Development and Initial Psychometric Properties of the Cannabidiol Outcome Expectancies Questionnaire (CBD-OEQ)

Abstract

Cannabidiol (CBD), a nonpsychoactive cannabinoid, is used by many individuals to treat medical and mental health conditions, despite limited support for the efficacy of CBD for these conditions. Identification of CBD-related outcome expectancies (i.e., beliefs concerning the anticipated effects of CBD) could be useful in understanding the etiology and maintenance of CBD use and/or be useful in administration or clinical trial research. Although there are several measures of cannabis outcome expectancies, cannabis comprises several active compounds (e.g., tetrahydrocannabinol [THC], CBD). Thus, cannabis outcome expectancies may not reflect CBD-specific outcome expectancies. Yet, no known CBD-specific outcome expectancy measure exists. The present study used a three-phase, mixed-methods approach to develop and test the psychometric properties of the Cannabidiol Outcome Expectancy Questionnaire (CBD-OEQ). The CBD-OEQ assessed endorsement (i.e., how much an individual agrees/disagrees with an expected outcome) and desirability ratings (i.e., how desirable an expected outcome is). The initial item pool was administered to 600 adults who endorsed having heard of or using CBD products. Factor analyses supported a 60-item, six-factor structure. There was an initial support for internal consistency and convergent, discriminant, and incremental validity of the CBD-OEQ subscale scores in the present sample. Desirability ratings explained minimal additional variance in CBD variables for most subscales, but moderated the relationship between endorsement ratings and use behaviors for Global Negative Effects and No Effect subscales. The newly developed CBD-OEQ could be used as both a research and a clinical tool.

In recent years, there has been a rapid increase in public, scientific, and regulatory interest in cannabidiol (CBD), a naturally occurring cannabinoid with purported therapeutic effects (Rong et al., 2017). In a nationally representative sample of U.S. adults, approximately one in seven adults reported using CBD to treat specific medical or mental health conditions (e.g., sleep, anxiety, pain; Brenan, 2019). An estimated 40% of adults who have never used CBD express interest in trying it (Giandelone & Luce, 2019), and internet searches for CBD products have substantially increased since 2014 (Leas et al., 2019). Present human subjects research on CBD is limited, with mixed findings on therapeutic benefits (White, 2019). In response, the Food and Drug Administration is calling for more research on CBD to better understand CBD and to inform public health decisions (Hahn & Abernethy, 2019).

The National Institutes of Health’s funding for CBD research has more than tripled in the past 6 years (National Institutes of Health, 2021). Yet, present research on psychosocial factors related to CBD use is limited to a handful of studies that focus on demographic factors and reasons for use. People use CBD for pain, anxiety, relaxation, stress relief, and sleep (Corroon & Phillips, 2018; Wheeler et al., 2020), with high levels of perceived efficacy (Corroon & Phillips, 2018; Moltke & Hindocha, 2021). Some use CBD to replace other substances (e.g., cannabis; Wheeler et al., 2020) or prescription medications for medical and mental health conditions (Corroon & Phillips, 2018; Gill, 2019). Notably, among individuals who reported past 90-day use CBD-only products, tetrahydrocannabinol (THC)-only products, and/or CBD–THC combined products, those who used CBD-dominant products (i.e., containing mostly CBD and minimal to no THC) exclusively or more frequently than THC-dominant products (i.e., containing mostly THC and minimal to no CBD) reported using cannabis less often, in lower amounts, and for greater self-reported medical (rather than recreational) purposes than those who used more THC-dominant products (Fedorova et al., 2021). These findings suggest that those who prefer CBD products may have specific beliefs and expectations about what CBD products do.

Outcome expectancies refer to beliefs about the positive and negative effects of using a substance (e.g., alcohol, cannabis, tobacco; Goldman, 1994; Patel & Fromme, 2015). Outcome expectancy and expectancy–value theories are rooted in social learning perspectives, and posit that individuals engage in a certain behavior, in part, because they have expectations that particular reinforcing effects will occur as a result of use; these outcomes vary in desirability or value (Patel & Fromme, 2015). When applied to substance use behaviors, expectancy theory posits that substance use behaviors are influenced by an individual’s specific substance use outcome expectations (Jones et al., 2001). Although most of the research on substance-related outcome expectancies thus far has focused on alcohol and tobacco, there is a small but growing literature on other substance expectancies, including cannabis. Outcome expectancies can be positive (e.g., “Smoking marijuana makes me less tense or relieves anxiety”) or negative (e.g., “Marijuana can cause me to become depressed and disappointed with myself”; Schafer & Brown, 1991). Endorsement (i.e., how much an individual agrees/disagrees with an expected outcome), desirability (i.e., how desirable an expected outcome is), and the interaction between the two (i.e., expectancy–value interaction) are associated with substance use initiation, use frequency, and use-related problems (Patel & Fromme, 2015). Outcome expectancies can be measured regardless of use status, which is important for understanding potential risk and protective outcome expectancies related to use versus abstinence.

In the only known study of CBD-related expectancies, Spinella et al. (2021) conducted a randomized two-session cross-over study assessing whether expecting to receive CBD influenced self-reported stress, anxiety, and mood during a laboratory-based stress test and whether beliefs about CBD products (assessed prior to the administration session) moderated the relationship between expectancy condition and self-report outcomes. Beliefs about CBD were assessed via three items that asked participants to rate the extent to which they believed CBD “improves mood,” “reduces anxiety,” and “reduces stress” from 1 (not at all) to 10 (completely). Although participants received CBD-free oil in both sessions, individuals who endorsed higher baseline anxiety reduction expectancies reported significantly less subjective anxiety when led to expect CBD oil versus CBD-free oil. Importantly, this was not the case for individuals with low or moderate anxiety reduction expectancies, suggesting that baseline outcome expectancies may play a key role in the purported anxiolytic effects of CBD. Findings from this study suggest that the identification and measurement of CBD-related outcome expectancies could provide the valuable insight into CBD use behaviors and have important implications for clinical research and intervention efforts. However, the study is somewhat limited by using single items to assess expectancies and not assessing desirability ratings of the expectancies, which is problematic because isolating expectancy from value ratings and their interaction is important for understanding outcome expectancies. It may also be important to assess and understand other expectancies (e.g., sleep, pain) that are beyond the scope of the Spinella et al. (2021) study. Yet, no known CBD-specific outcome expectancy measure exists.

Thus, the present study used a three-phase, mixed-methods approach to develop and test the initial psychometric properties of the Cannabidiol Outcome Expectancy Questionnaire (CBD-OEQ), a self-report measure designed to assess endorsement (i.e., how much an individual agrees/disagrees an expected outcome) and desirability (i.e., how desirable an expected outcome is) of CBD-specific outcome expectancies. In Phase I, interviews were conducted with 21 individuals who had heard of and/or used CBD products to inform measure development. Phase II consisted of item selection and initial questionnaire development via additional literature review and a review panel of experts and laypersons. In Phase III, the 133-item CBD-OEQ was administered to a sample of 600 individuals. First, factor structure was determined using exploratory and confirmatory factor analyses (EFA and CFA) using randomly split-halves of the sample. Second, we tested initial psychometric properties (convergent, discriminant, concurrent, incremental validity) of the CBD-OEQ subscale scores in the present sample.

In line with other substance outcome expectancies work (Aarons et al., 2001; Schafer & Brown, 1991; Waddell et al., 2021), and to establish initial construct validity, we hypothesized that factor analyses would identify positive and negative factors and that positive endorsement subscales (i.e., how much an individual agrees or disagrees with a positive expected outcome) would be positively associated with CBD-related use behaviors, whereas negative subscales would be negatively associated with CBD-related use behaviors. We also hypothesized that CBD-OEQ endorsement subscales would explain additional variance in CBD-related outcomes above and beyond an existing cannabis outcome expectancy measure. Consistent with prior work with other substance outcome expectancies showing increased predictive utility when including valuation subscales (e.g., Buckner et al., 2013; Copeland & Brandon, 2002), we hypothesized that CBD-OEQ desirability subscales would be significantly, positively associated with CBD-related outcomes and explain unique additional variance above and beyond endorsement ratings. Finally, given that expectancy–value interactions do not appear to significantly predict cannabis use (Buckner et al., 2013), we hypothesized that the Endorsement × Desirability interaction would not significantly predict CBD use behaviors.

Phase I: Item Generation

The aim of the Phase I item generation stage was to generate all potentially relevant items through interviews to help ensure content validity (Clark & Watson, 2019). Consistent with methods used in the development of other substance-related outcome expectancy measures (Brown et al., 1980; Schafer & Brown, 1991), individual interviews were conducted to elicit expectations of the positive and negative effects of using CBD products among individuals with a wide range of exposure to CBD products (i.e., no use history to daily use).

Phase I: Method

Participants

Participants were recruited from a psychology research pool in the southern United States and via advertisements in the local community (e.g., flyers, Craigslist postings, email), including local CBD shops. Participants were eligible for the study if they were 18+ years of age and endorsed awareness of CBD products. Psychology students (n = 9) were compensated with research credits. Community-recruited participants (n = 12) were compensated $25. To ensure that we recruited individuals with a wide range of exposure to CBD products, we monitored the number of lifetime CBD use frequency responses per use category—that is, never, infrequent (i.e., less than 4 times in life to once to 8 times per year), monthly to weekly (once or twice a month to once to 4 times per week), and daily (almost every day to once or more every day). We aimed to recruit four to six individuals per use category.

