Opioid-sparing effect of cannabinoids for analgesia: an updated systematic review and meta-analysis of preclinical and clinical studies

Cannabinoid co-administration may enable reduced opioid doses for analgesia. This updated systematic review on the opioid-sparing effects of cannabinoids considered preclinical and clinical studies where the outcome was analgesia or opioid dose requirements. We searched Scopus, Cochrane Central Registry of Controlled Trials, Medline, and Embase (2016 onwards). Ninety-two studies met the search criteria including 15 ongoing trials. Meta-analysis of seven preclinical studies found the median effective dose (ED50) of morphine administered with delta-9-tetrahydrocannabinol was 3.5 times lower (95% CI 2.04, 6.03) than the ED50 of morphine alone. Six preclinical studies found no evidence of increased opioid abuse liability with cannabinoid administration. Of five healthy-volunteer experimental pain studies, two found increased pain, two found decreased pain and one found reduced pain bothersomeness with cannabinoid administration; three demonstrated that cannabinoid co-administration may increase opioid abuse liability. Three randomized controlled trials (RCTs) found no evidence of opioid-sparing effects of cannabinoids in acute pain. Meta-analysis of four RCTs in patients with cancer pain found no effect of cannabinoid administration on opioid dose (mean difference −3.8 mg, 95% CI −10.97, 3.37) or percentage change in pain scores (mean difference 1.84, 95% CI −2.05, 5.72); five studies found more adverse events with cannabinoids compared with placebo (risk ratio 1.13, 95% CI 1.03, 1.24). Of five controlled chronic non-cancer pain trials; one low-quality study with no control arm, and one single-dose study reported reduced pain scores with cannabinoids. Three RCTs found no treatment effect of dronabinol. Meta-analyses of observational studies found 39% reported opioid cessation (95% CI 0.15, 0.64, I2 95.5%, eight studies), and 85% reported reduction (95% CI 0.64, 0.99, I2 92.8%, seven studies). In summary, preclinical and observational studies demonstrate the potential opioid-sparing effects of cannabinoids in the context of analgesia, in contrast to higher-quality RCTs that did not provide evidence of opioid-sparing effects.


INTRODUCTION
Opioids are widely prescribed for chronic pain, but due to concerns related to harms, recommendations have been made to reduce reliance on higher doses [1]. One strategy to reduce opioid dose requirements has been through use of opioid-sparing medicines. Opioid-sparing medicines can (1) delay or prevent the initiation of treatment with opioid analgesics (2) decrease the duration of opioid treatment (3) reduce the total dosages of opioid used or (4) reduce opioid-related adverse outcomes, without causing an unacceptable increase in pain [2].
There is substantial interest in the opioid-sparing potential of cannabinoids in the context of pain management. Preclinical data have consistently demonstrated opioid-sparing effects [3]. Interest from policy makers has been further driven by ecological and epidemiological research [4]; however, highly publicized findings have recently been questioned [5].
The overlapping neuroanatomical distribution of opioid and cannabinoid receptors in the central and peripheral nervous system in areas involved with anti-nociception support potential opioid-sparing effects. Opioids and cannabinoids have comparable neurobiological properties with significant degree of functional interaction [6]. Opioid and cannabinoid receptors are G i/o -proteincoupled receptors with similar intracellular signaling mechanisms, including: inhibition of the adenylate cyclase activity, reduced activity of voltage-dependent calcium channels, activation of inwardly-rectifying potassium channels, and stimulation of the MAP kinase cascade. Cannabinoid type-1 (CB1) and mu receptors can interact directly as functional heterodimers when co-expressed in the same neuron [7] and cannabinoid administration may stimulate the synthesis and release of endogenous opioid peptides centrally and peripherally [8]. Each of these properties would predict a synergistic interaction between opioids and cannabinoids, yet further complexity is afforded by the pharmacological profile of the drug. For example, in the case of protean agonists the level of activation of cannabinoid receptors (both constitutive and stimulated) impacts upon the observed pharmacological effect [9,10], whilst partial agonists such as the endocannabinoid anandamide could act as an antagonist in the presence of a more efficacious agonist [11].
Our previous systematic review and meta-analysis found robust preclinical evidence supporting the opioid-sparing potential of delta-9-tetrahydrocannabinol (THC), but limited clinical research testing the opioid-sparing effects of cannabinoids [3]. With the proliferation of research in the past five years, this review aims to provide an updated synthesis of preclinical and clinical studies on the opioid-sparing effects of cannabinoids.

