：一旦發(fā)生繼發(fā)感染，死亡概率明顯增加。教授說(shuō)得有道理，應(yīng)該用抗生素。主治2：病因?yàn)榫凭?，理論上屬于化學(xué)性胰腺壞死，而非感染性壞死，現(xiàn)在也沒(méi)有證據(jù)表明已經(jīng)發(fā)生壞死感染，一旦過(guò)度使用抗生素，可能導(dǎo)致后期繼發(fā)真菌感染，更有可能增加死亡率，抗生素的使用應(yīng)該要謹(jǐn)慎。大查房時(shí)住院請(qǐng)示上級(jí)是否要使用抗生素EBMMethod

AssessyourpatientAskclinicalquestions對(duì)于還沒(méi)有發(fā)生胰腺壞死感染的重癥胰腺炎患者使用抗生素能否預(yù)防后期的感染、降低死亡率？Acutepancreatitis(AP)Mild(80%)Recoverywithin5to7days

APSevere

(20%)Highrateofcomplications&significantmortalityIsenmann.Pancreas2001CurrenttrendsImprovementintheintensivecareImprovementintheradiologicaltechniquesMorepatientssurvivedthefirst2wMortalityremainsstablefordecadesMODS2weekslaterFirst2weekCausesofmortalityInfectionHowtopreventinfectionEBMMethod

AcquirethebestevidenceAssessyourpatientAskclinicalquestions“Firstwedoanon-linesearch…”MethodIdentificationandSelectionofStudiesSelectioncriteriaStudydesign:randomized,controlledtrialPopulation:patientswithCTconfirmedANPIntervention:prophylacticantibioticsI.V.Comparison:placeboOutcomes:infectedpancreaticnecrosis&deathEarlyRCTs(1970s)FinchWT.AnnSurg.1976

InadequateseverityofdiseaseInadequatepancreaticpenetranceofampicillinSmallpatientsnumberInadequatestatisticalpowerHowesR.JSurgRes.1975CraigRM.AnnInternMed1975LaterRCTs(1980s)BegerHG.Gastroenterology.1986

AntibioticsCarbapenemsFluoroquinolonesCephalosporinsBuchlerM.Gastroenterology.1992Conflictingresults…ConflictingGuidelines…TakedaK.JHepatobiliaryPancreatSurg.2006YESNathensAB.CritCareMed.2004NONotsureWorkingPartyofBSG.Gut.2005Guideline1RCT1RCT2Guideline2RCT3ExpertBExpertAConfusedmind

What’sthetruth?EBMMethod

AcquirethebestevidenceAppraisetheevidenceAssessyourpatientAskclinicalquestionsQualitativeAnalysisAdequateUnclearNotGenerationofallocationsequenceAllocationconcealmentDoubleblindingCochraneHandbookforSystematicReviewsofInterventions4.2.5QuantitativeAnalysisReviewManagerUserGuide.2006TworeviewersindependentlyabstracteddataDifferenceswereexpressedasRRwithits95%CIStatisticalheterogeneityevaluatedbyCochrantestRandom-effectmodelwasusedwhenP<0.1PotentiallyrelevantarticlesidentifiedandscreenforretrievalofRCTs:n=359Studiesselectedformoredetailedevaluation:n=15Abstract/reviews/trialswereexcludedasnon-relevant:n=344Trialswereexcludedforvariousreasons:n=8RCTsselectedforinclusion:n=7AuthorYearSettingNo.BlindRiskofbiasDosagePederzoli1993Multi-center74singlehighImipenem0.5giv8hrlySainio1995Singlecenter60singlehighCefuroxime1.5giv8hrlySchwarz1997Singlecenter26singlehighOfloxacin0.2gb.d..&metronidazole0.5gb.d.Nordback2001Singlecenter39singlehighImipenem1giv8hrlyIsenmann2004Multi-center76doublelowCiprofloxacin0.4gb.d,&metronidazole0.5gb.dDellinger2007Multi-center100doublelowMeropenem0.5g8hrlyRokke2007Multi-center73nohighImipenem0.5g8hrlyCharacteristicsofincludedRCTs傳統(tǒng)的綜述AuthorYearSettingNo.BlindRiskofbiasResultsPederzoli1993Multi-center74singlehighPositiveSainio1995Singlecenter60singlehighPositiveRokke2007Multi-center73nohighPositiveNordback2001Singlecenter39singlehighPositiveIsenmann2004Multi-center76doublelowNegativeDellinger2007Multi-center100doublelowNegativeSchwarz1997Singlecenter26singlehighNegative傳統(tǒng)的綜述AuthorYearSettingNo.BlindRiskofbiasResultsPederzoli1993Multi-center74singlehighPositiveSainio1995Singlecenter60singlehighPositiveRokke2007Multi-center73nohighPositiveNordback2001Singlecenter39singlehighPositiveIsenmann2004Multi-center76doublelowNegativeDellinger2007Multi-center100doublelowNegativeSchwarz1997Singlecenter26singlehighNegative4項(xiàng)陽(yáng)性研究3項(xiàng)陰性研究4：3抗生素有效？辛普森悖論(Simpson'sParadox)秘書(shū)：今天有很多男生在校門(mén)口抗議，他們說(shuō)今年我們醫(yī)院女生錄取率是男生的兩倍，他們投訴我們學(xué)校錄取學(xué)生有性別歧視，院長(zhǎng)：我不是特別交代今年要盡量提升男生錄取率以免落人口實(shí)嗎？辛普森悖論(Simpson'sParadox)比賽100場(chǎng)足球以總勝率評(píng)價(jià)球隊(duì)實(shí)力強(qiáng)弱中國(guó)隊(duì)找20只歐洲強(qiáng)隊(duì)比賽，勝1場(chǎng)(5%)；找80只南亞弱隊(duì)比賽，勝40場(chǎng)(50%)；