Participants included 21 adults (61.9% female) aged 18–59 years (M = 28.43, SD = 13.37). The racial/ethnic composition of the sample was majority non-Hispanic/Latin White (90.5%). Lifetime CBD use frequency of the sample was as follows: never (n = 5), infrequent (n = 7), monthly to weekly (n = 4), and daily (n = 5). Those who reported lifetime CBD use (n = 16) endorsed using CBD products weekly or more (68.8%) and lifetime cannabis use (88.2%). The average age of first use of CBD was 24.0 years (SD = 10.3). Preferred methods of CBD consumption included concentrated oil (n = 8), topical (n = 8), tinctures (n = 7), pills/capsules (n = 5), edibles (n = 9), beauty products (n = 4), and others (e.g., CBD flower, dip pouches, cigarettes; n = 2). Among those who endorsed lifetime cannabis use, 68.7% endorsed using cannabis weekly or more in the past month. For past-month substance use, two individuals endorsed smoking, 38.1% endorsed electronic nicotine delivery systems (ENDS) use, and 76.2% endorsed alcohol use.

Procedure

Participants were first screened via Qualtrics, an online data collection website, for eligibility criteria and lifetime CBD use frequency. Ineligible participants were informed of their ineligibility and discontinued from the survey. Eligible participants were invited to attend one in-person appointment and scheduled an interview via an online study sign-up website. At the in-person appointment, participants provided written informed consent to participate in the study prior to data collection. The study protocol was approved by the University’s Institutional Review Board prior to data collection. Interviews and self-report measures (administered via Qualtrics) were counterbalanced to minimize response bias.

Trained undergraduate research assistants conducted the interviews by reading all interview questions directly from a script to reduce the likelihood of error. Adapted from prior work on cannabis-related outcome expectancies (Schafer & Brown, 1991), participants were asked “What effects would you expect from using CBD products?” and were asked to respond according to their own experiences with CBD or personal beliefs. Participants were asked to provide answers for positive and negative effects and to come up with as many outcome expectancies as possible. During the interview, research assistants took notes of participant responses and immediately following the interview, the interviewer entered participant responses into a spreadsheet. All interviews were audiotaped and if needed, the interviewer was able to listen to the audio recording to confirm responses. Research assistants who were not involved in data collection independently reviewed audiotapes and confirmed that there were no discrepancies in the interview questions or in data entered into the spreadsheet.

Measures

Screener Questions.

First, participants indicated their age via text response; individuals who responded 18 years or older were directed to the next set of questions. Awareness of CBD products was assessed using a one-item question (“Have you ever heard of or used any of the following products?”) and participants could select CBD products, hair products, skin care products, natural products, or essential oils. Individuals who did not select “CBD products” were screened out. Lifetime CBD use frequency was assessed using a question from the Cannabidiol Use Form (CBD-UF) developed for the present study, which was modified from the Marijuana Use Form (MUF; Buckner et al., 2007): “On average, how often have you used CBD products in your entire life? Please do not include use of marijuana or other marijuana products in your answer” on a 0 (never) to 6 (once or more every day) scale.

Laboratory Appointment Measures.

The CBD-UF assessed past 3-month (0 = none to 10 = 21 or more times a week) and past-month (0 = none to 10 = 21 or more times a week) CBD use frequency, preferred methods of CBD use, and age of first use. The MUF (Buckner et al., 2007) assessed lifetime cannabis use frequency (0 = never to 7 = once or more every day).

Past-month substance use frequency and quantity were assessed using the self-report version of the Timeline Follow-back (TLFB; Sobell & Sobell, 1992) on which important events (e.g., Halloween) and federal holidays (e.g., Fourth of July, Labor Day) were labeled. Per each day on the calendar, participants reported the number of standard alcoholic drinks consumed, “joints” (i.e., cannabis cigarettes as opposed to something cigar-sized or bigger) used, combustible cigarettes smoked, and the number of times they used an e-cigarette. TLFB scores have demonstrated acceptable validity and reliability for alcohol, cannabis, combustible smoking, and ENDS use in prior work (e.g., Carpenter et al., 2017; Litt et al., 2016; Robinson et al., 2014; Sobell & Sobell, 1992). A modified TLFB assessed CBD use by asking participants to report the number of times they used CBD products per day.

Data Processing and Analytic Strategy

Consistent with prior outcome expectancy questionnaire development (Brown et al., 1980; Schafer & Brown, 1991) and scale development best practice (Brod et al., 2009), we performed content analysis (i.e., a method of systematically coding and classifying qualitative data; Hsieh & Shannon, 2005) on individual statements taken from interviews to inform item generation (Hsieh & Shannon, 2005). First, the two primary reviewers independently read and reread the spreadsheet of participant outcome expectancy statements. Second, reviewers examined the outcome expectancy statements for common themes and grouped similar statements together. Third, reviewers worked conjointly to develop a codebook for the data using a standardized codebook and coding protocol (MacQueen et al., 1998). The codebook included a label, brief and full definitions, and examples for each outcome expectancy theme. Fourth, the codebook and outcome expectancies statements list were provided to two additional coders (i.e., undergraduate research assistants). Coders were instructed to independently choose an appropriate code for each outcome expectancy statement. Level of agreement between the two coders was quantified using kappa, a reliability statistic that corrects for chance agreement among coders. Interrater reliability was excellent (κ ≥ 0.99; McHugh, 2012). To minimize bias, primary reviewers and secondary coders did not conduct interviews. This study was not preregistered. Data and study materials are not available to other researchers.

Phase I: Results

Primary reviewers evaluated 238 outcome expectancy statements; there was some content overlap within the statements. For example, the majority of participants stated that they expected they would feel less anxious if they were to use CBD. Ten initial outcome expectancy themes were identified: positive physical effects (e.g., “helps muscle aches,” “pain relief,” “body high”), negative physical effects (e.g., “nausea,” “blurred vision”), sleep (e.g., “helps with sleep”), positive mood (“better mood,” “euphoria”), negative mood (e.g., “gets rid of negative feelings,” “helps with depression”), stress and anxiety management (e.g., “anxiety relief,” “stress relief”), positive mental benefits (e.g., “mental clarity”), broad negative effects on mental well-being (e.g., “trouble focusing,” “unproductive”), relaxation (e.g., “calms me down,” “chilled out”), and no effect (e.g., “no help,” “placebo effect may cause you to feel better”). The most common outcome expectancy theme statements included positive physical effects (57 statements), stress and anxiety management (33 statements), and relaxation (32 statements).

Phase II: Item Selection and Initial Questionnaire Development

The aims of Phase II were: (a) generate additional items via literature review, (b) organize items into categories that reflect the conceptualization and perception of the population of interest, which may or may not reflect factors derived from factor analysis in Phase III; and (c) have reviewers review the item pool for content and face validity and clarity and conciseness.

Phase II: Method

Procedure

Primary reviewers from Phase I and an expert in substance use expectancies (last author) discussed and refined the themes to create the initial item pool. Items were reworded into first-person in a Likert scale response format from 0 (disagree strongly) to 5 (agree strongly) for endorsement ratings and (0 = bad to 5 = good) for desirability ratings, as assessed in prior work (Buckner et al., 2013; Schafer & Brown, 1991). Instructions for endorsement and desirability CBD-OEQ ratings were modified from the Marijuana Effect Expectancy Questionnaire (Buckner et al., 2013; Schafer & Brown, 1991) by replacing the term “marijuana” with “CBD” (see Appendix in Supplemental Materials). The reviewer panel included one PhD-level professor, two graduate students in clinical psychology with expertise in substance use research, and two undergraduate research assistants. Reviewers were provided with a working definition for each outcome expectancy theme and a list of items for each theme. Reviewers rated the relevance of each item to its theme (0 = not at all relevant to 5 = extremely relevant) and were encouraged to provide feedback on additional potential items.

Phase II: Results

Two additional themes were identified through extant literature review: substance use reduction/cessation and social enhancement. Several themes were modified for clarity and specificity (e.g., the Phase I “positive mood” theme was modified to “desired/helpful mood” effects) and the initial “no effect” theme was split into two themes to ensure enough items were generated to capture both the placebo effect and no effect (i.e., expecting to not feel any different when using CBD) themes. In total, 13 themes and 138 initial items were generated from the Phase I interviews, literature review, and discussions with primary reviewers. On the basis of reviewer’s feedback, eight items with similar content were combined (e.g., items “CBD helps me manage my skin issues, like acne” and “CBD helps me manage my eczema” were combined to “CBD helps me manage my skin issues, like acne or eczema”), removing four items from the initial item pool. Additional items were modified for clarity and conciseness. Two items received mean relevance ratings less than 4; one was removed and the second modified to improve relevance. After these five items were removed, the final item pool contained 133 items.

Phase III: Examination of Factor Structure and Initial Test of Psychometric Properties

Phase III: Method

Participants and Procedures

Participants (N = 600) were recruited for a 30–60-min online study (administered using Qualtrics) about CBD via several recruitment sources, including Qualtrics Panel, a secure service operated by Qualtrics, Inc. (n = 312), the university psychology department research pool (n = 77), and through local and online sources via flyers, Craigslist postings, social media posts, and email blasts sent by CBD stores (n = 211). Data collected through Qualtrics Panel has been shown to be reliable and of similar or better quality than other panel source data or undergraduate samples (Heen et al., 2014; Kees et al., 2017); meta-analytic findings demonstrate that data from online panels approximates conventionally recruited data (Walter et al., 2019). Of note, Qualtrics Panel has been successfully used to recruit participants for studies on substance use behaviors (e.g., Morean & Lederman, 2019).

Participants were eligible for the study if they were 18+ years of age and endorsed awareness of CBD products. As in Phase I, participants completed a question on lifetime CBD use frequency and responses were used to monitor participants’ use patterns throughout recruitment to ensure the sample had a wide range of exposure to CBD products. Qualtrics Panel recruited participants via email based on their answers to the demographics survey that participants completed when they signed up to be a panelist. Participants who completed the survey via Qualtrics Panel were compensated independently through their panel (gift cards, cash, etc.) in a predetermined amount not to exceed $8. Participants who completed the survey through the psychology research pool were compensated with research credits. Community recruited participants were entered in a raffle for a chance to win one of eight $25 prizes.

