Benefits and harms of medical cannabis: a scoping review of systematic reviews

Abstract

Background: There has been increased interest in the role of cannabis for treating medical conditions. The availability of different cannabis-based products can make the side effects of exposure unpredictable. We sought to conduct a scoping review of systematic reviews assessing benefits and harms of cannabis-based medicines for any condition.

Methods: A protocol was followed throughout the conduct of this scoping review. A protocol-guided scoping review conduct. Searches of bibliographic databases (e.g., MEDLINE®, Embase, PsycINFO, the Cochrane Library) and gray literature were performed. Two people selected and charted data from systematic reviews. Categorizations emerged during data synthesis. The reporting of results from systematic reviews was performed at a high level appropriate for a scoping review.

Results: After screening 1975 citations, 72 systematic reviews were included. The reviews covered many conditions, the most common being pain management. Several reviews focused on management of pain as a symptom of conditions such as multiple sclerosis (MS), injury, and cancer. After pain, the most common symptoms treated were spasticity in MS, movement disturbances, nausea/vomiting, and mental health symptoms. An assessment of review findings lends to the understanding that, although in a small number of reviews results showed a benefit for reducing pain, the analysis approach and reporting in other reviews was sub-optimal, making it difficult to know how consistent findings are when considering pain in general. Adverse effects were reported in most reviews comparing cannabis with placebo (49/59, 83%) and in 20/24 (83%) of the reviews comparing cannabis to active drugs. Minor adverse effects (e.g., drowsiness, dizziness) were common and reported in over half of the reviews. Serious harms were not as common, but were reported in 21/59 (36%) reviews that reported on adverse effects. Overall, safety data was generally reported study-by-study, with few reviews synthesizing data. Only one review was rated as high quality, while the remaining were rated of moderate (n = 36) or low/critically low (n = 35) quality.

Conclusions: Results from the included reviews were mixed, with most reporting an inability to draw conclusions due to inconsistent findings and a lack of rigorous evidence. Mild harms were frequently reported, and it is possible the harms of cannabis-based medicines may outweigh benefits.

Systematic review registration: The protocol for this scoping review was posted in the Open Access (https://ruor.uottawa.ca/handle/10393/37247).

Keywords: Cannabis; Marijuana; Medical marijuana; Scoping review; Systematic review.

Fig. 1

PRISMA-style flow diagram of the review selection process

Fig. 2

Review coverage of the various cannabis-based interventions

Fig. 3

Conditions or symptoms across reviews that were treated with cannabis. IBD inflammatory bowel disease, MS multiple sclerosis, RD rheumatic disease

Fig. 4

Indications for cannabis use across included reviews

Fig. 5

Outcomes

Fig. 6

Cannabis vs. placebo. Authors’ presentations of the findings were mapped using the categorization shown in Table 2. According to the reviews’ intended scope for the condition being treated, outcomes were mapped into “pain,” “non-pain outcomes,” and “adverse events.” For each condition and outcome pair (i.e., each row in the grid), the number of reviews reporting findings is shown according to the results categorization. For pain, reviews numbered in different categories signal discordant findings across those reviews. For non-pain outcomes, reviews presenting findings in the different categories would signal different results for different outcomes, as well as discordant findings within and across reviews. Adverse events are grouped as a whole and “favors intervention” would be interpreted as a decrease in events with cannabis when compared with the control group. Favors int = favors intervention; Favors Ctrl = favors control; Not stat sig = not statistically significant

Fig. 7

Cannabis vs. placebo, high and moderate quality reviews. Authors’ presentations of the findings were mapped using the categorizations shown in Table 2. According to the reviews’ intended scope for the condition being treated, outcomes were mapped into “pain,” “non-pain outcomes,” and “adverse events.” For each condition and outcome pair (i.e., each row in the grid), the number of reviews reporting findings is shown according to the results categorization. For pain, reviews numbered in different categories signal discordant findings across those reviews. For non-pain outcomes, reviews presenting findings in the different categories would signal different results for different outcomes, as well as discordant findings within and across reviews. Adverse events are grouped as a whole and “favors intervention” would be interpreted as a decrease in events with cannabis when compared with the control group. Favors int = favors intervention; Favors Ctrl = favors control; Not stat sig = not statistically significant

Fig. 8

Cannabis vs. active drugs. Authors’ presentations of the findings were mapped using the categorizations shown in Table 2. According to the reviews’ intended scope for the condition being treated, outcomes were mapped into “pain,” “non-pain outcomes,” and “adverse events.” For each condition and outcome pair (i.e., each row in the grid), the number of reviews reporting findings is shown according to the results categorization. For pain, reviews numbered in different categories signal discordant findings across those reviews. For non-pain outcomes, reviews presenting findings in the different categories would signal different results for different outcomes, as well as discordant findings within and across reviews. Adverse events are grouped as a whole and “favors intervention” would be interpreted as a decrease in events with cannabis when compared with the control group. Favors int = favors intervention; Favors Ctrl = favors control; Not stat sig = not statistically significant

