RESEARCHOpenAccessUpdatingtheevidencefortheroleofcorticosteroidsinseveresepsisandsepticshock:aBayesianmeta-analyticperspectiveJohnLMoran1*,PetraLGraham2,SueRockliff3,AndrewDBersten4AbstractIntroduction:Currentlow(stress)dosecorticosteroidregimensmayhavetherapeuticadvantageinseveresepsisandsepticshockdespiteconflictingresultsfromtwolandmarkrandomisedcontrolledtrials(RCT).
Wesystematicallyreviewedtheefficacyofcorticosteroidtherapyinseveresepsisandsepticshock.
Methods:RCTswereidentified(1950-September2008)bymultipledata-baseelectronicsearch(MEDLINEviaOVID,OVIDPreMedline,OVIDEmbase,CochraneCentralRegisterofControlledtrials,Cochranedatabaseofsystematicreviews,HealthTechnologyAssessmentDatabaseandDatabaseofAbstractsofReviewsofEffects)andhandsearchofreferences,reviewsandscientificsocietyproceedings.
Threeinvestigatorsindependentlyassessedtrialinclusionanddataextractionintostandardisedforms;differencesresolvedbyconsensus.
Results:Corticosteroidefficacy,comparedwithcontrol,forhospital-mortality,proportionofpatientsexperiencingshock-resolution,andinfectiveandnon-infectivecomplicationswasassessedusingBayesianrandom-effectsmodels;expressedasoddsratio(OR,(95%credible-interval)).
Bayesianoutcomeprobabilitieswerecalculatedastheprobability(P)thatOR≥1.
FourteenRCTswereidentified.
High-dose(>1000mghydrocortisone(equivalent)perday)corticosteroidtrialswereassociatedwithanull(n=5;OR0.
91(0.
31-1.
25))orhigher(n=4,OR1.
46(0.
73-2.
16),outlierexcluded)mortalityprobability(P=42.
0%and89.
3%,respectively).
Low-dosetrials(1,000mgperday)[29].
Bayesianmeta-regression[21]wasusedtodeterminetherelationbetweenlogoddsmortalityand(i)averagepatientageand(ii)control-armrisk,aslog-oddsmortal-ity[17,24].
Theslope(b)with95%CrIandtheprobabil-itythatb≥0(Pb)werepresented.
Heterogeneitywaspresentedasthestandarddeviation,τ,betweenstudies[30];τcloseto0indicateslittleheterogeneity,τ=0.
5indicatesmoderateandτ>1reflectssubstantialhetero-geneity[18].
Forheuristicpurposesweseparatelyestimated:(i)pooledestimateswiththeSchumer[31]andCooperativeStudyGroup(CSG)[32]studiesremovedinasensitivityanalysisduetopreviousidentificationoftheformerasapotentialoutlier[7]andtheremotenessofthelatter1963trialfromcurrenttherapeuticregimens;(ii)certainparametersofclinicalimportintheriskdifferencemetric[21],albeitthismetricmaysufferfrompotentialbiaswithvaryingtimetoevent[24];(iii)themortalityORandprobability(P)thattheORwas1ormoreinthepredictivedistribution(thatis,inthenext'new'study);(iv)themortalityORforhypothesizedstudiesofsize2,000and4,000patients;(v)theBayesianpredictiveP-valuethattheCORTICUStrial[8]wasinconsistentwiththeothertrialsofthelow-dosecorticosteroidgroup;thatis,theCORTICUSstudywasomittedfromanalysis(leavingn=7trials)andareplicatestudyofthesamesizeastheCORTICUSstudywasdrawn,withareplicatebaseline,andanewtreatmenteffectwasestablishedbaseduponthepredictivedistribution.
ABayesianpredictiveP-valuewassubsequentlyobtained,expressingtheprobabilitythatthefuturestudywouldbeas'extreme'asthatobserved.
Publicationandtheassociatedphenomenonofsmall-studybiaswereaddressedusingtheapproachofPetersandcolleagues[33]viacontour-enhancedfunnelplots;formalquantitativetestingforsmall-study-biaswasper-formedusingtheapproachofHarbordandcolleagues[34],whichhaseffectivepropertiesinthepresenceofappreciableheterogeneity.
ImplementationwasviatheRpackage'meta'[35]anduser-writtenroutines.
ResultsUsingmultipleelectronicsearches,1,843abstractsofpublishedpaperswereidentified(includingduplicates).
Areviewoftheseabstracts(JLM,PLG)identified115papersofpotentialinterestincludingreviewpapers.
Thepublishedtextof31'randomized'clinicaltrials,includ-ingsevenabstractsfromproceedingsofscientificmeet-ings,werefurtherreviewed(JLM,PLG,AB):twowereexcludedonthebasisofreportingfromprevioustrials,onereportednomortalityoutcomedataandoneusedpseudo-randomization.
Afurther13studies,includingfourabstracts-onlywereexcludedforreasonsgiveninTable1.
Thefinalcohortwasof14trials[8,31,32,36-46],includingtwoabstractsfromthereportsofscientificmeetings;11ofthestudieshadbeenconsid-eredbypreviousmeta-analyses[4-6,10]andthethreeremainingstudies[8,42,44]werepost-2004,thepublica-tiondateofthetwocomparatormeta-analyses[6,7](Figure1andTable2).
ThetrialpatientsizevariedMoranetal.
CriticalCare2010,14:R134http://ccforum.
com/content/14/4/R134Page3of15from28[44]to499[8]andthetotalnumberofpatientswas1,991,ofmeanage55yearsand66%male.
Totalcorticosteroiddosageinthehigh-dosecohortrangedfrom7,000to42,000hydrocortisone-equivalentmgoveronetothreedays,whereasinthelow-dosecohort,dosagewas856to2,175hydrocortisone-equivalentmgover2to10daystreatmentwith0to14daysoftaper-ing(Table3).
Averagehigh-andlow-dosecontrolarmmortalitieswere47%and54%,respectively.
FurthercharacteristicsofthetrialsaregiveninTables2and3.
Theprimaryoutcomeofhospitalmortalitywasavail-ableinsixstudies[8,32,36,39,43,46];theotherstudieshadrecorded28-or30-daymortalityandonestudyrecorded14-daymortality(Table2).
Sepsisandshockdefinitionsoftrialscompletedbefore1992[31,32,38,41,43,46]weregenerallyconsistentwithdefini-tionsoftheAmericanCollegeofChestPhysiciansandSocietyofCriticalCareMedicineConsensusConferenceonsepsisandorganfailure,albeitthetwotrialspublishedin1971[41]and1963[32]used'lifethreateninginfec-tions'asacriteria(Table2).
Ofinterest,trialsbefore1998werepredominantlyreportedfromtheUSA;after1997,theywerefromEuropeanandothernon-USAsites.
Trialpatientdatabyoutcomes(hospitalmortality;shock-reversal;corticotrophin-responsiveness;shockreversalbycorticotrophin-responsiveness;andsecondarycomplications,asinfectious,gastro-intestinalbleedingandnew-onsethyperglycemia)areshowninTable4.
MortalityoutcomeNeitherthelow-dosenorhigh-dosecohortshowedasig-nificantsteroidtreatmenteffectforthemortalityOR,althoughtherewasmodestevidenceofbenefitinthelow-dosecohort(P=20.
