Rabkin 1999.

Methods	Randomized controlled trial. Randomization method: computer‐generated block randomization; 2:1 for fluoxetine:placebo. Power: not reported. Analysis: ITT analysis reported for some outcomes but not all (CGI only).
Participants	Country: USA. Setting: unknown. Recruitment: not specified. Inclusion criteria: known HIV seropositive for ≥ 2 months; physically healthy except HIV‐related conditions; aged 18‐70 years; people with AIDS defining condition had to be receiving treatment with primary care provider; DSM‐IV diagnosis of major depression or dysthymia (or both). Exclusion criteria: psychosis, bipolar disorder, past 6 months substance abuse, panic disorder, current risk of suicide, cognitive impairment, other antidepressant within last 2 weeks, psychotherapy within last 4 weeks, HIV wasting syndrome, significant diarrhoea, unstable health. Failure of previous antidepressant regimens was not an exclusion criterion in this study. Number enrolled: 120. Number dropped out: 33. Age (mean): 39 years. Baseline HAM‐D score (mean): Fluoxetine: 20 (SD 4.7) Placebo: 19 (SD 5.1) CD4 T‐cell count (mean): 295 cells/mm3 (SD 287). ART: 47%. Gender: not reported. Ethnicity: African American (20%), Latino (15%), white (65%). Socioeconomic details: 36% receiving disability benefits.
Interventions	Experimental arm: fluoxetine 20 mg for 8 weeks, increased every two weeks by 20 mg if response poor. Comparison arm: placebo. Duration: participants maintained randomized assignment for 8 weeks, then treatment responders continued treatment and follow‐up to week 26.
Outcomes	All outcomes measured at: baseline, 4 weeks and 8 weeks (and 26 weeks for treatment responders). Primary outcomes: improvement in depression: 21‐item HAM‐D score, CGI score, BSI, Beck Hopelessness Scale; number of dropouts. Secondary outcomes: immunological recovery: CD4 lymphocyte count; quality of life measures: QLESQ; adverse effects: SAFTEE.
Notes	Date of study: 1993. Funding: not reported. Declaration of conflict of interest by authors: not reported.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Patients were randomly assigned through computer‐generated blocks of six in a 2:1 ratio to fluoxetine or placebo."
Allocation concealment (selection bias)	Unclear risk	Block randomization may lead to lack of allocation concealment as the next assignment may be predictable.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not reported.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not reported.
Incomplete outcome data (attrition bias) All outcomes	High risk	High dropout rate (27.5%). There were systematic differences between dropouts and completers, dropouts had milder depressive symptoms at baseline. Of 33 dropouts, 24 were on fluoxetine and 9 were on placebo. Difficult to draw conclusions about the impact of this on results.
Selective reporting (reporting bias)	High risk	Some reported outcomes were ITT, others were not. This suggests possible selection of reported results.
Other bias	Low risk	None noted.