This review addresses the following Key Questions (KQs):

KQ 1: In adults or adolescents with chronic or subacute pain, what are the benefits of cannabinoids for treatment of chronic or subacute pain?

KQ 2: In adults or adolescents with chronic or subacute pain, what are the harms of cannabinoids for treatment of chronic or subacute pain?

KQ 3: In adults or adolescents with chronic or subacute pain, what are the benefits of kratom or other plant-based substances for treatment of chronic or subacute pain?

KQ 4: In adults or adolescents with chronic or subacute pain, what are the harms of kratom or other plant-based substances for treatment of chronic or subacute pain?

Organizing principle of cannabis-related studies based on ratios of THC to CBD

Definitions of effect sizes

Implications for Clinical Practice, Education, Research, or Health Policy

The implications of the present findings for clinical practice and policy are mixed. Our results suggest that select individuals with chronic neuropathic pain may experience small short-term improvements in pain when using high-THC to CBD cannabis products (synthetic or plant-derived, synthetic equivalent THC). The impact of this intervention on moderate or long-term outcomes is unknown. Extracted cannabis products with a comparable THC to CBD ratio may also result in small improvements in pain severity. On the rother hand, cannabis products with a low-THC to CBD ratio (oral synthetic or purified CBD alone or combined purified THC and synthetic CBD in a ratio of ~1:2) may not result in improvements in pain or function. Those who take products containing comparable or high ratios of THC or combined THC plus CBD in a ~1:2 ratio are also at increased risk for adverse events, including dizziness, sedation and nausea. The expected benefit of cannabis products compared to placebo appears comparable to those observed with prescription opioids, several nonopioid medications, and nonpharmacological interventions. However, comparing effects of different interventions based on cross-trial comparisons must be done with extreme caution, due to potential differences across studies in the populations studied, interventions evaluated, and outcomes assessed; head-to-head trials are needed to understand the comparative effectiveness of cannabis products versus other interventions. The evidence on adverse events with cannabis-related products is much less robust than the evidence on similar outcomes with opioids or nonopioid medications. The risk of sedation and dizziness appears similar with cannabis-related products, opioids, and the anticonvulsants pregabalin and gabapentin, while the risk for nausea may be larger with opioids and the antidepressant duloxetine than with cannabis-related products. However, these comparisons are qualitative and indirect and based on limited evidence on cannabis products relative to the other drugs and require confirmation. Evidence is too limited to compare effects on serious and long-term harms, even indirectly, though respiratory depressant effects of opioids potentially resulting in death are well-known. Understanding how cannabis products' adverse event profiles compare with other available treatments for chronic pain, particularly opioid and non-opioid medications, is essential to determine the benefit to harm ratio. At this time, the strength of this evidence is mostly low, and more data are needed to confidently recommend this as a treatment for various chronic pain-related conditions or for patients with diverse demographic or clinical characteristics. No studies are available to inform decisions regarding use of cannabis for subacute pain or in adolescents.

As noted in the limitations above, baseline use of opioids for pain and the impact of cannabinoids on the use of opioids for pain were very poorly reported. In an effort to more effectively and safely manage chronic pain, a prominent goal of current research is to identify alternative treatments with equal or better benefits for pain while avoiding potential unintended consequences that could result in harms. Unfortunately, much of the findings to date are low SOE or insufficient evidence, and more high-quality studies are needed. Furthermore, the unavailability or unclear availability of studied cannabis products in specific settings may reduce the generalizability of findings.

Our synthesis of the evidence also suggests several important additional questions that might be suitable to be addressed in a clinical practice guideline, based on an assessment of potential benefits and harms, as well as uncertainties in the evidence. Examples of questions that could be addressed through a guideline process include: At what point in the treatment decision tree should cannabis-based medicines be considered? How should patient preferences be taken into account? What are pragmatic dosing guidelines? To further inform guidelines, additional studies on the comparative effects on costs of care would be useful.

Implications for Future Research

The gaps in the research evidence that are outlined above lead to specific recommendations for conducting future studies that will improve the strength of the conclusions that can be drawn, and provide better guidance for policymakers, clinicians, and patients alike.

This review does not provide cost information.