4%)(Table5andFigure2).
Theoddsofmortality(fourstudies[8,36,42,45]),forbothcor-ticotrophinrespondersandnon-responderswasnotsig-nificantlydifferentcomparedwithcontrol(Table5).
Table1StudyexclusionsStudyYearpublishedReasonforexclusionWagnerandcolleagues[78]1955Descriptionofpneumoniatherapyonlywithnoseveritystratification.
Allocationby'historynumber'Thompsonandcolleagues[79]1976Abstract;nineof60patientswithcardiogenicshock;nosubsetanalyses.
Post-randomizationexclusionof4patientsLucasandLedgerwood[80]1984Open-labelstudy;pseudo-randomizationbyhospitalnumberVASSCS[81]1987Predominantlysepsispatientswithnosubgroupofshockedpatients.
NotimingoffluidboluswithrespecttoreportedhypotensionSchattnerandcolleagues[82]1997pseudo-randomizationofpatientswith'earlysepsis'Kehandcolleagues[60]2003Cross-overplacebostudyinsepticshockConfalonieriandcolleagues[83]2005Communityacquiredpneumoniastudy;nosubsetanalysesforshockedpatientsRinaldiandcolleagues[84]2006Postrandomizationexclusionof15patients;3withsepticshockHuhandcolleagues[85]2006Abstract;twohydrocortisonearms;noconcurrentplaceboarmreportedLoisaandcolleagues[86]2007Twohydrocortisonearms;noconcurrentplacebogroupNawabandcolleagues[87]2007Abstract;severecommunityacquiredpneumonia,nosubsetanalysis;outcomestoday-7onlyCicarelliandcolleagues[88]2007Unspecifiedpost-randomizationexclusionof'allpatientswhoprogressedtorefractorysepticshock'Kurugundlaandcolleagues[89]2008Abstract;ICUoutcomesreportedonlyVASSCS,VeteransAdministrationSystemicSepsisCooperativeStudy.
Figure1Flowchartforidentificationofstudiesoncorticosteroidsinseveresepsisandsepticshock;numberoftrialsevaluatedateachstageofthesystematicreview.
Moranetal.
CriticalCare2010,14:R134http://ccforum.
com/content/14/4/R134Page4of15Table2FinalstudycohortStudyYearpublishedYearcompletedTrialoriginTrialReportedasPaper#/abstractDesignAllocationconcealmentEffectandsamplesizecalculationEarlystoppingSepsis/shockdescriptionPredominantpatienttypePrimaryoutcomeCooperativeStudyGroup[32]1963NAUSAMulticenterPaperDouble-blindYesNoNo'Lifethreateninginfections'MedicalHospitalmortalityKlasterskyandcolleagues[41]19711970BelgiumSinglecenterPaperDouble-blindYesNoNo'Lifethreateninginfections'Cancer30-daymortalitySchumer[31]19761975USASinglecenterPaperDouble-blindNANoNoSeptichistory,fallingbloodpressureandpositivebloodculturesSurgical28-daymortalitySprungandcolleagues[43]19841982USATwo-centersPaperOpen-labelYesNoNoSBP0)=21.
9%);andforeighttrials(CSGtrialexcluded[32])was-0.
072(95%CrI=-0.
202to0.
018;P(RD>0)=5.
3%),similartothe6.
6%reportedbyAnnaneandcolleagues[48].
ThemortalityORinthepredictivedistribution(fromeighttrials)was0.
703(95%CrI=0.
156to2.
198;P(OR>1)=19.
9%).
Forhypothe-sizedstudiesofsize2,000and4,000patients,themor-talityORswerepredictedtobe0.
724(95%CrI=0.
184to2.
108)and0.
726(95%CrI=0.
184to2.
096),respec-tively.
TheBayesianpredictiveP-value,reflectingtheinconsistencyoftheCORTICUSstudy[8]withtheremainingtrials(n=7;CSGtrialexcluded[32])was0.
074.
DiscussionDespitethedisappointmentoftheCORTICUS[8]trial,ourreviewsuggestsamodesttohighprobability(80%to98%)ofefficacyforlow-dosesteroidswithrespecttobothmortalityandshockreversal;themortalityeffectbeingrisk-related(Table5).
TheseprobabilitiesaretobeinterpretedinthecontextofCrIspanningthenullforallestimates(seeStatisticalanalysis,above).
WefoundnostrongevidenceforthedeterminacyofACTHresponsivenessnorcomplicationsofcorticosteroidther-apy.
Thisbeingsaid,itisofinteresttonotetheadmoni-toryimpactoftheCORTICUSstudyonrecentsummarystatementsofsepsismanagement[2,3,13,29,49].
Consistentwithpreviousmeta-analyses[6,7]wefoundnulloradverseeffectsofhigh-dosester-oids;theprobabilityoftherapeuticcomplicationsbeinglow(Table5).
Theuseofprolongedlow-dosecorticosteroidwasjustifiedinthelandmarkAnnaneandcolleaguestrialonthebasisthat"severesepsismaybeassociatedwithrelativeadrenalinsufficiencyorsystemicinflammation-inducedglucocorticoidreceptorresistance.
.
.
"[36].
Aproposofthisstatement,itisinstructivetonotethattheprimaryaimoftheCORTICUSstudywas28-daymortalityinpatientsnotrespondingtocorticotrophin[8].
Arecentreviewofcorticosteroidinsufficiencyinthecriticallyillhassuggestedthatinstateswheresuchinsufficiency[50]islocated"withinthetissueitself.
.
.
theadrenalglandfunctioncouldbenormal.
.
.
itwouldbeimpossibletodiagnosethisstateonthebasisofserumoreventissuelevelsofglucocorticoids.
.
.
[and].
.
.
treatmentwouldrequiresupraphysiologicallevelsofTable4Trialpatientdatabyoutcome(Continued)Annaneandcolleagues[36]95/160103/15060/15140/14936/15034/14918/3618/3465/11446/11522/15027/15011/1508/149NANATandanandcolleagues[44]11/1413/145/143/14NANANANANANANANANANANANAOppertandcolleagues[42]7/1811/2313/1818/236/189/23NANANANANANANANANANASprungandcolleagues[8]111/251100/245200/251184/248118/243136/244100/118104/13698/12576/10878/23461/13215/23413/232186/234161/232#,mortalitystatisticsforChawlaandcolleagues[47]wereabstractedfromtheAnnaneandcolleaguesmeta-analysis[6].
##,dataforhyperglycaemiaforBollaertandcolleagues[37]wasabstractedfromtheAnnaneandcolleaguesmeta-analysis[6].
GIS,gastrointestinal;NA,notavailable.
Moranetal.
CriticalCare2010,14:R134http://ccforum.
com/content/14/4/R134Page9of15glucocorticoids"[51].
Theinabilityinthecurrentmeta-analysistodemonstratetreatmentefficacywithrespecttomortalityandshock-reversalbaseduponcorticotrophinresponsivenessisinagreementwithMinneciandcolleagues[7]andsuggestsboththattestsofthelattertodirecttreatmentregimensaremis-placedandthatthenotionofadrenalinsufficiencyinseveresepsisandsepticshockisproblematic[52];a".
.
.
hardlydefinablediseaseentityorsyndrome.
.
.
"[53].
