We included 23 RCTs. Most were of cross-over design, on adults undergoing a variety of chemotherapeutic regimens ranging from moderate to high emetic potential for a variety of cancers. The majority of the studies were at risk of bias due to either lack of allocation concealment or attrition. Trials were conducted between 1975 and 1991. No trials involved comparison with newer anti-emetic drugs such as ondansetron.
Comparison with placebo
People had more chance of reporting complete absence of vomiting (3 trials; 168 participants; RR 5.7; 95% CI 2.6 to 12.6; low quality evidence) and complete absence of nausea and vomiting (3 trials; 288 participants; RR 2.9; 95% CI 1.8 to 4.7; moderate quality evidence) when they received cannabinoids compared with placebo. The percentage of variability in effect estimates that was due to heterogeneity rather than chance was not important (I2 = 0% in both analyses).
People had more chance of withdrawing due to an adverse event (2 trials; 276 participants; RR 6.9; 95% CI 1.96 to 24; I2 = 0%; very low quality evidence) and less chance of withdrawing due to lack of efficacy when they received cannabinoids, compared with placebo (1 trial; 228 participants; RR 0.05; 95% CI 0.0 to 0.89; low quality evidence). In addition, people had more chance of 'feeling high' when they received cannabinoids compared with placebo (3 trials; 137 participants; RR 31; 95% CI 6.4 to 152; I2 = 0%).
People reported a preference for cannabinoids rather than placebo (2 trials; 256 participants; RR 4.8; 95% CI 1.7 to 13; low quality evidence).
Comparison with other anti-emetics
There was no evidence of a difference between cannabinoids and prochlorperazine in the proportion of participants reporting no nausea (5 trials; 258 participants; RR 1.5; 95% CI 0.67 to 3.2; I2 = 63%; low quality evidence), no vomiting (4 trials; 209 participants; RR 1.11; 95% CI 0.86 to 1.44; I2 = 0%; moderate quality evidence), or complete absence of nausea and vomiting (4 trials; 414 participants; RR 2.0; 95% CI 0.74 to 5.4; I2 = 60%; low quality evidence). Sensitivity analysis where the two parallel group trials were pooled after removal of the five cross-over trials showed no difference (RR 1.1; 95% CI 0.70 to 1.7) with no heterogeneity (I2 = 0%).
People had more chance of withdrawing due to an adverse event (5 trials; 664 participants; RR 3.9; 95% CI 1.3 to 12; I2 = 17%; low quality evidence), due to lack of efficacy (1 trial; 42 participants; RR 3.5; 95% CI 1.4 to 8.9; very low quality evidence) and for any reason (1 trial; 42 participants; RR 3.5; 95% CI 1.4 to 8.9; low quality evidence) when they received cannabinoids compared with prochlorperazine.
People had more chance of reporting dizziness (7 trials; 675 participants; RR 2.4; 95% CI 1.8 to 3.1; I2 = 12%), dysphoria (3 trials; 192 participants; RR 7.2; 95% CI 1.3 to 39; I2 = 0%), euphoria (2 trials; 280 participants; RR 18; 95% CI 2.4 to 133; I2 = 0%), 'feeling high' (4 trials; 389 participants; RR 6.2; 95% CI 3.5 to 11; I2 = 0%) and sedation (8 trials; 947 participants; RR 1.4; 95% CI 1.2 to 1.8; I2 = 31%), with significantly more participants reporting the incidence of these adverse events with cannabinoids compared with prochlorperazine.
People reported a preference for cannabinoids rather than prochlorperazine (7 trials; 695 participants; RR 3.3; 95% CI 2.2 to 4.8; I2 = 51%; low quality evidence).
In comparisons with metoclopramide, domperidone and chlorpromazine, there was weaker evidence, based on fewer trials and participants, for higher incidence of dizziness with cannabinoids.
Two trials with 141 participants compared an anti-emetic drug alone with a cannabinoid added to the anti-emetic drug. There was no evidence of differences between groups; however, the majority of the analyses were based on one small trial with few events.
Quality of the evidence
The trials were generally at low to moderate risk of bias in terms of how they were designed and do not reflect current chemotherapy and anti-emetic treatment regimens. Furthermore, the quality of evidence arising from meta-analyses was graded as low for the majority of the outcomes analysed, indicating that we are not very confident in our ability to say how well the medications worked. Further research is likely to have an important impact on the results.