Does blinding of readers affect the results of meta-analyses?