Forty-three studies were identified. There were 40 studies of efficacy (n=8,691, range 42 to 646): 31 RCTs (n=6,837) and 12 observational studies (n=1,854) that included eight open-label studies with no controls. One of the studies of efficacy provided data on abuse and misuse. Three studies provided data solely on abuse and misuse. The four studies that assessed misuse and abuse outcomes included 16,098 patients. Mean quality score for RCTs was 10 (range 9 to 13). Twenty-four RCTs and seven observational studies had an excellent quality score (median quality score 13, range 9 to 15).
Efficacy (RCTs only): Opioid versus placebo only (18 RCTs) gave a significant reduction in pain (SMD -0.557, p<0.001; 18 studies) and physical function (SMD -0.432, p<0.001; nine studies). The effect was not significant for improved sleep (six studies). For quality of life, there was a significant decrease in the mental component score (SMD -0.220, p=0.04; four studies) and no significant effect on the physical component score (four studies). Sensitivity analyses found the effect size for pain reduction was greater for patients with neuropathic pain (SMD -0.907, p<0.001; four studies) than for those with osteoarthritis (SMD -0.457, p<0.001; 14 studies). There was no association between pain and physical function outcomes and study quality score or mean participant age.
Opioid versus active treatment (three studies): Head-to-head comparison studies found no significant differences in effect for opioid compared to non-steroidal anti-inflammatory or long acting tramadol.
Abuse and misuse (four studies): Two studies found older patients were significantly less likely to misuse or abuse opioids (adjusted odds ratio (AOR) 0.95, 95% confidence intervals (CI) 0.90 to 0.99) and (AOR 0.94, 95% CI 0.89 to 0.99). All four studies found low levels of abuse and misuse.
Adverse events (41 studies): For opioid-treated patients, occurrence rates for the most common events were constipation 30% (range 12% to 52%), nausea 28% (range 12% to 61%), dizziness 22% (range 10% to 37%) and somnolence 21% (range 10% to 39%). Events were mostly rated as mildly or moderately severe and were lower than in placebo control groups. Numbers needed to harm and data for age effects were reported.
There was no evidence of publication bias.