Introduction

The global increase in legal access to cannabis and cannabinoid-based products for medical and non-medical purposes has been paralleled by their widespread promotion and use. In particular, North America has the highest rate of cannabis use in the world, where some of the first jurisdictions to legalize cannabis exist [1,2,3]. In recent years, North America has been reported to have the highest prevalence of past-year cannabis use compared to other sub-regions globally at 14.5% [4]. Of the emerging cannabinoids available, cannabidiol (CBD), a principal cannabinoid presumed to be non-euphoric and non-intoxicating, has the highest prevalence of past-year use [5]. Past-year CBD use was reported to be 26.1% in the United States and 16.2% in Canada [6]. Common reported reasons for use include medical indications such as the management of pain, anxiety, and depression [6].

This surge in consumption and access to cannabis and cannabinoid products has sparked concerns regarding their careful use during ‘safety-sensitive’ work or activities (e.g., operating motor vehicles or machinery) [7, 8]. To date, a majority of research and public policy has focused on identifying and mitigating cannabis impairment risk related to delta-9-tetrahydrocannabinol (Δ9-THC), the main psychoactive cannabinoid known to produce acute deficits in cognitive performance and driving ability [9, 10]. In contrast, little attention has been given to CBD due to the general belief that it is non-impairing. Although the available evidence has pointed to a lack of cognitive, psychomotor, or subjective effects with oral and vapourized CBD even at high or supratherapeutic doses [11,12,13], there has yet to be a comprehensive, systematic review of the literature to synthesize data on the performance effects of acute CBD exposure, or evaluation of potential moderating factors that may impact sensitivity to performance effects.

This lack of clarity surrounding the effects of CBD on daily functioning presents several concerns. A primary concern is the potential public health consequences for traffic safety if people using CBD are operating motor vehicles under the assumption that it is non-impairing. It is equally as important to consider the implications of CBD-related impairment on workplace health, safety, and policy. At present, many workplaces err on the side of caution and treat cannabis as a single entity, subjecting CBD to the same restrictions as Δ9-THC [14, 15]. These restrictions are particularly relevant to people who use CBD for medical purposes to help manage symptoms or a condition, such as chronic pain, epilepsy, or anxiety when Δ9-THC cannot be used safely [16, 17]. In this context, CBD use may afford individuals the ability to engage in daily activities and workplace duties which they may otherwise be unable to do. Hence, efforts to clarify the risk of CBD-associated impairment are greatly needed to inform public health legislation, as well as workplace policy and practice.

The current meta-analysis synthesized and critically evaluated available evidence from human laboratory studies assessing the potential for CBD to impair cognitive and psychomotor performance. These effects were compared to a placebo control group and a positive control of Δ9-THC. Moderators were also evaluated to determine individual difference and product-specific factors that may alter the magnitude of effect.

Methods

Search strategy

This review was registered on PROSPERO (The International Prospective Register of Systematic reviews) (CRD42021247522) and reported in accordance with PRISMA guidelines (eTable 1) [18]. A systematic search in AMED, EMBASE, CENTRAL, PsychINFO, CINAHL, Clinicaltrials.gov, Medline, MedRxiv, and Web of Science was completed on June 22nd, 2022 and updated again on January 4th, 2023. Literature searches using the keywords associated with cannabidiol or CBD paired with cogniti*, or impair*, or domain-specific keywords (e.g., memory) were independently conducted by two reviewers. As an example, studies were identified by the following expression: (cogniti* OR driving OR coordinat* OR processing speed OR reaction time OR executive function OR memory OR “task performance and analysis” OR attention OR learn* OR task switching OR intoxic* OR motor OR impair* OR perform*) AND (cannabidiol* OR CBD OR Epidiolex OR Epidyolex). The star symbol (*) was used to capture derivatives of search terms (by suffixation) and enclosed quotation marks were used to capture exact phrases. See eTable 2 for full search strategies.