MATERIALS AND METHODS Search
We conducted an updated systematic literature search in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) recommendations [12]. The initial searches conducted on October 29, 2015, had no date limits and the findings have been reported earlier, along with the methods (in lieu of a published/registered protocol) [3]. The updated searches were conducted on December 20, 2020 within Scopus, Cochrane Central Registry of Controlled Trials, Medline, and Embase databases and results were combined with the earlier search. A combination of search terms relating to opioids (e.g., analgesics, opioid*, opiate), cannabinoids (e.g., cannabis, sativex, nabiximol, cannabidiol, tetrahydrocannabinol) and outcomes of interest (e.g., pain, opioid sparing, opioid dose, antinociceptive) were used, consistent with the initial search (Appendix 1). Additional targeted searches of reference lists from identified studies and review articles were conducted to find additional studies not identified by the main searches.

Study eligibility
Eligible studies included: (i) human or animal studies; (ii) for human studies, controlled clinical and preclinical studies where cannabinoids were administered within a medical or clinical therapeutic framework and the study outlined details of cannabinoid administration; (iii) documented concurrent administration of opioids and cannabinoids; (iv) an outcome of either pain/analgesia (including acute, chronic, cancer and non-cancer and experimental pain studies) or opioid requirements/opioid-sparing.
Studies were excluded based on the following criteria: (i) wrong intervention (e.g., cannabinoid use not defined, no cannabinoid administered, non-concurrent opioid and cannabinoid use, non-therapeutic opioid use); (ii) wrong study design (e.g., case reports, epidemiological studies, reviews, letters without empirical data, commentary or news article); (iii) no outcome measure of interest (i.e., pain/analgesia or opioid dose); (iv) full text unavailable; (v) duplicate manuscript; (vi) abstract where full paper published; (vii) unable to confirm eligibility details, or access required data from authors (Appendix 2).
Titles and abstracts, and full texts were screened independently by two authors (SN, LMP, JM, BM, GC, MG, LP and K-EK) using Covidence software [13]. Where inconsistencies were identified, the authors were able to reach consensus on each occasion.

Data extraction and outcomes
The same data extraction forms used in the initial review were used. All data were extracted by one of the authors (SN, LMP and BW, BM) and checked by a second author (SN, LP, BM, JM, MG or K-EK). These same authors reviewed and resolved any inconsistencies. For abstracts without a full text, and missing data, attempts were made to contact authors for additional information.

Outcome measures
For preclinical studies, the primary outcome was the dose of opioid required to give an equivalent antinociceptive effect in the presence and absence of cannabinoids.

Analysis
Preclinical studies. Data were extracted and, where studies that were sufficiently similar in design and outcome measures, meta-analysis was undertaken. For the residual studies, a narrative review was conducted.
To prepare the data for the meta-analysis, the ED 50 and either confidence limits or standard error were extracted from the relevant literature. ED 50 is calculated on the log 10 scale. Therefore, to meet the assumption of normality, the log 10 c ED 50 must be used in the meta-analysis. The log 10 of the confidence limits must also be determined to calculate the standard deviation (SD) of the log 10 c ED 50 : SD log 10 c ED 50 ¼ log 10 UL À log 10 c ED 50 1:96 where UL is the upper confidence limit.
When only standard error was reported, the confidence limits were calculated using the method of Litchfield and Wilcoxon [14] and the above procedure was repeated to calculate the standard deviation. This method also allowed for the inclusion of studies that did not report exact sample sizes for all treatment groups, as sample size was not required for the calculation of standard deviation.
Data for the meta-analysis were analyzed using Review Manager 5.4 (Cochrane Collaboration, Oxford, UK). When calculating the continuous outcome of an equally effective opioid dose (e.g., the log 10 ED 50 for morphine when administered alone versus when administered with a cannabinoid), the inverse variance statistical method and random effects model were used to compensate for study heterogeneity.
No statistical difference was found in outcomes between the studies that used different rodent species or nociceptive assays. Therefore, the mean difference of log 10 ED 50 and the corresponding 95% confidence intervals were calculated. Due to the nature of log calculations, the mean differencewhen back-transformed to the original units-represents the response ratio. For easier interpretation, we present the reciprocal of the response rate.
Clinical studies. The outcomes of interest in clinical studies were: (1) reduction in total opioid doses, (2) reductions in pain through the addition of a cannabinoid, (3) adverse events, and (4) evidence of abuse liability. A broad range of study designs were considered. Where studies used sufficiently similar methods and outcome measures, meta-analyses were conducted.
Clinical trials. Meta-analysis for clinical trials was conducted with Revman 5.4, where medians and interquartile ranges were required to be converted into means and standard deviations to allow inclusion in meta-analyses, we used methods established by Luo et al. [15] and Wan et al. [16].