Oftheseventrialsreportingshock-reversal[8,36,37,39,40,42,44],timetothelatterend-pointwastheprimarystudyend-pointinthree[37,39,42].
Allpub-lishedstudiesusedtime-to-eventanalysisbaseduponconventionalKaplan-Meierestimates,censoringthosewhodiedand/orthoseinwhomvasopressortherapycouldnotbewithdrawnattimeofassessment.
However,suchanalysesareproblematic,becausetheyignorethecompetingriskofthosewhodiedand/orthoseinwhomvasopressortherapycouldnotbewithdrawn.
Inthepre-senceofcompetingrisksKaplan-Meierestimatescannotbeinterpretedasprobabilities[54,55].
Undertheconditionsofcompetingrisks,theprobabilityofaneventismoreappropriatelyestimatedbythecumulativeincidencefunction,which,fortheparticulareventofinterest,isafunctionofthehazardsofallthecompetingeventsandnotsolelyofthehazardoftheeventtowhichitrefers.
Hypothesistestsforthecumulativeinci-dencefunctiondonotnecessarilyequatewiththefamil-iarlog-ranktest[56].
Howthenarewetounderstandthesefavourableeffectsoflow-dosecorticosteroidsGlucocorticoidactiononinflammation[57],vascularreactivity[58]andinteractionsbetweencorticosteroidsand'signallingpath-ways'[59]mayexplainthesalutaryeffectsinsepsis[60];anti-inflammatoryandcoagulationeffectswouldappeartobedifferentiallydosedependent[61].
Loworstressdosesofhydrocortisone,ascurrentlyprescribed,arenotreplacementorphysiologicaldoses;theygenerateplasmacortisollevelsgreaterthan2,500nmol/l,inexcessoftheusualupperlimits(1,000to1,500nmol/l)ofpatientsinsepticshock[42,60,62].
Thepresumedimmune-modulation[63]oftheseprolongedlow-doseTable5OutcomeeffectestimatesOutcomeNOR(95%CrI)P(%)τ(95%CrI)b(95%CrI)Pb(%)MortalityHighdose50.
912(0.
313to1.
253)42.
01.
00(0.
42to1.
89)HighdoseexcludingSchumer[31]41.
406(0.
727to2.
614)89.
30.
25(0.
01to1.
40)Lowdose90.
796(0.
396to1.
386)20.
40.
65(0.
23to1.
44)LowdoseexcludingCSG[32]80.
706(0.
371to1.
096)5.
80.
39(0.
04to1.
15)Corticotrophinresponders*40.
882(0.
285to2.
073)36.
40.
49(0.
02to1.
78)Corticotrophinnon-responders*40.
831(0.
334to1.
971)28.
00.
43(0.
02to1.
69)Shock-reversalHighdose21.
078(0.
227to6.
311)54.
91.
39(0.
06to1.
93)Lowdose71.
999(1.
069to4.
55)98.
20.
57(0.
04to1.
62)Corticotrophinresponders*31.
830(0.
499to7.
845)86.
70.
87(0.
05to-1.
92)Corticotrophinnon-responders*31.
845(0.
637to7.
267)91.
90.
55(0.
02to1.
86)Meta-regression(logoddsmortality)AverageageHighdose40.
777(0.
285to2.
426)27.
30.
72(0.
04to1.
87)0.
60(-0.
23to1.
51)94.
52ExclSchumer[31]31.
390(0.
399to4.
872)77.
00.
66(0.
03to1.
90)0.
10(-1.
57to1.
74)58.
05Lowdose60.
658(0.
334to1.
223)7.
60.
36(0.
02to1.
51)0.
05(-0.
10to0.
18)80.
53UnderlyingriskHighdose50.
943(0.
292to3.
049)45.
41.
14(0.
46to1.
49)0.
23(-1.
71to2.
58)60.
98ExclSchumer[31]41.
372(0.
596to3.
249)82.
90.
38(0.
01to1.
74)-0.
09(-1.
31to1.
42)41.
47Lowdose90.
752(0.
389to1.
291)14.
50.
57(0.
17to1.
37)-0.
49(-1.
14to0.
27)7.
80ExclCSG[32]80.
676(0.
347to1.
076)4.
90.
40(0.
03to1.
23)-0.
28(-0.
88to0.
50)19.
08Oddsofthefollowingcomplications(corticosteroidsversuscontrol)SuperinfectionHighdose41.
127(0.
364to3.
924)62.
20.
55(0.
02to2.
85)Lowdose60.
955(0.
388to1.
749)43.
60.
46(0.
03to1.
62)GIbleedingHighdose30.
824(0.
167to3.
186)37.
30.
74(0.
03to1.
90)Lowdose51.
103(0.
379to3.
031)59.
60.
58(0.
02to1.
84)HyperglycemiaHighdose31.
012(0.
244to4.
266)50.
80.
64(0.
03to1.
88)Lowdose31.
430(0.
155to3.
985)57.
40.
87(0.
05to1.
93)*allstudieswerelowdose;CI,confidenceinterval;CSG,CooperativeStudyGroup;GI,gastro-intestinal;N,numberofstudiesreportingdataforthatendpoint;NA,notapplicable;OR,oddsratio.
ExclSch=ExcludingSchumer[31];ExclCSG=ExcludingCSG[32]Moranetal.
CriticalCare2010,14:R134http://ccforum.
com/content/14/4/R134Page10of15Figure2Corticosteroidmortalityeffect(OR),stratifiedbyhigh(upperpanel)orlow(lowerpanel)dosesteroidregimen;forestplotrepresentationoftheeffect.
Theverticalstraightlinedenotesnulleffect(oddsratio(OR)=1).
TheindividualpointsdenotetheORforeachstudyandthelinesoneithersidethe95%Bayesiancredibleintervals(CrI).
Figure3Contour-enhancedfunnelplotofmortalityoddsversusstandarderrorforalltrials(n=14).
Verticalaxis,standarderror;horizontalaxis,mortalityodds(logscale).
The'contours',baseduponatwo-sidedPvalue,aretheconventionallevels(not'pseudo'confidenceintervals)ofstatisticalsignificance(<0.
01,<0.
05,<0.
1)fortheprimarystudiesandareindependentofthepooledestimate(ifthepooledestimateisbiased,thecontoursarenotaffected)[33].
Figure4Contour-enhancedfunnelplotofmortalityoddsversusstandarderrorforlow-dosecorticosteroidtrials(n=9).
Verticalaxis,standarderror;horizontalaxis,mortalityodds(logscale).
The'contours',baseduponatwo-sidedPvalue,aretheconventionallevels(not'pseudo'confidenceintervals)ofstatisticalsignificance(<0.
01,<0.
05,<0.
1)fortheprimarystudiesandareindependentofthepooledestimate(ifthepooledestimateisbiased,thecontoursarenotaffected)[33]Figure5Corticosteroidshock-reversaleffect(OR),stratifiedbyhigh(upperpanel)orlow(lowerpanel)dosesteroidregimen;forestplotrepresentationoftheeffect.
Theverticalstraightlinedenotesnulleffect(oddsratio(OR)=1).
TheindividualpointsdenotetheORforeachstudyandthelinesoneithersidethe95%Bayesiancredibleintervals(CrI).
Moranetal.
