Gana 2006.

Methods	Multicenter, randomized, double‐blind, placebo‐controlled, fixed‐dose, parallel‐group clinical trial *We also included data from another publication by Kosinski which was a secondary analysis of the trial conducted by Gana and coworkers (Gana 2006).
Participants	1011 men/women ages 18–74 years with radiographically confirmed ACR Functional Class I–III OA of the knee or hip. Number of participants: tramadol 100 mg: 202; tramadol 200 mg: 201; tramadol 300 mg: 201; tramadol 400 mg: 202; control group: 205 Participants were required to have baseline index joint pain of at least 40 mm on 100‐mm pain VAS (0 = no pain, 100 = extreme pain) after the washout period. % women: tramadol 100 mg: 62.4%; tramadol 200 mg: 63.7%; tramadol 300 mg: 59.2%; tramadol 400 mg: 57.9%; placebo: 68.8%
Interventions	Active group 1: tramadol 100 mg once daily Active group 2: tramadol 200 mg once daily Active group 3: tramadol 300 mg once daily Active group 4: tramadol 400 mg once daily Control group: placebo once daily Treatment for 12 weeks.
Outcomes	From the Gana publication: PI: WOMAC Osteoarthritis Index for index and non‐index joints in the past 48 hours using 100‐mm VAS (0 'no pain' to 100 'extreme pain'), overall pain rated daily at approximately 8:00 p.m. using a 100‐mm VAS in response to the question "Overall, how much pain have you experienced in your study joint today?" Other: WOMAC Physical Function subscale (0–1700 mm), participant and physician global assessments of disease activity on a 100‐mm VAS, participants responded to the following sleep‐related questions using a 100‐mm VAS: (trouble falling asleep, need for sleep medication, how often they were awakened by pain during the night and how often they were awakened by pain in the morning). Participants assessed the overall quality of sleep using a 100‐mm VAS in response to the question, "Over the past week, how would you rate the overall quality of your sleep?"; SF‐36 Health Survey; adverse events (either spontaneously or in response to non‐directed questioning; results of physical exams, vital signs, clinical laboratory tests and electrocardiograms at study visits); and 16‐item questionnaire to record the presence or absence of common symptoms of physical dependence. Extracted physical function outcome: WOMAC Physical Function subscale (0–1700 mm) at 12 weeks, with lower values indicating benefit From the Kosinski publication: PI: arthritis PI during the past 48 hours in the index joint using a 100‐mm VAS with anchors of 0 (no pain) and 100 (extreme pain). Other: CPSI, which consists of 5 questions about severity of sleep impairment during previous week; adverse events, withdrawal symptoms, and physical dependence after abrupt discontinuation of study treatment. Extracted pain outcome: Arthritis Pain Intensity scale (0–100 mm) at 12 weeks, with lower value indicating less pain
Notes	From the Gana publication: Supported by Biovail Laboratories International SRL From the Kosinski publication: Supported by a grant from Ortho‐McNeil Janssen Scientific Affairs, LLC
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "A randomization schedule was generated with permuted blocks of 10 subjects. Each site received study medication kits that were marked with the randomization numbers. Investigators used an interactive voice‐response system to assign randomization numbers to subjects." (p.1392)
Allocation concealment (selection bias)	Low risk	Quote: "Each site received study medication kits that were marked with the randomization numbers. Investigators used an interactive voice‐response system to assign randomization numbers to subjects. Eligible subjects were randomly assigned to a 1:1:1:1:1 ratio." (p.1392)
Blinding of personnel	Unclear risk	Authors did not explicitly report blinding of personnel.
Blinding of participants	Low risk	Quote: "To preserve blinding, study medication tablets were similar in appearance and size." (p.1392)
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Since this was a participant‐reported outcome and participants were blinded, the outcome assessment can be considered blinded.
Incomplete outcome data for pain and physical function	High risk	Quote: "Analyses were conducted on an intent‐to‐treat (ITT) population, defined as all randomized subjects who took at least one dose of study medication, using the last‐observation‐carried‐forward approach to replace missing post‐baseline efficacy data." (p.1393) 558/1020 (55.2%) of participants completed 12 weeks of treatment. However, 1011/1020 participants were included in the ITT population. Those omitted from ITT did not receive treatment. Withdrawals due to adverse events: 22.7% with tramadol vs 10.2% with placebo. Withdrawals for reasons other than adverse events: 22.4% with tramadol vs 15.3% with placebo. The imbalance in withdrawals due reasons other than adverse events was likely to also impact the outcome data despite an ITT analysis. *Gana 2006 provided outcome data for physical function while Kosinski 2007 provided outcome data for pain.
Incomplete outcome data for adverse effects All outcomes	High risk	Quote: "Analyses were conducted on an intent‐to‐treat (ITT) population, defined as all randomized subjects who took at least one dose of study medication, using the last‐observation‐carried‐forward approach to replace missing post‐baseline efficacy data." (p.1393) 558/1020 (55.2%) of participants completed 12 weeks of treatment. However, 1011/1020 participants were included in the ITT population. Those omitted from ITT did not receive treatment. Withdrawals due to adverse events: 22.7% with tramadol vs 10.2% with placebo. Withdrawals for reasons other than adverse events: 22.4% with tramadol vs 15.3% with placebo. The imbalance in withdrawals due reasons other than adverse events was likely to also impact the outcome data despite an ITT analysis. *Gana 2006 provided outcome data for any adverse events, withdrawals due to adverse events and serious adverse events.
Selective reporting (reporting bias)	Unclear risk	All prespecified outcomes (pain, physical function and stiffness, and sleep quality) appeared to have been reported in the results, but we did not have access to the protocol for verification.
Other biases	Low risk	No other sources of bias.