總勝率41%(1+40/100)韓國(guó)隊(duì)找80只歐洲強(qiáng)隊(duì)比賽，勝20場(chǎng)(25%)；找20只南亞弱隊(duì)比賽，勝15場(chǎng)(75%)；

總勝率35%(20+15/100)。中國(guó)隊(duì)比韓國(guó)隊(duì)強(qiáng)？辛普森悖論(Simpson'sParadox)量與質(zhì)是不等價(jià)的，但量比質(zhì)容易測(cè)量，所以人們總是習(xí)慣用量來(lái)評(píng)定好壞。SARS患者死亡率輕癥(n)死亡(率)重癥(n)死亡(率)中醫(yī)院28%80例8例(10%)20例20例(100%)西醫(yī)院41%20例1例(5%)80例40例(50%)辛普森悖論(Simpson'sParadox)簡(jiǎn)單的將分組資料相加匯總，是不一定能反映真實(shí)情況的。為了避免辛普森悖論出現(xiàn)，就需要斟酌個(gè)別分組的權(quán)重，以一定的系數(shù)去消除以分組資料基數(shù)差異所造成的影響。傳統(tǒng)的綜述無(wú)法完成的，必須采用meta分析一類(lèi)的方法。Whatismeta-analysis?TheNationalLibraryofMedicinedefinemeta-analysisasa“quantitativemethodofcombiningtheresultsofindependentstudiesandsynthesizingsummariesandconclusionswhichmaybeusedtoevaluatetherapeuticeffectiveness,plannewstudies,etc.,withapplicationchieflyintheareasofresearchandmedicine.”Meta-analysismaybebroadlydefinedasthequantitativereviewandsynthesisoftheresultsofrelatedbutindependentstudies.綜合同類(lèi)研究，有效地增加研究樣本量，提高統(tǒng)計(jì)學(xué)效能研究某一療法有效或有害的具體程度探討和分析某些單項(xiàng)臨床試驗(yàn)未能涉及的問(wèn)題，并為開(kāi)展新的臨床試驗(yàn)提供研究方向和科學(xué)依據(jù)Meta-analysis的目的注意數(shù)據(jù)能否合并統(tǒng)計(jì)異質(zhì)性：可以由統(tǒng)計(jì)軟件計(jì)算處理臨床異質(zhì)性：由研究者確定并進(jìn)行處理StudydesignPopulationInterventionComparisonOutcomesHowtodoameta-analysis?常用軟件ReviewManager(完全免費(fèi))Stata(收費(fèi)但破解版很多)

其他軟件(大多收費(fèi)).Howtodoameta-analysis?丁香園中文教程Infectedpancreaticnecrosisisnotreduced0.11100.75(0.37,1.51)Sainio(1995)0.62(0.32,1.22)Schwarz(1997)1.50(0.58,3.90)Dellinger(2007)

0.81(0.54,1.22)