CriticalCare2010,14:R134http://ccforum.
com/content/14/4/R134Page11of15regimensunderpinstherationaleofcriticalillness-relatedcorticosteroidinsufficiency[14,29].
Thisbeingsaid,theAnnaneandcolleagues[36]trialusedafixedseven-daysteroidcoursewithouttaperingandclaimedefficacyandnodifferenceinthecomplicationrateswasevidentbetweenthehigh-andlow-dosecohortsinboththecurrentandAnnaneandcolleagues'meta-analyses[6].
Asmentionedincommentary[64],differencesincontrolgroupmortalitiesoftheAnnaneandcolleagues[36]andCORTICUS[8]trialsmayexplaindifferingout-comesonthebasisofrisk-relatedtreatmenteffects.
Thelatterwerepersuasivelydemonstratedinthecurrentmeta-analysis.
Theestimateofmortalityriskatwhichlow-dosecorticosteroidsbegantoexhibitatreatmenteffect,44%,wasclinicallyplausiblegiventherangeofcontrol-armmortalitiesof30to93%.
Suchdemonstra-tion,usingappropriateBayesianmethodology[17,24],representsanovelinsightintocriticalcaretherapeuticefficacy.
CritiqueofmethodologyOuranalyticapproachwastoconsiderthetwotreat-mentcohorts,high-andlow-dosecorticosteroid,sepa-rately;wedidnotproduceanoveralltreatmenteffectonthebasisthatboththetreatmentintentionandeffective(daily)corticosteroiddoseofthetwocohortswerequitedisparate.
Analternateapproachwouldhavebeentoconsideralltrials(n=14)withtotalhydrocortisonedoseorcalendaryearaseffect-moderators.
Intheabsenceofindividualpatientdata,suchanalyses,withonly14studies,havelowpower.
Secondaryoutcomeanalysiswasbesetbyselectionbiasinreporting[65],aswitnessedbystudynumbersinTable5;parameterestimatesmaybebiasedundersuchcircumstances.
Thestudylistaddressinglow-dosecorti-costeroidmortalityefficacy(n=9)includedasinglestudy[32]in1963,theothersbeingfromtheperiod1996to2005(Figure2).
Plausibleestimatesofcurrenttherapeuticefficacywouldsuggestanalysisexcludingtheformerstudy,theresultofwhichwastoreducehetero-geneityofthemortalityeffectby40%andtorevealaprobabilityofcorticosteroidefficacyof94.
2%(Table5).
Thesingle-investigatorsingle-centreSchumerstudy,conductedoveraprolongedeight-yearperiod,hasbeenpreviouslysubjecttosubstantivecritique[7]andrecentcautionsregardingextendedrecruitmenttime[66]andinferencefromsingle-centerstudies[67]meritsitscon-siderationasanoutlier.
Thattheinclusionofthelargebutnull-effectCORTI-CUStrial[8]inthecurrentmeta-analysisdidnotextin-guishaprobabletreatmenteffectdeservescomment.
Theimpactofthesinglelargetrialisundoubted,buttheevidenceproducedbysuchatrialmaybe"lessreli-ablethanitsstatisticalanalysissuggests"[68].
Weadoptedarandomeffectsmethodology[69]inthepre-senceofmoderatebetweenstudyheterogeneity(τ,Table5);undertheseconditionslargestudiesmayhavelittleimpactuponameta-analysis[70]andtheremaybevirtuein(clinical)heterogeneity[71].
Thedegreeofasymmetryofthecontour-enhancedfunnelplotinthelow-dosecohort(seeResults,Mortalityoutcome,above)raisesconcernsaboutarandomeffectsmethodology[69],buttherewasnoquantitativeevidenceofsmall-studyeffects(atthe0.
1level)andthenumberofstudieswassmall.
Inthepresenceofsparsedataandmoderateheterogeneity(Table5),theinterpretationoffunnelplotasymmetryisproblematic[34,72]andexplorationofthereasonsforsuchheterogeneityisthepreferredanalyticfocus[34].
Withrespecttotheefficacyofcorticosteroidsinseveresepsisandsepticshock,thedivergentpositionsrepresentedbytheAnnaneandcolleagues[36]andCORTICUS[8]trialsremainunresolved.
Tworecent(calendaryear2009)updates[48,73]ofpreviousmeta-analyses[6,7]alsomeritcomment.
Bothoftheupdatedmeta-analyses,usingfrequentistmethodology,foundefficacyoflow-doseprolongedcorticosteroidswithrespecttothemortalityeffect,Annaneandcolleagues[48]foundarelativeriskof0.
84(95%confidenceinter-val(CI)=0.
72to0.
97;P=0.
02)andMinneciandcol-leagues[73]foundanORof0.
64(95%CI=0.
45to0.
93;P=0.
02),andshockreversal,thelattereffectcon-sistentwiththeestimatesofthecurrentstudy(Table5).
Studyinclusionsinthesemeta-analysesdifferedandwerenotthesameasinourmeta-analysis,whichadoptedarigorousexclusionpolicy(Table1).
Thefre-quentistmeta-regressionmethodsusedbybothmeta-analyses[48,73]toestimatetherisk-relatedtreatmentefficacyofsteroidsareproblematic[17,24].
Althoughsuchmethodsmayidentifyputativeriskrelatedtreat-menteffectsinmeta-analysestheyfailtoallowforbothregressiontothemean(thedifferencebetweenoutcomeandbaselinebeingcorrelatedwithbaseline)andthesto-chasticnatureofthecontrolrate(regressiondilutionbias).
Thestochasticcharacteristicofthecontrolrateisalsonotaddressedastheexpectedresponsein(ordin-ary)linearregressionisconditionaluponindependent(fixed)variablesandthereisnoinherentaccountingfortherandomerrorinestimationofthiscontrolrate.
SuchproblemsareovercomebytheuseofBayesianmethods[17,24].
Bothmeta-analyseswerejudiciousintheirconclusionsabouttreatmentefficacyandthiswasreiteratedbyanaccompanyingeditorial[74].
However,neitherstudywasabletoattendtothisuncertaintyinatangiblemanner.
ThisispreciselywhatourBayesiananalysisquantifies:whatwastheprobabilityoftreatmentefficacy.
Forexample,ouranalysisdemonstratedthattheprobabilityMoranetal.
CriticalCare2010,14:R134http://ccforum.
com/content/14/4/R134Page12of15ofadversemortalityoutcomewithlow-dosecorticoster-oids(outlierexcluded)was5.
8%(Table5).
Theomissionofsuchaprobabilitystatementcannotbejustifiedbyanappealto"thenominalPvaluesfortheseoutcomeswereverycloseto0.
05.
.
.
.
"[48].
Wehavepreviouslycau-tionedtheagainstinterpretationof95%CI(andasso-ciatedfrequentistPvalues)asprobabilitystatements[75].
Furthermore,neithermeta-analysisreportedexplorationofestimatesfromapredictivedistribution,whichmaybeconsideredasamoreappropriatefuturetreatmentsummarythanthemeaneffect[18].
SuchacapacityrecommendsBayesianmethodology,althoughmeta-analyticpredictionintervals,whichaddressthe".
.
.
dispersionoftheeffectsizes.
.
.
"arecomputablefromafrequentistperspective[76].
WithrespecttoreservationsexpressedregardingthestatusoftheCORTICUSstudy[29,74],wefoundnocompellingevidence(Bayesianpre-dictiveP-value0.
