The prevalence of cannabidiol (CBD) use in North America and Europe: A meta‐analysis
CANNABIDIOL IN NORTH AMERICA AND EUROPE
Weidberg et al.
Addictive Behaviors Research Group (GCA), Department of Psychology University of Oviedo Oviedo Spain
Neuroscience Institute of Principado de Asturias (INEUROPA), Faculty of Psychology University of Oviedo Oviedo Spain
Faculty of Health International University of La Rioja (UNIR) Logroño Spain
* CorrespondenceClara Iza‐Fernández, Addictive Behaviors Research Group (GCA), Department of Psychology, University of Oviedo, Plaza Feijoo s/n 33003, Oviedo, Spain.
Email: izaclara@uniovi.es
Abstract
Background and aims
The global cannabidiol (CBD) market has expanded in recent years. Several studies suggest CBD use, whether for medical or recreational purposes, is on the rise; however, no systematic assessments of its prevalence have been conducted so far. This meta‐analysis aimed to estimate the prevalence of CBD use in North America and Europe independently.
Methods
Searches were conducted in PubMed, PsycINFO and Web of Science on 25 March 2025. A total of 43 studies (48 distinct samples; n = 388 447; 57.52% female) from North America (k = 30; n = 353 088) and Europe (k = 13; n = 35 359) were included in the analyses. The prevalence was estimated at five time periods (i.e. lifetime, past 12 months, past 30 days, past 7 days and daily use) separately for each continent region. Differences in prevalence estimates by sex, year of data collection and sample type (i.e. clinical versus community) were examined. Publication bias was estimated using several indicators, and the methodological quality of the studies was appraised using the Joana Brigg's Institute (JBI) checklist.
Results
In preliminary analyses of heterogeneity there was evidence of differences in prevalence between North America and Europe across all time periods (P‐value range: < 0.001–0.057). Consequently, the pooled prevalence estimates were calculated for both continental regions separately. In Europe, the pooled lifetime prevalence of CBD use was 12.8% [95% confidence interval (CI) = 0.06–0.25], 17.6% (95% CI = 0.11–0.28) in the past 12 months, 7.2% (95% CI = 0.02–0.21) in the past month, 4.3% (95% CI = 0.01–0.13) in the past week and 2.1% (95% CI = 0.01–0.08) daily. In North America, lifetime pooled estimates were 28.9% (95% CI = 0.2–0.39), 19.5% in the past 12 months (95% CI = 0.11–0.32), 12% in the past month (95% CI = 0.07–0.2), 10.5% in the past week (95% CI = 0.01–0.47) and 6.4% daily (95% CI = 0.03–0.13). Of the moderators tested, sample type statistically significantly influenced the meta‐analyses. In Europe, clinical samples (25.6%) reported higher past‐year prevalence of CBD use in comparison with community samples (11.6%), whereas in North America there was higher past‐year use of CBD among community samples (26.1%) compared with clinical samples (4.1%).
Conclusions
Cannabidiol use appears to be more prevalent in North America compared with Europe.
Article notes
Weidberg S , Iza‐Fernández C , Alemán‐Moussa L , Krotter A , González‐Roz A . The prevalence of cannabidiol (CBD) use in North America and Europe: A meta‐analysis. Addiction. 2026;121(7):1653–1670. 10.1111/add.70360 PMC1329108241724499
Footnote Group
INTRODUCTION
Cannabidiol (CBD) is a non‐psychoactive cannabis compound that, along with tetrahydrocannabinol (THC), constitutes the two most abundant cannabinoids present in the cannabis sativa plant [1]. The World Health Organization has placed CBD as a safe cannabis compound, highlighting that there is no evidence for any public health‐related problems associated with the use of CBD, both recreational and/or medical [2]. This assumed safety profile has been commercially exploited; it is estimated that there are over 1000 CBD products commercialized in several forms of administration (e.g. aqueous beverages, shampoos, oil extracts, vape pens and edibles) [3]. Between 2017 and 2021, the three most popular CBD companies in the United States (US) spent nearly $1 500 000 in CBD advertising [4]. The global CBD market is expected to grow at a compound annual rate of 15.8% from 2024 to 2030 to reach USD 22.05 billion by 2030 [5]. Such rising market share of CBD opposes the slight decline in per capita alcohol consumption in the United States [6]. In contrast, the broader cannabis market, although also growing, faces tighter regulation [7]. This comparison underscores CBD as a high‐growth market driven by evolving policy landscapes.
