Objectives There is increased interest in cannabinoids for cancer pain management and legislative changes are in progress in many countries. This study aims to determine the beneficial and adverse effects of cannabis/cannabinoids compared with placebo/other active agents for the treatment of cancer-related pain in adults.
Methods Systematic review and meta-analysis to identify randomised controlled trials of cannabinoids compared with placebo/other active agents for the treatment of cancer-related pain in adults to determine the effect on pain intensity (primary outcome) and adverse effects, including dropouts. Searches included Embase, MEDLINE, PsycINFO, Web of Science, ClinicalTrials.gov, Cochrane and grey literature. Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines were followed.
Results We identified 2805 unique records, of which six randomised controlled trials were included in this systematic review (n=1460 participants). Five studies were included in the meta-analysis (1442 participants). All had a low risk of bias. There was no difference between cannabinoids and placebo for the difference in the change in average Numeric Rating Scale pain scores (mean difference −0.21 (−0.48 to 0.07, p=0.14)); this remained when only phase III studies were meta-analysed: mean difference −0.02 (−0.21 to 0.16, p=0.80). Cannabinoids had a higher risk of adverse events when compared with placebo, especially somnolence (OR 2.69 (1.54 to 4.71), p<0.001) and dizziness (OR 1.58 (0.99 to 2.51), p=0.05). No treatment-related deaths were reported. Dropouts and mortality rates were high.
Conclusions Studies with a low risk of bias showed that for adults with advanced cancer, the addition of cannabinoids to opioids did not reduce cancer pain.
Trial registration number CRD42018107662.
- other cancer
- systematic review
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Cancer-related pain is common, occurring in up to 60% of patients undergoing anti-cancer therapy and 90% of those with advanced disease.1 There is an increased recent interest in cannabinoids (including cannabis) for pain management along with more permissive legislative changes in many countries.2 3 The medicinal use of cannabis is already legal in 40 countries and 29 US states.4 The WHO guidelines for the pharmacological and radiotherapeutic management of cancer pain in adults and adolescents suggest that data analysis is needed on cannabinoids for cancer pain.5
Patients with cancer use cannabinoids. An anonymous survey (2040 out of 3138 surveys (65%) were returned) in Canada showed that 356 (18%) patients reported cannabis use within the preceding 6 months. Of these, 80% acquired cannabis through friends and 46% of patients used it for cancer-related pain.6 In another anonymous survey of adult patients with cancer in a cancer centre in a US state with legalised cannabis, random urine testing of sampled participants was used.7 The response rate was 34% (926/2737); of these, 21% had used cannabis in the last month, most frequently for pain.7
A systematic review was performed to identify all randomised controlled trials (RCTs) of cannabinoids compared with placebo or other active agents for the treatment of cancer-related pain in adults. A meta-analysis was performed to determine cannabinoid effectiveness and adverse effects, including dropouts. A recent systematic review and meta-analysis that assessed the efficacy, tolerability and safety of medical cannabis and cannabis-based medicines for cancer pain reported very low quality evidence for a non-significant 50% reduction in pain (p=0.82).8 This work supplements the systematic review by Häuser et al.8 The current systematic review has a broader search strategy, and authors were contacted to provide additional findings and information on study design. The primary outcome in this systematic review was the absolute change in mean pain intensity, which is a more sensitive outcome than a dichotomous outcome, for example, proportion of participants who report a pain relief of 50% or greater from baseline to end of study.9 10 The aim was to determine the beneficial and adverse effects of cannabinoids compared with placebo or other active agents for the treatment of cancer-related pain in adults from RCTs.
This systematic review was prepared according to the recommendations in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocol (PRISMA-P) statement11 and was conducted/reported following an a priori protocol according to the PRISMA guidelines.12
Strategies were devised to be inclusive of all potentially relevant studies using both Medical Subject Heading (MeSH) terms and text word searches to increase the search sensitivity. Terms for “cannabis/cannabinoids”, “cancer/neoplasms” and “pain” were combined to identify relevant studies. The search terms for cannabinoids included individual drug names and generic terms “cannabinoids” and “cannabis”. The cancer search included the MeSH term “exp neoplasms/” and text word searches for synonyms for cancer. The “pain” search included terms and synonyms for pain. The Embase search strategy is included as an online supplementary file. Search strategies from all other databases are available on request from the authors.
