Cannabis for Chronic Pain: How Large the Benefit Actually Is

The systematic review that underpins current guidance, published in The BMJ in 2021, found that non-inhaled medical cannabis probably increases the proportion of patients achieving a minimally important improvement in pain by about 10 percentage points compared with placebo, at moderate certainty.

That translates to roughly one additional patient in ten experiencing a meaningful reduction in pain. It is a real effect. It is also a modest one, and it sits alongside a dizziness risk that rose from 9 percent in trials under three months to 28 percent in trials of three months or longer.

Patients arrive with expectations set by testimonials at one extreme and by dismissal at the other. Neither prepares them for what the trials show, which is a modest average benefit with meaningful variation between individuals and a real side effect burden that grows with duration.

Setting the expectation correctly at the start changes what happens next. A patient told to expect a small improvement who gets one will continue sensibly. A patient told to expect transformation who gets a small improvement will escalate the dose, which is where most of the avoidable harm in this field occurs.

The most rigorous synthesis available is a systematic review and meta-analysis published in The BMJ in 2021, which included 32 randomized trials and 5,174 adults with chronic pain. Twenty-nine of those trials compared medical cannabis or cannabinoids with placebo. Medical cannabis was given orally in 30 trials and topically in two. Twenty-eight trials studied chronic non-cancer pain and four studied cancer-related pain, with follow-up ranging from one to 5.5 months.

For pain relief, the review found a weighted mean difference of 0.50 cm on a 10 cm visual analogue scale, with a confidence interval from 0.25 to 0.75 cm, at moderate certainty. Because a 1 cm change is the accepted minimally important difference, the authors modelled this as a 10 percentage point increase in the proportion of patients achieving a meaningful improvement, with a range from 5 to 15 points.

For physical functioning, the improvement was very small but rated high certainty: 1.67 points on the 100-point SF-36 physical functioning scale, modelled as a 4 percentage point increase in patients reaching the 10-point threshold. For sleep quality, a small improvement at high certainty, modelled as a 6 percentage point increase.

Emotional, role, and social functioning did not improve, at high certainty. That is an underappreciated finding. Cannabis moved pain, function, and sleep in these trials. It did not move how people felt about their lives.

An international guideline panel including patients, clinicians, and methodologists reviewed that evidence and published a recommendation in The BMJ alongside it. The recommendation was to offer a trial of non-inhaled medical cannabis or cannabinoids, in addition to standard care and management, for people living with chronic cancer or non-cancer pain when standard care is not sufficient.

The recommendation was graded weak, and the panel explained why in terms worth repeating: the balance between benefits and harms is close. A weak recommendation in the GRADE framework means that most informed patients would choose the intervention but many would not, and that the decision therefore requires shared decision making rather than a default.

This is the clearest statement available of where cannabis sits in chronic pain care. It is a reasonable option to add after other measures have been tried, for patients who place high value on small improvements in pain, function, and sleep and who accept a modest risk of mostly transient side effects. It is not a first-line therapy, and the guideline does not present it as one.

A Cochrane review published in 2018 examined cannabis-based medicines specifically for chronic neuropathic pain in adults, including 16 studies with 1,750 participants over durations of two to 26 weeks. Most used an oromucosal THC and CBD spray; two used nabilone, two used inhaled herbal cannabis, and two used dronabinol.

For substantial pain relief of 50 percent or more, 21 percent of participants responded on cannabis-based medicines against 17 percent on placebo, a risk difference of 5 percentage points with a number needed to treat of 20, at low quality evidence. For moderate relief of 30 percent or more, 39 percent responded against 33 percent, a risk difference of 9 percentage points with a number needed to treat of 11, at moderate quality.

The harms in the same review are the part patients rarely hear. Withdrawal due to adverse events occurred in 10 percent on cannabis-based medicines against 5 percent on placebo, giving a number needed to harm of 25. Nervous system adverse events occurred in 61 percent against 29 percent, a number needed to harm of 3. Psychiatric adverse events occurred in 17 percent against 5 percent, a number needed to harm of 10.

The reviewers concluded that the potential benefits might be outweighed by the potential harms, and noted that the trials excluded people with a history of substance misuse and significant comorbidity, which describes a substantial share of real chronic pain populations.

A 2022 systematic review in Annals of Internal Medicine, conducted for the Agency for Healthcare Research and Quality, categorized products by THC to CBD ratio and by source rather than pooling them together. It included 18 randomized placebo-controlled trials with 1,740 participants and seven cohort studies with 13,095.

Synthetic products with very high THC content, above 98 percent THC, may be associated with moderate improvement in pain severity and in response rates of 30 percent or greater, with increased sedation and a probable large increase in dizziness. Sublingual spray with a comparable THC to CBD ratio of about 1.1 to 1 is probably associated with small improvement in pain severity and overall function, with a possible large increase in dizziness and sedation and a moderate increase in nausea.

Extracted products with high THC to CBD ratios, ranging from 3 to 1 up to 47 to 1, may be associated with a large increased risk of withdrawal from the study due to adverse events and of dizziness, without a corresponding efficacy signal in this analysis.

For every other product category, including the whole-flower products that dominate dispensary sales, the review found the evidence insufficient or not reported. That gap is the central honesty problem in this field: the products with trial evidence are largely not the products patients buy.

Almost all the randomized evidence is short. Follow-up in the BMJ review ranged from one to 5.5 months; the Cochrane neuropathic review ran two to 26 weeks. Chronic pain is measured in years.