Despite its commercial proliferation as a cannabis compound, CBD use is currently a topic of both legal and scientific debate [8]. Reviews of empirical studies highlight the lack of sufficient evidence to support CBD safety and its positive impact on physical health [9, 10]. From a legal standpoint, in the European Union (EU) there is no homogeneous law on CBD use [11]. Indeed, the criminal or administrative response toward CBD use is under the responsibility of each EU Member State [12]. Nowadays, the only CBD product that is officially labelled as a medical one by the Food and Drug Administration and the European Medicines Agency is Epidiolex, which was approved to treat seizures of two rare forms of epilepsy that begin in childhood [13, 14].
Currently, the vast majority of CBD use relies on self‐medication, nourished by consumer's expectations of health benefits that are not supported by scientific evidence [15]. Nevertheless, frequent use of CBD products may carry certain long‐term effects, such as diarrhea, fatigue, somnolence and increased hepatic enzymes [16, 17]. Another concern of CBD use relates to the real content of cannabinoids in over‐the‐counter CBD products [18]. In accordance with European legislation [19], the generally accepted cannabinoid profile of hemp is 1.5% to 3% CBD and less than 0.2% THC. In native and produced hempseed, THC content can reach up to 0.3% and up to 0.75% in hempseed oil. CBD concentrations can exceed 10% in the United States, but THC concentrations cannot be higher than 0.3% on a dry‐weight basis [20]. Nevertheless, labels on CBD and product content are usually inaccurate or restricted at best and missing at worst [21, 22, 23]. Consequently, public pressure exists on pharmaceutical authorities to establish clear and concise regulatory frameworks for CBD products [23].
CBD consumption is often associated with other substance use, especially cannabis, but also with tobacco and e‐cigarette, among others [8]. In the United States, 63% of last year CBD consumers also reported cannabis use [24]. Nevertheless, CBD has proven to interact with THC by potentiating its pharmacological effects via CB1R‐dependent mechanism [25, 26]. Concerns also exist about other drug–drug interactions beyond cannabis, as CBD is metabolized by cytochrome P450 (CYP), with their interaction leading to potential changes in the co‐administered substance [8, 27]. This interaction is especially dangerous when it comes to prescription drugs with a narrow therapeutic index, as they are also metabolized by CYP, leading to potential adverse effects [28]. Adulteration of CBD products with synthetic cannabinoids [29] and contamination with heavy metals, residual solvents or pesticides have also been reported [30].
Previous studies assessing CBD prevalence have mainly used cross‐sectional designs. In non‐clinical samples, US estimates show a lifetime prevalence of 39.7% [31] and past year prevalence of 20.6% among adults [24]. In Europe, 11.4% of German adults reported lifetime over‐the‐counter CBD use [32], and 1.1% were current users [33]. Lifetime use was 16.4% in France [8] and past‐year use reached 18.5% among young Swiss men [34]. In the United Kingdom (UK), lifetime and current use were 18% and 8%, respectively [35]. Among clinical samples, lifetime prevalence reached 24% in patients with hip and knee osteoarthritis [36], 25.2% with spine complaints [37], 34.3% in substance use disorders [38] and 61.9% in fibromyalgia [39]. Some studies suggest higher CBD use among women [24], although findings are mixed [33]. This potential sex difference in CBD use is important because, in non‐clinical samples, females reach greater concentrations of CBD compared to males [40] and exhibit greater relative exposure to CBD metabolites in plasma over time [41], which ultimately could lead to harm to health.
To our best knowledge, there are no previous efforts aimed at synthetizing CBD prevalence use and no cross‐national studies comparing prevalence between countries have been conducted. This constitutes an important research gap, especially when considering the heterogeneity in the prevalence estimates provided and how they are treated, with some studies merging them (e.g. past and current use) [33] while others isolate them [35]. Moreover, because there is a lack of study comparisons, a systematic assessment of potential differences in prevalence estimates by socio‐demographic characteristics (i.e. sex) and sample type (clinical vs. community) is warranted.
Against this background, the aim of this study was to conduct a meta‐analysis of CBD prevalence in North America and Europe, as well as to describe potential differences in CBD prevalence when considering socio‐demographic correlates (sex), sample type (clinical and non‐clinical) and date of data collection (past 5 years vs. preceding period).
METHODS
This study was conducted in accordance with the Preferred Reporting Items for Systematic reviews and Meta‐Analyses literature search extension (PRISMA‐S) statement [42] and the study protocol was pre‐registered in the International Prospective Register of Systematic Reviews (PROSPERO) (ID: CRD420250655105).
