Patient-friendly summary
If you read nothing else
Bottom line
Cannabinoids produced only small reductions in chronic pain and small sleep improvements that fell below clinically meaningful thresholds, had no effect on acute or cancer pain, and increased non-serious side effects; the authors concluded the harms seem to outweigh the benefits for pain.
Moderate evidencePublished
Evidence hierarchy
Study participants
People with any type of pain: chronic/neuropathic pain (44 trials, 4306 participants), cancer pain (9 trials, 1681), acute pain (10 trials, 965), and fibromyalgia-related pain (2 trials, 65).
Study Summary
This systematic review with meta-analysis and Trial Sequential Analysis pooled 65 randomised placebo-controlled trials (7017 participants) evaluating cannabinoids for any type of pain. Cannabinoids reduced chronic pain (including neuropathic pain) and improved quality of sleep, but both effect sizes fell below the authors' predefined minimal important differences, so the clinical importance is questionable. Cannabinoids showed no benefit for acute pain or cancer pain, and significantly increased the risk of non-serious adverse events (number needed to harm of seven). Fifty-nine of the 65 trials were at high risk of bias, certainty of evidence was low to very low, and the authors concluded that the harms of cannabinoids for pain seem to outweigh the potential benefits.
Key Findings
| Finding | Detail | Impact |
|---|---|---|
| Chronic pain reduced, but below the clinically meaningful threshold | Across 16 trials (2030 participants), cannabinoids reduced chronic pain versus placebo (mean difference NRS -0.43; 98% CI -0.72 to -0.15; P = 0.0004), but the effect was below the predefined minimal important difference of 1 point on NRS. Low certainty of evidence. | High |
| No effect on acute pain | Four trials (530 participants) showed no evidence of a difference between cannabinoids and placebo (mean difference NRS 0.52; 98% CI -0.40 to 1.43; P = 0.19), with high, unresolvable heterogeneity (I2 = 90%). Very low certainty of evidence. | Medium |
| No effect on cancer pain | Six trials (1550 participants) showed no evidence of a difference (mean difference NRS -0.13; 98% CI -0.33 to 0.06; P = 0.1). Trial Sequential Analysis had enough information to reject a clinically relevant reduction. Low certainty of evidence. | Medium |
| Increased risk of non-serious adverse events | Across 29 trials (5536 participants), cannabinoids increased non-serious adverse events (RR 1.20; 95% CI 1.15 to 1.25; P < 0.001), corresponding to a number needed to harm of seven. Dizziness, fatigue, vertigo, nervous system disorders, and gastrointestinal disorders were specifically increased. Very low certainty of evidence. | High |
| Quality of sleep improved, but below the clinically meaningful threshold | Seventeen trials (3291 participants) showed cannabinoids improved quality of sleep (mean difference NRS -0.42; 95% CI -0.65 to -0.20; P = 0.0003), but the effect size was below the predefined minimal important difference. Low certainty of evidence. | Medium |
| No difference in all-cause mortality, serious adverse events, or quality of life | All-cause mortality (7 trials, 2073 participants: RR 1.20; 98% CI 0.85 to 1.67; P = 0.22), serious adverse events (18 trials, 3980 participants: RR 1.18; 98% CI 0.95 to 1.45; P = 0.07), and quality of life (4 trials, 548 participants: MD -1.38; 98% CI -11.8 to 9.04) showed no statistically significant difference; Trial Sequential Analysis indicated information was insufficient to confirm or reject these effects. | Medium |
| No trials reported on dependence or psychosis | None of the 65 included trials reported analyzable data on cannabinoid dependence or psychosis, leaving these long-term safety outcomes unaddressed. | Medium |
Across 16 trials (2030 participants), cannabinoids reduced chronic pain versus placebo (mean difference NRS -0.43; 98% CI -0.72 to -0.15; P = 0.0004), but the effect was below the predefined minimal important difference of 1 point on NRS. Low certainty of evidence.
Four trials (530 participants) showed no evidence of a difference between cannabinoids and placebo (mean difference NRS 0.52; 98% CI -0.40 to 1.43; P = 0.19), with high, unresolvable heterogeneity (I2 = 90%). Very low certainty of evidence.
Six trials (1550 participants) showed no evidence of a difference (mean difference NRS -0.13; 98% CI -0.33 to 0.06; P = 0.1). Trial Sequential Analysis had enough information to reject a clinically relevant reduction. Low certainty of evidence.
Across 29 trials (5536 participants), cannabinoids increased non-serious adverse events (RR 1.20; 95% CI 1.15 to 1.25; P < 0.001), corresponding to a number needed to harm of seven. Dizziness, fatigue, vertigo, nervous system disorders, and gastrointestinal disorders were specifically increased. Very low certainty of evidence.
Seventeen trials (3291 participants) showed cannabinoids improved quality of sleep (mean difference NRS -0.42; 95% CI -0.65 to -0.20; P = 0.0003), but the effect size was below the predefined minimal important difference. Low certainty of evidence.
All-cause mortality (7 trials, 2073 participants: RR 1.20; 98% CI 0.85 to 1.67; P = 0.22), serious adverse events (18 trials, 3980 participants: RR 1.18; 98% CI 0.95 to 1.45; P = 0.07), and quality of life (4 trials, 548 participants: MD -1.38; 98% CI -11.8 to 9.04) showed no statistically significant difference; Trial Sequential Analysis indicated information was insufficient to confirm or reject these effects.
None of the 65 included trials reported analyzable data on cannabinoid dependence or psychosis, leaving these long-term safety outcomes unaddressed.
Strengths
- Predefined, published protocol following PRISMA guidelines
- Used Trial Sequential Analysis and adjusted significance thresholds to control for random errors from sparse data and repeated testing
- Included more trials (65) than any previous review on this topic, increasing power and precision
- Formal risk-of-bias assessment of all trials and GRADE certainty rating for every outcome
- Predefined minimal important differences for all outcomes (with a deliberately lenient 1-point NRS threshold) to judge clinical relevance, not just statistical significance
- Engaged Danish patient associations at the protocol stage to select patient-relevant outcomes
Limitations
- 59 of 65 trials were at high risk of bias, so benefits may be overestimated and harms underestimated
- Only 35 of 65 trials provided data usable in the meta-analyses (many cross-over trials lacked first-phase data)
- Several outcomes had high, unresolvable statistical heterogeneity (e.g., acute pain I2 = 90%, quality of life I2 = 86%)
- Cannabinoids were compared only with placebo, not with other analgesics such as opioids
- VAS and NRS scores were combined by conversion, which may lose some information
- Subgroup analyses relied on aggregate trial-level data, risking study-level confounding
- Certainty of evidence was low to very low across outcomes; no trials reported on dependence or psychosis
- The funnel plot for non-serious adverse events showed signs of small-study effects
Key Takeaways for Patients
What This Means for You
- 01For long-lasting (chronic) pain, including nerve pain, cannabinoids may produce a small reduction, but the researchers judged the size of that benefit too small to be clearly meaningful for most people.
- 02Cannabinoids did not appear to help with short-term (acute) pain or cancer pain in this analysis.
- 03Cannabinoids may slightly improve sleep quality, but again the effect was smaller than what the researchers considered clinically important.
- 04Cannabinoids increased the chance of bothersome (non-serious) side effects such as dizziness, fatigue, vertigo, and stomach problems; about 1 in 7 people experienced an added side effect.
- 05The included trials were mostly low quality and did not report on the risk of dependence or psychosis, so important long-term safety questions remain unanswered.
- 06The authors concluded that for pain, the potential harms of cannabinoids seem to outweigh the potential benefits.