In August 2018, the following electronic databases were searched: Embase (Ovid); Ovid MEDLINE(R) In-Process & Other Non-Indexed Citations; PsycINFO (Ovid); Conference Proceedings Citation Index–Science (Web Of Science; Thomson Reuters, New York City, NY); ClinicalTrials.gov (US NIH); ISRCTN registry (BMC); Cochrane Database of Systematic Reviews (Wiley); Cochrane Central Register of Controlled Trials (Wiley); Database of Abstracts of Reviews of Effect (Wiley). All searches were repeated on 1 August 2019 to ensure that there were no further publications since the original searches.
Searches were also conducted for grey literature using the following online databases: the Bielefeld Academic Search Engine (BASE) (https://www.base-search.net/), OpenGrey (http://www.opengrey.eu/) and Mednar (https://mednar.com/).
In addition to the electronic search, reference lists from reviews on cannabis/cannabinoids to treat cancer pain were manually searched as were identified publications. Experts in the field were consulted to ensure that no articles were missed. Unpublished studies were also included in the search. When only a conference abstract was available and the full study was unpublished, authors were contacted to try to ascertain further information. No language date or publication type restrictions were applied to the search.
Inclusion, exclusion and selection criteria
Studies were included if they were RCTs which assessed the effect of cannabinoids (THC:CBD, THC extract, nabiximols, Sativex, medical cannabis) compared with placebo or other active agents for the treatment of cancer-related pain in adults, with pain as the primary outcome (table 1).
Cochrane protocols determining studies for inclusion were followed, only including studies where the whole patient population had cancer pain. If this was not the case but results were presented separately for the cancer pain subgroup, the study and extracted data for the target subgroup were included.
Studies were excluded if they did not meet the eligibility criteria (table 1). Studies conducted in patients undergoing surgery, healthy volunteers or animals were excluded from this systematic review as these groups have different cannabinoid usage (duration, administration schedule) compared with patients on cannabinoids for cancer pain. Studies other than RCTs potentially have too much bias to be included. Studies not having pain as the primary outcome were not included as they would not be designed or powered to determine the effect of cannabinoids on pain.
Two authors (EGB and JWB) independently reviewed all titles and abstracts (in duplicate) to assess their relevance for inclusion. Full-text papers were retrieved for those fulfilling the criteria and also for those publications for which the ability to assess their eligibility could not be assessed on the basis of the titles and abstracts alone. EGB and JWB then independently assessed the full texts of all potentially relevant studies.
Disagreement at all stages was resolved by consensus and with recourse to a third review author (MIB). If a study was rejected at the full-text stage, a reason was given. The results of these searches and selections are shown in the PRISMA flow diagram (figure 1).12
Two authors (EGB and JWB) independently extracted data from each included paper regarding study aims/objectives, design, patient population, intervention (cannabinoid used and dose), comparator, clinical outcome measures (eg, pain) and results (association between cannabinoid use and pain and reported adverse events). Disagreement was resolved by consensus and with recourse to a third review author (MIB). When data were not reported in full, authors were contacted for additional information.
The primary outcome of interest was absolute mean change from baseline to the end of treatment in average pain on a Numeric Rating Scale (NRS). Secondary outcomes were adverse effects and study dropouts.
Quality assessment of data
Assessment of risk of study bias was independently assessed by two authors (EGB and JWB) using the Cochrane Collaboration risk of bias tool for RCTs which graded the risk of bias as high, low or unclear in six domains (Selection bias: random sequence generation and allocation concealment; Performance bias: blinding of participant and personnel; Detection bias: blinding of outcome assessment; Attrition bias: incomplete outcome data; Reporting bias: selective reporting).14 Disagreement at all stages was resolved by consensus and with recourse to a third review author (MIB). When this information was not available in the publication, authors were contacted.