Observational studies
For observational studies, meta-analyses on proportions reporting changes in opioid dose outcomes were conducted using a random effect model in Stata (metaprop, code available on request). A pooled prevalence was calculated with 95% confidence intervals for each of the identified outcomes that were comparable; (i) reduced opioid use, (ii) ceased opioid use. For remaining outcomes, a narrative synthesis was conducted.
Clinical studies were scored for quality using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) criteria [17]. Quality ratings were not applied to preclinical studies. As all meta-analyses had less than ten studies funnel plots were not used to assess bias [18].

RESULTS
Ninety eligible publications representing data from 92 studies were identified; 29 in the initial searches and 63 in the updated searches. Forty preclinical (21 since 2016) and 37 clinical studies (controlled trials n = 20 [12 since 2016] and observational n = 17 [13 since 2016]) were identified for inclusion (see Appendix 3). Fifteen registered clinical trials, where data were not yet available were also identified.
Measures of abuse liability. Six studies reported on measures of abuse liability including intracranial self-stimulation (ICSS) [38], conditioned place preference [43,44], oxycodone selfadministration [50], and drug discrimination [32,33]. None provided evidence that cannabinoids increased abuse liability. CP55,940 had no effect on ICSS with morphine or tramadol [38], JWH105 when co-administered with morphine reduced conditioned place preference, and LY2828360 when administered with morphine blocked condition place preference [43,44]. THC reduced oxycodone self-administration [50], and attenuated the discriminative stimulus effect of morphine and heroin in nondependent monkeys, but not in dependent monkeys [33]. CP55,940 and WIN55,212 reduced the discriminative stimulus effect of morphine and decreased heroin self-administration, both effects were reversed by the CB1 receptor inverse agonist rimonabant [32].
Meta-analysis identified an opioid-sparing effect with morphine and delta-9-THC co-administration with one study [47] added to the previous meta-analysis, Z = 4.46, p < 0.001 (mean difference in log 10 ED 50 = -0.54 [-0.78, -0.31]). As there was significant heterogeneity in the data (I 2 = 99%), a random effects model was used. When back-transformed to the original units, the response ratio was 3.5 (95% CI 2.04, 6.03) indicating that the median effective dose (ED 50 ) of morphine was 3.5 times lower when administered with delta-9-THC compared to when administered alone.
Results from clinical studies Thirty-five eligible publications representing 37 clinical studies with 5180 participants provided data relevant to the research question ( Table 2).
Clinical trials-experimental pain. Five laboratory-based studies in healthy volunteers (n = 82) examined pain responses with coadministered opioids and cannabinoids using double-blind within-patient study designs (Table 2a). Four studies examined oral dronabinol (2.5-20 mg) [59][60][61][62] and one examined smoked cannabis [63]. Inconsistent outcomes were observed; two studies found evidence of increased pain, two found some measures of decreased pain, and one study found effects of cannabinoids on pain "unpleasantness" but not pain ratings. One study found low dose dronabinol (2.5 mg) decreased the analgesic effects of oxycodone as measured with a pressure algometer with no effect of 5 or 10 mg dronabinol on analgesic outcomes [61]. Another study noted potentially hyperalgesic effects of cannabinoids [59]. This was in contrast to the analgesic effect observed on pain threshold and tolerance with a cold pressor test when smoked cannabis was administered with 5 mg oxycodone compared oxycodone or cannabis alone, although effects were not found on measure of outcomes of pain intensity or bothersomeness [63]. Dunn et al. [62] demonstrated analgesic effects from dronabinol 2.5 mg when co-administered with hydromorphone on thermal pain measures, but not with higher doses of dronabinol, or on other measures of pain. Roberts et al. [60] found that the coadministration of dronabinol and morphine resulted in reduced pain "unpleasantness" compared to either drug alone. Three experimental studies included measures of abuse liability, and found that smoked cannabis and dronabinol may increase the abuse liability ratings of oxycodone and hydromorphone using measures such as ratings of feeling high and drug liking [61][62][63].
Clinical trials-acute pain. Three double-blind randomized controlled trials (n = 545) examined the opioid-sparing effects of CBD in acute pain [64][65][66]. Nabilone and dronabinol were examined in acute post-operative pain and CBD in acute low back pain (<30 days duration). No benefit on opioid dose requirements or analgesic outcomes was identified (Table 2b).