Eligibility criteria and study selection

For inclusion, studies had to meet the following criteria: (1) involve adult participants; (2) placebo-controlled experimental design; (3) report route of cannabinoid administration and dose schedule; (4) measures of self-report, researcher observation, or objective neurocognitive or psychomotor assessments within 0–8 h of CBD administration. CBD administration was defined as administration of any form of CBD either in isolation or with a THC content of <1%. Self-reported/subjective measures of neurocognitive performance were restricted to those with specific constructs (e.g., alert, sedation). Subjective ratings of drug high or intoxication were excluded due to lack of specificity. Only full-length, English-language original research articles were accepted. Studies were excluded if: (1) performance test(s) were not administered within 8 h of CBD administration; (2) either the dose of CBD administered, or the length of time between CBD administration and the performance test(s), was not reported and could not be estimated (e.g., in regard to dose, reporting the number of ‘puffs’ smoked from a cannabis cigarette was not considered adequate to estimate dose); (3) there was no confirmation of a ≥24-h abstinence period for intoxicating substances (e.g., cannabis, alcohol, other recreational drugs) before performance assessments. See eTable 3 for PICOS statement. Three authors (LL, LE, and AC) assessed study eligibility and quality blinded, and resolved any disagreement by consensus. Authors LL, CP, and LE screened titles and abstracts. Authors LL and AC assessed full texts for eligibility and quality.

Data extraction and outcome measures

Studies were required to have measured driving performance, a discrete cognitive skill (e.g., information processing), and/or subjective cognitive or psychomotor function. These performance outcomes were used as markers of potential impairment. Each cognitive test used in the included studies was categorized into a performance domain (Table 1). Categorizations of measures by cognitive function/domain were based on previous meta-analyses, to allow for greater comparability across the literature [9, 19]. All outcome measures of neurocognitive function or psychomotor performance on objective or subjective assessments were extracted for each domain (e.g., reaction time, accuracy of responses, mean score etc.). Additional variables were extracted including study participant characteristics, dose, product type, method of administration, concomitant drugs, comorbidities, cannabis experience, and type of performance assessment. The primary outcome was the peak mean difference in acute performance measures between CBD and placebo, as quantified by Hedges’ g. The secondary outcome was the peak mean difference in acute performance measures between CBD and Δ9-THC. Eligible effect estimates for the peak mean difference in studies with multiple time points were constrained to 0–120 minutes post-inhaled cannabis and 30–240 min post-oral cannabis consumption given the pharmacokinetics of each route of administration [13, 20, 21]. See eMethods for further details.

Table 1 Cognitive performance tests and associated domains.
Full size table

Effect size computation

Hedges’ g effect estimates were calculated from the standardized mean difference (SMD) between matched intervention groups (CBD, Δ9-THC, placebo). Hedges’ g was used to provide a more unbiased estimate for small sample sizes [22]. All effect sizes were recorded such that positive Hedges’ g values indicated a greater magnitude of impaired performance. Effect sizes were interpreted using the convention (g = 0.2 [small], 0.5 [medium], 0.8 [large]) [23]. In order to compute Hedges’ g, Cohen’s d was first computed using the formula [24]:

$$d=\frac{{\bar{Y}}_{{diff}}}{{S}_{{within}}}=\frac{{\bar{Y}}_{1}-{\bar{Y}}_{2}}{{S}_{{within}}}$$

The standard deviation within groups was imputed from the standard deviation of the difference using the formula:

$${S}_{{within}}=\frac{{S}_{{diff}}}{\sqrt{2\left(1-r\right)}}$$

where r is the correlation between pairs of observations. If r was not reported or unable to be calculated from raw data, the standard 0.5 assumption was used. Cohen’s d was then converted to Hedges’ g using the formula:

$$g=J\times d$$

The J conversion factor was computed using the formula:

$$J=1-\frac{3}{4{df}-1}$$

Meta-analytic methods

Omnibus effect estimates and moderation analyses were conducted using a robust variance estimation (RVE) meta-regression approach. The RVE approach allows for the incorporation of dependent effect size measurements (e.g., multiple effect sizes from crossover studies or studies with multiple outcome measures for the same participants) without violating independence assumptions by using robust standard errors based on heteroskedasticity-robust estimates and clustered methods (see refs. [25, 26] for details). This analysis utilized a modified RVE method for small-sample size adjustments [26]. Moderator analyses were carried out for the primary outcome of peak mean difference between CBD and placebo in a series of one-covariate analyses. Model outputs were not interpreted if the degrees of freedom were <4, as recommended by Hedges et al., 2010.

Sensitivity analyses were conducted using a standard random-effects meta-regression approach. In studies with multiple effects, effect estimates were averaged to produce a single effect. Traditional publication bias measures (e.g., Egger’s plot for funnel asymmetry) were conducted on the average effect size model as they have not yet been widely validated for RVE models.

All analyses were carried out in R using the robumeta [27] and Metafor packages [28].