Strengths and Limitations

The evidence base on cannabis and other plant-based treatments for chronic and subacute pain has multiple important limitations. Fifty-four percent of trials enrolled patients with chronic pain due to a neuropathic cause (7 in patients with multiple sclerosis, 7 with a mix of conditions or not specified, 2 with diabetic neuropathy, and 1 each with chemotherapy, HIV, back pain, or spinal cord injury). There is little or no evidence on other types of chronic pain, including low back pain, osteoarthritis, fibromyalgia, and inflammatory arthritis. Although inclusion criteria were revised to also address subacute pain, no eligible studies were identified. In terms of age, there is limited evidence on younger and older populations, with most patients being middle-aged (mean age 54 years). Inclusion criteria were expanded in 2023 to include adolescents; however, no eligible studies have identified. Studies generally excluded patients with a history of psychiatric disorders other than prior history of depression or anxiety. Importantly, there was either no evidence or inadequate evidence to evaluate important patient populations based on sex/gender, race/ethnicity, age, or pregnancy/lactating status

Another limitation is the lack of consistent nomenclature and details regarding the interventions and products studied. For example, products are described as plant-derived in some studies, but without a consistent way of describing the process or the resulting purity of the products. Other studies used words such as "standardized" to describe the amount of THC in a whole-plant cannabis product, again with lack of description of how this was defined or determined. Studies that evaluated products that contained THC and CBD did not consistently report the ratio of THC to CBD, other than for Sativex oromucosal spray (close to a 1 to 1 ratio). One trial that evaluated a topical product did not describe details regarding extraction methods or product purity, and it was unclear if the product was intended to provide local or systemic effects. When necessary, we attempted to obtain additional details regarding products from study authors, not all authors responded to queries. Although we attempted to categorize products accurately, some misclassification is possible. Other limitations include the complete lack of evidence on other plant-based compounds such as kratom, no RCT evidence on whole-plant cannabis products, and only a single, small study each for topical or sublingual CBD or cannabinoids other than THC or CBD.

Change in pain severity was the most commonly reported outcome. Other important outcomes were mainly not reported or inconsistently reported or defined. Pain response, defined as a 30 percent or greater improvement in pain, was reported in 10 of 37 studies (27%); 24 of 37 studies (65%) reported on overall function (including pain interference) or disability. The studies poorly reported baseline use of opioids for pain, and only observational studies (7 studies) reported the impact of cannabis interventions on changes to prescription opioid use. While almost all studies reported the number of patients who withdrew from studies due to adverse events, 49 percent did not report SAEs, and 59 percent did not report the overall adverse events, particularly by group. When SAEs were reported, studies either used a unique definition, or did not provide one. In reporting on specific adverse events, not all studies were clear about whether the events were the number of individuals with at least one event, or if a single patient could contribute to an event more than once. Other adverse events that have been reported in noncomparative observational studies and were prioritized for this review (development or exacerbation of psychosis, CUD, and cognitive deficits) were not reported.

Trials were limited by study design and small sample sizes (range 5 to 339; mean 88), particularly for assessing harms. The SOE of the findings was very commonly downgraded due to imprecise estimates (see Appendix G). There were also differences in some key baseline characteristics, including baseline pain scores, which were frequently not adjusted for in study analyses. Another methodologic concern is that many conclusions in the included studies were drawn from post-hoc analyses. Study durations of included RCTs were primarily short-term and included less than 6 months followup (1 RCT reported intermediate followup durations of 47 weeks); 40 percent of trials were 4 to 6 weeks long. This is a key limitation, as pain severity in patients with chronic pain may vary substantially in the short-term and may be influenced temporarily by an intervention or treatment; it is most useful to understand the enduring impact of a treatment on pain severity. Similarly, adverse events such as CUD, cognitive deficits, and SAEs may take time to develop, and longer studies are required to capture such events. Two RCTs, utilized an enriched enrollment randomized withdrawal design in which all patients receive cannabis in a run-in phase; only patients who respond to cannabis and tolerate it are randomized, to continuation or withdrawal. Such trials are intentionally designed to select for patients who respond to and tolerate cannabis, potentially exaggerating treatment effects and underestimating harms compared to patients not selected based on these features., However, our findings were similar when the enriched enrollment randomized withdrawal trials were excluded from analyses. Well-designed head-to-head studies comparing a plant-based product with a standard of care treatment for chronic pain are lacking. The current evidence consists only of small, poorly designed, crossover or observational studies.

