Page 44 - ESHRE2019
P. 44
Bias vs generalisability
• Bias and generalisability are different things.
Internal validity
Does the study provide an unbiased estimate of the treatment effect?
This is a matter of degree – slight bias won’t make us get wrong conclusion
• Butthereisnoexternalvaliditywithoutinternalvalidity! • Rememberthiswhenpeopletalkabout‘realworlddata’.
Does RoB matter?
Meta-analyses that include biased trials may present a misleading conclusion about the effect of a treatment
Empirical evidence
• Between 30–50% exaggeration of treatment efficacy in results from trials at high-risk of bias (Wood 2008)
• Intervention effect estimates were exaggerated by 17% in the trials with inadequate or unclear allocation concealment and 7% in non-blinded compared with blinded trials (Moher 1998)
External validity
Will the estimated treatment effect apply to different settings?
Are the results generalisable?
42
39