Risk of bias and quality assessment

All studies were assessed for risk of bias using the revised Cochrane Risk of Bias (RoB) tool [29]. The RoB 2.0 comprises five domains, including the randomization process, deviation from intended interventions, missing data, measurement of the outcome, selective outcome reporting, and “other sources of bias”. Two independent assessors (LL and AC) performed the risk of bias assessments, with any disagreement resolved by consensus. A decision around the interpretability of the available evidence was made by categorizing studies by the research question and rating them based on their quality.

Results

Study characteristics

Figure 1 shows the PRISMA flowchart of study selection. Given the limited literature base, a broad search strategy was adopted in an attempt to capture all possible studies (See eMethods for more details). A total of 15,990 records were identified from database searches. After the removal of duplicates, 11,355 records were screened, of which 508 documents were reviewed for eligibility by full text. A total of 20 studies were included, where 16 studies were included in the quantitative analysis [12, 13, 30,31,32,33,34,35,36,37,38,39,40,41,42,43] and an additional four studies were included in the qualitative synthesis due to insufficient data for quantitative synthesis [44,45,46,47]. Among the 16 studies included in the quantitative analysis, there was minimal missing outcome data, with only one timepoint of an outcome missing in a single study. Additionally, supplementary and/or raw data were received from eight of the 16 studies.

Fig. 1
figure 1

PRISMA flow diagram.

Full size image

The characteristics and key findings of the 20 included studies are presented in Table 2 and eTable 3. Outcome measures and dependent variables for each study included in the quantitative synthesis are presented in eTables 4 and 5. Seventeen studies were double-blind, randomized, placebo-controlled cross-over designs and three were double-blind, randomized, placebo-controlled parallel-group trials. Of the eligible studies, 14 included healthy adult participants [12, 13, 30,31,32,33,34,35,36, 38, 39, 42, 46, 47]; one study included adults with social anxiety disorder [44]; one study was comprised of participants with psychosis [41]; one study was comprised of participants at high-risk of psychosis [45]; one study was participants with nicotine dependence [37]; one study included healthy adults with low and high Schizotypy Personality Questionnaire scores, but no clinically diagnosed schizophrenia or psychosis [40]; and one study included adults with chronic pain and fibromyalgia [43]. The majority of study populations were cannabis-naive or had few lifetime exposures. Only three studies included participants who had a recent history of occasional or frequent cannabis use [40, 42, 45]. Cannabinoids were primarily administered through an oral route (N = 14, 70%) or via vapourization (N = 5, 25%) alone, with one study administering both oral and vapourized cannabinoids (N = 1, 5%) [13]. Doses of oral CBD ranged from 15 mg to 4500 mg and from 12.5 to 400 mg for vapourized CBD. Doses of oral Δ9-THC ranged from 10 mg to 30 mg and vapourized doses ranged from 8 mg to 30 mg Δ9-THC. All included studies used a single-dose regimen. Qualitative findings are presented in the eResults.

Table 2 Study characteristics and key findings.
Full size table

Quantitative findings

Omnibus meta-analysis of peak performance effects of acute CBD exposure compared to placebo

The omnibus RVE meta-analysis indicated a significant, but small effect size for impaired performance following acute CBD consumption compared to placebo (Hedges’ g = 0.122, 95% CI: 0.023–0.221, p = 0.019). Moderate heterogeneity was observed among studies (I2 = 38.24%). A consistent omnibus estimate was observed when collapsing effect sizes into a single average estimate for each study, Hedges’ g = 0.113, 95% CI: 0.014–0.212, p = 0.026. Model results are presented in Table 3.

Table 3 Summary of primary meta-analytic findings.
Full size table

Moderator analyses

Results of moderator analyses are presented in Table 3. A significant moderator effect was observed for measure type. This effect reflected larger mean differences between CBD and placebo when subjective measures were used (Hedges’ gSubjective = 0.288 versus Hedges’ gObjective = 0.048). Objective measures were not found to be significantly different than 0 when changed to the reference group to test the significance of the intercept.

A significant moderation effect was observed for cognitive function. This reflected a significantly larger mean difference for measures of subjective sedation/tiredness (Hedges’ g = 0.329) compared to episodic memory (Hedges’ g = 0.066) and working memory (Hedges’ g = 0.026). Information processing measures were not observed to have a significantly different mean difference compared to subjective sedation/tiredness measures (p = 0.157). However, only subjective sedation/tiredness was observed to be significantly different from 0. Comparisons to measures of divided attention, driving, executive function and subjective alertness could not be made due to insufficient degrees of freedom.

