Elliott 1998.

Methods	Randomized controlled trial. Randomization method: not specified. Power: no power calculation reported. Analysis: not ITT analysis.
Participants	Country: USA. Setting: outpatient psychiatric and medical clinics in Seattle, USA. Recruitment: responses to advertisements at outpatient clinics. Inclusion criteria: PLWH with a DSM‐III‐R diagnosis of major depression according to SCID interview and HAM‐D score ≥ 18. Exclusion criteria: alcohol or substance abuse in previous month confirmed by urine drug testing, organic brain syndrome, dementia, severe concurrent HIV‐related physical illness, < 12 years of education, high suicide risk or a history of bipolar disorder, traumatic head injury or psychosis. Failure of previous antidepressant regimens was not an exclusion criterion in this study. Number randomized: 75. Number dropped out: 19 (25%) dropped out by 4 weeks, 41 (58%) dropped out by 12 weeks. Age: not reported. Gender: 70 (93%) men. Baseline HAM‐D score (mean): 24.33 (SD 5.66). CD4 T‐cell count (mean): 368 cells/mm3 (SD 307). ART: 19 (25%). Ethnicity: 56 (75%) white race. Socioeconomic details: 56(75%) 'single,' 22 (29%) employed.
Interventions	Experimental arm: paroxetine started at 10 mg daily, increased to 20 mg by week 1 and then 40 mg by week 2 if tolerated. Comparison arm 1: imipramine started at 50 mg daily, steadily increased to 100 mg by week 1 and 200 mg by week 2 if tolerated. Comparison arm 2: placebo. Duration: 12 weeks.
Outcomes	All outcome assessments were carried out at 2, 4, 6, 8 and 12 weeks. Primary outcomes: improvement in depression: SCID/21‐item HAM‐D, CGI, BSI; number of dropouts. Secondary outcomes: health‐related quality of life: SAS and QLESQ (only measured for 41 who completed a 12‐week study period and analysis only compared composite outcome of antidepressant 'responders' vs antidepressant 'non‐responders'); adverse effects: SAFTEE tool.
Notes	Date of study: not reported. Funding: not reported. Declarations of conflict of interest by authors: not reported. Other: USD10 payment per visit, only offered after screening.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Not reported.
Allocation concealment (selection bias)	Unclear risk	Not reported.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"Participants were randomly assigned to blind treatment." Uncertain if personal were blinded.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"All ratings were made by one of two authors who were blinded to study drug assignment."
Incomplete outcome data (attrition bias) All outcomes	High risk	25% had dropped out at 4 weeks and by 12 weeks 58% had dropped out. 8‐week analysis was conducted using LOCF for participants who had dropped between 4 and 8 weeks. The reasons for dropout were unevenly distributed between the 3 groups; however, numbers were too small to make meaningful conclusions about which direction this might bias the results. 4 weeks of antidepressant therapy may not be adequate to evaluate true antidepressant effect. Analysis was conducted with LOCF from 4 weeks for a portion of study population. This would likely bias towards no effect of antidepressants.
Selective reporting (reporting bias)	High risk	HAM‐D score was dichotomized with no mean (SD) reported. BSI results not reported but not a major outcome of interest so unlikely to bias study outcome. Authors report quality of life outcome measures for 'responders' and 'non‐responders,' but these results were not given for the individual treatment arms (placebo vs antidepressant).
Other bias	Low risk	None noted.