Goldenberg 1996.

Methods	Study setting: Single center study in US Study design: cross‐over Study duration: (1) Duration of wash‐out not reported, (2) each study period 6 weeks, separated by two weeks of wash‐out, (3) no follow‐up
Participants	Inclusion criteria: Patients who met ACR criteria (1990). Age within 18‐60 years. Willingness to discontinue all central nervous system active medications, nonsteroidal antiinflammatory drugs, and analgesics (other than acetaminophen) for at least 1 week prior to the study. Pain VAS ≥ 30, Hamilton Rating Scale <= 18 Exclusion criteria: No current or past history of systemic illness (cardiac, kidney, haematological or liver disease) Total sample (N = 31): Age: 43 (SD 9.1) years; 90% female; 100 white; Pain baseline 68.4 (SD 20.4); FIQ total baseline 57.3 (17.6)
Interventions	Amitryptiline: 25 mg p.o. fixed Fluoxetine : 20 mg p.o. fixed Amitriptyline 25mg + fluoxetine 20 mg p.o.: fixed Placebo: p.o. Rescue medication: acetaminophen Co‐therapies allowed or prohibited: no information provided
Outcomes	At least 30% pain reduction: Calculated by imputation method Global improvement: Composite score (global well‐being, pain, tender point score, physician global VAS): >= 25% improvement on a 100‐point scale Fatigue: VAS 0‐100 Sleep problems: VAS 0‐100 Depression: Beck Depression Inventory Drop out due to adverse: Reported Serious adverse events: Not reported Pain: VAS 0‐100 Disease‐related quality of life: FIQ total score VAS 0‐80 Physical functioning: Not reported Anxiety: Not assessed Tenderness: Tender point score
Notes	High attrition rate (39%); withdrawals for any reason: amitriptyline 1/22 (4.5%), fluoxetine 4/26 (15%), amitrptyline + fluoxetine 5/24 (21%), placebo: 1/20 (5%). No details on adverse events reported. Carry‐over effects possible. Authors included a factor for periods effects in the analysis Funding: Public funding
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	The pharmacy assigned patients using randomisation tables
Allocation concealment (selection bias)	Unclear risk	No details reported
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"All tablets were identical in appearance"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"All evaluations were done by a physician who had no prior contact to the patients"
Incomplete outcome data (attrition bias) All outcomes	High risk	Completer analysis
Selective reporting (reporting bias)	Unclear risk	Insufficient information about the protocol to permit judgment
Group similarity at baseline (selection bias)	Low risk	Identical demographic and clinical data because of cross‐over design