Participant demographics appear in Table 1. The sample was 59.3% female, 74.4% non-Hispanic/Latin White, with a mean age of 37.35 years (SD = 17.86, range: 18–88). The majority of the sample endorsed using cannabis (68.3%) and alcohol (60.5%) in the past month. Lifetime CBD use frequency of the sample was as follows: never (n = 146), less than 4 times in life to once to 8 times per year (n = 156), once or twice a month to once to 4 times per week (n = 157), and almost every day to once or more every day (n = 141). Among individuals who reported lifetime CBD use, 65.6% reported using CBD products weekly or more; 75.6% endorsed lifetime cannabis use. Average age of first use of CBD was 29.5 years (SD = 15.7). Preferred methods of CBD consumption included tinctures/oils/concentrates/capsules (n = 128), edibles (n = 85), beauty/topical products (n = 71), and vaping/combustible methods (n = 63).

Table 1.

Phase I and Phase III Demographics and Characteristics and Differences Between Phase III Split-Half Samples

	Phase I Total (N = 21)	Phase III
Variable		Total (N = 600)	EFA (n = 300)	CFA (n = 300)	F or χ2	p	d or Cramer’s V
Recruitment source (% community)	57.1%	48.0%	47.7%	48.3%	0.03	.870	.01
Age	28.43 (13.37)	37.35 (17.86)	37.26 (18.0)	37.43 (18.0)	0.01	.911	.01
Sex assigned at birth (% female)	61.9%	59.3%	59.0%	59.7%	0.03	.868	.01
Race	—	—	—	—	2.40	.935	.06
White (%)	90.5%	81.7%	81.6%	81.9%	—	—	—
African American/Black (%)	9.5%	8.0%	8.4%	7.7%	—	—	—
Asian/Asian American (%)	—	2.7%	2.7%	2.7%	—	—	—
Asian Indian (%)	—	0.8%	1.0%	0.7%	—	—	—
American Indian/Alaskan Native (%)	—	1.7%	1.3%	2.0%	—	—	—
Native Hawaiian/Pacific Islander (%)	—	0.3%	0.3%	0.3%	—	—	—
Multiracial (%)	—	3.5%	4.0%	3.0%	—	—	—
Not listed (%)	—	1.2%	0.7%	1.7%	—	—	—
Ethnicity (% non-Hispanic/Latin)	100.0%	89.0%	90.0%	88.0%	0.61	.434	.03
CBD use intentionsa	5.38 (2.26)	3.50 (1.74)	3.36 (1.79)	3.62 (1.79)	1.02	.314	.15
Lifetime CBD frequency	0.81 (0.42)	2.55 (2.06)	2.53 (2.03)	2.57 (2.10)	0.05	.828	.02
Substance useb		—	—	—	—	—	—
Past 30-day CBD frequency	16.82 (10.97)	18.28 (10.67)	18.06 (10.96)	18.53 (10.34)	0.14	.707	.04
Past 30-day CBD quantity	17.18 (10.82)	35.94 (38.23)	36.32 (40.61)	35.47 (35.26)	0.03	.853	.05
Past 30-day cannabis frequency	10.67 (11.74)	15.18 (11.59)	15.07 (11.93)	15.30 (11.30)	0.02	.884	.02
Past 30-day cannabis quantity	17.58 (28.46)	28.23 (31.67)	27.46 (31.30)	29.00 (31.30)	0.12	.726	.05
Past 30-day alcohol frequency	5.63 (3.32)	10.0 (8.60)	9.79 (8.60)	10.16 (8.62)	0.17	.683	.04
Past 30-day alcohol quantity	19.11 (20.70)	22.25 (23.35)	22.67 (26.0)	21.81 (20.33)	0.12	.731	.04
Past 30-day smoking frequency	4.50 (3.54)	23.57 (9.94)	24.79 (9.49)	22.27 (10.33)	1.74	.190	.25
Past 30-day smoking quantity	8.50 (7.78)	189.31 (215.85)	201.61 (202.19)	176.08 (230.91)	0.38	.542	.12
Past 30-day ENDS frequency	18.50 (11.80)	18.68 (11.45)	18.85 (12.06)	18.41 (10.56)	0.30	.865	.04
Past 30-day ENDS quantity	105.25 (102.92)	114.61 (175.87)	136.52 (207.14)	79.00 (100.56)	2.15	.147	.35

Note. EFA = exploratory factor analyses; CFA = confirmatory factor analyses; CBD = cannabidiol; ENDS = electronic nicotine delivery systems.

^aCBD use intentions measure administered only to those who never used CBD products.

^bPast 30-day use variables only assessed among individuals who endorsed past-month use of substance listed (outliers excluded).

Measures

As in Phase I, questions from the CBD-UF and MUF assessed lifetime CBD and cannabis use frequency, respectively. The TLFB assessed past-month substance use frequency and quantity (alcohol, cannabis, combustible cigarettes, ENDS, CBD). Individuals who denied past-month use of a substance on screeners (e.g., “Have you used marijuana in the past-month?” 1 = yes, 0 = no) did not complete the past-month TLFB for that substance and were coded as “0” for TLFB frequency and quantity variables. TLFB CBD use frequency was strongly correlated with the single past-month CBD use item, r = .71, p < .001, similar to the magnitude found in cannabis research (r = .73; Norberg et al., 2012), and evinced moderate-to-large correlations with other TLFB frequency measures: cannabis (r = .52, p < .001), cigarettes (r = .40, p < .001), and ENDS (r = .29, p = .022).

Intentions to use CBD were assessed by asking participants who denied lifetime CBD use were asked to rate, “How likely is it that you will use CBD, even once or twice, over the next 12 months?” from 1 (I definitely will not) to 4 (I definitely will). Individuals with scores of two or greater were asked, “How likely is it that you will use CBD nearly every month for the next 12 months?” using the same scale. Scores of “1” (i.e., “I definitely will not”) from the first item were carried forward and coded as “1” for the second item. Items were modified from prior work on cannabis use intentions that demonstrated high predictive validity for later cannabis use (Skenderian et al., 2008). Items demonstrated acceptable internal consistency (α = .84).

The 48-item Marijuana Effect Expectancy Questionnaire (MEEQ; Schafer & Brown, 1991) was used to assess cannabis-specific outcome expectancies. Participants rate the level of agreement for each effect as a result of using cannabis from 0 (disagree strongly) to 5 (agree strongly). The present study used the MEEQ’s six lower order scale scores, which demonstrated acceptable internal consistency in prior work (Buckner & Schmidt, 2008), and in the present sample: Cognitive and Behavioral Impairment (α = .86), Relaxation and Tension Reduction (α = .86), Social and Sexual Facilitation (α = .72), Perceptual and Cognitive Enhancement (α = .75), Global Negative (α = .86), and Craving and Physical Effects (α = .77).

Three instructed response attention check questions (Meade & Craig, 2012) were included throughout the survey (e.g., “In order to show us that you are following the instructions, please choose ‘agree’ as your answer for this question”) to detect carelessly invalid responses.

Participants who answered all three attention check questions incorrectly (n = 48) were excluded and not compensated.

Data Analytic Strategy

Factor Analyses.

The present study used guidelines recommending a minimum EFA sample size of at least 300 participants (Field, 2013; Tabachnick & Fidell, 2013). Thus, the final sample of 600 participants was split randomly in half and differences were tested using Pearson’s chi-square for categorical data (e.g., sex assigned at birth, race, ethnicity) and independent t tests for continuous data (e.g., lifetime CBD use frequency). There were no significant differences across the split-half samples (Table 1). Factor analyses were conducted on CBD-OEQ endorsement ratings per prior work (Copeland & Brandon, 2002). EFA on the initial pool of 133 items was conducted using oblique rotation in R Studio’s psych package (Revelle, 2021) with the first randomly split-half sample (n = 300). Parallel analysis, a Monte Carlo method that compares eigenvalues of the sample data to randomly generated eigenvalues, was used to determine the number of factors to extract because it is considered the most accurate factor extraction method (Floyd & Widaman, 1995; Horn, 1965). Factors were retained if the sample data eigenvalues are greater than the randomly generated eigenvalues.

Items were retained if they had a primary factor loading greater than .40 and did not cross-load onto multiple factors (i.e., an item’s second-highest loading was not greater than .30; Field, 2013; Floyd & Widaman, 1995). Once simple structure was achieved, we conducted a reliability analysis for each retained factor and eliminated items if: (a) the item-total correlation was less than .30 or (b) if the Cronbach’s alpha if deleted value was greater than the overall alpha (Field, 2013). As scale development experts recommend including a broader and more comprehensive range of items in the initial item pool, including items with similar wording and content (DeVellis, 2016), several redundant items were included initially. However, to reduce participant burden and improve parsimony, retained items with content or wording similarities were reviewed by Phase I reviewers and removed based on Cronbach’s alpha if deleted and factor loadings. EFA was repeated using oblique rotation to confirm factor structure results if items were removed. Retained items were used to conduct a CFA with the second split-half sample (n = 300) using the R Studio laavan package (Rosseel, 2012). Model fit was evaluated using several model of fit indices: (a) the chi-square statistic (if chi-square is nonsignificant, the model adequately represents the data; Browne & Cudeck, 1992), (b) root mean squared error of approximation (RMSEA; values of .01, .05, .08 indicate excellent, good, and mediocre fit, respectively; MacCallum et al., 1996), (c) comparative fit index (CFI ≥ .95 is considered acceptable; Hu & Bentler, 1999; Kline, 2015), and (d) standardized root mean square residual (SRMR value ≤ .08 is considered acceptable; Hu & Bentler, 1999; Kline, 2015).

