Fig. 2

Difference in treatment ES estimates, stratified by meta-analyses characteristics, between: (a) trials with presence and lack of patient blinding; (b) trials with presence and lack of assessor blinding; (c) trials with presence and lack of care-provider blinding; (d) trials with and without principal-investigator blinding. Square, proportional to weight used in meta-meta-analysis; horizontal arrow/line, a 95% confidence interval; solid vertical line, line of no difference in treatment ES estimate