Literature search procedure and eligibility criteria
The literature search was conducted using PubMed, PsycINFO and Web of Science since inception to 25 March 2025. The following search string was entered: (‘prevalence’ OR ‘epidemiology’) AND (‘CBD’ OR ‘cannabidiol’). Studies were included if they were published in English in a peer‐reviewed journal and provided data on the use at any time of products specifically marketed for their CBD content. Studies on over‐the‐counter CBD use and use under prescription were included. Studies were excluded when (1) they did not provide CBD prevalence; (2) they did not provide the specific period of consumption; (3) all sample participants used CBD; (4) CBD prevalence was reported in time periods other than lifetime, past 12 months, past 30 days, past 7 days or daily; and (5) they were conference proceedings.
Study selection and data collection
The screening process was independently conducted by two reviewers, who first examined the titles and abstracts and subsequently assessed the full‐text articles. In instances of disagreement between both reviewers, consensus was reached through consultation with a third reviewer. Inter‐rater reliability was evaluated using Cohen's κ coefficient [43], which was 0.60, indicating moderate inter‐rater reliability according to standard conventions [44]. The following variables were extracted: author(s), country of study, year of recruitment, study population (i.e. community or clinical sample), mean age and SD, age range, sample size, percentage of females and the prevalence data (i.e. lifetime, past 12 months, past 30 days, past 7 days and daily), type of CBD use (prescribed by a healthcare professional or bought over the counter without a prescription) and mode of administration [i.e. oral (edibles and beverages), sublingual (CBD oil/tinctures), topical (CBD cream, lotion, patch or oil applied to the skin) and other methods (e.g. dabbing waxes, crystals, shatter, concentrates, strips, lozenges, sprays, pills and capsules)] assessed across different timeframes (i.e. lifetime, past 12 months and past 30 days). Data extracted from included studies are available on reasonable request.
A total of 2638 records were initially identified. After removing duplicates, 1680 studies were screened, and 43 of them met the inclusion criteria (including 48 distinct samples). Of these, 13 studies took place in Europe (n = 35 359), whereas the remaining 30 were performed in North America (n = 353 088). The PRISMA flow‐chart for the review process is shown in Figure 1.
Statistical analysis
Analyses were performed using the Metafor package [45] in R, version 4.5.1. Preliminary analyses aimed at detecting potential outliers. To assess the influence of individual studies across time frames and continents, externally studentized residuals (i.e. residuals standardized by the variance estimated after removing each study from the model fit) were computed applying a Bonferroni correction (two‐tailed α = 0.05) to control for type I error. Residuals with a value greater than 1.96 are considered potential outliers [46]. Influence diagnostics were obtained using Cook's distance, a measure of the impact of deleting each study on the model parameters. Leave‐one‐out analyses were also performed, sequentially excluding one study at a time to assess the influence of each individual study on the corresponding pooled prevalence.
As a sensitivity analysis, meta‐analysis was first performed using untransformed proportions. However, because of potential outliers within the dataset, a logit transformation was then applied. This method is well‐suited for non‐extreme proportions, because it stabilizes the variance and approximates the normality of the effect size distribution [47]. We adopted a random‐effects approach using the restricted maximum likelihood estimator (REML) [48] and the Knapp‐Hartung adjustment [49] because of substantial heterogeneity across the studies in terms of CBD product (over‐the‐counter vs. prescribed), mode of administration (e.g. oral, topical, smoked, etc.) and population (adolescents, young adults, people with medical conditions, etc.). The pooled prevalence of CBD use and its corresponding 95% CI was calculated at five time periods: lifetime, past 12 months, past 30 days, past 7 days and daily. Cochran's Q, I2 and tau2 (τ2) were computed to characterize heterogeneity. I2 ≤25% suggests low heterogeneity, ~50% suggests moderate heterogeneity and ≥75% suggests high heterogeneity across studies [50]. Additionally, a 95% prediction interval (PI) was computed as an indicator of uncertainty in the summary effect if a new study were included.
Last, a set of moderation analyses using the logit‐transformed data were conducted. Meta‐regressions tested the moderating effect of sex (percentage of women) and year of data collection (2018–2025), whereas mixed effects analyses examined the moderating effects by sample type (clinical vs. community) and data collection period (past 5 years vs. preceding period).
Publication bias
The analysis of publication bias was conducted based on five different tests, including: (1) Egger's regression intercept to examine the asymmetry of the funnel plot with a value of the linear regression intercept close to zero indicating the absence of publication bias [51]; (2) the Begg and Mazumdar rank correlation test, which is based on Kendall's τ‐test [it calculates the rank order correlation (Kendall's τ b) between the effect size and the standard error] [52]; (3) the fail‐safe N‐test to determine the number of potentially missing studies that would significantly alter the estimates [53]; (4) Tweedie's trim and fill [54] to impute potentially missing studies and re‐compute the overall estimate; and (5) funnel plots adjusted using Tweedie's trim and fill procedure.