Initial Psychometric Properties of Endorsement Ratings.

As in prior work (e.g., Richards et al., 2021), psychometric properties of the CBD-OEQ were tested among the full sample (N = 600). First, the internal consistency of the CBD-OEQ subscales was examined using Cronbach’s alpha (Cronbach, 1951). Second, convergent validity was tested using bivariate correlations with four outcome variables: intentions to use CBD (among those who had never used CBD), lifetime CBD use, and past-month CBD use frequency and quantity (among those who reported past-month CBD use). Third, we tested discriminant validity by comparing bivariate correlations between the CBD-OEQ subscales, past 30-day CBD use frequency and quantity, and past 30-day other substance use frequency and quantity (i.e., alcohol, cannabis, combustible smoking, ENDS) among all participants. William’s modification of the Hotelling test was calculated using R Studio’s cocor package (Diedenhofen & Musch, 2015), which tests for statistically significant differences between two nonindependent correlations (Williams, 1959). Discriminant validity was indicated by significant pairs of correlations (p < .05, one-tailed). Incremental validity of the CBD-OEQ was via a series of hierarchical linear regression models. Separate regressions were conducted for each of the four CBD-related outcome variables. MEEQ subscales were entered at Step 1, and CBD-OEQ subscales were entered at Step 2 to ensure that the effect at Step 2 cannot be attributed to shared variance with variables in Step 1 (Tabachnick & Fidell, 2007). All predictor variables were centered to reduce multicollinearity.

Utility of CBD-OEQ Desirability Ratings.

First, bivariate correlations between the CBD-OEQ desirability ratings, endorsement ratings, and CBD-related outcomes were conducted to assess convergent validity. Second, to test the incremental validity of desirability ratings, separate regressions were conducted for each pair of endorsement and desirability rating subscales and each of the four CBD-related outcomes, with endorsement entered in Step 1 and desirability entered in Step 2. Third, to test whether desirability ratings moderated the relationship between endorsement ratings and CBD-related outcomes, moderation analyses were conducted using the PROCESS macro for SPSS (Hayes, 2013). Separate models were constructed for each CBD outcome variable and CBD-OEQ subscale. The endorsement rating was entered as the predictor and the corresponding desirability rating as the moderator. Significant interactions were probed by testing low, mean, and high values of the moderator (i.e., 1 SD below the sample mean, mean, and 1 SD above the mean), and estimating the conditional effect of the focal predictor on the outcome at that specific value (Hayes, 2013).

Phase III: Results

EFA

Bivariate correlations between items were all less than .80. Bartlett’s test was significant and indicated correlation adequacy, χ²(8,778, N = 300) = 37342.12, p < .001. The Kaiser-Meyer-Olkin; KMO test (KMO = 0.95) was well above the minimum recommended statistic of 0.5 for sampling adequacy (Field, 2013). Parallel analysis suggested retaining seven factors. In the seven-factor model, the seventh factor did not have any items that loaded above 0.40 after removing 48 items with low factor loadings and/or cross-loadings. Thus, we also evaluated a six-factor model; EFA was repeated until the model achieved simple structure with each item loading onto only one factor, and all six-factors had at least three items with factor loadings above .40. To avoid confirmation bias, we also evaluated a five-factor model, however, items important for content validity (e.g., items related to “no effect” expectancies) were not able to be retained due to cross-loading or low primary factor loading. In the six-factor model, 85 items remained after removing items that did not meet criteria. Primary reviewers from Phase I removed 14 additional items because their wording and content were similar to other items with higher factor loadings. Factor reliability analyses indicated that Cronbach’s alpha if removed was higher or did not improve measure reliability if retained for seven items. After these items were removed, three additional items with low primary factor loading/cross-loading were removed. The final 60-item, six-factor EFA model indicated adequate model fit: RMSEA = 0.037, 90% CI [0.038, 0.044], SRMR = 0.03, CFI = 0.95, χ²(1,425, N = 300) = 2146.25, p < .001.

Factor loadings of the final pool of items are in Table 2. Factor 1 (“Positive Mood and Mental Health Effects”) included 19 items that assessed positive CBD expectancies for mood and mental health (α = .97). Factor 2 (“Global Negative”) included 17 items that measured negative CBD expectancies for mood, mental health, physical health, and other negative side effects (α = .95). Factor 3 (“Pain Relief”) included 10 items that assessed pain reduction expectancies (α = .95). Factor 4 (“Harm Reduction”) included six items measuring CBD expectancies associated with substance use reduction and cessation (α = .93). Factor 5 (“Sleep”) contained five items related to improved sleep (α = .92). Factor 6 (“No Effect”) was comprised of three items related to no effect expectancies (α = .82).

Table 2.

Exploratory Factor Analysis Results

Factor/item	Factor loading	M (SD)	α
Positive mood and mental health—19 items	—	66.21 (17.90)	.97
I feel more confident in social situations when I use CBD	0.89	3.41 (1.22)	—
CBD makes social situations more fun and enjoyable	0.79	3.19 (1.20)	—
CBD makes it easier for me to connect with others	0.79	3.38 (1.21)	—
I worry less when I use CBD	0.78	3.58 (1.15)	—
When I use CBD, negative feelings go away	0.77	3.26 (1.19)	—
CBD makes me feel less anxious in social situations	0.74	3.57 (1.21)	—
I have a good, happy feeling when I use CBD	0.72	3.67 (1.12)	—
CBD improves my mental health	0.69	3.49 (1.17)	—
CBD makes me feel less depressed	0.69	3.38 (1.20)	—
CBD gives me a strong sense of well-being	0.69	3.43 (1.20)	—
CBD gets rid of my anxiety	0.68	3.58 (1.21)	—
When I use CBD, I feel like I can cope with my problems	0.68	3.50 (1.17)	—
When I’ m going through a hard time, CBD helps me deal	0.66	3.39 (1.16)	—
My emotions are easier to control when I use CBD	0.66	3.40 (1.22)	—
I feel less paranoid when I use CBD	0.62	3.32 (1.20)	—
CBD makes my mood more stable	0.60	3.57 (1.15)	—
CBD calms me down and makes me feel more mellow	0.58	3.69 (1.12)	—
My mood is better when I use CBD	0.58	3.81 (1.09)	—
I feel less panicky after I use CBD	0.52	3.61 (1.15)	—
Global negative effects—17 items		40.93 (14.80)	.94
I have a hard time focusing when I use CBD	0.87	2.44 (1.23)	—
When I use CBD, I have difficulty remembering things	0.78	2.36 (1.20)	—
My body feels numb when I use CBD	0.76	2.44 (1.23)	—
I feel less sharp when I use CBD	0.76	2.63 (1.25)	—
CBD makes me feel too sleepy during the day	0.75	2.58 (1.21)	—
CBD causes me to feel dizzy or lightheaded	0.74	2.45 (1.26)	—
CBD makes my mental health worse	0.72	2.12 (1.20)	—
I feel unproductive and less motivated when I use CBD	0.72	2.30 (1.18)	—
When I use CBD, sometimes I feel “high” when I do not want to feel high	0.68	2.40 (1.27)	—
My vision gets blurry when I use CBD	0.68	2.16 (1.16)	—
When I use CBD, I feel nauseous	0.68	2.18 (1.16)	—
I am less alert and not as sensitive to my surroundings when I use CBD	0.67	2.61 (1.23)	—
CBD causes me to have a dry throat or cough	0.67	2.51 (1.25)	—
When I use CBD, I have a poor appetite and do not want to eat	0.60	2.28 (1.16)	—
CBD negatively interacts with my prescription medications	0.59	2.35 (1.17)	—
My mouth feels dry when I use CBD	0.54	2.70 (1.22)	—
CBD causes me to lose weight even though I did not want to	0.49	2.46 (1.15)	—
Pain relief–10 items		37.25 (8.79)	.94
CBD helps relieve my pain	0.85	3.88 (1.02)	—
I have less muscle pain, soreness, or stiffness when I use CBD	0.83	3.79 (1.09)	—
CBD lessens my joint pain and arthritis pain	0.82	3.68 (1.12)	—
CBD helps relieve my back pain	0.79	3.71 (1.13)	—
Minor aches and pains go away when I use CBD	0.78	3.86 (1.03)	—
I have less pain when I take CBD	0.76	3.84 (1.06)	—
I can manage my pain better when I use CBD	0.76	3.80 (1.06)	—
CBD reduces my worst or most severe pain	0.71	3.59 (1.10)	—
I can move easier when I use CBD	0.55	3.49 (1.15)	—
CBD helps relieve my headaches or migraines	0.49	3.62 (1.14)	—
Harm reduction-6 items	—	18.58 (6.01)	.92
CBD makes it easier for me to cut back on alcohol, cigarettes, or other drugs	0.84	3.16 (1.18)	—
CBD helps me quit using other substances, like alcohol, drugs, or cigarettes	0.83	3.05 (1.21)	—
My cravings for alcohol, cigarettes, or other drugs are easier to deal with when I use CBD	0.74	3.14 (1.22)	—
When I use CBD, I have fewer withdrawal symptoms when I am trying to cut back on alcohol, cigarettes, or other drugs	0.73	3.09 (1.10)	—
CBD causes me to drink less alcohol, smoke fewer cigarettes, and/or use less drugs	0.73	3.10 (1.21)	—
CBD helps me stay sober from alcohol or other drugs	0.68	3.04 (1.22)	—
Sleep-5 items		18.72 (4.68)	.91
When I use CBD, I fall asleep faster	0.78	3.71 (1.11)	—
CBD improves the quality of my sleep	0.73	3.84 (1.08)	—
I sleep better when I use CBD	0.72	3.81 (1.07)	—
When I use CBD, I sleep through the night	0.72	3.75 (1.11)	—
Taking CBD before bed helps me feel rested and refreshed in the morning	0.66	3.62 (1.14)	—
No effect—3 items		7.44 (3.18)	.81
I do not feel any different when I use CBD	0.75	2.48 (1.24)	—
CBD has no effect on the way I feel	0.72	2.48 (1.25)	—
I do not notice any effects from using CBD	0.72	2.48 (1.26)	—

Note. CBD = cannabidiol.

