Transcription of [email protected] …
1 ShijieRen, Jeremy E. Oakley and John W. StevensUniversity of OfHealth AndRelated (SD) beliefsonthe range oftreatmenteffectsinordertoinfertheprior distributionforthebetween-studySD ,wherethe range ,where isP( < ),Lfor low heterogeneity isP( < < ),Mfor moderate isP( < <1),Hfor high isP( >1),EHfor extremehigh Discussion , : A manuscript Incorporating genuine prior information about between-study heterogeneity in random effects pairwise and network meta-analyses is in submission. The pre-print can be downloaded from :Comparisonofresultsobtainedfromfixedeff ectandrandomeffectsmodelswiththepriordis tributionforheterogeneityasuniform[0,5], lognormalproposedbyTurneretal(2012), 1: confirmation of the need for a random effects model Canyoubecertainthatthetreatmenteffectsac rossthestudieswillbeidentical,ignoringwi thin-studysamplingvariability? Stage 2: consideration of an upper bound for Let ?Denotingthislimitby ,thismeansthatyouwouldthinkvaluesof above aretooimplausibletobecontemplated.
2 Stage 3: consideration of a full distribution for Turner RM, et al. Predicting the extent of heterogeneity in meta-analysis, using empirical data from the Cochrane Database of Systematic Reviews. IntJ Epidemiol. 2012;41(3):818 27. Canyoujudgessomevaluesintherange[1, ]tobemorelikelythanothers? Ifsheisnotabletomakesuchjudgements,thenw eproposeusingexternalinformationsuchasth epriordistributionsproposedbyTurneretal( 2012),butnowtruncatedto0,log( ) Ifsheisabletomakesuchjudgements, the data from NICE TA163 of infliximab for treating acute exacerbations in adults with severely active ulcerative colitis. Data were available from 4 studies of 3 treatments (placebo, infliximab and ciclosporin) and two studies to inform each treatment : Colectomy rate at 3 monthstreatment effect on log ORscaleOR, median (95% CrI) ciclosporinvs. placeboOR , median (95% CrI) infliximab vs. placebo ( , ) ( , )0000 Randomeffects with [0,5] (0, ) (0, ) ( , ) (0, ) wide CrIRandomeffectswith 2~log ( , ) ( , ) ( , ) ( , ) ( , ) extreme heterogeneityRandom effects with 2~truncatedlog , (0, ) ( , ) ( , ) ( , ) ( , ) the possibility of extreme heterogeneityRandomeffectswith ( 1)~ ( , )and =log +1 ( , ) ( , ) ( , ) ( , ) the possibility of extreme heterogeneityGeneral elicitation framework