Methodological quality assessment
Risk of bias of the studies reviewed was appraised using the Checklist for Analytical Cross Sectional Studies by the Joanna Briggs Institute (JBI) freely available online (https://jbi.global/critical-appraisal-tools) [55]. This tool measures the methodological quality of a study by examining the description of the inclusion/exclusion criteria and the study's design, the measurement of the study's exposure and outcomes and the adequacy of the statistical analysis. It comprises eight items and response options include: ‘yes’, ‘no’, ‘unclear’, ‘not applicable’. Given the epidemiological nature of this meta‐analysis, items 5 (i.e. identification of confounding factors) and 6 (i.e. adjustment for confounding factors) of the scale were excluded from the methodological quality assessment. Because of this, total scores on this scale ranged between 0 and 6, the latter being indicative of a higher methodological quality. To better interpret the results, total scores were transformed into percentages. Ratings less than 50% were indicative of high risk of bias, 50% to 69% of moderate risk of bias and greater than 69% of low risk of bias. A two‐step process was undertaken to assess risk of bias whereby two reviewers conducted an independent evaluation of the 43 meta‐analyzed studies and came to a consensus if discrepancies arose. This process yielded an inter‐rater agreement of 100%.
RESULTS
Study characteristics
Table 1 summarizes the characteristics of the studies, comprising 388 447 individuals, with sample sizes ranging from 83 to 65 336 participants. Participants were predominantly female (57.5%) with a mean age of 31.3 years (SD = 9.1). Eleven studies (26%) included representative sample weights, whereas 32 (74%) did not.
| Study | Country | Year of recruitment | Sample type | Sample size (% females) | Age (M/SD) |
|---|---|---|---|---|---|
| Europe | |||||
| Kasper et al. [35] | UK | 2019–2020 | Community | 517 (0) | 25 (5) |
| Bertholet et al. [34] | Switzerland | 2019–2020 | Community | 5233 (0) | 28.2 (1.3) |
| Hua et al. [56] | UK | 2020 | Community | 83 (61.4) a | 37.41 (18.72) |
| Hua et al. [56] | UK | 2020 | Clinical | 83 (61.4) a | 37.39 (15.46) |
| Alayli et al. [33] b | Germany | 2020–2021 | Community | 4026 (52.2) | NR |
| Casanova et al. [57] b | France | 2021 | Community | 1969 (53.8) | 51.7 (18.5) |
| Geppert et al. [32] b | Germany | 2020 | Community | 1011 (50.5) | 49.4 (19.8) |
| Geppert et al. [32] b | Germany | 2021 | Community | 2000 (52.9) | 43.6 (16.2) |
| Hotham et al. [58] | UK | 2019–2020 | Community | 6672 (NR) | 13–14 c |
| Roser et al. [38] | Switzerland | 2019 | Clinical | 469 (27.5) | 40.3 (12.7) |
| Sund et al. [59] | UK | 2021 | Community | 2478 (53.9) a , c | 45 (35–57) d |
| Barré et al. [8] b | France | 2022 | Community | 3229 (51.5) | 47 (33–60) d |
| Duval et al. [60] | France | 2020–2021 | Clinical | 183 (29) | 40.6 (11.8) |
| Ragusa et al. [61] | France | 2023 | Clinical | 1041 (82.5) a | 55.2 (11.2) |
| Gonzalez‐Roz et al. [62] | Spain | 2023–2025 | Community | 6365 (55.5) | 19.4 (2.6) |
| North America | |||||
| Jehangir and Parkman[63] | USA | 2018 | Clinical | 197 (80.7) a | 44.9 (15.9) |
| Braley et al. [64] | USA | 2019–2020 | Clinical | 1217 (80.4) | 51.2 (12.3) |
| Wheeler et al. [31] | USA | 2018–2019 | Community | 340 (76.5) c | NR |
| Mathur et al. [65] | USA | 2018 | Clinical | 371 (91.4) | 49 (38–59) d |
| Boehnke et al. [39] | USA | 2020 | Clinical | 2701 (95) a | 56.5 (12) |
| Carrubba et al. [66] | USA | 2019 | Clinical | 96 (100) | 44.1 (13.9) |
| Deckey et al. [36] | USA | NR | Clinical | 200 (56) | 67 (21) |
| Lovecchio et al. [37] | USA | 2019 | Clinical | 214 (42) | 58 (14.5) |
| Yu et al. [67] | USA | 2019–2020 | Clinical | 109 (62) | NR |
| Deckey et al. [68] | USA | 2020–2021 | Clinical | 823 (45.4) | 50.9 (16.2) |
| Dunbar et al. [69] | USA | 2019–2020 | Community | 2534 (54.2) a | 22.6 (0.8) |