074)thatthistrialwasinconsistentwiththeremaining(n=7)trials.
Continuedcontroversyandconventionalwisdom[77]wouldappeartomandatetheconductofalarge-(mega)-trialofthistherapyinwell-definedpatientsub-sets;anabsolutetreatmenteffectof7.
2%,controlarmriskof54%and90%powerwouldsuggestatotalpatientnumberofgreaterthan2,000.
Thisbeingsaidourpre-dictiveestimateswereunabletosuggestefficacyforfuture'large'trials,albeitthetrialbasefromwhichtheseestimatesweremadewassmall.
ConclusionsAlthoughanulleffectformortalitytreatmentefficacyoflow-dosecorticosteroidtherapyinseveresepsisandsep-ticshockcouldnotbeexcluded,thereappearstobecredibleevidenceforshockreversalefficacy.
Similarly,althoughanulleffectwasnotexcluded,advantageouseffectsoflow-dosesteroidshadahighprobabilityofdependenceuponpatientageandunderlyingrisk.
Low-dosesteroidefficacywasnotdemonstratedincortico-trophinnon-responders.
Bayesianmethodsareappositetoexpressuncertaintyinefficacyestimatesfrommeta-analyses.
KeymessagesTheefficacyofcorticosteroidsinpatientswithseveresepsisandsepticshockisuncertaindespiterecentmeta-analyticreviews.
Bayesianmethodsareappositetoexpressuncer-taintyinefficacyestimatesfrommeta-analyses.
Theefficacyoflow-dosecorticosteroidshadahighprobabilityofdependenceuponpatientageandunderlyingrisk;low-dosesteroidefficacywasnotdemonstratedincorticotrophinnon-responders.
Bayesianmeta-analyticpredictiveestimateswereunabletosuggestefficacyforfuturelargetrials.
Anulleffectformortalitytreatmentefficacyoflow-dosecorticosteroidtherapyinseveresepsisandsepticshockcouldnotbeexcluded.
AdditionalmaterialAdditionalfile1:Electronicsearchstrategy.
DetailedsearchstrategyofelectronicdatabasesAbbreviationsACCP:AmericanCollegeofChestPhysicians;ACTH:adreno-corticotrophinhormone;CI:confidenceinterval;CORTICUS:CorticosteroidTherapyofSepticShock;CrI:credibleintervals;CSG:CooperativeStudyGroup;OR:oddsratio;SCCM:SocietyofCriticalCareMedicine.
Authordetails1DepartmentofIntensiveCareMedicine,TheQueenElizabethHospital,28WoodvilleRoad,Woodville,SouthAustralia5011,Australia.
2DepartmentofStatistics,FacultyofScience,MacquarieUniversity,BalaclavaRoad,NorthRyde,NSW2109,Australia.
3DepartmentofLibraryServices,TheQueenElizabethHospital,28WoodvilleRoad,Woodville,SouthAustralia5011,Australia.
4DepartmentofCriticalCareMedicine,FlindersMedicalCentreandSchoolofMedicine,FlindersUniversity,SturtRoad,BedfordPark,SouthAustralia5042,Australia.
Authors'contributionsThestudywasconceivedbyJLM,PLGandAB.
SRconstructedthesearchtermsandconductedtheelectronicsearch.
JLM,PLGandABreviewedstudiesfulfillinginclusioncriteriaandpre-definedvariables.
JLM,andPLGconductedthequalityassessmentandstatisticalanalysis.
Allauthorscontributedtothewritingofthepaper,criticalreviewandfinalapproval.
CompetinginterestsTheauthorsdeclarethattheyhavenocompetinginterests.
Received:16September2009Revised:25May2010Accepted:13July2010Published:13July2010References1.
WeitzmanS,BergerS:Clinicaltrialdesigninstudiesofcorticosteroidsforbacterialinfections.
AnnInternMed1974,81:36-42.
2.
KehD,Weber-CarstensS,AhlersO:Adjunctivetherapiesinseveresepsisandsepticshock:Currentplaceofsteroids.
CurrentInfectiousDiseaseReports2008,10:354-361.
3.
MesottenD,VanhorebeekI,VandenBergheG:Thealteredadrenalaxisandtreatmentwithglucocorticoidsduringcriticalillness.
NatClinPractEndocrinolMetab2008,4:496-505.
4.
CroninL,CookDJ,CarletJ,HeylandDK,KingD,LansangMA,FisherCJJr:Corticosteroidtreatmentforsepsis:Acriticalappraisalandmeta-analysisoftheliterature.
CritCareMed1995,23:1430-1439.
5.
LeferingRM,NeugebauerEAMP:Steroidcontroversyinsepsisandsepticshock:Ameta-analysis.
CritCareMed1995,23:1294-1303.
6.
AnnaneD,BellissantE,BollaertPE,BriegelJ,KehD,KupferY:Corticosteroidsforseveresepsisandsepticshock:asystematicreviewandmeta-analysis.
BMJ2004,329:480.
7.
MinneciPCM,DeansKJM,BanksSMP,EichackerPQM,NatansonCM:Meta-analysis:theeffectofsteroidsonsurvivalandshockduringsepsisdependsonthedose.
AnnInternMed2004,141:47-56.
8.
SprungCL,AnnaneD,KehD,MorenoR,SingerM,FreivogelK,WeissYG,BenbenishtyJ,KalenkaA,ForstH,LaterrePF,ReinhartK,CuthbertsonBH,PayenD,BriegelJ,CORTICUSStudyGroup:Hydrocortisonetherapyforpatientswithsepticshock.
NEnglJMed2008,358:111-124.
9.
AnnaneD,BriegelJ,KehD,MorenoR,SingerM,SprungCL,CorticusStudyCoordinators:Clinicalequipoiseremainsforissuesofadrenocorticotropichormoneadministration,cortisoltesting,andtherapeuticuseofhydrocortisone.
CritCareMed2003,31:2250-2251.
Moranetal.
CriticalCare2010,14:R134http://ccforum.
com/content/14/4/R134Page13of1510.
MinneciPC,DeansKJ,BanksSM,EichackerPQ,NatansonC:CorticosteroidsforSepticShock.
AnnInternMed2004,141:742-743.
11.
AnnaneD:Cortisolreplacementforseveresepsisandsepticshock:whatshouldIdoCriticalCare2002,6:190-191.
12.
VincentJL:Steroidsinsepsis:anotherswingofthependuluminourclinicaltrials.
CriticalCare2008,12:141.
13.
DellingerRP,LevyMM,CarletJM,BionJ,ParkerMM,JaeschkeR,ReinhartK,AngusDC,Brun-BuissonC,BealeR,CalandraT,DhainautJF,GerlachH,HarveyM,MariniJJ,MarshallJ,RanieriM,RamsayG,SevranskyJ,ThompsonBT,TownsendS,VenderJS,ZimmermanJL,VincentJL,InternationalSurvivingSepsisCampaignGuidelinesCommittee;AmericanAssociationofCritical-CareNurses;AmericanCollegeofChestPhysicians;AmericanCollegeofEmergencyPhysicians;CanadianCriticalCareSociety;EuropeanSocietyofClinicalMicrobiologyandInfectiousDiseases;EuropeanSocietyofIntensiveCareMedicine;EuropeanRespiratorySociety;InternationalSepsisForum;JapaneseAssociationforAcuteMedicine;JapaneseSocietyofIntensiveCareMedicine;SocietyofCriticalCareMedicine;SocietyofHospitalMedicine;SurgicalInfectionSociety;WorldFederationofSocietiesofIntensiveandCriticalCareMedicine:SurvivingSepsisCampaign:internationalguidelinesformanagementofseveresepsisandsepticshock:2008.
