Log in with your email address username.


[Correspondence] Network meta-analysis of antidepressants

Cipriani and colleagues1 used re-sponse rate (a reduction of ≥50% of the total score) and dropout rate as their primary outcomes. The use of ranking scales for response rates is a misleading way to analyse data,2 which can inflate small differences between groups.3 The effect on the ranking scales was only a standardised mean difference of 0·30 (95% credible interval 0·26–0·34), which is far less than what is clinically relevant4 and is similar to the estimates reported in previous meta-analyses.