Despite limitations in the evidence base, our review has several strengths. First, the living systematic review approach allows us to add new studies soon after they are published, thereby providing an opportunity to update conclusions in a timely fashion. This may be important as cannabis and other plant-based treatments become more readily available to patients, providers and researchers. For this update, new evidence on high-THC to CBD ratio products resulted in reduced effect size for improvement in pain intensity and downgrading of the SOE from low to insufficient for pain response SOE, underscoring the importance of incorporating new evidence. Another strength of our approach is using a framework that categorizes cannabis-related products by both their THC and CBD ratios and type (synthetic or purified and extracted), providing a way to conceptualize the evidence on these two prominent cannabinoids that is consistent with how they are available to consumers. These categories were determined a priori, with the input of a Technical Expert Panel convened for this review. A final strength that separates this review from others is the exclusion of very short-term studies (e.g., a small number of dosing sessions), improving the applicability of findings to chronic pain, where use is likely to be longer term.

There are also some limitations to our review process. We excluded non-English language publications and study results published only as abstracts. For almost all analyses, there were too few trials to apply graphical or statistical methods to detect small sample effects, a potential marker of publication bias. Although there were sufficient trials when combining synthetic or purified and extracted high-THC to CBD ratio products versus placebo to apply such methods, which indicated potential small sample effects, results are difficult to interpret because a statistical test for small sample effects was above the threshold for statistical significance, and there was potential heterogeneity introduced by combining synthetic and extracted products. We did not identify unpublished trials of high-THC to CBD products; one unpublished trial of patients with diabetic neuropathy (n=294) found no difference between nabiximols (comparable THC to CBD ratio oromucosal spray) for diabetic neuropathy (0 to 10 scale, mean difference -0.12, 95% CI -0.60 to 0.36), suggesting that additional unpublished trials could attenuate estimates of effect. Another potential limitation is that we categorized nabilone as a synthetic high-THC product, although it may be more accurately described as a synthetic cannabinoid—a chemical analog to THC, which could have differing effects than THC/dronabinol. To address this possibility, we performed stratified analyses among outcomes that were pooled for synthetic high-THC interventions. The effect size for change in pain severity was larger with nabilone than with dronabinol, but the difference between the effect sizes was not statistically significant. We also grouped purified THC (dronabinol) or CBD with synthetic dronabinol and CBD, because they are chemically identical to the synthetic product. As noted above, when synthetic THC was analyzed separately from purified THC, benefits were present with synthetic THC but not with purified THC. However, results should be interpreted with caution because they are not based on head-to-head comparisons, and there were potential sources of heterogeneity that could invalidate indirect comparisons. For example, all trials of purified THC evaluated dronabinol, but almost all trials of synthetic THC evaluated nabilone. Meta-analyses were based on relatively small numbers of trials (less than 20), which can result in overly narrow estimates using standard random effects approaches, including the profile likelihood model., Therefore, we conducted sensitivity analyses using the Bartlett's correction. Although the Bartlett's correction resulted in wider CIs for pooled estimates, it did not change overall conclusions regarding the statistical significance of findings. The exception was high-THC products and increased risk of sedation, which was imprecise and no longer statistically significant using the Bartlett's correction. We did not apply the Bartlett's correction when there were fewer than five studies, because it may result in overly conservative (wide) CIs in this situation. Meta-analyses based on fewer than five studies should be interpreted with caution, as CIs based on the profile likelihood method may be overly precise. Our inclusion criteria required that the study population have chronic or subacute pain, or have subgroup analyses for this group, which may be why we did not find evidence related to kratom. As in other recent systematic reviews of interventions to treat chronic pain, we grouped the magnitude of effects into small, moderate, and large effects, rather than according to published minimal clinically important difference (MCID) thresholds. Defining clinical significance in chronic pain is difficult because it is subjective and difficult to correlate with real-life experiences of patients. For example, the MCID for improvement in pain is 15 points on a 0 to 100 scale. However, interventions commonly used for chronic pain, including opioids and nonsteroidal anti-inflammatory drugs do not achieve this level of reduction., The typical reduction with opioids, nonopioid medications, nonpharmacological interventions, and cannabinoids is small, 5 to 10 points and may be considered a clinically important effect by patients and clinicians. In addition, when the pooled estimate is close to a pre-defined effect size threshold, a relatively small change in pooled estimates can result in re-categorization. In this update, for synthetic high-THC to CBD ratio products versus placebo, the effect size for pain intensity was re-classified from moderate to small, based on the addition of one new trial that decreased the pooled estimate from -1.10 to -0.95, which could suggest a more substantial change in the estimate than actually observed. On the other hand, using pre-defined thresholds provides consistency in classifying effect size. With more evidence, pooled estimates should become more stable and less likely to require re-categorization. Since this is a living systematic review, new evidence will be incorporated into the review and findings updated on a regular basis.