| Gaiha et al. [70] | USA | 2021 | Community | 6131 (56.3) a , c | 21.9 (6.8) |
| Goodman et al. [71] b | Canada | 2019 | Community | 15 042 (49.9) | 40.1 (14.7) |
| Goodman et al. [71] b | USA | 2019 | Community | 30 288 (50.2) | 40.8 (14.7) |
| Wysota et al. [72] | USA | 2020 | Community | 2464 (57.4) | 24.7 (4.7) |
| Nguyen et al. [73] | USA | 2021 | Community | 1158 (50.5) | 43 (32–62) d |
| Brasky et al. [74] | USA | 2021–2022 | Clinical | 934 (55.2) c | 63 (54–70) d |
| Dai et al. [75] b | USA | 2022 | Community | 28 291 (48.9) | 14.5 (2) |
| De Genna et al. [76] | USA | 2019–2021 | Community | 186 (100) | 19.7 (1.7) |
| Egan and Cox [77] | USA | 2021 | Community | 1042 (NR) | 18–64e |
| Liu et al. [78]b | USA | 2022 | Community | 28 291 (48.9) | NR |
| McClure et al. [79] | USA | 2019 | Clinical | 1388 (50.5) | NR |
| Osaghae et al. [80] | USA | 2021–2022 | Clinical | 1886 (56.3) | NR |
| Wilson‐Poe et al. [81]b | USA | 2023 | Community | 1142 (51.7) | 48 (33–63)d |
| Seltzer et al. [82] | USA | NR | Clinical | 119 (38.7) | NR |
| Bhatia et al. [83] | Canada and USA | 2019–2021 | Community | 1096 (100) | 16–65e |
| Bhatia et al. [83] | Canada and USA | 2019–2021 | Community | 65 336 (100) | 16–65e |
| Choi et al. [24]b | USA | 2022 | Community | 47 100 (51.5) | NR |
| Hill et al. [84]b | USA | 2022 | Community | 47 100 (51.2)a | NR |
| Hill et al. [84]b | USA | 2022 | Community | 357 (41.7)a | NR |
| McCauley et al. [85] | USA | 2021 | Community | 5779 (59.7) | 13–40e |
| Tillman et al. [86] | USA | 2024 | Clinical | 96 (NR) | NR |
| Park [87]b | USA | 2022 | Community | 59 060 (NR) | NR |
Most of the studies were carried out in North America (30/43; 69.8%), with all of them having participants from the United States (30/43; 69.8%). Two of these studies (2/43; 4.7%) included participants from both the United States and Canada. The remaining studies were conducted in Europe (13/43; 30.2%), specifically in the United Kingdom (4/43; 9.3%), France (4/43; 9.3%), Germany (2/43; 4.7%), Switzerland (2/43; 4.7%) and Spain (1/43; 2.3%).
A total of 62.8% (27/43) of the studies included community samples, whereas 39.5% (17/43) involved clinical samples. In particular, 13 of 43 studies (35.1%) included samples with medical health conditions, specifically joint problems (3/43; 7%), cancer (3/43; 7%), chronic pain diseases (2/43; 4.7%), autoimmune problems (2/43; 4.7%), cerebral palsy (1/43; 2.3%), gastric diseases (1/43; 2.3%) and cystic fibrosis (1/43; 2.3%). A total of four of 43 (9.3%) included individuals diagnosed with substance use disorders (2/43; 4.7%), fibromyalgia (1/43; 2.3%) and autism spectrum disorder (1/43; 2.3%). One study [56] included both community and clinical samples and another [84] two different community samples.
Only seven of 43 (16.3%) studies provided data on whether CBD was obtained over the counter or by medical prescription. Of these, four (9.3%) collected data specifically on commercial CBD products, whereas the remaining three (7%) differentiated between prescribed and over‐the‐counter CBD use.
A total of 20 of 43 (46.5%) studies provided data on lifetime CBD mode of administration. The most common form of administration was sublingual (e.g. oil tincture) (61.4%), followed by vaping (52.8%), smoking (43%), oral (e.g. edibles and drinkables) (33%), topical (31.9%) and other methods (13.7%). Among the 18.6% (8/43) studies reporting past‐year prevalence, smoking (56.9%) was the most frequently reported route, followed by sublingual (41.9%), topical (31.8%), oral (27.5%), vaping (26.2%) and other methods (15.5%). Of the 9.3% (4/43) studies reporting past‐month prevalence, vaping (78.4%) was chosen as the most common route, followed by oral (66%), topical (42.8%), sublingual (33.7%), other methods (15.4%) and smoking (6.9%). None of the studies reviewed providing data on weekly and daily use informed about mode administration.