CBD dose and route of administration were not significant moderators (ps > 0.5).

Comparisons to Δ9-THC

Two secondary analyses were conducted to assess the difference in peak performance effect between CBD and Δ9-THC (Table 4). The first analysis compared peak mean difference in performance measures between cannabis and placebo, with cannabinoid (CBD or Δ9-THC) as a moderator. Eight studies that had a CBD and a Δ9-THC arm were included. Cannabinoid type was a significant moderator of effect sizes. This effect reflected larger mean differences between Δ9-THC and placebo (Hedges’ g = 0.356, 95% CI: 0.059–0.398, p = 0.016) compared to CBD and placebo (Hedges’ g = 0.128).

Table 4 Summary of secondary meta-analyses.
Full size table

The second analysis provided a direct comparison of the peak mean difference in performance measures between Δ9-THC consumption compared to CBD consumption. Eight studies that had a CBD and Δ9-THC arm were included. The omnibus RVE meta-analysis indicated a significantly greater effect on performance for Δ9-THC compared to CBD. This effect reflected a moderate effect size for impaired performance following Δ9-THC consumption compared to CBD (Hedges’ g = 0.416, 95% CI: 0.017–0.816, p = 0.043).

Quality of evidence

The quality of available evidence was deemed moderate-to-high (See eFig. 1 and eResults). Of the 20 clinical trials analyzed, three (15%) were deemed to have an overall ‘low risk’ of bias, 16 (76%) were assessed as having ‘some concerns’, and one (5%) was identified as ‘high risk’ of bias.

Egger’s test for funnel plot asymmetry was not statistically significant, consistent with the funnel plot visual (See eFig. 2 and eResults).

Discussion

The results of this meta-analysis indicate that acute CBD consumption had a small but statistically significant effect on performance as assessed by all outcomes in aggregate, compared to placebo. Moderator analyses revealed this effect was significant only for subjective ratings of sedation/drowsiness, and no significant effects were observed for objective task performance on domains including memory, psychomotor ability, driving performance, information processing, attention, or higher order cognitive functioning. Dose and route of administration were not significant moderators in this analysis. As expected, acute doses of Δ9-THC produced significantly greater impaired performance than CBD relative to placebo and in direct comparison to CBD under the same experimental conditions. It is important to note that this sample was composed of primarily naive or infrequent cannabis users. It is unknown if these findings would translate to individuals with consistent cannabis product use, generally, or CBD use, specifically (e.g., medical cannabis patients). Additionally, the small, statistically significant effect size for the primary comparison of performance on cognitive and psychomotor measures between CBD to placebo may not translate to functional impairment, particularly given that these differences were limited to subjective feelings of sedation or tiredness.

This evidence synthesis supports that acute CBD consumption does not negatively impact neurocognitive function, as assessed by objective neurocognitive measures, consistent with findings from earlier trials and reviews [11, 48, 49]. It is important to note that these findings are from a sample of primarily infrequent cannabis consumers and may not represent the actual population of individuals who use CBD chronically. Infrequent cannabis consumers would most likely have the highest risk of impairment compared to individuals who use CBD chronically. Additionally, this sample was primarily in healthy adults. The effect of CBD may be different in different clinical populations. The small effect of subjective sedation noted in the current study has been reported inconsistently within previous literature. Somnolence and sedation are noted as potential side effects in Epidiolex prescribing information [50]. However, it has been proposed that CBD-related sedation in the context of the treatment of epilepsy may be due to drug interactions rather than CBD itself [51, 52].

Discrepancies between subjective and objective indicators of impairment have been noted previously. Some evidence suggests that people who use cannabis may overestimate their level of sedation and other indicators of impairment [53, 54], while others may compensate for expected impairment-related effects [55]. Drug expectancy may also contribute to this phenomenon. The expectation of receiving a certain drug can produce subjective and behavioral effects similar or opposite to those related to the drug, even in the absence of the drug itself. Such expectations can be formed by verbal information about the content and supposed effects of the drug, prior experience, and observational learning [56]. Metrik et al. [55, 57] have shown that the expectancy of receiving Δ9-THC produces greater subjective effects, including euphoria and sedation. CBD expectancy may also impact subjective and drug effect ratings [58]. Given that cannabis expectancy seems to affect self-reported reactions and drug responses, this calls into question the level of functional impairment associated with the small effect size obtained from this synthesis.

