The evidence we have from randomised trials shows cannabinoids to be slightly better than conventional antiemetics for treating chemotherapy induced emesis, and patients prefer them. They are also more toxic. Two extreme positions could be taken, perhaps using the following arguments.
Arguments for and against
The optimistic position favours cannabinoids. Overwhelmingly, patients preferred cannabinoids for future chemotherapy, even though cannabinoids were only slightly more effective than other antiemetics and only for moderately emetogenic chemotherapy. Patients' subjective view on preference is more important than the scientifically evaluated efficacy of that intervention. Although side effects occur more often with cannabinoids, these may be concentrated in a fairly small number of patients so that most patients find cannabinoids effective without undue adverse effects. There are even some potentially beneficial side effects. Of 100 cancer patients undergoing chemotherapy who received a cannabinoid 30 more would be sedated, 20 would feel a sensation of a “high,” and 15 would feel euphoric compared with 100 who received a conventional antiemetic. Some patients may perceive a degree of sedation or somnolence as useful during chemotherapy. Thus, further clinical trials with cannabinoids in chemotherapy are justified.
The pessimistic position favours conventional antiemetics, as cannabinoids are not much better, and their toxicity is unacceptably high (dizziness, dysphoria, hallucinations, paranoia). The toxic effects may lead to study withdrawal. There were no comparisons of cannabinoids with a serotonin (5-HT3) receptor antagonist, the best comparator for prevention of acute emesis in highly emetogenic chemotherapy. It is, however, unlikely that cannabinoids would be more effective and less toxic than a 5-HT3 receptor antagonist.
The correct position is probably somewhere in the middle. Undoubtedly, most patients preferred cannabinoids for future chemotherapy cycles. One in two compared with placebo, and one in three compared with conventional antiemetics would have preferred to receive cannabinoids again. We do not know whether the incidence of cannabinoid related “beneficial” side effects was related to this overwhelming preference for cannabinoids for future chemotherapy.
Efficacy and safety
Before a chemical compound can be recommended for medical use, both its efficacy and safety must be proved. Cannabinoids were more effective than conventional antiemetics (prochlorperazine, metoclopramide). Of 100 cancer patients treated with oral cannabinoids during chemotherapy, 16 will not be nauseated (number needed to treat 6.4) and 13 will not vomit (8) who would have done so had they all received a conventional antiemetic. Compared with placebo, cannabinoids were obviously better, although a placebo may not be an adequate comparator in patients having chemotherapy. We could not establish a dose-response relation, mainly because there were insufficient quality data from the original trials. In some trials, dose was adjusted during the trial.34
The relation between plasma concentration of a cannabinoid and its antiemetic efficacy is unclear. In one trial, antiemetic efficacy was related to the plasma concentration of dronabinol.35
Another trial found no correlation between dronabinol serum levels and efficacy or adverse reactions.41
Defining an intervention's usefulness includes estimates of the likelihood for harm. The physical and neuropsychiatric adverse effects of long term use of cannabis are well established, based mainly on observations from long term marijuana smokers.62
Our systematic review shows clearly that cannabinoids are toxic for many patients even when taken orally and acutely (for 24 hours). Some adverse effects occurred almost exclusively with cannabinoid exposure. For instance, 5% of patients had paranoia, 6% had hallucinations, and almost 13% had dysphoria or depression (table ). The number of patients withdrawing from the studies due to intolerable side effects is the most reliable parameter of the severity of cannabinoid related toxicity. One in eleven patients treated with cannabinoids will stop treatment who would not have stopped treatment had they taken a placebo or another antiemetic. This is an important new message for doctors, policy makers, and patients.
These results should make us think hard about the ethics of clinical trials of cannabinoids when safe and effective alternatives are known to exist and when efficacy of cannabinoids is known to be marginal. The trials analysed here are likely to be the largest subgroup on the medical use of cannabinoids and therefore the single most important source of information on their potential for harm.
Effect of bias
This meta-analysis is open to some biases, and they all have the potential to overestimate the efficacy and to underestimate the harm of cannabinoids. The trials we included were of acceptable quality according to the Oxford quality scale, with 25 of 30 trials scoring 3 or 4. In 70% of trials an adequate method of blinding was described. Most crossover trials used a double dummy design. Cannabinoids were given as tablets or intramuscular injection, so any psychological effect of smoking a joint was not a factor. However, cannabinoids showed specific adverse effects that control treatments did not, and their incidence was high. In one trial of oral nabilone, many patients identified which drug they received because of the adverse effects experienced.59
In a series of 100 blinded dronabinol and placebo treatments, nurses correctly identified the active treatment in 85% and patients in 95%; seven of the 10 errors were made by patients on the first drug trial of the study.63
We must therefore assume that most of these trials had some degree of observer bias.
Some trials studied selected groups of patients who either had not responded to conventional antiemetic prophylaxis during previous chemotherapy cycles (“high risk” patients) or regularly used cannabis. Both subgroups introduce a bias in favour of cannabinoids. The selection of refractory patients introduces bias against the regimens that do not include cannabinoids.4
Younger age and previous marijuana exposure have been identified as factors that predicted response to antiemetic treatment.64
Regular cannabis users may be more likely to believe in its antiemetic efficacy. They may also experience adverse effects of cannabis more positively than patients who have not previously taken cannabis (for instance, a “high” or somnolence). In one trial, where more than 50% of patients were regular cannabis users, only 6% did not believe that cannabis would reduce their nausea and vomiting symptoms.60
In view of the numerous biases in these trials, “preference” and “satisfaction” are likely to be surrogate end points. These endpoints alone do not allow to judge with confidence on the usefulness of cannabinoids in the control of chemotherapy related emesis.
Finally, we have the problem of size. Small trials can be greatly affected by the random play of chance.65
Of the 30 studies available for analysis, only nine had over 50 analysable patients and only four more than 100. Small size has been shown to overestimate treatment effects in other circumstances,66
and it is not possible to rule out similar effects here.
The research agenda needs to be clear. Priority should go to trials of cannabinoids for indications where there are few competing drugs, such as spasticity in multiple sclerosis. In chemotherapy, the combination of weak antiemetic efficacy with potentially beneficial side effects (sedation, euphoria) raises the question whether further trials should be designed to establish the usefulness of cannabinoids as adjuncts to modern antiemetics (for instance, 5-HT3 receptor antagonists). Minimal effective doses would then be needed. Identification of patients who are most likely to profit from the antiemetic effect of cannabinoids and least likely to suffer from neuropsychiatric adverse effects is needed.
In conclusion, the cannabinoids reviewed here were slightly superior to conventional antiemetics after chemotherapy, and patients preferred them. However, potentially serious adverse effects, even when the drugs are taken short term orally or intramuscularly, are likely to limit their widespread use. In selected patients, cannabinoids may be useful as mood enhancing adjuvants for the control of chemotherapy related sickness.