Publication bias
All except two estimators of publication bias (see Table 4) suggested the absence of publication bias. Specifically, the trim and fill procedure suggested one potential missing study in the meta‐analysis of European past‐year prevalence of CBD and four in the lifetime estimate of North America. Funnel plots suggested minimal publication bias (Figure S1).
| Model | Egger's test intercept (P value) | Begg and Mazumdar rank correlation—τ (P value) | Fail‐safe n | Duval and Tweedie's trim and fill |
|---|---|---|---|---|
| Europe | ||||
| Lifetime | −1.033 (0.395) | −0.111 (0.728) | 41 | 0 |
| Past year | −1.806 (0.480) | 0.143 (0.720) | 35 | 1 |
| Past month | −1.683 (0.393) | −0.2 (0.719) | 15 | 0 |
| Past week | −3.127 (0.987) | −0.067 (1) | 22 | 0 |
| Daily | −2.336 (0.481) | −0.067 (1) | 28 | 0 |
| North America | ||||
| Lifetime | −0.523 (0.234) | −0.232 (0.165) | 52 | 4 |
| Past year | −0.832 (0.066) | −0.091 (0.761) | 36 | 0 |
| Past month | −1.864 (0.671) | 0.007 (0.967) | 154 | 0 |
| Past week | −2.541 (0.648) | 0.333 (0.750) | 1 | 0 |
| Daily | −2.964 (0.607) | 0.111 (0.728) | 84 | 0 |
Methodological quality assessment of the reviewed studies
Table 5 depicts the results from the methodological quality assessment conducted with the JBI Checklist for Analytical Cross Sectional Studies. On average, studies scored 82.2% on the checklist. Of the 43 studies reviewed, nine studies scored six of six (100%) and 21 studies scored five of six (83.3%). These were classified as low risk of bias. Moderate risk of bias was attributed to 11 studies scoring four of six (66.7%). The lowest scoring was three of six (50%), which occurred in three cases and was indicative of high risk of bias. Most studies were appraised negatively on item 1 (‘Were the criteria for inclusion in the sample clearly defined?’), with 29 of 43 studies being rated as ‘no’. Additionally, a substantial number of studies (15/43; 34.88%) also scored negatively on item 6 (‘Was appropriate statistical analysis used?’).
| Study | Items | Total score (Y) | Percentage (Y) | Risk of bias | |||||
|---|---|---|---|---|---|---|---|---|---|
| 1 | 2 | 3 | 4 | 5 | 6 | ||||
| Alayli et al. [33] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| Barré et al. [8] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| Bertholet et al. [34] | Y | Y | Y | Y | Y | N | 5/6 | 83.3 | Low |
| Bhatia et al. [83] | N | Y | Y | Y | Y | N | 4/6 | 66.7 | Moderate |
| Boehnke et al. [39] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| Braley et al. [64] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| Brasky et al. [74] | Y | Y | Y | Y | Y | Y | 6/6 | 100 | Low |
| Carrubba et al. [66] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| Casanova et al. [57] | Y | Y | Y | Y | Y | Y | 6/6 | 100 | Low |
| Choi et al. [24] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| Dai et al. [75] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| De Genna et al. [76] | Y | Y | Y | Y | Y | N | 4/6 | 66.7 | Moderate |
| Deckey et al. [36] | N | N | Y | Y | Y | Y | 4/6 | 66.7 | Moderate |
| Deckey et al. [68] | Y | N | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| Dunbar et al. [69] | N | Y | Y | Y | Y | N | 4/6 | 66.7 | Moderate |
| Duval et al. [60] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| Egan and Cox [77] | Y | Y | Y | Y | Y | N | 5/6 | 83.3 | Low |
| Gaiha et al. [70] | N | Y | Y | Y | Y | N | 4/6 | 66.7 | Moderate |
| Geppert et al. [32] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| Gonzalez‐Roz et al. [62] | N | Y | Y | Y | Y | N | 4/6 | 66.7 | Moderate |
| Goodman et al. [71] | Y | Y | Y | Y | Y | Y | 6/6 | 100 | Low |
| Hill et al. [84] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| Hotham et al. [58] | Y | Y | Y | Y | Y | N | 5/6 | 83.3 | Low |
| Hua et al. [56], sample 1 (community) | N | Y | Y | Y | Y | N | 4/6 | 66.7 | Moderate |
| Hua et al. [56], sample 2 (clinical) | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| Jehangir and Parkman [63] | Y | Y | Y | Y | Y | Y | 6/6 | 100 | Low |
| Kasper et al. [35] | N | Y | Y | Y | Y | N | 4/6 | 66.7 | Moderate |
| Liu et al. [78] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| Lovecchio et al. [37] | Y | Y | Y | Y | Y | Y | 6/6 | 100 | Low |
| Mathur et al. [65] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| McCauley et al. [85] | Y | N | Y | Y | Y | N | 4/6 | 66.7 | Moderate |
