BIAS - KU Leuven偏置鲁汶大学.pptVIP

* * * * * * * * * Evaluation of the Validity of Reviews Harald Herkner, MD MScAssociate Professor of Emergency MedicineEditor Cochrane Anaesthesia GroupMedical University Vienna, Austria Menu What Reviews are – and what they are not Bias at different levels Heterogeneity Is the interpretation appropriate? Example of a critical appraisal tool Reviews are .. EBM top level evidence source Low bias scientific method High precision statistical tool Sometimes a Cochrane review Always man-made Reviews are .. … not a simple entity Standard parallel group RCT Cluster RCT Non inferiority RCT Observational studies Diagnostic accuracy studies … not free of limitations of included studies A Systematic Review … is a scientific method with numerous assumptions, judgements and sources of error. Must withstand scientific scrutiny Needs critical appraisal Bias at individual study level Examples Selection at group allocation Differential loss to follow up Differential outcome measurement Allocation Intervention Control RCT: Random Allocation Allocation Intervention Control RCT: Selection Bias Relative Risk BIAS Bias at the between-study level Reporting bias “A distortion that arise if dissemination of studies is affected according to the size or direction of their effect.” BIAS Reporting Bias Types Publication bias Time-lag bias Language bias Retrieval bias Double publication bias Outcome reporting bias Reporting Bias? Peters JCE 08 Funnel Plot Asymmetry! Peters JCE 08 Reporting Bias! Contour enhanced funnel plot Reporting Bias Types Publication bias: Handsearch? Registries? Time-lag bias: Update reviews? Language bias: No language restrictions? Retrieval bias: Medline Embase minimum? Double publication bias: Diligent reviewers? Outcome reporting bias: Study protocols? Another example Peters JCE 08 Asymmetry but no reporting bias Peters JCE 08 Asymmetry, no reporting bias Peters JCE 08 ROB assessment? Sensitivity analysis! Heterogeneity? Data entry error? Play of c