For the meta-analysis, the difference in the mean change from the randomisation baseline to the end of treatment in average pain NRS score was calculated and 95% CI was calculated for each study. Data on the numbers of patients experiencing adverse events for each group, the OR and 95% CI were calculated for each study adverse event. The mean difference or ORs were pooled using a fixed-effect model or random-effects model (the Mantel-Haenszel method) and the corresponding 95% CIs were calculated.
Where the analysis indicated significant heterogeneity, a random-effects model was chosen, otherwise a fixed-effects model was applied. Statistical heterogeneity was assessed using Cochran’s Q test. Cochran’s Q tests the presence versus the absence of heterogeneity and the p value is stated. The I2 index describes the percentage of variation across studies that is due to heterogeneity rather than chance. Interpretation is as follows: low, moderate and high to I2 values of 25%, 50% and 75%, respectively.15 The importance of the observed value of I2 depends on (1) magnitude and direction of effects and (2) strength of evidence for heterogeneity (eg, p value from the χ2 test or a CI for I2). A funnel plot was used to test for publication bias.
We identified 2805 unique records of which six RCTs were included in this systematic review. Due to the heterogeneous nature of some of these studies (in study design, duration/dose of cannabinoid administered, timing of outcome measurement), five studies were included in a meta-analysis (representing a total of 1442 participants) and six studies were included in a narrative analysis (representing a total of 1460 participants).
From the six included RCTs (two were reported in a single publication), one was a small cross-over pilot randomised study, two were phase II studies and three were phase III studies (table 2). From the two early randomised double-blind phase II studies in patients with advanced cancer and pain unrelieved by opioids,16 17 one reported that cannabinoids had analgesic effects,16 and the primary outcome of the other was negative.17 Subsequent to these studies, three phase III placebo RCTs with a similar methodology have been reported. Data from two RCTs were reported in a single publication, with the primary efficacy endpoints (percent improvement (study 1) and mean change (study 2) in average daily pain NRS scores).18 Neither these nor the third RCT (primary endpoint: per cent change in the average pain NRS score)19 reported a positive effect of nabiximols compared with placebo on their primary endpoints. These studies had a low risk of bias.
The small cross-over pilot randomised study (n=18) assessed nabiximols versus placebo for use for treatment of chemotherapy-induced neuropathic pain and reported no statistically significant difference between nabiximols and placebo on the NRS for pain intensity: mean pretreatment score=6.75; and at the end of 4 weeks, nabiximols group score=6.00 while placebo group score=6.380.20 However, further analysis in five patients who responded to treatment showed an average decrease of 2.6 on an 11-point NRS for pain intensity.20
Studies used a pump-action oromucosal spray for medication delivery which used 1:1 THC:CBD extract versus placebo. Some studies had additional arms, for example, THC extract.16 Dose titration differed between studies. Patients self-titrated to the optimal dose16 20 or were randomly assigned to different doses.17 In the phase III studies, patients titrated medication according to a pre-specified dose escalation protocol until they achieved pain relief, developed adverse events or reached the maximum dose of 10 sprays/day.18 19
Quality assessment of included studies was performed using the Cochrane Risk of Bias Tool (online supplementary table 1). The studies included were at low risk of bias. Although the studies were funded (or had medication supplied) by industry, and publication bias is more common when most of the published studies are funded by industry, taken in the context of the results, these are overall negative studies making publication bias less likely. The funnel plot (online supplementary figure 1) showed that distribution was roughly symmetrical, indicating that publication bias was not likely to be present.