| McClure et al. [79] | Y | Y | Y | Y | Y | Y | 6/6 | 100 | Low |
| Nguyen et al. [73] | N | N | Y | Y | Y | N | 3/6 | 50 | High |
| Osaghae et al. [80] | Y | Y | Y | Y | Y | Y | 6/6 | 100 | Low |
| Park [87] | Y | Y | Y | Y | Y | Y | 6/6 | 100 | Low |
| Ragusa et al. [61] | Y | Y | Y | Y | Y | Y | 6/6 | 100 | Low |
| Roser et al. [38] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| Seltzer et al. [82] | N | N | Y | Y | Y | Y | 4/6 | 66.7 | Moderate |
| Sund et al. [59] | N | N | Y | Y | Y | N | 3/6 | 50 | High |
| Tillman et al. [86] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| Wheeler et al. [31] | N | N | Y | Y | Y | N | 3/6 | 50 | High |
| Wilson‐Poe et al. [81] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
| Wysota et al. [72] | N | Y | Y | Y | Y | N | 4/6 | 66.7 | Moderate |
| Yu et al. [67] | N | Y | Y | Y | Y | Y | 5/6 | 83.3 | Low |
DISCUSSION
This meta‐analysis is the first to estimate CBD prevalence in North America and Europe. Rates were consistently higher in North America, with the largest gap in daily use (6.4% vs. 2.1%). Although differences narrowed over longer time frames, they remained notable. Past‐week and lifetime use were over twice as high in North America (10.5% vs. 4.3%; 28.9% vs. 12.6%), and past‐month use was nearly double (12% vs. 7.2%). Sex and data collection year had no effect on the prevalence estimates, while sample type (clinical vs. community) moderated several estimates, with higher past‐year use being observed among community samples in North America and among clinical samples in Europe.
Our findings align with research showing that the proliferation of CBD is partly driven by increasing marketing campaigns, predominantly launched by the cannabis industry [88, 89]. Regardless of the media channel (e.g. print, online, social media), the most prominently marketed theme typically focuses on therapeutic health claims (e.g. pain reduction, anxiety, sleep disorders and stress) [88, 90]. It is likely that users' perceptions of CBD's therapeutic benefits exceed its actual clinical effectiveness for many of the conditions for which it is used [71, 91]. Compared to Europe, North America offers a wider range of marketed CBD products, increasing its availability—particularly because CBD edibles (e.g. gummies, biscuits and chocolate) cannot legally be marketed in Europe [92]. In fact, edibles have shown to have a lower risk perception than cannabis smoking, especially when socially consumed [93], which could account for the prevalence difference between the two continents. In the United States and Canada, edibles increased their overall market share from 10.6% in 2019 to 11.1% in 2020 [94]. Furthermore, North American cannabis companies are investing in the development of faster‐acting edibles, which may arguably be more attractive to potential users [95].
Another empirical account for the differences observed in prevalence rates by continental region may be related to the THC content in the finished CBD products. Most cannabis users prefer to consume THC‐containing products in CBD form [96], which suggests products with higher THC concentrations may be more appealing. Most countries restrict the THC limit based on the content of plant material from which the product is derived (for most European countries, this limit ranges between 0.2% and 0.3%, and in the United States and Canada is 0.3% of dry weight) [19, 20]. Nevertheless, this legal limit offers no clear guidelines on the THC concentrations for finished CBD products, once the plant biomass has been removed [97]. Beyond this general legal framework, legislation of hemp‐derived products in the US relies on the individual states [98], leading to an inconsistent outlook of laws that generates confusion for both consumers and producers. Emerging evidence suggests that US local and national brands of hemp‐derived oil products are taking advantage of this situation. An analysis of 80 US unregulated CBD products revealed that 21 of them labelled as THC‐free are not, and 65% of these products had THC levels above the legal limit [97]. A similar study found that 12 of 14 European CBD oils tested contained THC, but 93% of them were below the legal limit of 0.2% [99]. These results support the idea that CBD users, especially those from the United States, are at risk of unintended or overconsumption of THC, leading to adverse health effects (e.g. overdose, potential abuse and driving while intoxicated).
