Silverfield 2002.

Methods	Parallel, multicenter, double‐blind RCT
Participants	Participants ages 35–75 years with symptomatic OA of hip or knee received stable doses of NSAID or COX‐2 Number of participants: tramadol/acetaminophen group: 197; control group: 111 % women: active group: 76.6%; control group: 63.1%
Interventions	Active group: tramadol 37.5 mg/acetaminophen 325 mg 1 or 2 pills QID Control group: placebo Treatment for 10 days in addition to ongoing NSAID or COX‐2‐selective inhibitor therapy. Number of pills/day increased up to 8. Participants continued receiving NSAID or COX‐2 at the same doses taken before study entry.
Outcomes	PI and pain relief evaluated using a 4‐point adjective scale (none, mild, moderate, severe) % participants in each relief category WOMAC Index score Participants who received tramadol had less pain than participants who received placebo (data not used in the pooling because of 4‐point scale) Extracted pain outcome: WOMAC Pain subscale (0–5) at 10 days, with lower values indicating benefit Extracted physical function outcome: WOMAC Physical Function subscale (0.5) at 10 days, with lower values indicating benefit
Notes	We contacted 1 of the coauthors and obtained the requested information (percentage of participants with moderate pain relief). Funding source not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "patients were assigned sequentially according to a randomization schedule" (p.286) Quote: "A list of unique medication code numbers was prepared using a computerized random‐number generator to ensure that any given patient was assigned randomly to 1 of 3 initial treatment groups." (p.286)
Allocation concealment (selection bias)	Unclear risk	No mention of allocation concealment.
Blinding of personnel	Low risk	Quote: "Treatment assignments were not revealed to patients, investigators, clinical staff, or monitors until all patients had completed treatment and database was finalized." (p.286)
Blinding of participants	Low risk	Quote: "Study medication consisted of identical‐appearing tablets containing tramadol/acetaminophen or matching placebo." (p.286)
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "Treatment assignments were not revealed to patients, investigators, clinical staff, or monitors until all patients had completed treatment and database was finalized." (p.286)
Incomplete outcome data for pain and physical function	High risk	Quote: "Efficacy summaries were based on the intent‐to‐treat population, defined as all patients randomized to receive study medication who took at least 1 dose and had post randomization efficacy data. Safety summaries were based on all randomized patients who took at least 1 dose of study medication." (p.287) Quote: "If a patient dropped out within the first 4 hours after taking the first dose, the baseline‐observation‐carried forward approach was used to impute missing pain assessments. All other missing assessments were imputed using the last‐observation‐carried‐forward approach." (p.287) All randomized participants were included in ITT analysis and were evaluable for safety. Withdrawals due to adverse events: 12.7% (25/197) with tramadol vs 5.4% (6/111) with placebo. Withdrawals for reasons other than adverse events (discontinued prematurely, lack of efficacy, protocol violations, other): 7.6% (15/197) with tramadol vs 0% with placebo. Last or baseline observations carried forward are hardly adequate imputation techniques. In addition, there was differential dropout since the adverse events predominantly occurred in the tramadol group.
Incomplete outcome data for adverse effects All outcomes	Low risk	Quote: "Efficacy summaries were based on the intent‐to‐treat population, defined as all patients randomized to receive study medication who took at least 1 dose and had post randomization efficacy data. Safety summaries were based on all randomized patients who took at least 1 dose of study medication." (p.287) Quote: "If a patient dropped out within the first 4 hours after taking the first dose, the baseline‐observation‐carried forward approach was used to impute missing pain assessments. All other missing assessments were imputed using the last‐observation‐carried‐forward approach." (p.287) All randomized participants were included in ITT analysis and were evaluable for safety. Withdrawals due to adverse events: 12.7% (25/197) with tramadol vs 5.4% (6/111) with placebo. Withdrawals for reasons other than adverse events (discontinued prematurely, lack of efficacy, protocol violations, other): 7.6% (15/197) with tramadol vs 0% with placebo. Last or baseline observations carried forward are hardly adequate imputation techniques. In addition, there was differential dropout since the adverse events predominantly occurred in the tramadol group.
Selective reporting (reporting bias)	Unclear risk	All prespecified outcomes (pain, physical function, and WOMAC Index score) 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.