Change in pain intensity was the primary outcome of interest in this systematic review. Change in pain intensity was the primary outcome in the studies of Johnson et al,16 Fallon et al 18 and Lichtman et al,19 and a secondary outcome in Portenoy et al.17 Lynch et al measured change in the NRS for pain intensity and reported that there was no statistically significant difference between the treatment and the placebo groups, but as this study only included people with chronic neuropathic pain and was a small exploratory study, it was not included in the meta-analysis.20
The meta-analysis is shown in figure 2. There was no difference between cannabinoids and placebo for the difference in the change in average NRS pain scores: mean difference −0.21 (−0.48 to 0.07, p=0.14). Including only phase III studies in the meta-analysis, there was no benefit from cannabinoid use: mean difference −0.02 (−0.21 to 0.16, p=0.80) (figure 3).18 19 The change in pain intensity was a secondary outcome in Portenoy et al; their primary outcome (30% reduction in baseline pain) was not statistically different between cannabinoids and placebo (p=0.59).17 In Portenoy et al, data were not available for the mean pain difference of all three doses combined,17 so only the low dose (1–4 sprays) was used in the meta-analysis as this was the most effective dose.
All studies reported on adverse events (table 3). Dizziness, nausea, vomiting, somnolence and fatigue were the main reported adverse events. In general, cannabinoids were reported to have a higher risk of adverse events compared with placebo. Fallon et al, Lichtman et al and Portenoy et al reported only the adverse events in ≥5% of patients.17–19 In Johnson et al, it is only those reported in three or more patients.16 Lynch et al reported more adverse events compared with placebo, but as this study only included people with chronic neuropathic pain and was a small pilot study, it was not included in the meta-analysis.20 In the meta-analysis, only the low dose (1–4 sprays) was used from Portenoy et al for consistency with the pain score meta-analysis.
The meta-analysis shows a higher odds of somnolence (OR 2.69 (1.54 to 4.71), p<0.001) and dizziness (OR 1.58 (0.99 to 2.51), p=0.05) in the cannabinoid group (figure 4).16–19 There was also a higher odds of nausea (OR 1.41 (0.97 to 2.05), p=0.08) and vomiting in the cannabinoid group (OR 1.34 (0.85 to 2.11, p=0.21)), but these were not statistically significant (figure 4).16–19
Dropouts due to adverse events
In Johnson et al, dropouts due to adverse events were 16.7% in the THC:CBD group and 5% in the placebo group.16 In Portenoy et al, adverse event discontinuations were dose related: 19.8% in all patients on nabiximols and 17.6% in the placebo group.17 In study 1 by Fallon et al, 19% Sativex patients and 14.6% placebo patients discontinued due to adverse events.18 In study 2 by Fallon et al, during the 2-week single-blind Sativex titration period, 17.5% patients discontinued Sativex due to adverse events.18 In the treatment period, 20.4% withdrew from the Sativex group and 12.6% withdrew from the placebo group.18 In Lichtman et al, discontinuation due to adverse events was 20.1% in the Sativex group and 17.7% in the placebo group.19 No treatment-related deaths were reported in any study. Figure 5 shows the dropouts due to adverse events which shows a higher odds of dropouts due to adverse events in the cannabinoid group (OR 1.33 (0.95 to 1.85, p=0.10)), but not statistically significant. In the meta-analysis, only the low dose (1–4 sprays) was used from Portenoy et al for consistency with the pain score meta-analysis.
Studies with a low risk of bias showed that for adults with advanced cancer, the addition of cannabinoids to opioids did not reduce cancer pain compared with placebo. This work complements and builds on the systematic review by Häuser et al.8 Although the same overall conclusions were attained, this systematic review and meta-analysis is based on additional methodological information and thus supported by higher-quality evidence (as included studies were deemed to have lower risk of bias). Furthermore, the primary outcome in this systematic review is a more sensitive outcome to detect minimal changes in pain.9 This systematic review provides good evidence that cannabinoids do not have a role in cancer-related pain.
In all the included RCTs, pain was the primary reason for administering cannabinoids and change in pain score or pain intensity was the primary outcome. Five RCTs were included in the meta-analysis (n=1442) where cannabinoids were given as an adjuvant treatment in addition to their existing stable dose of opioids. In the meta-analysis, the two phase II studies and three phase III studies included patients with chronic cancer pain (average pain duration of all studies of 1.2–2.0 years), with an average pain ≥4 and ≤8 on 0–10 NRS pain score, who were on regular opioids, randomised to the same THC:CBD medication and had a placebo comparator.