Across continental regions, sublingual mode of administration was the most prevalent throughout lifetime, whereas CBD inhalation (i.e. smoking, vaping) was the most common form of administration when shorter prevalence points (past 12 months and past 30 days) were considered. This supports the notion that, although generally advertised as a health and well‐being product, CBD recreational marketing is booming and that, even when it may be medically prescribed, it is usually self‐prescribed for conditions with limited or no evidence of efficacy [15, 71, 100].
Although CBD use was generally comparable between clinical and community samples across most prevalence estimates, it was higher for past‐year use among European clinical samples and North American community samples. The more restrictive regulation of cannabis, including medical products, in Europe compared to North America [20, 101, 102] may lead European consumers to shift from cannabis to CBD to alleviate their clinical conditions, which could arguably account for this finding. In particular, the higher prevalence rates in clinical samples are concerning, because they suggest consumers have poor awareness of the health risks associated with CBD, including its adverse effects when interacting with other psychoactive substances and/or medications [28]. Corroon and Phillips [103] showed that most users were learning about CBD from the internet and other informal sources (i.e. friends or family members), whereas less than one in 10 received information from their physician. When self‐prescribing CBD for mental health problems, such as alleviating depressive symptoms, only 49% of users had informed their mental health professional about it and less than one in five had consulted their physician or pharmacist before doing so. Because CBD containing products, especially vaped or smoked, are largely unregulated and often mislabeled [104], there is no precise way for users to know the exact concentration of CBD, or other constituents, which may be present in the final product.
There were no significant differences in the CBD prevalence by sex. This result contrasts with other studies reporting sex differences in patterns and motivations for cannabis use, including the use for therapeutic purposes [105], with men consistently showing higher rates of cannabis use among both clinical [106] and community samples [107]. The absence of sex differences in CBD prevalence might be accounted by powerful marketing campaigns, especially in terms of online advertising of cannabis products. Social media exposure and promotion by influencers have shown to influence the purchase of cannabis products [108, 109]. Cannabis influencers are changing the stereotypical features of illegal cannabis culture traditionally dominated by men, to one where cannabis products are mainstream and acceptable for women and viewed as a desirable accessory in certain female lifestyles [110]. Nevertheless, a recent study showed sex differences in the pharmacokinetics (i.e. drug bioavailability, clearance) of CBD, leading to a lower required dose to reach a certain drug level and clinical effect in females [111].
The results of this meta‐analysis should be interpreted considering certain limitations. First, estimates should be viewed with caution because of the limited number of high‐quality epidemiological surveys available. Only over 26% included representative sample weights and nearly one‐third of the studies (32.6%) were classified as having either moderate (25.6%) or high (7%) risk of bias, so the pooled prevalence estimates should be interpreted carefully. Second, this study did not differentiate between prescribed and over‐the‐counter CBD products. Unfortunately, only seven of 43 studies provided such data, and a meta‐regression could not be computed. Third, generalizability to other continental regions is not warranted, as North American studies, especially those from the United States, are overrepresented (nearly 70% of the ones included). Moreover, although relevant prevalence moderators were considered, others were not, such as age group, the legal normative in each country or other substance use, including THC.
In conclusion, this is the first study to show that CBD consumption is highly prevalent. As such, several recommendations relevant to the legislative framework for CBD can be articulated. Policy makers should ensure label accuracy and establish clear regulatory guidelines for THC concentrations in CBD products, because discrepancies often exist between source plant material and final products [97]. A harmonized international legal framework is also needed given that many CBD products are purchased online [31, 112]. European consumers can easily access mislabeled North American products, especially regarding THC concentration [97]. Legal gaps in the promotion, packaging and marketing of CBD, especially those appealing to minors, must also be addressed [75, 89]. From a prevention perspective, CBD packaging could adopt plain labeling with warnings about CBD/THC content, adverse effects and evidence‐based health information [62, 113]. Mirroring universal prevention strategies to address problematic cannabis use [114], working on myths and education on CBD contrasted evidence (e.g. differentiating between medical cannabis preparations that have a proven therapeutic effect and recreational CBD products that lack evidence of efficacy) is encouraged. Additionally, rising awareness of potential risks, such as drug interactions, and misleading marketing claims could help consumers make more informed decisions about CBD use. Last, the present results highlight the need for large epidemiological studies (e.g. the National Survey on Drug Use and Health and the ESTUDES/EDADES surveys in Europe) to include potential moderators of CBD use, such as frequency of consumption, motives or awareness of health risks, which could be incorporated into future meta‐analyses.
DECLARATION OF INTERESTS
None.
Supporting information
ACKNOWLEDGEMENTS
None.
DATA AVAILABILITY STATEMENT
Data extracted from included studies are available upon reasonable request.