Applicability

A number of factors could impact the applicability of our findings. The evidence currently is most applicable to patients with neuropathic pain with mostly moderate to severe pain (mean baseline score in RCTs was 6.6 on a 0 to 10 scale, with a range of 4.0 to 8.67). There is also considerable variability within the included studies among the types of neuropathic pain patients experience, and treatment effects might be different depending on the specific neuropathic pain condition. As described above, two RCTs, utilized an enriched enrollment randomized trial design, which is purposefully designed to selectively randomized patients who respond to cannabis and tolerate it.

The evidence applies to adults with chronic pain. The evidence base addressed similar proportion of men and women, with around 58 percent of enrolled participants across all studies being female. While the age range across studies was broad, with mean study ranges of 45 to 68 years, the evidence is mainly applicable to middle-aged patients (overall mean age 54 years). Therefore, the applicability of findings to older patients, in whom the balance of benefits to harms may be very different, with potentially more harms, is uncertain. No study evaluated adolescents and no study enrolled persons with subacute pain. Non-White persons were not well-represented in the studies. It is also unclear how the evidence applies to patients currently taking prescription opioids to treat chronic pain or patients with serious mental illness or other comorbidities who are often excluded from trials. In terms of interventions, this evidence is most applicable to extracted comparable THC to CBD ratio oromucosal spray, high-THC (THC or THC analogue only) synthetic and purified (dronabinol) oral products, and low-THC (CBD only) synthetic oral products. The evidence for comparable THC to CBD oral spray is applicable to mean dosing of 8.4 sprays per day (23 mg THC/21 mg CBD). The evidence for high-THC to CBD ratio synthetic and plant-derived, synthetic equivalent products applies to dosing that was titrated upward, with a maximum dose of 13 to 25 mg per day of dronabinol and 0.25 to 4 mg per day of nabilone (mean doses not reported). For extracted high-THC to CBD products, the evidence was too heterogeneous and limited (2 RCTs) to describe a generally applicable dose. For low-THC to CBD products, findings are most applicable to oral synthetic CBD at a dose of up to 20 to 50 mg/day and combined oral purified THC and synthetic CBD (synthetic) at a dose of 22.5 to 25 mg THC and 45 to 50 mg CBD daily. The applicability of evidence to other products including topical or sublingual low-THC to CBD ratio products and whole plant cannabis is very low or non-existent.

Another factor impacting applicability is that availability of the studied cannabis products varies depending on regulatory and other factors. For example, purified THC (dronabinol) was manufactured in the Netherlands and Denmark and may be available in some European countries, but is not approved by the U.S. Food and Drug Administration (FDA) at this time. Nabiximols are manufactured and available in Canada and some European countries, but are not approved by the FDA. In the United States, multiple whole-plant CBD products are available, but their composition varies, none are FDA approved, and availability varies from state to state depending on laws regarding cannabis use. Although our intervention categories were based on THC to CBD ratio and intended to group together interventions more likely to have similar effect, the generalizability of one cannabis product within a particular category to others is uncertain.

This evidence applies to short-term treatment and mainly informs the impact on mean changes in pain severity and common adverse events. The outcomes after longer term treatment may be different and could influence other outcomes not considered in short-term studies included here (e.g. psychosis, CUD, cognitive deficits). Other information for assessing applicability, such as settings for recruiting participants or the number randomized relative to the number eligible, was lacking.

Only 14 percent of studies were conducted in the United States, with the majority being from Europe (57%). We were unable to assess the impact of country of study or other geographic location characteristics (e.g., rural, metropolitan) on the applicability of specific results.

A number of evidence gaps or limitations in the evidence potentially impacted the applicability of our findings including lack of evidence on extracted whole-plant or purified products, whole-plant cannabis, low-THC to CBD ratio non-oral products, and kratom.