CritCareMed2008,36:296-327.
14.
MarikPE,PastoresSM,AnnaneD,MeduriGU,SprungCL,ArltW,KehD,BriegelJ,BeishuizenA,DimopoulouI,TsagarakisS,SingerM,ChrousosGP,ZalogaG,BokhariF,VogeserM,AmericanCollegeofCriticalCareMedicine:Recommendationsforthediagnosisandmanagementofcorticosteroidinsufficiencyincriticallyilladultpatients:ConsensusstatementsfromaninternationaltaskforcebytheAmericanCollegeofCriticalCareMedicine.
CritCareMed2008,36:1937-1949.
15.
ShojaniaKG,SampsonM,AnsariMT,JiJ,DoucetteS,MoherD:HowquicklydosystematicreviewsgooutofdateAsurvivalanalysis.
AnnInternMed2007,147:224-233.
16.
FinferS:Corticosteroidsinsepticshock.
NEnglJMed2008,358:188-190.
17.
SharpSJ,ThompsonSG:Analysingtherelationshipbetweentreatmenteffectandunderlyingriskinmeta-analysis:comparisonanddevelopmentofapproaches.
StatMed2000,19:3251-3274.
18.
SpiegelhalterDJ,AbramsKR,MylesJP:Bayesianapproachestoclinicaltrialsandhealth-careevaluationChichester:JohnWiley&Sons,Ltd2004.
19.
BoneRC,BalkRA,CerraFB,DellingerRP,FeinAM,KnausWA,ScheinRM,SibbaldWJ:Definitionsforsepsisandorganfailureandguidelinesfortheuseofinnovativetherapiesinsepsis.
TheACCP/SCCMConsensusConferenceCommittee.
AmericanCollegeofChestPhysicians/SocietyofCriticalCareMedicine.
Chest1992,101:1644-1655.
20.
PeterJV,JohnP,GrahamPL,MoranJL,GeorgeIA,BerstenA:Corticosteroidsinthepreventionandtreatmentofacuterespiratorydistresssyndrome(ARDS)inadults:meta-analysis.
BMJ2008,336:1006-1009.
21.
WarnDE,ThompsonSG,SpiegelhalterDJ:Bayesianrandomeffectsmeta-analysisoftrialswithbinaryoutcomes:methodsfortheabsoluteriskdifferenceandrelativeriskscales.
StatMed2002,21:1601-1623.
22.
LunnDJ,ThomasA,BestN,SpiegelhalterD:WinBUGS-ABayesianmodellingframework:Concepts,structure,andextensibility.
StatisticsandComputing2000,10:325-337.
23.
SpruanceSL,ReidJE,GraceM,SamoreM:Hazardratioinclinicaltrials.
AntimicrobAgentsChemother2004,48:2787-2792.
24.
MoranJ,SolomonP,WarnD:Methodologyinmeta-analysis:astudyfromCriticalCaremeta-analyticpractice.
HealthServicesandOutcomesResearchMethodology2004,5:207-226.
25.
SuttonAJ,CooperNJ,AbramsKR,LambertPC,JonesDR:ABayesianapproachtoevaluatingnetclinicalbenefitallowedforparameteruncertainty.
JClinEpidemiol2005,58:26-40.
26.
WijeysunderaDN,AustinPC,HuxJE,BeattieWS,LaupacisA:Bayesianstatisticalinferenceenhancestheinterpretationofcontemporaryrandomizedcontrolledtrials.
JClinEpidemiol2009,62:13-21.
27.
WoolcottJC,RichardsonKJ,WiensMO,PatelB,MarinJ,KhanKM,MarraCA:Meta-analysisoftheimpactof9medicationclassesonfallsinelderlypersons.
ArchInternMed2009,169:1952-1960.
28.
AitkinM,FrancisB,HindeJ,DarnellR:Statisticalmodellingandinference.
StatisticalModellinginROxfordOX26DP,UK:OxfordUniversityPress2009,28-96.
29.
MarikPE:Criticalillness-relatedcorticosteroidinsufficiency.
Chest2009,135:181-193.
30.
RuckerG,SchwarzerG,CarpenterJ,SchumacherM:UnduerelianceonI^2inassessingheterogeneitymaymislead.
BMCMedResMethodol2009,8.
31.
SchumerW:Steroidsinthetreatmentofclinicalsepticshock.
AnnSurg1976,184:333-341.
32.
CooperativeStudyGroup:Theeffectivenessofhydrocortisoneinthemanagementofsevereinfections.
JAMA1963,183:462-465.
33.
PetersJL,SuttonAJ,JonesDR,AbramsKR,RushtonL:Contour-enhancedmeta-analysisfunnelplotshelpdistinguishpublicationbiasfromothercausesofasymmetry.
JClinEpidemiol2008,61:991-996.
34.
HarbordRM,EggerM,SterneJA:Amodifiedtestforsmall-studyeffectsinmeta-analysesofcontrolledtrialswithbinaryendpoints.
StatMed2006,25:3443-3457.
35.
SchwarzerG:meta:Meta-analysis.
[http://cran.
r-project.
org/web/packages/meta/index.
html].
36.
AnnaneD,SebilleV,CharpentierC,BollaertPE,FrancoisB,KorachJM,CapellierG,CohenY,AzoulayE,TrochéG,Chaumet-RiffaudP,BellissantE:Effectoftreatmentwithlowdosesofhydrocortisoneandfludrocortisoneonmortalityinpatientswithsepticshock.
JAMA2002,288:862-871.
37.
BollaertPE,CharpentierC,LevyB,DebouverieM,AudibertG,LarcanA:Reversaloflatesepticshockwithsupraphysiologicdosesofhydrocortisone.
CritCareMed1998,26:645-650.
38.
BoneRC,FisherCJJr,ClemmerTP,SlotmanGJ,MetzCA,BalkRA:Acontrolledclinicaltrialofhigh-dosemethylprednisoloneinthetreatmentofseveresepsisandsepticshock.
NEnglJMed1987,317:653-658.
39.
BriegelJ,ForstH,HallerM,SchellingG,KilgerE,KupratG,HemmerB,HummelT,LenhartA,HeyduckM,StollC,PeterK:Stressdosesofhydrocortisonereversehyperdynamicsepticshock:aprospective,randomized,double-blind,single-centerstudy.
CritCareMed1999,27:723-732.
40.
ChawlaK,KupferY,TesslerS:Prognosticvalueofcortisolresponseinsepticshock.
JAMA2000,284:309.
41.
KlasterskyJ,CappelR,DebusscherL:Effectivenessofbetamethasoneinmanagementofsevereinfections.
Adouble-blindstudy.
NEnglJMed1971,284:1248-1250.
42.
OppertM,SchindlerR,HusungC,OffermannK,GrafKJ,BoenischO,BarckowD,FreiU,EckardtKU:Low-dosehydrocortisoneimprovesshockreversalandreducescytokinelevelsinearlyhyperdynamicsepticshock.