Five trials from four publications in the 1970s (including a total of 128 participants) were excluded as these were single-dose studies, assessing short-term effects of cannabinoids at 6–7 hours.21–24 Four of these studies evaluated delta-9-tetrahydrocannabinol (THC) or nitrogen-containing benzopyran derivative, modification of delta-1-trans-tetrahydrocannabinol (NIB).21 22 24 The fifth study used the cannabinoid benzopyranoperidine.23 Of these five single-dose studies assessing efficacy at 6–7 hours, three used THC or NIB and reported no different in efficacy compared with codeine.21 22 24 The fifth study used the cannabinoid benzopyranoperidine and reported that about 30% of patients had increased pain intensity with this drug.23
Cannabinoids are associated with short-term adverse effects including drowsiness, dizziness, confusion, hallucinations, euphoria, nausea and vomiting, and diarrhoea.25 A systematic review evaluating the adverse effects of medical cannabinoids reported patients using medical cannabinoids had 1.86 times higher risk of non-serious adverse effects compared with controls while there was no significant difference between serious adverse effects.26
Our analysis echoed this, showing that in general cannabinoids were reported to have a higher risk of adverse events compared with placebo with somnolence and dizziness reaching statistical significance.
Strengths and limitations
This is a rigorously conducted systematic review that included ‘grey’ literature and authors were contacted when data and methodological information was not included in the publication. This enabled the included studies to be considered at low risk of bias. The studies included were RCTs that assessed clinically relevant cannabinoids as an adjuvant to opioid medications in patients with advanced cancer that had mixed aetiologies of pain due to their cancer. Change in pain score was used as the primary outcome to assess if cannabinoids had an effect on pain as this is more sensitive to changes compared with 30% or 50% decrease in pain.
Despite the detailed search strategy, it is possible that not all relevant studies were included. There were inconsistencies between studies in the patients included, the interventions, comparators and outcomes. In the meta-analysis, a secondary outcome was used for Portenoy et al (as this was the primary outcome for this systematic review).16 17
The included studies had several potential limitations. Self-reported NRS pain score might not be the best measure for such trials, as this simple instrument does not capture the complexity of pain especially when it has been long-standing problem. The fidelity of the use of the oromucosal spray, which affects absorption and pharmacokinetic factors, was not assessed and this might also affect the effectiveness of the medication used and the outcome measured. Some of the included studies had kept the maintenance doses of opioid and other medications the same throughout the trial; ways to decrease doses when appropriate should be considered as this might also have an impact on adverse effects. The negative results from some of the RCTs could be due to a relatively high number of patient withdrawals and high mortality rate.16–19 Publication bias is more common when most of the published studies are funded by industry. However, the primary outcome for most of these studies was negative, making publication bias less likely for these studies. Aside from lack of therapeutic efficacy, the negative results from some of the RCTs could also be due to a relatively high number of patient withdrawals from studies, and also high mortality rate and increased number of lost patients.16–19
For a medication to be useful, there needs to be a net overall benefit, with the positive effects (analgesia) outweighing adverse effects. None of the included phase III studies show benefit of cannabinoids. One of the phase II studies showed benefit in their primary outcome16; the other was negative in its primary outcome, although a secondary outcome was positive.17 When statistically pooled, there was no decrease in pain score from cannabinoids. There are, however, significant adverse effects and dropouts reported from cannabinoids. Based on evidence with a low risk of bias, cannabinoids cannot be recommended for the treatment of cancer-related pain.
We would like to thank Helen Elwell, Clinical librarian, for her guidance and assistance on the use and refinement of search terms and searches.
Contributors EGB, MIB and JWB contributed to study design. EGB and JWB performed the searches and contributed to data collection and data analysis. JWB and EGB drafted the article. VA undertook the statistical analysis. MIB contributed to writing of the article. All authors were responsible for approval of the final report.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement All data relevant to the study are included in the article or uploaded as online supplementary information.