CFA

A 60-item, six-factor CFA model was tested with the second randomly split-half sample (n = 300). The six-factor CFA model indicated adequate model fit across most fit indices: RMSEA = 0.06, 90% CI [0.05, 0.06], SRMR = 0.07; CFI = 0.90, χ²(1,315, N = 300) = 2561.02, p < .001. See Appendix in Supplemental Materials for CBD-OEQ items and instructions.

Initial Tests of Psychometric Properties of CBD-OEQ Endorsement Ratings

Data Screening.

Seventeen participants (2.8%) answered two attention check items incorrectly; 119 participants (20.4%) answered one attention check item incorrectly. Data were screened for outliers (z scores > 3.29 standard deviations above the mean; Tabachnick & Fidell, 2007). Outliers were identified among individuals who used each substance and excluded from analyses concerning the outlying response for past 30-day CBD use quantity (n = 5), cannabis use quantity (n = 3), and alcohol use quantity (n = 7). Some substance use variables were not normally distributed (i.e., skew > 3, kurtosis > 10; Kline, 2015), which is often the case with substance use variables (e.g., Buckner et al., 2017). Among those who reported past 30-day CBD use, ENDS use quantity was slightly skewed (skew = 3.21); among the entire sample, several 30-day use quantity variables were skewed and/or kurtotic (cannabis: skew = 3.14, kurtosis = 10.84; combustible smoking: skew = 4.42, kurtosis = 21.14; ENDS: skew = 7.86, kurtosis = 73.87); these variables were corrected using a log(x + 1) transformation.

Convergent Validity.

Means, standard deviations, and bivariate correlations among endorsement subscales and CBD-related variables are given in Table 3. Among those who had never used CBD products, CBD use intentions were significantly, positively associated with Positive Mood, Sleep, Pain Relief, and Harm Reduction subscales and significantly, negatively associated with the Global Negative subscale. Similarly, lifetime CBD use frequency was significantly, positively associated with Positive Mood, Sleep, Pain Relief, and Harm Reduction subscales and significantly, negatively associated with the Global Negative subscale. Among those who reported past-month CBD use, past 30-day CBD use quantity/use frequency were significantly, positively associated with Positive Mood, Sleep, Pain, and Harm Reduction. Past 30-day quantity was significantly, positively associated with No Effect expectancies.

Table 3.

Bivariate Correlations Among CBD-OEQ Endorsement Subscales and CBD-Related Outcomes

Variable	1	2	3	4	5	6	7	8	9	10
1. Positive mood	—
2. Global negative effects	.15**	—
3. Pain	.70**	.03	—
4. Sleep	.73**	.07	.70**	—
5. Harm reduction	.71**	.24**	.57**	.55**	—
6. No effect	−.20**	.37**	−.22**	−.29**	−.04	—
7. Intentions to use CBDa	.30**	−.16*	.26**	.24**	.24**	.13	—
8. Lifetime CBD use frequency	.33**	−.11**	.27**	.34**	.31**	−.05	c	—
9. Past 30-day CBD frequencyb	.27**	.06	.22**	.29**	.25**	.07	c	.67**	—
10. Past 30-day CBD quantityb	.29**	.09	.21**	.30**	.20**	.12*	c	.46**	.68**	—
M	66.21	40.93	37.25	18.58	18.72	7.44	3.50	2.55	18.28	35.94
(SD)	(17.90)	(14.80)	(8.79)	(6.01)	(4.68)	(3.18)	(1.74)	(2.06)	(10.67)	(38.23)

Note. CBD-OEQ = cannabidiol outcome expectancy questionnaire; CBD = cannabidiol.

^aAdministered only to those who never used CBD products (n = 157).

^bOnly individuals who used CBD in the past month (n = 287).

^cCannot be computed—CBD use intentions measure administered only to those who never used CBD products.

^*p < .05.

^**p < .01.

Discriminant Validity.

Results of Hotelling–William’s tests comparing significant bivariate correlations between the CBD-OEQ subscales, past 30-day CBD use, and past 30-day other substance use (cannabis, alcohol, combustible smoking, ENDS) among all participants are presented in Table 4. Overall, most correlations between CBD-OEQ subscales and past 30-day CBD use frequency/quantity were significantly stronger than the corresponding correlations between CBD-OEQ subscales and other substance use frequency/quantity. The correlation between the Positive Mood subscale and CBD use frequency and the correlation between the Harm Reduction subscale and CBD use frequency were not significantly stronger than the corresponding correlations with cannabis use frequency. Correlations did not significantly differ between CBD use quantity and cannabis use quantity for any of the CBD-OEQ subscales.

Table 4.

Hotelling–Williams t-Test Results Testing for Statistically Significant Differences Between the CBD-OEQ Endorsement Subscales and Past 30-Day Use Among Entire Sample (N = 600)

	CBD frequency	Cannabis frequency				Alcohol frequency				Smoking frequency				ENDS frequencya
Subscale	r 12	r 13	r 23	t	p	r 13	r 23	t	p	r 13	r 23	t	p	r 13	r 23	t	p
Positive mood	.32	.26	.45	1.48	.070	.12	.28	4.28	<.001	.14	.26	3.82	.0001	.14	.26	2.26	.012
Pain	.29	.19	—	2.43	.008	.08	—	1.89	.030	.16	—	2.74	.003	.16	—	—	—
Sleep	.31	.21	—	2.27	.012	.10	—	4.44	<.001	.11	—	4.21	<.001	.11	—	4.27	<.001
Harm reduction	.34	.28	—	1.39	.082	.16	—	3.82	<.001	.20	—	3.01	.001	.20	—	4.72	<.001
	CBD quantity	Cannabis quantitya				Alcohol quantity				Smoking quantitya				ENDS quantitya
Positive mood	.32	.28	.44	1.00	.162	—	—	—	—	.14	.25	3.80	.0001	.09	.10	4.39	<.001
Pain	.27	.23	—	1.00	.168	—	—	—	—	.16	—	2.29	.011	—	—	—	—
Sleep	.27	.24	—	0.73	.234	—	—	—	—	.11	—	3.30	.0001	.09	.10	4.39	<.001
Harm reduction	.32	.30	—	0.50	.310	—	—	—	—	.19	—	2.76	.003	—	—	—	—

Note. CBD-OEQ = cannabidiol outcome expectancy questionnaire; CBD = cannabidiol; ENDS = electronic nicotine delivery systems. r₁₂ = Pearson’s correlation coefficient between CBD-OEQ subscale and CBD use frequency/quantity; r₁₃ = Pearson’s correlation coefficient between CBD-OEQ subscale and cannabis, alcohol, combustible smoking, or ENDS frequency/quantity; r₂₃ = Pearson’s correlation coefficient between CBD use frequency/quantity and cannabis, alcohol, combustible smoking, or ENDS frequency/quantity.

^aLog transformed variable.

Incremental Validity.

Multicollinearity was within normal limits for all models (variance inflation factor; VIF values < 3.0). With CBD-OEQ and MEEQ subscales entered into the model, CBD-OEQ subscales significantly accounted for additional variance in CBD use intentions and use frequency, as well as past 30-day use frequency and quantity among those who reported past-month CBD use (Table 5). Not all of the CBD-OEQ endorsement subscales remained significant predictors of use behaviors, and several MEEQ subscales remained significant predictors of use behaviors—for example, the CBD-OEQ Sleep, Harm Reduction, and Global Negative endorsement subscales and MEEQ Cognitive and Behavioral Impairment, Global Negative, and Perceptual and Cognitive Enhancement remained significant predictors of lifetime CBD use frequency.

Table 5.

Hierarchical Linear Regressions of CBD-OEQ Endorsement Subscales and MEEQ Subscales Predicting CBD Use Outcomes