OveralleffectBetterantibioticsBetterplacebo0.40(0.15,1.06)Pederzoli(1993)0.22(0.03,1.71)Nordback(2001)1.20(0.42,3.43)Isenmann(2004)0.44(0.12,1.57)Rokke(2007)Author(Year)RR(95%CI)BaiYu.AmJGastroenterol,2007.Mortalityisnotreduced0.11100.14(0.02,1.09)Sainio(1995)0.20(0.01,3.80)Schwarz(1997)1.11(0.49,2.50)Dellinger(2007)0.70(0.42,1.17)OveralleffectBetterantibioticsBetterplacebo0.60(0.15,2.51)Pederzoli(1993)0.53(0.11,2.50)Nordback(2001)0.64(0.15,2.67)Isenmann(2004)0.77(0.19,3.20)Rokke(2007)Author(Year)RR(95%CI)BaiYu.AmJGastroenterol,2007.OutcomeSubgroup(n)RR(CI)PSinglecenterIPNMortality3Articles(n=144)3Articles(n=144)0.83[0.45,1.53]0.30[0.10,0.95]0.550.04MulticenterIPNMortality4Articles(n=323)4Articles(n=323)0.78[0.39,1.53]0.86[0.49,1.54]0.460.62SingleblindedIPNMortality4Articles(n=218)4Articles(n=218)0.69[0.38,1.24]0.40[0.16,0.96]0.210.04DoubleblindedIPNMortality2Articles(n=176)2Articles(n=176)1.35[0.67,2.75]0.97[0.48,1.97]0.400.94SubgroupanalysesBaiYu.AmJGastroenterol,2007.OutcomeSubgroup(n)RR(CI)PBeta-lactamIPNMortality5Articles(n=365)5Articles(n=365)0.65[0.38,1.13]0.74[0.42,1.30]0.130.29Quinolone+imidazoleIPNMortality2Articles(n=102)2Articles(n=102)1.16[0.66,2.03]0.51[0.14,1.85]0.610.31LowriskofbiasIPNMortality2Articles(n=176)2Articles(n=176)1.35[0.67,2.75]0.97[0.48,1.97]0.400.94HighriskofbiasIPNMortality5Articles(n=291)5Articles(n=291)0.66[0.40,1.10]0.48[0.22,1.02]0.110.06SubgroupanalysesBaiYu.AmJGastroenterol,2007.DiscussionAntibiotics:amuchdebatedtopicwithcontroversyPreviousmeta-analysis:GenerallyfavorantibioticsArethesemeta-analysisthefinalconclusion?Includedstudiesofbothoralandintravenousantibiotics,andearliertrialsusingampicillinDiscussionInvolvedonly3trialswith160patients,whichmayleadtotheindefiniteconclusionDiscussionIncludedthetrialvaluatingtheeffectivenessofprophylacticantibioticswithselectivegutdecontaminationinANPDiscussionDidnotincludethelargestRCTonthistopicDiscussionMeta-analysisbeforeusBothoralandintravenousrouteincludedSmallsamplesizeofeachtrialTrialswithmethodologicallimitationsHighriskofoverestimatedresults!!!BaiYu.AmJGastroenterol,2007.DiscussionWhatisnewinourmeta-analysis?SamplesizeincreasedIncludedcurrentbestevidenceWellconductedandanalyzedRisksminimizedBaiYu.AmJGastroenterol,2007.DiscussionWhatislackinginourmeta-analysis?Patients’age&etiology&organfunctionTimingofinitiationoftreatmentTypeIIerrormaybepossibleRisksremainedBaiYu.AmJGastroenterol,2007.OutcomeSubgroup(n)RR(CI)PSinglecenterIPNMortality3Articles(n=144)3Articles(n=144)0.83[0.45,1.53]0.30[0.10,0.95]0.550.04MulticenterIPNMortality4Articles(n=323)4Articles(n=323)0.78[0.39,1.53]0.86[0.49,1.54]0.460.62SingleblindedIPNMortality4Articles(n=218)4Articles(n=218)0.69[0.38,1.24]0.40[0.16,0.96]0.210.04DoubleblindedIPNMortality2Articles(n=176)2Articles(n=176)1.35[0.67,2.75]0.97[0.48,1.97]0.400.94SubgroupanalysesBaiYu.AmJGastroenterol,2007.OutcomeSubgroup(n)RR(CI)PSinglecenterMortality3Articles(n=144)0.30[0.10,0.95]0.04SingleblindedMortality4Articles(n=218)0.40[0.16,0.96]0.04P=0.04marginalsignificance

AGAACGProphylacticantibioticscannotreduceinfectedpancreaticnecrosisandmortalityinpatientswithacutenecrotizingpancreatitisFuturehigh-qualitylarge-scaletrialsinhigh-riskptsarewarrantedSummaryBaiYu.AmJGastroenterol,2007.Changingtrends1990s2000s2008常規(guī)應(yīng)用