As expected, Δ9-THC produced significantly higher magnitudes of impaired performance compared to CBD. This adds validation for detecting and examining impaired cognitive and psychomotor performance for CBD and THC using the same experiments and designs. However, the question of whether concurrent CBD and Δ9-THC consumption increases or decreases the magnitude of impairment remains. Many cannabis products contain both CBD and Δ9-THC, including whole-plant CBD-dominant products. Additionally, CBD may be co-administered with Δ9-THC preparations with the expectation that CBD can ameliorate Δ9-THC-related cognitive impairment, anxiety, and sedation while also offering a range of therapeutic benefits [59,60,61]. Evidence from both experimental and naturalistic studies suggest that the addition of CBD to Δ9-THC produces differential dose-dependent effects, which may depend on the ratio of CBD:Δ9-THC and route of administration [30, 42, 43, 62, 63]. One study found that low-dose vapourized CBD (4 mg) enhanced impairment relative to Δ9-THC (8 mg) alone, whereas high-dose CBD (400 mg) reduced impairment across objective and subjective measures [42]. Other studies have reported that vapourized Δ9-THC/CBD-equivalent cannabis (13.75 mg Δ9-THC + 13.75 mg CBD) is no less impairing than Δ9-THC-dominant cannabis (13.75 mg Δ9-THC), and in some cases CBD may actually exacerbate Δ9-THC-induced acute impairment, as measured by psychomotor assessments and simulated driving performance [47, 62]. Pharmacokinetic data from the available research has also shown that peak plasma concentrations of Δ9-THC appear to be higher when CBD is co-administered [30, 43, 62], although several studies have also found no evidence of changes [64,65,66]. CBD can inhibit the metabolism of Δ9-THC and other drugs, and these interactions are more likely to occur after oral ingestion of CBD than with inhalation [67,68,69,70]. Thus, it is imperative to consider the potential for CBD to increase impairment when combined with other drugs, even if acute doses of CBD alone are not associated with functional impairment in controlled research studies.

The majority of participants in the current investigation were naive to cannabis or had few lifetime exposures. It has previously been observed that people who regularly use cannabis experience less cannabis-associated impairment compared to those with occasional use [9, 71]. As such, it is unknown how these findings would translate to populations with more frequent cannabis use (e.g., medical cannabis patients). However, it could be predicted that the small effect on subjective sedation observed in the current study may be diminished with frequent CBD use, in line with what has been observed in studies assessing Δ9-THC-associated impairment in frequent cannabis users [9, 72]. Further, some evidence suggests that CBD may improve cognitive function with prolonged use [73].

Future directions

The available literature on the acute performance effects of CBD consumption only allowed for assessment of performance in certain domains of cognitive function and in certain contexts of use (e.g., naive to cannabis consumption). Of key importance, there is a need to examine the impact of frequent, long-term CBD use on neurocognitive function to examine if tolerance diminishes the observed effect. Particularly for common safety sensitive tasks completed by the general population, such as driving, to gain a more robust picture of real-world risk. Finally, the majority of the trials in this study used CBD isolate products. In the real-world, full spectrum CBD-dominant products (which include other major and minor cannabinoids [including low levels of Δ9-THC] and terpenes), balanced CBD:Δ9-THC products, and lower CBD to Δ9-THC ratio products are commonly used. Effects on neurocognitive performance associated with these products should be further investigated as other cannabinoids and terpenes may contribute to impairing effects.

Limitations

This meta-analysis had several limitations. There was insufficient data, due to the sparse number of studies that included frequent cannabis users, to examine the potential difference between infrequent and frequent cannabis users. As such, these findings may not translate to populations who consistently use CBD. Additionally, although moderation analyses were conducted to assess variability, there are undoubtedly other variables that may impact an individual’s magnitude or risk of impaired neurocognitive performance (e.g., comorbidities, concomitant medications) that were not addressed in the included studies.

Conclusion

This meta-analysis suggests acute CBD consumption may be associated with a small increase in subjective sedation compared to placebo in infrequent cannabis users, but does not significantly impact performance across a range of cognitive domains. These results are consistent with previous evidence supporting that CBD consumption does not impact neurocognitive function. As such, acute use of CBD in the absence of THC or other drugs is unlikely to lead to functional impairment. Further research is warranted to investigate the risk of impaired neurocognitive function with daily CBD consumption, in addition to assessing performance in alternative domains.