Data were combined for all antidepressant classes, ie, SNRIs, SSRIs and TCAs. It was assumed that all Canadian drugs within a class are essentially the same and, in equipotent doses used continuously over several weeks, act similarly. Subgroup analyseswere done forindividualdrugswhenever possible.
Discontinuation rates were extracted by a third reviewer. The articles retrieved for the efficacy study were examined, and placebo controlled studies were selected. Two discontinuation rates were calculated – those due to ADRs and those due to perceived (by the patient) lack of effect. Both rates were summarized in the same manner as that for efficacy.
Quality analysis of the retrieved studies was not undertaken because inclusion criteria were comprehensive and accounted for most of the items that would appear on a quality assessment checklist. Hence, all studies are expected to have comparable (high) quality.
Statistical significance between rates was determined by calculating individual Z scores for each rate,