Clinical trials-cancer pain. Seven controlled trials (1795 participants) investigated the opioid-sparing effect of cannabinoids in patients with different forms of cancer pain. One small, nonrandomized study found a non-significant effect of cannabis on pain control [67], and a second pilot found no effect of medical cannabis on pain, but an increase in opioid dose in a group that received delayed cannabis [68] (Table 2c). The remaining studies were all larger single or double-blind randomized trials. Five randomized controlled trials (reported in four publications) examined THC and nabiximols compared to placebo in patients with cancer pain who were taking opioids [69][70][71][72]. Two studies found improved analgesia with nabiximols compared to the placebo. Johnson et al. [69] found no effect of nabiximols on breakthrough opioid dose requirements. Portenoy et al. [70] conducted a dose-ranging study, and a significant analgesic effect was only found in the lowest dose group, with poorer tolerability observed for higher doses. The remaining three studies found no benefit of adding cannabinoids on their primary outcome of analgesia. Although Lichtman et al. [72] did not find a significant effect of cannabinoids on pain in an intention to treat analysis, the per-protocol analysis did find a significant effect (Table 2c). Four of seven studies required maintenance opioid doses to be kept stable [70][71][72]; five studies measured breakthrough opioid doses requirements as an outcome with no evidence of a difference found [69][70][71][72]. No cancer pain studies included measures of abuse liability. Meta-analyses were possible on the outcomes of change in mean total oral morphine equivalent daily dose (OMEDD) from baseline (n = 4 studies), percent change in pain score from baseline (n = 4 studies) and adverse events (n = 5 studies). Metaanalysis of four studies (n = 1119 participants) found no effect of Where opioid-sparing effect reported as not observed, results were additive rather than synergistic; or no increased analgesic effect was observed.
Clinical trials-chronic non-cancer pain. Five clinical trials (139 participants, Table 2d) examined the effects of dronabinol [73][74][75] and smoked cannabis [76,77] in patients with chronic non-cancer pain. Most studies had short observation periods (5 h to 5 days) [74][75][76][77], and used crossover designs [73][74][75][76]. Opioid dose was an outcome in one study, with no difference between smoked cannabis and placebo [76]. All five studies reported on analgesic outcomes with conflicting findings. A single-arm open-label study (with no comparison group) recruited people with mixed types of chronic non-cancer pain (n = 24) who were prescribed opioids, and found significant overall reductions from baseline pain ratings following co-administration of cannabinoids [77]. In contrast, a double-blind crossover study in sickle cell patients found no significant differences analgesia effects between placebo and vaporized cannabis [76]. Two studies recruited patients with chronic pancreatitis and found no effect of dronabinol on pain measures compared with placebo [73,74]. A sub-analysis in patients with chronic postsurgical abdominal pain found lower pain among those who received dronabinol compared with placebo [73]. A single-dose study in patients with mixed-chronic pain conditions, found dronabinol 10 and 20 mg was associated increased analgesia compared with placebo [75]. These studies did not include measures of abuse liability.
Clinical studies-observational. Seventeen observational studies (n = 2674) examined the opioid-sparing effects of cannabinoids; three small retrospective case series of two to three patients each [78][79][80], two retrospective cohort studies [81,82], two retrospective matched cohort studies [83,84], and ten prospective observational cohort studies [85][86][87][88][89][90][91][92][93], including two open-label extension studies [75,93] (see Table 2e). Two retrospective matched cohort studies examined acute analgesia with traumatic injury [83] and joint arthroplasty [84]. Both found no difference in pain scores, but reduced opioid consumption on at least one measure. For pain management following joint arthroplasty, there was no change in daily opioid dose with dronabinol administration, but a reduced total opioid consumption due to significantly shorter hospital stays in the dronabinol group [84]. One study compared those prescribed nabilone with those that had not received it, using propensity scoring to adjust for the greater severity of the nabilone prescribed group [89]. The remaining observational studies did not have control conditions and examined opioid use in patients with a range of different types of chronic non-cancer pain. Seven studies reported on the outcome of OMEDD after commencing medical cannabinoids, with reductions from 9 to 140 mg OMEDD reported (Table 2b). Four studies quantified the reduction in pain scores, which ranged from 12% to 70%, with two studies exceeding the minimum threshold of a 30% reduction in pain to be clinically meaningful. Meta-analysis was possible for studies that reported the proportion of patients who reported opioid reduction or cessation; eight studies reported the proportion of patients who ceased opioids (range 2-100%), with a pooled prevalence of 0. 39

Quality ratings of clinical studies
The clinical studies were rated using the GRADE criteria. Of the clinical trials, five laboratory studies provided moderate evidence, three clinical trials in acute pain provided high quality evidence, six clinical studies provided low-high quality evidence in cancer pain, and five studies in chronic non-cancer pain were assessed as low-moderate quality. The seventeen observational studies were assessed to be low to very-low-quality evidence ( Table 2).
Ongoing clinical trials. We identified 15 registered clinical trials which, based on published protocols and clinical trial registry entries, may provide important data for future updated reviews (Appendix 6).