CritCareMed2005,33:2457-2464.
43.
SprungCL,CaralisPV,MarcialEH,PierceM,GelbardMA,LongWM,DuncanRC,TendlerMD,KarpfM:Theeffectsofhigh-dosecorticosteroidsinpatientswithsepticshock.
Aprospective,controlledstudy.
NEnglJMed1984,311:1137-1143.
44.
TandanSM,GuleriaR,GuptaN:Lowdosesteroidsandadrenocorticalinsufficiencyinsepticshock:Adouble-blindrandomisedcontrolledtrialfromIndia.
AmJRespirCritCareMed2005,171:A43.
45.
YildizO,DoganayM,AygenB,GuvenM,KelestimurF,TutuuA:Physiological-dosesteroidtherapyinsepsis.
CriticalCare(London,England)2002,6:251-259.
46.
LuceJM,MontgomeryAB,MarksJD,TurnerJ,MetzCA,MurrayJF:Ineffectivenessofhigh-dosemethylprednisoloneinpreventingparenchymallunginjuryandimprovingmortalityinpatientswithsepticshock.
AmRevRespirDis1988,138:62-68.
47.
ChawlaK,KupferY,GoldmanI,TesslerS:Hydrocortisonereversesrefractorysepticshock.
CritCareMed1999,27:33A.
48.
AnnaneD,BellissantE,BollaertPE,BriegelJ,ConfalonieriM,DeGaudioR,KehD,KupferY,OppertM,MeduriGU:Corticosteroidsinthetreatmentofseveresepsisandsepticshockinadults:asystematicreview.
JAMA2009,301:2362-2375.
49.
SprungCL,AnnaneD:Corticosteroidsinsepticshock.
ControversiesinIntensiveCareMedicineBerlin:MedizinischWissenschaftlicheVerlagsgesellschaftKuhlenR,MorenoR,RanieriM,RhodesA2008,205-210.
50.
SchaafMJM,CidlowskiJA:Molecularmechanismsofglucocorticoidactionandresistance.
JSteroidBiochemMolBiol2002,83:37-48.
51.
CooperMS,StewartPM:AdrenalInsufficiencyinCriticalIllness.
JIntensiveCareMed2007,22:348-362.
52.
DicksteinG:Ontheterm"relativeadrenalinsufficiency"–orwhatdowereallymeasurewithadrenalstimulationtestsJClinEndocrinolMetab2005,90:4973-4974.
Moranetal.
CriticalCare2010,14:R134http://ccforum.
com/content/14/4/R134Page14of1553.
deJongMFC,BeishuizenA,GroeneveldAB:Definingrelativeadrenalinsufficiencyinthecriticallyill:TheACTHtestrevisited.
YearbookofIntensiveCareandEmergencyMedicineBerlin:Springer-VerlagVincentJL2006,539-551.
54.
PintilieM:Competingrisks:ApracticalperspectiveChichester,UK:JohnWiley&SonsLtd2006.
55.
SouthernDA,FarisPD,BrantR,GalbraithPD,NorrisCM,KnudtsonML,GhaliWA,APPROACHInvestigators:Kaplan-Meiermethodsyieldedmisleadingresultsincompetingriskscenarios.
JClinEpidemiol2006,59:1110-1114.
56.
WilliamsonPR,Kolamunnage-DonaR,SmithCT:Theinfluenceofcompetingriskssettingonthechoiceofhypothesistestfortreatmenteffect.
Biostat2007,8:689-694.
57.
RhenT,CidlowskiJA:Antiinflammatoryactionofglucocorticoids–newmechanismsforolddrugs.
NEnglJMed2005,353:1711-1723.
58.
YangS,ZhangL:Glucocorticoidsandvascularreactivity.
CurrentVascularPharmacology2004,2:1-12.
59.
RussellJA,WalleyKR,GordonAC,CooperDJ,HébertPC,SingerJ,HolmesCL,MehtaS,GrantonJT,StormsMM,CookDJ,PresneillJJ,DieterAyersfortheVasopressinandSepticShockTrialInvestigators:Interactionofvasopressininfusion,corticosteroidtreatment,andmortalityofsepticshock.
CritCareMed2009,37:811-818.
60.
KehD,BoehnkeT,Weber-CartensS,SchulzC,AhlersO,BerckerS,etal:Immunologicandhemodynamiceffectsof"low-dose"hydrocortisoneinsepticshock:adouble-blind,randomized,placebo-controlled,crossoverstudy.
AmJRespirCritCareMed2003,167:512-520.
61.
deKruifMD,LemaireLC,GiebelenIA,vanZoelenMA,PaterJM,vandenPangaartPS,GrootAP,deVosAF,ElliottPJ,MeijersJC,LeviM,vanderPollT:Prednisolonedose-dependentlyinfluencesinflammationandcoagulationduringhumanendotoxemia.
JImmunol2007,178:1845-1851.
62.
ArafahBM:Hypothalamicpituitaryadrenalfunctionduringcriticalillness:limitationsofcurrentassessmentmethods.
JClinEndocrinolMetab2006,91:3725-3745.
63.
MinneciP,DeansK,NatansonC,EichackerPQ:Increasingtheefficacyofanti-inflammatoryagentsusedinthetreatmentofsepsis.
EuropeanJournalofClinicalMicrobiology&InfectiousDiseases2003,22:1-9.
64.
SeamN:Corticosteroidsforsepticshock:correspondence.
NEnglJMed2008,358:2068-2069.
65.
WilliamsonPR,GambleC,AltmanDG,HuttonJL:Outcomeselectionbiasinmeta-analysis.
StatMethodsMedRes2005,14:515-524.
66.
AnnaneD:Improvingclinicaltrialsinthecriticallyill:uniquechallenge–sepsis.
CritCareMed2009,37:S117-S128.
67.
BellomoRMFF,WarrillowSJM,ReadeMCM:Whyweshouldbewaryofsingle-centertrials.
CritCareMed2009,37:3114-3119.
68.
BormGF,LemmersO,FransenJ,DondersR:Theevidenceprovidedbyasingletrialislessreliablethanitsstatisticalanalysissuggests.
JClinEpidemiol2009,62(62(7)):711-715,e1.
69.
HigginsJPT,ThorntonA,SpiegelhalterDJ:Are-evaluationofrandom-effectsmeta-analysis.
JRStatSocSerAStatSoc2009,172:137-159.
70.
SuttonAJ,CooperNJ,JonesDR,LambertPC,ThompsonJR,AbramsKR:Evidence-basedsamplesizecalculationsbaseduponupdatedmeta-analysis.
StatMed2007,26:2479-2500.
71.
ShrierI,PlattRW,SteeleRJ:Mega-trialsvs.
meta-analysis:Precisionvs.
heterogeneityContemporaryClinicalTrials2007,28:324-328.
72.
TerrinN,SchmidCH,LauJ:Inanempiricalevaluationofthefunnelplot,researcherscouldnotvisuallyidentifypublicationbias.
JClinEpidemiol2005,58:894-901.
73.
MinneciPC,DeansKJ,EichackerPQ,NatansonC:Theeffectsofsteroidsduringsepsisdependondoseandseverityofillness:anupdatedmeta-analysis.
