Connor 1999a.

Study characteristics
Methods	Design: single‐centre, randomised, placebo‐controlled, parallel, flexible dose, double‐blind Duration of intervention: 12 weeks Placebo run‐in: not specified
Participants	Setting: Veterans Administration Medical Center Sample size: 54 randomised to fluoxetine and placebo Mean age: 37 (32.44) (median age in years (quartiles)) Sex: 5 men and 49 women, none combat‐related Diagnostic measure: DSM‐III and SCID Inclusion criteria: Quote: “Individuals aged 18‐55 were included if they met DSM‐III‐R criteria for PTSD according to the Structured Clinical Interview for DSM‐III‐R and were civilians" Exclusion criteria: Quote: “Exclusion criteria, also determined by SCID, comprised a history of psychosis, bipolar disorder, antisocial personality disorder, current or recurrent or recent risk of suicide, homicide, and drug or alcohol abuse disorder within the previous six months” Comorbidity: None Dropouts: 16/54 (6/27 in the fluoxetine and 11/27 in the placebo group)
Interventions	Pharmacological intervention: Quote: “Drug or placebo was initiated at a daily dose of 10 mg or its equivalent and increased at a rate of 10 mg per week to a maximum of 60 mg/day as needed and tolerated"
Outcomes	Primary outcome: DGRP (change, severity) Secondary outcomes: SIP, DTS, SDS, VS Time points: Quote: “Assessments were conducted at pre‐treatment baseline (week 0) and at weeks 1, 2, 3, 4, 6, 8, 10 and 12" Data estimation: ITT and LOCF
Notes	Dates of trial: Not stated Industry‐funded: Yes Medication provided by industry: Yes. Quote: "Fluoxetine and matching placebo capsules were provided by Lilly Research Laboratories, and packaging was performed by the Duke University Medical Centre Pharmacy" Any of the authors work for industry? No
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Randomisation was achieved by means of computer‐generated group assignment, whose results were available only to the hospital pharmacy"
Allocation concealment (selection bias)	Low risk	Quote: "Fluoxetine and matching placebo capsules were provided by Lilly Research Laboratories, and packaging was performed by the Duke University Medical Centre Pharmacy"
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Quote: "Our threefold approach to measurement – with the physician, interviewer and patient rating ‐ minimised the likelihood of halo effaces between scales yet still resulted in differences favouring drug for all three approaches. This strategy may have also helped to offset any possible breaches in the double‐blind – which may occur in any trial – although study blinding was not rigorously assessed"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "The Duke was assessed by the treating physician, who was blind to the other outcome measures. The SIP was rated by the study coordinator, while the DTS, SDS, and VS measures were completed by the patient"
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	More participants withdrew from the placebo group (11/27; 41%) compared to the fluoxetine group (6/27; 22%). Reasons for withdrawal were explained. No information was provided on sample characteristics at endpoint. Data estimation was based on ITT and LOCF values. Quote: "The results indicate broad comparability between the treatment groups with respect to demographic features, age at onset of PTSD, and number of traumatic events. Of the 17 early dropouts, 11 (65%) came from the placebo group, whereas six (36%) came from the fluoxetine group., a non‐significant difference. Reasons for drop‐out in the drug group included lack of efficacy (n=1), loss to follow‐up (n=3), relocation (n=1) and protocol violation (n=1: reported non‐attendance). Subjects in the placebo group dropped out owing to lack of efficacy (n=5), loss to follow‐up (n=1), relocation or job concerns (n=3), and protocol violation (n=2: noncompliance; initiation of treatment with prohibited non‐protocol antidepressant). For all analyses, missing values were filled in by carrying forward the score from the previous week. Outcome analyses are presented here as an intention‐to‐treat analysis, in which the last observation was carried forward to Week 12. However, we conducted an analysis of completer patients at Week 12, with essentially similar results"
Selective reporting (reporting bias)	Unclear risk	The study protocol was not available for this study
Other bias	Low risk	No other sources of bias were identified. Medication was provided by industry, but no authors work for industry.