Stuppaeck 1994.

Methods	Six‐week, randomised, double‐blind study.
Participants	Inclusion criteria: DSM III major depression, melancholic subtype, Hamilton rating scale for depression (HDRS) 18+ Exclusion criteria: senile dementia (DSM‐III) alcohol or drug addiction, patients with a high risk for suicide, or ECT during the last 3 months, as well as relevant somatic diseases. Patients were also excluded if they received long‐term treatment with lithium. Age: 18‐65 years. Country: Austria, Germany. Setting: inpatients.
Interventions	Paroxetine: 78 participants. Amitriptyline: 75 participants. Paroxetine dose range: 20‐50 mg/day (mean daily dose: 33.3). Amitriptyline dose range: 50‐250 mg/day (mean daily dose: 166).
Outcomes	Montgomery and Asberg Depression Rating Scale (MADRS), Clinical Global Impression (CGI). Total dropout. Number of patients experiencing at least one side effect, side‐effect profile.
Notes	Funding: paroxetine manufacturer. quality rating: 21
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: patients were randomly allocated". No further details.
Allocation concealment (selection bias)	Unclear risk	No information provided.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "double blind (...) paroxetine was administered as a single daily dose in the morning, whereas amitriptyline was given each morning, noon, and evening. Patients in the paroxetine group received matching placebo tablets at noon and in the evening".
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Quote: "double blind".
Incomplete outcome data (attrition bias) All outcomes	High risk	Quote: "analyses were performed on the Intention‐to‐treat (ITT) population with the Last Observation Carried Forward (LOCF) and visit wise datasets". More than 20% of participants in both arms abandoned the study prematurely.
Selective reporting (reporting bias)	High risk	Only most frequent adverse effects reported.
Other bias	Unclear risk	Sponsorship bias cannot be ruled out.