Variable	B	SE	β	t	p	sr 2	ΔR2	Adj. R2	F
DV: CBD use intentionsa
Step 1—MEEQ	—	—	—	—	<.001	—	.16	.13	5.59
Cognitive & behavioral impairment	−0.03	0.03	−0.14	−1.29	.199	−0.09	—	—	—
Relaxation & tension reduction	0.01	0.03	0.02	0.16	.876	0.01	—	—	—
Social & sexual facilitation	0.04	0.03	0.14	1.27	.206	0.09	—	—	—
Perceptual & cognitive enhancement	0.05	0.04	0.14	1.22	.223	0.08	—	—	—
Global negative effects	−0.01	0.03	−0.03	−0.33	.742	−0.02	—	—	—
Craving & physical effects	−0.02	0.04	−0.04	−0.37	.713	−0.03	—	—	—
Step 2—CBD-OEQ	—	—	—	—	<.001	—	.06	.17	4.10
Positive mood	0.01	0.01	0.08	0.58	.565	0.04	—	—	—
Global negative effects	-0.03	0.01	-0.17	-2.04	.043	-0.14	—	—	—
Sleep	0.04	0.05	0.10	0.85	.395	0.06	—	—	—
Pain	−0.01	0.02	−0.03	−0.27	.791	−0.02	—	—	—
Harm reduction	0.03	0.03	0.09	0.97	.334	0.07	—	—	—
No effect	0.12	0.05	0.17	2.32	.021	0.16	—	—	—
DV: Lifetime CBD use frequency
Step 1—MEEQ	—	—	—	—	.033	—	.08	.07	8.93
Cognitive & behavioral impairment	-0.03	0.02	-0.13	-2.14	.033	-0.08	—	—	—
Relaxation & tension reduction	−0.03	0.02	−0.10	−1.74	.083	−0.06	—	—	—
Social & sexual facilitation	0.00	0.02	0.01	0.18	.854	0.01	—	—	—
Perceptual & cognitive enhancement	0.07	0.02	0.18	3.04	.002	0.11	—	—	—
Global negative effects	0.07	0.01	0.27	5.02	<.001	0.18	—	—	—
Craving & physical effects	−0.02	0.02	−0.04	−0.85	.397	−0.03	—	—	—
Step 2—CBD-OEQ					<.001		.16	.25	26.0
Positive mood	0.01	0.01	0.10	1.39	.165	0.05	—	—	—
Global negative	−0.05	0.01	−0.35	−7.58	<.001	−0.27	—	—	—
Sleep	0.12	0.03	0.27	4.58	<.001	0.16	—	—	—
Pain	−0.01	0.01	−0.04	−0.64	.523	−0.02	—	—	—
Harm reduction	0.04	0.02	0.13	2.34	.020	0.08	—	—	—
No effect	0.06	0.03	0.09	2.04	.041	0.07	—	—	—
DV: Past 30-day CBD use frequencyb
Step 1—MEEQ					.075		.04	.02	1.94
Cognitive & behavioral impairment	−0.14	0.12	−0.11	−1.15	.250	−0.06	—	—	—
Relaxation & tension reduction	−0.03	0.16	−0.02	−0.22	.830	−0.01	—	—	—
Social & sexual facilitation	−0.03	0.16	−0.02	−0.18	.858	−0.01	—	—	—
Perceptual & cognitive enhancement	0.05	0.20	0.02	0.24	.811	0.01	—	—	—
Global negative effects	0.18	0.11	0.15	1.64	.103	0.09	—	—	—
Craving & physical effects	0.18	0.20	0.08	0.93	.354	0.05	—	—	—
Step 2—CBD-OEQ					<.001		.09	.09	3.42
Positive mood	0.01	0.06	0.02	0.18	.855	0.01	—	—	—
Global negative	−0.11	0.06	−0.17	−2.02	.044	−0.11	—	—	—
Sleep	0.36	0.23	0.13	1.54	.126	0.09	—	—	—
Pain	0.09	0.12	0.06	0.75	.456	0.04	—	—	—
Harm reduction	0.33	0.15	0.19	2.20	.028	0.12	—	—	—
No effect	0.44	0.23	0.14	1.87	.062	0.11	—	—	—
DV: Past 30-day CBD use quantityb
Step 1—MEEQ	—	—	—	—	.026	—	.05	.03	2.44
Cognitive & behavioral impairment	−0.07	0.42	−0.02	−0.16	.874	−0.01	—	—	—
Relaxation & tension reduction	−0.37	0.55	−0.06	−0.67	.502	−0.04	—	—	—
Social & sexual facilitation	0.37	0.58	0.06	0.63	.529	0.04	—	—	—
Perceptual & cognitive enhancement	0.60	0.71	0.08	0.84	.399	0.05	—	—	—
Global negative effects	−0.19	0.40	−0.04	−0.48	.635	−0.03	—	—	—
Craving & physical effects	−0.23	0.70	−0.03	−0.33	.742	−0.02	—	—	—
Step 2—CBD-OEQ	—	—	—	—	<.001	—	.08	.10	3.53
Positive mood	0.23	0.23	0.10	0.98	.327	0.06	—	—	—
Global negative effects	−0.21	0.19	−0.09	−1.07	.287	−0.06	—	—	—
Sleep	0.27	0.81	0.03	0.33	.740	0.02	—	—	—
Pain	0.37	0.42	0.07	0.88	.379	0.05	—	—	—
Harm reduction	1.04	0.53	0.17	1.96	.051	0.11	—	—	—
No effect	2.06	0.82	0.19	2.51	.013	0.14	—	—	—

Note. sr² = squared semiparital correlation coefficient; Adj. R² = adjusted R²; CBD-OEQ = cannabidiol outcome expectancy questionnaire; MEEQ = marijuana effect expectancy questionnaire; CBD = cannabidiol; significant results (p < .05) are bolded.

^aAdministered only to those who never used CBD products.

^bOnly individuals who used CBD in the past month.

Utility of CBD-OEQ Desirability Ratings

Internal Consistency of Desirability Subscales.

Internal consistency was adequate for all desirability subscales: Positive Mood (α = .95), Global Negative (α = .95), Pain (α = .95), Harm Reduction (α = .90), Sleep (α = .87), and No Effect (α = .77).

Correlations Among Endorsement and Desirability Subscales.

The Global Negative (r = .50, p < .001) desirability subscale was strongly correlated with its corresponding endorsement subscale. Positive Mood (r = .41, p < .001), No Effect (r = .32, p < .001), Sleep (r = .31, p < .001), and Pain (r = .29, p < .001) desirability subscales were moderately correlated with their corresponding endorsement subscales. The Harm Reduction (r = .06, p = .137) desirability subscale was not significantly correlated with its corresponding endorsement subscale.

Convergent Validity of Desirability Subscales.

Descriptive and bivariate correlations between the CBD-OEQ desirability subscales and CBD use outcomes are given in Table 6. Consistent with hypotheses, Positive Mood, Global Negative, and No Effect desirability subscales were significantly, positively correlated with CBD use behaviors. Pain, Sleep, and Harm Reduction desirability subscales were not significantly correlated with any CBD outcomes.

Table 6.

Bivariate Correlations Among CBD-OEQ Desirability Rating Subscales and CBD-Related Outcomes

Variable	1	2	3	4	5	6	7	8	9	10
1. Positive mood	—
2. Global negative effects	−.33**	—
3. Pain	.83**	−.51**	—
4. Sleep	.80**	−.44**	.85**	—
5. Harm reduction	.76**	−.46**	.77**	.73**	—
6. No effect	−.16**	.50**	−.21**	−.16**	−.12**	—
7. Intentions to use CBDa	.12**	.35**	.05	.08	−.03	.18*	—
8. Lifetime CBD use frequency	.13**	.31**	.00	−.01	.03	.13**	c	—
9. Past 30-day CBD frequencyb	.09	.21**	−.03	−.05	.01	.11	c	.66**	—
10. Past 30-day CBD quantityb	.06	.22**	−.07	−.09	.01	.13*	c	.46**	.68**	—
M	80.18	35.87	44.75	21.93	25.79	8.59	3.50	2.55	18.28	35.94
(SD)	(13.00)	(15.21)	(6.85)	(3.54)	(4.73)	(2.51)	(11.74)	(2.06)	(10.67)	(38.22)

Note. CBD-OEQ = cannabidiol outcome expectancy questionnaire; CBD = cannabidiol.

^aAdministered only to those who never used CBD products (n = 157).

^bOnly individuals who used CBD in the past month (n = 287).

^cCannot be computed—CBD use intentions measure administered only to those who never used CBD products.

Incremental Validity of Desirability Subscales.

Multicollinearity was within normal limits for all models (VIF values < 3.0). See Table 7 for significant models and Supplemental Tables 1–4 for nonsignificant models. Inconsistent with hypotheses, only Global Negative and No Effect desirability ratings remained significant positive predictors of use intentions, whereas Global Negative, Pain, Sleep, and No Effect desirability ratings remained significant positive predictors of lifetime use frequency. Global Negative and Sleep desirability ratings remained significantly related to past 30-day use frequency and quantity.

Table 7.

Statistically Significant Hierarchical Linear Regressions of Desirability and Endorsement Subscales Predicting CBD-Related Outcomes

Subscale	B	SE	β	t	p	sr 2	ΔR2	Adj. R2	F
CBD use intentionsa (n = 157)
Global negative effects
Step 1—Endorsement	−0.04	0.01	−0.26	−3.74	<.001	−0.25	0.03	0.02	4.57
Step 2—Desirability	0.05	0.01	0.42	5.95	<.001	0.40	0.16	0.18	20.43
No effect
Step 1—Endorsement	0.07	0.05	0.10	1.31	.191	0.10	0.02	0.01	3.31
Step 2—Desirability	0.12	0.06	0.15	2.05	.042	0.15	0.02	0.03	3.67
Lifetime CBD use frequency (n = 600)
Global negative effects
Step 1—Endorsement	−0.05	0.01	−0.36	−8.39	<.001	−0.31	0.01	0.01	7.70
Step 2—Desirability	0.07	0.01	0.49	11.48	<.001	0.42	0.18	0.19	70.57
Pain relief
Step 1—Endorsement	0.07	0.01	0.29	7.05	<.001	0.28	0.07	0.07	45.11
Step 2—Desirability	−0.03	0.01	−0.09	−2.09	.038	−0.08	0.01	0.07	24.85
Sleep
Step 1—Endorsement	0.17	0.02	0.38	9.43	<.001	0.36	0.12	0.11	77.37
Step 2—Desirability	−0.08	0.02	−0.13	−3.23	.001	−0.12	0.02	0.13	44.53
Global negative effects
Step 1—Endorsement	−0.11	0.05	−0.18	−2.18	.030	−0.13	0.00	0.00	0.97
Step 2—Desirability	0.21	0.05	0.33	4.11	<.001	0.24	0.06	0.06	8.95
Sleep
Step 1—Endorsement	0.77	0.16	0.30	4.89	<.001	0.28	0.06	0.06	18.52
Step 2—Desirability	−0.45	0.19	−0.15	−2.40	.017	−0.14	0.02	0.07	12.29
Past 30-day CBD use quantityb (n = 287)
Global negative effects
Step 1—Endorsement	−0.30	0.19	−0.13	−1.60	.113	−0.09	0.01	0.01	2.47
Step 2—Desirability	0.72	0.19	0.32	3.85	<.001	0.22	0.05	0.05	8.70
Pain relief
Step 1—Endorsement	1.26	0.31	0.25	4.13	<.001	0.24	0.05	0.05	13.44
Step 2—Desirability	−0.77	0.36	−0.13	−2.17	.031	−0.13	0.02	0.06	9.16
Sleep
Step 1—Endorsement	2.40	0.57	0.26	4.24	<.001	0.245	0.04	0.04	11.84
Step 2—Desirability	−1.95	0.67	−0.18	−2.90	.004	−0.168	0.03	0.06	10.29

Note. sr² = squared semiparital correlation coefficient; Adj. R² = adjusted R²; CBD = cannabidiol. Nonsignificant models can be found in Supplemental Materials. Step 2 change in R² significant at p < .05 in bold.

