web only notes 11/2/48

As a result of the small number of participants in arms that allowed comparison of cognitive therapy with pharmacotherapy, the study has reduced power. Authors’ post hoc power calculation suggests the study had power to detect only large differences between treatments in remission rates (over 20% difference). For the primary outcome, authors used a conservative intention to treat analysis (assumed that people with missing HAM-D scores did not achieve remission). The study did not provide a comparison between augmentation and switch strategies. The design of the STAR*D study may have led to bias in the participating sample, as the proportion of people who accepted psychotherapy was lower than expected.