Fig. 9

Cannabis vs. active drugs, high and moderate quality reviews. Authors’ presentations of the findings were mapped using the categorizations shown in Table 2. According to the reviews’ intended scope for the condition being treated, outcomes were mapped into “pain,” “non-pain outcomes,” and “adverse events.” For each condition and outcome pair (i.e., each row in the grid), the number of reviews reporting findings is shown according to the results categorization. For pain, reviews numbered in different categories signal discordant findings across those reviews. For non-pain outcomes, reviews presenting findings in the different categories would signal different results for different outcomes, as well as discordant findings within and across reviews. Adverse events are grouped as a whole and “favors intervention” would be interpreted as a decrease in events with cannabis when compared with the control group. Favors int = favors intervention; Favors Ctrl = favors control; Not stat sig = not statistically significant

Fig. 10

Cannabis vs. placebo + active drug. Authors’ presentations of the findings were mapped using the categorizations shown in Table 2. According to the reviews’ intended scope for the condition being treated, outcomes were mapped into “pain,” “non-pain outcomes,” and “adverse events.” For each condition and outcome pair (i.e., each row in the grid), the number of reviews reporting findings is shown according to the results categorization. For pain, reviews numbered in different categories signal discordant findings across those reviews. For non-pain outcomes, reviews presenting findings in the different categories would signal different results for different outcomes, as well as discordant findings within and across reviews. Adverse events are grouped as a whole and “favors intervention” would be interpreted as a decrease in events with cannabis when compared with the control group. Favors int = favors intervention; Favors Ctrl = favors control; Not stat sig = not statistically significant

Fig. 11

Cannabis vs. placebo + active drug, high and moderate quality reviews. Authors’ presentations of the findings were mapped using the categorizations shown in Table 2. According to the reviews’ intended scope for the condition being treated, outcomes were mapped into “pain,” “non-pain outcomes,” and “adverse events.” For each condition and outcome pair (i.e., each row in the grid), the number of reviews reporting findings is shown according to the results categorization. For pain, reviews numbered in different categories signal discordant findings across those reviews. For non-pain outcomes, reviews presenting findings in the different categories would signal different results for different outcomes, as well as discordant findings within and across reviews. Adverse events are grouped as a whole and “favors intervention” would be interpreted as a decrease in events with cannabis when compared with the control group. Favors int = favors intervention; Favors Ctrl = favors control; Not stat sig = not statistically significant

Fig. 12

One cannabis formulation vs. other. Authors’ presentations of the findings were mapped using the categorizations shown in Table 2. According to the reviews’ intended scope for the condition being treated, outcomes were mapped into “pain,” “non-pain outcomes,” and “adverse events.” For each condition and outcome pair (i.e., each row in the grid), the number of reviews reporting findings is shown according to the results categorization. For pain, reviews numbered in different categories signal discordant findings across those reviews. For non-pain outcomes, reviews presenting findings in the different categories would signal different results for different outcomes, as well as discordant findings within and across reviews. Adverse events are grouped as a whole and “favors intervention” would be interpreted as a decrease in events with cannabis when compared with the control group. Favors int = favors intervention; Favors Ctrl = favors control; Not stat sig = not statistically significant

Fig. 13

One cannabis formulation vs. other, high and moderate quality reviews. Authors’ presentations of the findings were mapped using the categorizations shown in Table 2. According to the reviews’ intended scope for the condition being treated, outcomes were mapped into “pain,” “non-pain outcomes,” and “adverse events.” For each condition and outcome pair (i.e., each row in the grid), the number of reviews reporting findings is shown according to the results categorization. For pain, reviews numbered in different categories signal discordant findings across those reviews. For non-pain outcomes, reviews presenting findings in the different categories would signal different results for different outcomes, as well as discordant findings within and across reviews. Adverse events are grouped as a whole and “favors intervention” would be interpreted as a decrease in events with cannabis when compared with the control group. Favors int = favors intervention; Favors Ctrl = favors control; Not stat sig = not statistically significant

Fig. 14

Cannabis vs. all comparators combined. Authors’ presentations of the findings were mapped using the categorizations shown in Table 2. According to the reviews’ intended scope for the condition being treated, outcomes were mapped into “pain,” “non-pain outcomes,” and “adverse events.” For each condition and outcome pair (i.e., each row in the grid), the number of reviews reporting findings is shown according to the results categorization. For pain, reviews numbered in different categories signal discordant findings across those reviews. For non-pain outcomes, reviews presenting findings in the different categories would signal different results for different outcomes, as well as discordant findings within and across reviews. Adverse events are grouped as a whole and “favors intervention” would be interpreted as a decrease in events with cannabis when compared with the control group. Favors int = favors intervention; Favors Ctrl = favors control; Not stat sig = not statistically significant