Seventeen trials (5,174 participants, range nine to 803) were included. Six trials were reported to be at a low risk of bias overall; other studies were limited by lack of reporting of randomisation sequence generation and allocation concealment methods or by lack of an intention-to-treat analysis.
In 14 trials of μ-opioid receptor antagonists treatment was found to be more effective than placebo (RR 0.69, 95% CI 0.63 to 0.76; Ι²=51%). Among the specific treatments, methylnaltrexone (RR 0.67, 95% CI 0.54 to 0.84; Ι²=72%; six trials), naloxone (RR 0.64, 95% CI 0.56 to 0.72; Ι²=0; four trials) and alvimopan (RR 0.71, 95% CI 0.65 to 0.78; Ι²=11%; four trials) were all found to be more effective than placebo. Data were limited for other treatments.
The μ-opioid receptor antagonists increased adverse events by 11% compared to placebo (RR 1.11, 95% CI 1.04 to 1.20; 10 trials). Increases in risk were statistically significant for diarrhoea (RR 1.61, 95% CI 1.21 to 2.13; 11 trials) and abdominal pain (RR 1.63, 95% CI 1.06 to 2.51; 11 trials) but were not statistically significant for other adverse events. For each specific treatment a general trend towards increased incidence of adverse events was reported but these results were not statistically significant. There was no statistically significant difference between treatment and placebo for reversal of analgesia.
There was some evidence of possible publication bias.