A companion systematic review of non-randomized studies, covering 39 studies and 12,143 adults with chronic pain, attempted to fill that gap. Adverse events were common, with a pooled prevalence of 26.0 percent, and psychiatric adverse events occurred in 13.5 percent, both at very low certainty. Serious adverse events, discontinuation, cognitive adverse events, accidents and injuries, and dependence or withdrawal syndrome each typically affected fewer than one in twenty patients.

Two findings from that review deserve emphasis. Studies with 24 weeks or more of cannabis use reported more adverse events than shorter studies, with a statistically significant interaction. And the reviewers found insufficient evidence comparing the harms of medical cannabis with the harms of other pain management options, including opioids.

That last point should end a common talking point. The claim that cannabis carries a safer risk profile than opioids for chronic pain is frequently asserted and has not been established by direct comparison. It may well be true on several dimensions. It has not been demonstrated by the evidence that exists.

One further signal comes from the BMJ review itself: dizziness risk was 9 percent above placebo in trials shorter than three months and 28 percent above placebo in trials of three months or longer, at high certainty, with the interaction test reaching statistical significance. Whatever else changes with duration, dizziness does.

For the central question of whether non-inhaled cannabis beats placebo for chronic pain, the evidence is moderate to high certainty and the answer is yes by a small margin. That conclusion rests on a large body of randomized trials assessed with GRADE by independent methodologists and converted into a clinical guideline through a formal process. It is among the better-characterized questions in cannabis medicine.

For nearly every downstream question, the evidence is weak. Which product, which ratio, which route, which patient, and what happens after six months are all poorly addressed. The certainty ratings drop sharply as soon as the question becomes specific enough to guide an individual prescription.

Blinding is a persistent problem in cannabis trials. THC produces recognizable subjective effects, so participants and sometimes assessors can infer allocation, which inflates apparent benefit on subjective outcomes. Pain is entirely subjective. The reviews acknowledge this and it is the single largest threat to the effect estimates above.

The trials excluded the patients clinicians most often see. The Cochrane reviewers noted explicitly that participants with a history of substance misuse and other significant comorbidities were excluded, which limits how far the tolerability findings transfer to real chronic pain populations.

The evidence is overwhelmingly about oral and oromucosal preparations. Thirty of 32 trials in the BMJ review used oral products. Inhaled cannabis, which is what a large share of patients actually use, is barely represented, and the Cochrane subgroup analysis of herbal cannabis was rated very low quality.

These reviews do not show that cannabis works better than any specific alternative. There is no adequately powered head-to-head trial against gabapentinoids, duloxetine, topical agents, or structured exercise and psychological therapy for chronic pain.

They do not show that cannabis is safer than opioids for chronic pain. The systematic review of long-term harms found insufficient evidence to make that comparison, which means the claim is unsupported rather than refuted.

They do not identify who responds. No trial has validated a predictor of response, whether by pain phenotype, genotype, prior cannabis experience, or baseline severity. Selection of candidates remains clinical judgment.

They also do not describe what happens beyond six months, which is the timeframe most patients with chronic pain are actually planning for.

Chronic pain care has few good options and many partial ones. Read against that background, a small improvement in pain with moderate certainty is not a dismissible result. It is comparable in magnitude to several accepted pharmacologic options for chronic non-cancer pain, most of which also produce modest average benefits with meaningful side effect burdens.

What makes cannabis different is the asymmetry between the evidence and the marketplace. Few other treatments have this large a gap between the products that have been tested and the products that are sold, or between the size of the demonstrated effect and the confidence of the claims made for it. Closing that gap is mostly a research and regulatory problem rather than a clinical one.

The number I give patients is roughly one in ten. Ten percentage points more people reach a meaningful improvement in pain on cannabis than on placebo. Some patients hear that and decide it is not worth the effort, which is a reasonable conclusion. Others have been living with pain long enough that a one in ten chance of meaningful improvement is the best offer on the table, which is also reasonable.

What I will not do is tell someone that cannabis is going to fix their pain. I have watched patients escalate for months chasing a result the evidence never promised, and the escalation is where the dizziness, the sedation, the falls in older patients, and the dependence show up.

The dizziness finding is the one I act on most. Nine percent in short trials and 28 percent in trials past three months is not a minor detail for a 74-year-old with osteoarthritis and a history of falls. Duration changes the risk calculation, and the plan should be revisited rather than renewed.

I would also retire the line that cannabis is safer than opioids for chronic pain. I suspect it is true in several respects, and I have no trial that shows it. Saying so plainly costs a talking point and buys credibility, which is the better trade.

Cannabis produces a small but real improvement in chronic pain, a very small improvement in physical function, and a small improvement in sleep, with roughly one additional patient in ten reaching a meaningful pain response compared with placebo. Guideline panels rate this a weak recommendation to offer as an addition to standard care when standard care is insufficient. Set the expectation at the start, choose a product and route deliberately, monitor for dizziness and sedation, and reassess rather than renewing by default, particularly past three months.

Carry forward the effect size and the certainty rating together. A 10 percentage point increase in meaningful pain response at moderate certainty is worth offering and not worth overselling. Carry forward the duration signal as well, because the side effect profile that looks acceptable at eight weeks looks different at six months. And treat any claim about cannabis versus opioids as unsupported until someone runs the comparison.