Narushima 2002.

Study characteristics
Methods	Study design: randomised (3‐arm parallel randomised controlled clinical trial) Sample size: 347 (USA: 343; Argentina: 4) Method of randomisation: no information specified Method of concealment: no information specified Blinding: double‐blind Analysis: intention‐to‐treat Date the study was conducted: July 1991 ‐ June 1997
Participants	Location: USA and Argentina Total N randomised: 48 Intervention n: 32 fluoxetine 17 (88% male; no information on age available) nortriptyline 15 (47% male; no information on age available) Control n: 16 (75 % male; no information on age available) Inclusion criteria: thromboembolic stroke or intracerebral haemorrhage; no depression at baseline; age 18‐85 Diagnostic criteria ‐ long‐term physical condition: thromboembolic stroke or intracerebral haemorrhage (assessed by neurologists or neuroradiologists who evaluated computerized tomography or magnetic resonance scans) Diagnostic criteria ‐ depression: depression diagnosed according to DSM‐IV assessed by the PSE Exclusion criteria: severe comprehension deficit (unable to answer correctly part 1 of the token test); use of antidepressant; contraindication to nortriptyline or fluoxetine; acute stroke (within 6 month of study entry) History of depression: no information specified
Interventions	Intervention: fluoxetine ≤ 3 weeks: daily dose of 10 mg weeks 4 to 6: daily dose of 20 mg weeks 7 to 9: daily dose of 30 mg weeks 10‐12: daily dose of 40 mg nortriptyline ≤ 1 week: daily dose of 25 mg weeks 2 and 3: daily dose of 50 mg weeks 3 to 6: daily dose of 75 mg weeks 6 to 12: daily dose of 100 mg Control: matched placebo Duration of treatment: 3 months Length of follow‐up: 24 month (at baseline, 3, 6, 9, 12, and 24 months) [no follow‐up for participants from Argentina (duration of participation: 3 months)] Other study arms: no
Outcomes	Primary outcome: incidence of depression ‐ diagnosis: diagnosis according to DSM‐IV assessed by the PSE adverse events: N/A Secondary outcome: incidence of depression ‐ cut‐off: N/A severity of depression: HAM‐D‐17 (means were not reported but presented in a figure) cost‐effectiveness: N/A cost‐utility: N/A
Notes	Decreasement of medication in case of severe side effects Language: English Funding: National Institute of Mental Health Grants MH‐40355, MHH‐52879, MH‐53592 and Research Scientist Award MH‐00163 (RGR) Declaration of interest: no information specified
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	no information specified "Patients were randomly assigned to 12 weeks of treatment with fluoxetine, nortriptyline, or placebo unless one of the active drugs was contraindicated. Nortriptyline was contraindicated in patients with cardiac conduction abnormalities, and fluoxetine was contraindicated in patients with intracerebral hemorrhage. Of the 48 nondepressed patients, 7 had a contraindication to nortriptyline and 5 had a contraindication to fluoxetine. These patients were randomly assigned to receive noncontraindicated medication or placebo. Thus, 75% of the patients were randomly assigned to nortriptyline or fluoxetine, whereas all patients were randomly assigned to active or placebo medication."
Allocation concealment (selection bias)	Unclear risk	no information specified
Blinding of participants and personnel (performance bias) All outcomes	High risk	"The effectiveness of the blind was determined by asking patients and raters to guess whether they were taking active or placebo medication. For active medication, patients were correct 59 % and raters 65 % of the time. For placebo, patients were correct 55 % and raters 46 % of the time. There was no statistically significant difference from a random distribution of guesses for either patients or raters." (p. 298) While blinding for participants and outcome assessment received low risk of bias ratings, further study personnel appeared to be aware of study allocation: "Doses were decreased if side effects were severe, which was the case for six nondepressed patients. Of these six patients, four were treated with nortriptyline and two with fluoxetine. To maintain the double‐blind design, doses were decreased for equal numbers of placebo patients. Nortriptyline‐treated patients were monitored for serum drug concentrations throughout the treatment period to adjust the drug level within the therapeutic window."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"The effectiveness of the blind was determined by asking patients and raters to guess whether they were taking active or placebo medication. For active medication, patients were correct 59 % and raters 65 % of the time. For placebo, patients were correct 55 % and raters 46 % of the time. There was no statistically significant difference from a random distribution of guesses for either patients or raters." (p. 298)
Incomplete outcome data (attrition bias) All outcomes	High risk	numbers randomised: fluoxetine: n = 17; nortriptyline: n = 15; placebo: n = 16; total: N = 48 number of drop‐outs at 3 months: fluoxetine: n = 2; nortriptyline: n = 2; placebo: n = 1; total: N = 5 numbers analysed post‐intervention: fluoxetine: n = 15; nortriptyline: n = 13; placebo: n = 15; total: N = 43 reasons for drop‐out (pp. 298‐299): fluoxetine: gastrointestinal symptoms, refused treatment (note: the authors initially name n = 4 drop‐outs for the treatment study, however, among these n = 2 who became depressed, and therefore are not considered drop‐outs for the prevention task but incidence events); nortriptyline: sedative effects, refused treatment placebo: rash
Selective reporting (reporting bias)	Unclear risk	no protocol available
Other bias	High risk	"Patients were randomly assigned to 12 weeks of treatment with fluoxetine, nortriptyline, or placebo unless one of the active drugs was contraindicated. Nortriptyline was contraindicated in patients within cardiac conduction abnormalities, and fluoxetine was contraindicated in patients with intracerebral hemorrhage. Of the 48 nondepressed patients, 7 had a contraindication to nortriptyline and 5 had a contraindication to fluoxetine. These patients were randomly assigned to receive noncontraindicated medication or placebo. Thus, 75 % of the patients were randomly assigned to nortriptyline or fluoxetine, whereas all patients were randomly assigned to active or placebo medication" (p. 297) no information on baseline characteristics, e.g. age no declaration of interest