DISCUSSION
The current update represents the largest synthesis of studies examining the opioid-sparing effects of cannabinoids, with double the number of preclinical studies, four times as many clinical studies and more than six times the number of participants (>5000) compared to our earlier review [3], reflecting the rapid growth of clinical research in this area. Most preclinical studies found synergistic effects with opioids and cannabinoids co-administration, predominantly with mixed CB1/CB2 agonists such as delta-9-THC, though effects varied with different cannabinoids, opioids and pain assays. Meta-analyses (with one addition preclinical study since 2015) demonstrated that morphine dose required to produce an equivalent analgesic effect was 3.5 times lower when co-administered with delta-9-THC, consistent with the previous review [3]. This effect would be clinically meaningful if replicated in well-controlled clinical studies. However, preclinical studies often have larger effect sizes, attributed to the reduced heterogeneity compared to clinical populations [94]. This body of preclinical research may help to identify specific cannabinoids and mechanisms that underlie an opioid-sparing effect, with the most consistent effects observed with mixed CB1/CB2 agonists, and evidence of potential antagonistic effects between CB1 agonist and mu receptor agonists in models of mechanical hyperalgesia.
A rapidly growing number of clinical studies have measured opioid-sparing endpoints, though findings were inconsistent. The highest quality studies were conducted in patients with cancer pain, where meta-analysis of four studies did not find significant effects on opioid dose or analgesia. Conflicting findings were found in studies of experimental pain, and in patients with chronic non-cancer pain. Further studies are needed to clarify the results found here given the small number of studies.
A limited number of controlled studies demonstrated benefits of combining cannabinoids with opioids for analgesia. Experimental pain studies found cannabinoids improved [62,63] and worsened [61] analgesia. These effects were not dose dependent, with significant effects seen with lower but not higher doses of delta-9-THC. Opioid-sparing effects were not seen in wellconducted RCTs with acute pain, or in meta-analyses of RCTs in cancer pain, and studies that did find positive effects have important limitations such as no control group [77], small sample sizes [67,75], and the mixed quality of the study design. Furthermore, some RCTs instructed patients to continue their pain medication in the same doses, which may preclude identifying a change in opioid dose [70][71][72][73]77], although changes in breakthrough opioid requirements were a secondary outcome in six studies [69][70][71][72]75]. Some clinical studies demonstrated beneficial effects of opioid and cannabinoid co-administration on other outcomes such as sleep, and functioning in chronic pain patients [75,77]. Conflicting results were found between preclinical studies and clinical trials on measure of abuse liability. Evidence of reduced abuse liability was found in some animal models, which contrasted directly with evidence of increased drug liking and subjective effects in human studies.
Finally, observational studies had methodological concerns including small sample sizes (several observational studies included in meta-analysis had two to three patients), no control groups or blinding, selection bias, and were likely to have been impacted by expectancy effects.
Although our review is much broader, we have drawn similar conclusions to earlier reviews. For example, a review of crosssectional surveys and cohort studies, representing lower quality evidence, found large reductions in opioid doses, though study designs prevented the drawing of causal conclusions [95]. A later review with five randomized trials with patients with chronic pain and 12 observational studies further concluded that there was uncertainty in the evidence [96], although this review considered a substantially smaller number of clinical trials than we consider. Future studies may benefit from focusing on populations with higher opioid tolerance, or higher motivation to reduce opioid doses, where clinical benefits may be greatest [97]. Standardization of outcomes for opioid-sparing research may assist with harmonization of outcome measures and support meta-analysis with future clinical trials [2].
Despite the inclusion of a larger number of studies, and the increased size and quality of clinical trials in recent years, our conclusions have not changed substantially from our earlier review. Nevertheless, we did identify 15 registered clinical trials indicating that this continues to be an active area of research in which the science is likely to continue to evolve.
In conclusion, preclinical studies support the opioid-sparing effect of delta-9-THC and other mixed CB1/CB2 agonists. Observational studies support the opioid-sparing potential of cannabinoids. However, findings from clinical trials provide conflicting results that may highlight important areas for future research. These include identifying optimal doses and populations who may experience benefits with cannabinoids. With numerous clinical trials currently underway, we will update our review, as higher-quality data may enable stronger conclusions to be made.