Eighteen RCTs (n=809 patients) were included in the review (12 cross-over trials and six parallel trials); sample sizes ranged from 10 to 177 patients. Two trials scored 5, two trials scored 4, six trials scored 2, and eight trials scored 3 points on the Jadad quality scale. Most of the trials did not report on allocation concealment or randomisation method. The adequacy of blinding was not assessed in any trial. Control of attrition bias was addressed in only five trials. There was variability across the trials in the reporting of outcomes. Drop-outs ranged from zero to 20.
Efficacy: A significant reduction in pain intensity was associated with cannabis-based interventions (SMD -0.61, 95% CI -0.84 to -0.37; I2=0%; seven trials). There was no evidence of publication bias.
Harms: A significant increase was reported in risks of central nervous system-related adverse events following cannabis-based interventions; specifically, euphoria (OR 4.11, 95% CI 1.33 to 12.72; I2=0%; NNH 8, 95% CI 5 to 19; four trials), harms related to alterations in perception (OR 4.51, 95% CI 3.05 to 6.66; I2=2.8%; NNH 7; nine trials), events concerning motor function (OR 3.93, 95% CI 2.83 to 5.47; I2 0%; NNH 5; eight trials), and events relating to cognitive function (OR 4.46, 95% CI 2.37 to 8.37; I2=0%; NNH 8; five trials). There were no significant differences for dysphoria.
An increased risk for gastro-intestinal adverse events was reported, but a statistical synthesis was not conducted due to high heterogeneity.
Subgroup analyses did not materially alter the main results.