Average differences in observed log odds ratios between underpowered () compared to adequately powered studies, in subset A of 1,107 meta-analyses, overall and within medical specialties, outcome types and intervention-comparison types.

1Other medical specialties, semi-objective outcomes, subjective outcomes and non-pharmacological interventions defined in footnotes to Table 2.
2Comparison is less meaningful when comparing two active interventions since the a priori “better” active intervention is not taken into account.