^aAdministered only to those who never used CBD products.

^bOnly individuals who used CBD in the past month (n = 287).

Moderation Analyses.

Interaction terms were not significant for Positive Mood, Pain, Sleep, or Harm Reduction models (Supplemental Tables 5–10). The Endorsement × Desirability interaction significantly predicted CBD use behaviors for the Global Negative and No Effect subscales. For the Global Negative models, Endorsement × Desirability interaction accounted for significant variance in CBD use intentions, lifetime CBD use frequency, and past-month CBD frequency (Table 8). Global Negative endorsement ratings were negatively associated with use intentions (Figure 1a) at low, b = −0.064, SE = 0.014, 95% CI [−0.091, −0.037] and mean levels of desirability, b = −0.040, SE = 0.010, 95% CI [−0.060, −0.019], but not high, b = −0.015, SE = 0.013, 95% CI [−0.042, 0.011]. Global Negative endorsement ratings were negatively associated with past 30-day CBD use frequency (Figure 1b) at low, b = −0.270, SE = 0.070, 95% CI [−0.410, −0.135] and mean levels of desirability, b = −0.155, SE = 0.053, 95% CI [−0.259, −0.050], but not high, b = −0.036, SE = 0.056, 95% CI [−0.147, 0.075]. Global Negative endorsement ratings were negatively associated with lifetime CBD use (Figure 1c) at low, b = −0.082, SE = 0.007, 95% CI [−0.096, −0.067]; mean, b = −0.052, SE = 0.006, 95% CI [−0.064, −0.041]; and high levels, b = −0.023, SE = 0.007, 95% CI [−0.037, −0.010].

Table 8.

Statistically Significant Regression Results for Moderation Models for the CBD-OEQ Global Negative Effects Subscale

Subscale	R 2	F	b	SE	t	p value	LLCI	ULCI	ΔR2
	DV: CBD use intentionsa
Model 8.1	0.22	16.61	—	—	—	<.001	—	—	—
Endorsement	—	—	−0.100	0.024	−4.09	<.001	−0.148	−0.051	—
Desirability	—	—	−0.036	0.034	−1.07	.287	−0.104	0.031	—
Interaction	—	7.47	0.002	0.001	2.73	.007	0.001	0.003	0.03
	DV: Lifetime CBD use frequency
Model 8.2	0.25	64.97	—	—	—	<.001	—	—	—
Endorsement			−0.121	0.012	−9.90	<.001	−0.145	−0.097
Desirability			−0.028	0.015	−1.85	.065	−0.059	0.002
Interaction	—	43.68	0.002	0.000	6.61	<.001	0.001	0.002	0.05
	DV: Past 30-day CBD use frequencyb
Model 8.3	0.09	9.80	—	—	—	<.001	—	—	—
Endorsement	—	—	−0.440	0.110	−4.00	<.001	−0.650	−0.220	—
Desirability	—	—	−0.096	0.106	−0.91	.363	−0.304	0.112	—
Interaction	—	10.88	0.007	0.002	3.30	.001	0.003	0.011	0.03

Note. CBD-OEQ = cannabidiol outcome expectancy questionnaire; LLCI = lower limit 95% confidence interval; ULCI = upper limit 95% confidence interval; CBD = cannabidiol; significant interactions at p < .05 in bold. Nonsignificant models can be found in Supplemental Materials.

^aAdministered only to those who never used CBD products (n = 157).

^bOnly individuals who used CBD in the past month (n = 287).

Figure 1. Global Negative Desirability Ratings Moderate Relationship Between Endorsement Ratings and CBD Use Outcomes.

Note. CBD = cannabidiol. Global Negative endorsement ratings were significantly negatively associated with CBD use intentions (a) and past 30-day CBD use frequency (b) at low and mean (but not high) levels of desirability ratings. Global Negative endorsement ratings were negatively associated with lifetime CBD use (c) at all levels of desirability.

For the No Effect models (Table 9), the Endorsement × Desirability interaction accounted for significant variance in use intentions, lifetime CBD use frequency, and past-month CBD quantity. No Effect endorsement ratings were significantly, positively associated with use intentions (Figure 2a) at high levels of desirability, b = 0.191, SE = 0.074, 95% CI [0.046, 0.337], but not at low, b = −0.036, SE = 0.070, 95% CI [−0.174, 0.102], or mean levels, b = 0.078, SE = 0.054, 95% CI [−0.028, 0.183]. No Effect endorsement ratings were significantly, positively associated with past 30-day use quantity (Figure 2b) at high levels of desirability, b = 1.967, SE = 0.846, 95% CI [0.301, 3.634], but not at low, b = −0.870, SE = 1.009, 95% CI [−2.856, 1.116], or mean levels, b = 0.549, SE = 0.742, 95% CI [−0.912, 2.010]. Endorsement ratings were negatively associated with lifetime use frequency (Figure 2c) at low, b = −0.175, SE = 0.036, 95% CI [−0.245, −0.105], and mean levels of desirability, b = −0.074, SE = 0.027, 95% CI [−0.127, −0.020], but not high levels, b = 0.028, SE = 0.034, 95% CI [−0.039, 0.094].

Table 9.

Regression Results for Moderation Models for the CBD-OEQ No Effect Subscale

Subscale	R 2	F	b	SE	t	p value	LLCI	ULCI	ΔR2
	DV: CBD use intentionsa
Model 9.1	0.07	4.40	—	—	—	.005	—	—	—
Endorsement	—	—	−0.336	0.180	−1.87	.063	−0.691	0.018	—
Desirability	—	—	−0.280	0.177	−1.58	.116	−0.629	0.069	—
Interaction	—	5.65	0.050	0.021	2.38	.019	0.008	0.091	0.03
	DV: Lifetime CBD use frequency
Model 9.2	0.06	13.14	—	—	—	<.001	—	—	—
Endorsement	—	—	−0.421	0.080	−5.24	<.001	−0.578	−0.263	—
Desirability	—	—	−0.186	0.077	−2.41	.016	−0.337	−0.034	—
Interaction	—	21.96	0.040	0.009	4.69	<.001	0.023	0.057	0.04
	DV: Past 30-day CBD use quantityb
Model 9.3	0.04	4.19	—	—	—	.006	—	—	—
Endorsement	—	—	−3.912	2.038	−1.92	.056	−7.925	0.100	—
Desirability	—	—	−2.567	1.750	−1.47	.144	−6.012	0.877	—
Interaction	—	6.36	0.504	0.200	2.52	.012	0.111	0.898	0.02

Note. CBD-OEQ = cannabidiol outcome expectancy questionnaire; LLCI = lower limit 95% confidence interval; ULCI = upper limit 95% confidence interval; CBD = cannabidiol; significant interactions at p < .05 in bold. Nonsignificant models can be found in Supplemental Materials.

^aAdministered only to those who never used CBD products (n = 157).

^bOnly individuals who used CBD in the past month (n = 287).

Figure 2. No Effect Desirability Ratings Moderate Relationship Between Endorsement Ratings and CBD Use Outcomes.

Note. CBD = cannabidiol. No Effect endorsement ratings were significantly, positively associated with CBD use intentions (a) and past 30-day CBD use quantity (b) at high levels of desirability ratings, but not low or mean levels, but negatively associated with lifetime CBD use frequency at low and mean, but not high levels of desirability (c).

Discussion

The present study used a three-phase, mixed-methods approach to develop and examine the initial construct validity and psychometric properties of the CBD-OEQ. Factor analyses supported a 60-item, six-factor structure: Positive Mood, Global Negative, Pain Relief, Sleep, Harm Reduction, and No Effect. Endorsement and desirability ratings of CBD-OEQ subscales demonstrated acceptable internal consistency, and there was evidence of convergent, discriminant, and incremental validity for some of the subscale scores. Desirability ratings explained minimal additional variance in CBD use behaviors above and beyond endorsement ratings for most subscales. Global Negative and No Effect desirability ratings moderated the relationship between endorsement ratings and CBD use outcomes, indicating that expectancy–desirability interactions may be more valuable predictors of CBD use behaviors for negative and no effect CBD outcome expectancies but appear to have limited utility for CBD positive outcome expectancies.

