Kennedy 2004.

Methods	Eight‐week, double‐blind, double‐dummy, parallel‐group, randomised study
Participants	Males and non‐pregnant females using adequate contraception were eligible if they were between 18 and 65. Participants had to have a diagnosis of moderate to severe major depressive disorder without psychotic features. Participants were excluded if they had a predisposition to seizures, serious suicidal risk, poor previous therapeutic response to antidepressant medication, significant DSM‐IV Axis II diagnosis, history of or current diagnosis of anorexia nervosa or bulimia, recent history of drug dependence or abuse, including alcohol. Age range: 18‐65 years.
Interventions	Paroxetine: 71 randomised. Bupropion: 69 randomised. Paroxetine dose: 20‐40 mg/day. Bupropion dose: 150‐300 mg/day.
Outcomes	Hamilton rating scale for depression (HDRS‐17), Clinical Global Impression (CGI), Hamilton rating scale for anxiety (HAM‐A), Quality of Life in Depression (QLDS). Total dropout, dropout due to side effects. Number of patients experiencing at least one side effect, side ‐ profile.
Notes	Poor therapeutic response is defined as at least two failed responses to antidepressant medications to from two different antidepressant classes, at adequate dose and duration. Funding: bupropion manufacturer.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "patients were randomly assigned". No further details.
Allocation concealment (selection bias)	Unclear risk	No information provided.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Quote: "double blind, double dummy". No further details.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Quote: "double blind, double dummy". No further details.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Quote: "we performed statistical analysis using the Last Observation Carried Forward (LOCF) method on all subjects who took at least one dose of medication and completed at least a second visit".
Selective reporting (reporting bias)	Unclear risk	Outcomes data not clearly reported (the study was primarily designed to evaluate sexual functioning during paroxetine or bupropion treatment).
Other bias	Unclear risk	Sponsorship bias cannot be ruled out.