ClinicalMicrobiologyandInfection2009,15:308-318.
74.
JaeschkeR,AngusDC:Livingwithuncertaintyintheintensivecareunit:shouldpatientswithsepsisbetreatedwithsteroidsJAMA2009,301:2388-2390.
75.
MoranJL,BerstenAD,SolomonPJ:Meta-analysisofcontrolledtrialsofventilatortherapyinacutelunginjuryandacuterespiratorydistresssyndrome:analternativeperspective.
IntensiveCareMed2005,31:227-235.
76.
BorensteinM,HedgesLV,HigginsJP,RothsteinHR:IntroductiontoMeta-analysisWestSussex,UK:JohnWiley&Sons,Ltd2009.
77.
FinferS:Corticosteroidsinsepticshock.
NEnglJMed2008,358:188-190.
78.
WagnerHN,BennettIL,LasagnaL,CluffLE,RosenthalMB,MirickGS:Theeffectofhydrocortisoneuponthecourseofpneumococcalpneumoniatreatedwithpenicillin.
BullJohnsHopkinsHosp1955,98:197-215.
79.
ThompsonWL,GurleyHT,LutzBA,JacksonBL,KyolsLK,MorrisIA:Inefficacyofglucocorticoidsinshock(double-blindedstudy).
ClinRes1976,24:258A.
80.
LucasCE,LedgerwoodAM:Thecardiopulmonaryresponsetomassivedosesofsteroidsinpatientswithsepticshock.
ArchSurg1984,119:537-541.
81.
TheVeteransAdministrationSystemicSepsisCooperativeStudyGroup:Effectofhigh-doseglucocorticoidtherapyonmortalityinpatientswithclinicalsignsofsystemicsepsis.
TheVeteransAdministrationSystemicSepsisCooperativeStudyGroup.
NEnglJMed1987,317:659-665.
82.
SchattnerA,el-HadorI,HahnT,LandauZ:Tripleanti-TNF-alphatherapyinearlysepsis:apreliminaryreport.
JIntMedRes1997,25:112-116.
83.
ConfalonieriM,UrbinoR,PotenaA,PiattellaM,ParigiP,PuccioG,DellaPortaR,GiorgioC,BlasiF,UmbergerR,MeduriGU:Hydrocortisoneinfusionforseverecommunity-acquiredpneumonia:apreliminaryrandomizedstudy.
AmJRespirCritCareMed2005,171:242-248.
84.
RinaldiS,AdembriC,GrechiS,DeGaudioAR:Low-dosehydrocortisoneduringseveresepsis:effectsonmicroalbuminuria.
CritCareMed2006,34:2334-2339.
85.
HuhJW,LimCM,KohY,HongSB:Effectoflowdosesofhydrocortisoneinpatientwithsepticshockandrelativeadrenalinsufficiency:3daysversus7daystreatment.
CritCareMed2006,34:A101.
86.
LoisaP,ParviainenI,TenhunenJ,HovilehtoS,RuokonenE:Effectofmodeofhydrocortisoneadministrationonglycemiccontrolinpatientswithsepticshock:aprospectiverandomizedtrial.
CritCare2007,11:R21.
87.
NawabQ,GoldenE,ConfalonieriM,UmbergerR,MeduriGU:Glucocorticoid(GC)TreatmentinSevereCommunity-AcquiredPneumonia(CAP):ComparisonofHydrocortisone[HC]vs.
Methylprednisolone[MP].
AmericanThoracicSociety:InternationalConference2007,A594.
88.
CicarelliDD,VieiraJE,BensenorFE:Earlydexamethasonetreatmentforsepticshockpatients:Aprospectiverandomizedclinicaltrial.
SaoPauloMedicalJournal2007,125:237-241.
89.
KurugundlaN,IrugulapatiL,KilariD,AmchentsevA,DevakondaA,GeorgeL,RaoofS:Effectofsteroidsinsepticshockpatientswithoutrelativeadrenalinsufficiency.
AmericanThoracicSociety:InternationalConference2008,A116.
doi:10.
1186/cc9182Citethisarticleas:Moranetal.
:Updatingtheevidencefortheroleofcorticosteroidsinseveresepsisandsepticshock:aBayesianmeta-analyticperspective.
CriticalCare201014:R134.
SubmityournextmanuscripttoBioMedCentralandtakefulladvantageof:ConvenientonlinesubmissionThoroughpeerreviewNospaceconstraintsorcolorgurechargesImmediatepublicationonacceptanceInclusioninPubMed,CAS,ScopusandGoogleScholarResearchwhichisfreelyavailableforredistributionSubmityourmanuscriptatwww.
biomedcentral.
com/submitMoranetal.
CriticalCare2010,14:R134http://ccforum.
com/content/14/4/R134Page15of15
legionbox怎么样?legionbox是一家来自于澳大利亚的主机销售商,成立时间在2014年,属于比较老牌商家。主要提供VPS和独立服务器产品,数据中心包括美国洛杉矶、瑞士、德国和俄罗斯。其中VPS采用KVM和Xen架构虚拟技术,硬盘分机械硬盘和固态硬盘,系统支持Windows。当前商家有几款大硬盘的独立服务器,可选美国、德国和瑞士机房,有兴趣的可以看一下,付款方式有PAYPAL、BTC等。...
Dynadot 是一家非常靠谱的域名注册商家,老唐也从来不会掩饰对其的喜爱,目前我个人大部分域名都在 Dynadot,还有一小部分在 NameCheap 和腾讯云。本文分享一下 Dynadot 最新域名优惠码,包括 .COM,.NET 等主流后缀的优惠码,以及一些新顶级后缀的优惠。对于域名优惠,NameCheap 的新后缀促销比较多,而 Dynadot 则是对于主流后缀的促销比较多,所以可以各取所...
点击进入亚云官方网站(www.asiayun.com)公司名:上海玥悠悠云计算有限公司成都铂金宿主机IO测试图亚洲云Asiayun怎么样?亚洲云Asiayun好不好?亚云由亚云团队运营,拥有ICP/ISP/IDC/CDN等资质,亚云团队成立于2018年,经过多次品牌升级。主要销售主VPS服务器,提供云服务器和物理服务器,机房有成都、美国CERA、中国香港安畅和电信,香港提供CN2 GIA线路,CE...
www.mfcclub.net为你推荐
bbs.99nets.com怎么打造完美SF刘祚天你们知道21世纪的DJ分为几种类型吗?(答对者重赏)百度关键词工具百度有关键字分析工具吗?Google AdWords有的www.qq530.com谁能给我一个听歌的网站?www.zjs.com.cn中通快递投诉网站网址是什么?百度指数词百度指数我创建的新词抓站工具公司网站要备份,谁知道好用的网站抓取工具,能够抓取bbs论坛的。推荐一下,先谢过了!杨丽晓博客明星的最新博文partnersonline我家Internet Explorer为什么开不起来ww.66bobo.com有的网址直接输入***.com就行了,不用WWW, 为什么?
已备案域名注册 国外vps租用 免费注册网站域名 最便宜虚拟主机 免费顶级域名 缓存服务器 好玩的桌面 刀片服务器的优势 cdn联盟 91vps 域名接入 qq对话框 中国电信宽带测速网 息壤代理 hkt 空间首页登陆 海外空间 监控服务器 lamp怎么读 酸酸乳 更多