Consistent with findings from the cannabis outcome expectancy literature (e.g., Buckner et al., 2013; Hayaki et al., 2010; Waddell et al., 2021), CBD-OEQ endorsement subscales concerning anticipated positive effects of CBD (i.e., Positive Mood, Sleep, Pain Relief, Harm Reduction) were significantly, positively associated with CBD-related outcomes. In contrast, the Global Negative subscale was significantly, negatively associated with CBD use behaviors. Higher negative cannabis-related outcome expectancies appear to be especially protective against the initiation of cannabis use (Aarons et al., 2001) and frequent cannabis use (e.g., Hapsari et al., 2017; Hayaki et al., 2010; Waddell et al., 2021). Comparably, findings from the present study suggest that negative expectancies for CBD could also reduce intentions to use CBD and CBD use frequency. However, mean item ratings for the Global Negative and No Effect subscales indicated that, on average, individuals were “uncertain” whether CBD produces negative or no noticeable effects. For the No Effect subscale, it may be that individuals are uncertain about “experiencing no effect” because most expect at least some degree of positive effects—higher positive expectancies were associated with lower No Effect expectancies, and overall, mean item ratings for positive subscale items were higher. For the Global Negative subscale, people may be uncertain whether CBD produces negative effects, such as dry mouth or nausea, because research suggests that CBD does not appear to elicit the same negative physiological or psychological consequences associated with THC (Rong et al., 2017).

Although outcome expectancies are typically defined as “anticipated positive or negative effects of a substance” (Patel & Fromme, 2015), the No Effect subscale (expecting not to feel any different from using CBD) consistently emerged as a unique predictor of CBD use patterns and moderated the relationship between endorsement ratings and CBD use behaviors. Specifically, at high levels of desirability, No Effect endorsement ratings were positively associated with both use intentions among individuals who never used CBD and past 30-day quantity among individuals who used CBD in the past month. Consistent with prior work on cannabis desirability ratings, effects were small (Buckner et al., 2013), but may be clinically meaningful (Abelson, 1985). Individuals who use higher doses of CBD (e.g., because they expect that CBD will not produce significant effects) may actually not be experiencing any effects, as some research indicates anxiolytic effects of CBD are reported at moderate doses of CBD, but not lower or higher doses (Zuardi et al., 2017). Thus, individuals who expect no effects from using CBD may be vulnerable to incorrectly dosing CBD products, which could prevent these individuals from achieving potential benefit from CBD.

Although the CBD-OEQ subscales discriminated between CBD use and alcohol, smoking, and ENDs use, it is notable that the Positive Mood and Harm Reduction CBD-OEQ subscales did not sufficiently discriminate between CBD and cannabis use patterns, which may be due to the high proportion of individuals in the sample who used both CBD and cannabis. Additionally, some of the cannabis-specific MEEQ subscales remained significantly associated with CBD use behaviors after accounting for variance attributable to CBD-OEQ subscales, suggesting that cannabis and CBD may be used as substitutes or compliments. For example, the MEEQ Perceptual and Cognitive Enhancement subscale remained negatively associated with lifetime CBD use frequency, whereas the MEEQ Global Negative Effects subscale remained positively associated with lifetime CBD use frequency. Individuals with higher positive expectancies for cannabis-related enhancement may be less likely to use CBD products because CBD is nonpsychoactive and does not produce perceptual changes associated with THC in cannabis (Rong et al., 2017). In contrast, those with higher negative cannabis expectancies may be more likely to use CBD products, as CBD does not appear to elicit the same negative consequences associated with THC (Rong et al., 2017). Future work directly testing these hypotheses will be an important next step in disentangling these relationships.

Although most CBD-OEQ desirability rating subscales were correlated with their corresponding endorsement subscales, results indicated that desirability ratings explained minimal additional variance in CBD use behaviors (0%–3% for most subscales), with the exception of the Global Negative subscale, which accounted for an additional 16% variance in CBD use intentions, 18% variance in lifetime CBD use, and 5% variance in past 30-day CBD use frequency and quantity. Contrary to expectation, the Endorsement × Desirability interaction accounted for additional variance in CBD use behaviors for Global Negative and No Effect subscales. Although these findings differ from prior work in which the interactions of endorsement and desirability ratings were found to be robust predictors of substance use outcomes (e.g., tobacco; Copeland & Brandon, 2002), they coincide with other findings suggesting somewhat limited utility of Endorsement × Desirability interactions in predicting substance use behaviors (e.g., alcohol; Nicolai et al., 2018), particularly in cannabis use behaviors (Buckner et al., 2013).

Endorsement × Desirability interactions of Global Negative and No Effect expectancies were associated with greater intentions to use CBD and more frequent CBD use. As individuals with higher cannabis-related negative outcome expectancies experience more cannabis use-related problems and greater dependence symptom severity (Buckner et al., 2013; Schuster et al., 2019), Global Negative and No Effect expectancies might play an important role in maintaining CBD use despite experiencing use-related problems (e.g., perceived dependence) or side effects. Although the present study did not assess negative CBD-related consequences or perceived dependence, up to 50% of individuals who use CBD report experiencing side effects (e.g., dry mouth, fatigue, appetite changes, dizziness) because of their CBD use (Corroon & Phillips, 2018; Wheeler et al., 2020), and nearly 11% of those who use CBD products reported difficulty quitting CBD (Wheeler et al., 2020), despite evidence that CBD products have low abuse or dependence potential (World Health Organization, World Health Organization Expert Committee on Drug Dependence, 2017). Thus, it may be possible that CBD outcome expectancies are also associated with experiencing more negative CBD-related consequences or perceived dependence symptoms. It will be important for future work to test whether the desirability of Global Negative and No Effect expectancies are associated with unintended negative consequences and/or perceived dependence symptoms.

Clinical Implications

Findings from the present study have important clinical and treatment implications. Increasing the intensity and number of negative outcome expectancies and reducing the intensity and number of positive expectancies have been implicated as a possible intervention to reduce substance use (Jones, 2004). Restructuring maladaptive CBD outcome expectancies could reduce use-related problems, side effects, or perceived dependence symptoms. For example, if a patient endorses high positive expectancies or higher desirability for Global Negative Effects, clinicians could provide psychoeducation and feedback (i.e., information regarding the limited research on CBD and anxiety) to temper patient expectancies and promote use of evidence-based treatments instead. If a patient endorses higher No Effect expectancies, they may incorrectly use a higher dose of CBD, which could prevent these individuals from achieving potential therapeutic benefit; providing individuals with psychoeducation around the actual (vs. theoretical) effects of CBD and CBD dosing may be an important intermediary step until better dosing guidelines are established.

The CBD-OEQ may also be useful for clinical trial or administration research, as emerging research suggests that preexisting CBD-related outcome expectancies influence subjective effects (e.g., anxiety reduction) of CBD regardless of actual pharmacology (Spinella et al., 2021). The CBD-OEQ could be used to expand upon the Spinella et al. (2021) administration study by testing whether CBD-OEQ subscales moderate the relationship between administration condition and other outcomes of interest. For example, researchers could test whether the CBD-related pain expectancy subscale (collected prior to administration of CBD) moderates the relationship between condition and subjective reports of pain. Researchers could also use the CBD-OEQ to assess and control for baseline outcome expectancies in randomized controlled clinical trials of the potential impacts of CBD, which could help provide a more accurate estimate of biological effects. Given the rapid increase in interest in the therapeutic potential of CBD, understanding the role of CBD effect expectancies could help disentangle medical effects from placebo effects produced by CBD outcome expectancies.

Limitations and Future Directions

The results from the present study should be interpreted in light of limitations that suggest potential areas for future research on CBD outcome expectancies. First, consistent with prior research (e.g., Fedorova et al., 2021), the majority of our samples were comprised of non-Hispanic/Latin White females. Although outcome expectancy measures tend to be invariant across race, sex assigned at birth, and substance use frequency (e.g., Waddell et al., 2021), future work will benefit from the formal testing of measurement invariance or differential item functioning of CBD-OEQ subscales and items by demographic factors or use frequency. Second, we did not explicitly assess whether participants could distinguish between CBD and THC products, which may be important given that THC and CBD are often used together (e.g., Fedorova et al., 2021). We aimed to distinguish between CBD and THC in our screener question on lifetime CBD use (i.e., “please do not include use of marijuana or other marijuana products in your answer”) but did not include information on differences between CBD and THC/cannabis due in the CBD-OEQ instructions due to concern of priming effects. Future work may consider testing whether cannabinoid knowledge and/or using CBD and THC as compliments impacts CBD expectancies.

Third, this study was cross-sectional—future work could use longitudinal assessments to determine the predictive utility of the CBD-OEQ. Fourth, the present study utilized guidelines recommending a minimum EFA sample size of at least 300 participants (Field, 2013; Tabachnick & Fidell, 2013); other recommended guidelines for EFA sample sizes (e.g., at least five participants per item) would require a larger sample size, and future work using larger samples will be an important next step in continued testing of the psychometric properties of this newly developed measure. Fifth, although the length of the CBD-OEQ is in line with existing expectancy measures (Schafer & Brown, 1991), a next step in this line of research could be to create a short form of the CBD-OEQ and determine whether it performs as well as the longer form, as has been done with other expectancy measures (Torrealday et al., 2008).

Despite these limitations, the present study is the first to develop a measure with promising psychometric properties to assess CBD-specific expectancies. Findings from this study suggest that individuals have CBD-related expectancies that are uniquely different from cannabis-related expectancies and are differentially related to CBD use outcomes. Further, cannabis-related outcome expectancy measures do not adequately assess important CBD expectancies (e.g., Pain Relief, Sleep, Harm Reduction, No Effect). As interest in and use of CBD products continues to grow, the CBD-OEQ could help identify individuals who do not presently use CBD who may initiate CBD use, those who may transition from infrequent to frequent CBD use, those who may be at greater risk for negative side effects or perceived dependence, and/or those for whom CBD may be a viable harm reduction strategy or potential symptom management strategy. The CBD-OEQ could also be used to assess and control for baseline CBD outcome expectancies in administration research or randomized controlled clinical trials.

Public Significance Statement.

We created a questionnaire to measure beliefs about the expected positive and negative effects of cannabidiol (CBD). Beliefs about the effects of CBD were related to CBD use patterns among adults and are unique from beliefs about cannabis. This newly developed measure could be used to help researchers better understand the effects of CBD and CBD use patterns among adults.