用or不用？不推薦常規(guī)應(yīng)用？EBMMethod

AcquirethebestevidenceAppraisetheevidenceApplyevidencetopatientcareAssessyourpatientAskclinicalquestions患者的病因（酒精性而非膽源性）.患者的一般情況較好（無(wú)重要臟器衰竭）.患者的胰腺壞死情況（CT上壞死面積<1/3）.治療方案與我們薈萃分析中入選的患者特征基本吻合，全組醫(yī)生討論后不建議使用抗生素?；颊唧w溫逐漸降至正常，兩周后基本痊愈，囑戒酒后出院。若患者的病因是膽源性，伴有膽管炎？若患者一般情況差，伴MODS?若患者的胰腺壞死面積>1/3?循證醫(yī)學(xué)不是刻舟求劍與薈萃分析中患者特征不再完全吻合，不可照搬其結(jié)論隨著新的高質(zhì)量RCT的出現(xiàn)，薈萃分析的結(jié)果將會(huì)更新WhatfutureRCTneedsDiscussionCTverifiedpancreaticnecrosisSamplesizemustbesufficientRecordage,etiologyandtiminghighriskforIPNRCTBaiYu.AmJGastroenterol,2007.Thereareavarietyofpotentialproblemssuchaspublicationbias.Itiswell-recognizedthatnegativeresultsarenotoftenpublished.Outsidefundedstudiesyieldstrongerpositiveresultsthannon-fundedstudies.Blindedrandomizedresultsaregenerallylesspositivethanunblindedrandomizedresults,whicharelesspositivethannon-randomizedresults.Softendpointsgenerallyyieldstrongerresultsthanhardendpoints.Softendpointspermitmoreflexibilityandsubjectivityinthereportingofresults.Note2、文獻(xiàn)閱讀抗生素與ERCP術(shù)后膽管炎奧曲肽與ERCP術(shù)后胰腺炎激素與ERCP術(shù)后胰腺炎別嘌呤醇與ERCP術(shù)后胰腺炎

硝酸甘油與ERCP術(shù)后胰腺炎硝酸甘油與ERCP術(shù)后胰腺炎80年代有報(bào)道硝酸甘油可松弛Oddi擴(kuò)約肌2001起RCT報(bào)道硝酸甘油可降低術(shù)后胰腺炎此后又有數(shù)項(xiàng)研究報(bào)道發(fā)表，結(jié)果不一致大多數(shù)研究均認(rèn)為其無(wú)預(yù)防術(shù)后胰腺炎作用術(shù)后胰腺炎的發(fā)生率約5-10%藥物干預(yù)使發(fā)生率下降一半，α=0.05,β=80%Samplesize>900cases沒(méi)有單獨(dú)一項(xiàng)研究可以達(dá)到如此樣本量硝酸甘油與ERCP術(shù)后胰腺炎Author[Ref.]YearRouteNo.(Pts)Wehrmann[16]2001Topical80Sudhindran[17]2001Sublingual186Moretó[18]2003Transdermal144Talwar[32]2005Topical104Kaffes[19]2006Transdermal318Beauchant[34]2008Intravenous208N?jgaard[33]2009Transdermal806Hao[35]2009Sublingual74大多數(shù)研究均顯示硝酸甘油有降低PEP的趨勢(shì)但大多數(shù)研究的結(jié)論認(rèn)為統(tǒng)計(jì)學(xué)差異不顯著樣本量較小，導(dǎo)致出現(xiàn)II類(lèi)錯(cuò)誤的概率加大硝酸甘油與ERCP術(shù)后胰腺炎Bai,etal.Endoscopy2009硝酸甘油可有效降低ERCP術(shù)后胰腺炎的發(fā)生率硝酸甘油與ERCP術(shù)后胰腺炎Bai,etal.Endoscopy2009小插曲確定好題目后盡可能快而好的完成如果將散在的金子有機(jī)結(jié)合就會(huì)成為一座金山如果將發(fā)霉垃圾聚集就會(huì)成為一堆更大的垃圾要解決實(shí)際問(wèn)題、而不是僅僅為了發(fā)表而寫(xiě)作入選研究的質(zhì)量的重要性遠(yuǎn)遠(yuǎn)高于研究的數(shù)量系統(tǒng)評(píng)價(jià)和薈萃分析RCT并非完美無(wú)瑕···Bai,etal.Hepatology2009