Discriminative-stimulus effects of cannabidiol oil in Sprague-Dawley rats

Abstract

Cannabidiol (CBD) is one of the major centrally active phytocannabinoid components of cannabis, and has been approved by the FDA only for the treatment of seizures associated with three rare disorders. It has also been touted as a potential treatment for anxiety in place of more traditional treatments like benzodiazepines. Although there is some evidence of anxiolytic effects of CBD, its suitability as a substitute for benzodiazepines is unknown. This experiment was designed to assess the extent to which CBD shares interoceptive discriminative-stimulus properties with the anxiolytic drug chlordiazepoxide (CDP), a benzodiazepine. In the present experiment, a range of doses (0-1569 mg/kg) of over-the-counter CBD oil was administered (i.g.) in male Sprague-Dawley rats trained to discriminate 5.6 mg/kg CDP from saline. Due to the long time-course effects of CBD, generalization tests were conducted at 90 and 120 min post-CBD administration. The two highest doses of CBD tested (1064 and 1569 mg/kg) were found to partially substitute for 5.6 mg/kg CDP, with mean percent responding on the CDP-associated lever reaching above 20% at time 2 (120 min post-CBD administration), suggesting that high doses of the over-the-counter CBD oils used in this experiment share interoceptive discriminative-stimulus properties to some degree with CDP. These results are novel in comparison to existing research into stimulus effects of CBD, in which substitution for benzodiazepines has not previously been observed.

Introduction

Cannabidiol (CBD) is one of the major psychoactive phytocannabinoid components of cannabis (Mechoulam, et al., 2002). In contrast to Δ⁹-tetrahydrocannabinol (THC), the phytocannabinoid responsible for the psychotropic effects of cannabis, CBD has been noted for its lack of a euphoric ‘high’ (Izzo, et al., 2009). The Food and Drug Administration (FDA) has approved CBD for use in the medication Epidiolex, which is administered orally at doses up to 25 mg/kg per day to treat seizures associated with Lennox-Gastaut syndrome, Dravet syndrome, and tuberous sclerosis complex (Corroon and Kight, 2018; Greenwich Biosciences, 2021). Cannabis and substances derived from cannabis remain classified as Schedule I drugs under federal law; however, CBD products derived from hemp ‘cultivated pursuant to a state’s pilot research program with Δ⁹-THC concentrations that do not exceed 0.3%’ in any part of the plant are currently legal under federal law to sell and purchase in all 50 states (Corroon and Kight, 2018, p. 191). CBD products such as topical lotions, edible candies and snacks, and multipurpose oils are widely available and marketed for use by adults, children, and pets. In surveys of self-reported CBD users, about 40% of respondents report using over-the-counter CBD to self-medicate anxiety (Corroon and Phillips, 2018; Moltke and Hindocha, 2021). However, CBD has not been evaluated or approved by the FDA for these purposes, nor are the pharmacological mechanisms responsible for these purported effects fully understood.

Evidence suggests that CBD may act as a positive allosteric modulator (PAM) at the α2 subunit of GABA_A receptors, resulting in enhanced GABA receptor activation (Bakas et al., 2017; Ruffolo et al., 2018). This proposed mechanism of action is supported by reports of CBD as an effective anticonvulsant in preclinical research (e.g. Uttl et al., 2021; Janisset et al., 2022) and clinical trials (Devinsky et al., 2017, 2018; Thiele et al., 2018; Patel et al., 2021). GABAergic activity has long been implicated in treatment of seizure disorders, as benzodiazepines (active at the GABA α2 subunit) have been used as anticonvulsants for decades (Goldenberg, 2010). Benzodiazepines’ net inhibitory effect via activation of the GABA_A receptor has also made them a popular treatment for anxiety disorders (Olkkola and Ahonen, 2008; Sigel and Ernst, 2018).

Earlier findings suggest that GABA_A receptors may be involved in CBD’s mechanism of action: Onaivi et al., (1990) found that CBD increased the time rats spent in the open arm of the elevated plus-maze, which is interpreted as a reduction in anxiety-like behavior, as known anxiolytics have similar effects in this assay. CBD’s anxiolytic effect was reduced by administration of flumazenil, a high-affinity benzodiazepine antagonist that blocks stimulation of the GABA_A receptor, suggesting that CBD exerts anxiolytic effects at least partially via activation of the GABA_A receptor. Additionally, Resstel et al., (2006) found that CBD attenuated conditioned-fear responses to a similar degree as diazepam, a benzodiazepine that acts as a PAM at the GABA_A receptor (Campo-Soria et al., 2006). These findings suggest that CBD’s purported anxiolytic effects could be mediated to some degree via GABA_A receptors, and CBD and benzodiazepines might be expected to share stimulus properties via their shared net effect of increasing GABA_A receptor activation. However, in a recent study by Gray et al., (2022) aimed at analyzing CBD’s abuse potential, pharmaceutical-grade CBD (0–150 mg/kg, i.g.) failed to substitute for 0.5 mg/kg midazolam (i.p.), another GABA_A PAM of the benzodiazepine class, with midazolam-lever responding remaining below 20% at all tested doses of CBD in female Lister Hooded rats. The present experiment builds on the existing literature with the inclusion of male subjects of a different strain, a less-potent benzodiazepine training drug, and over-the-counter CBD at higher doses.

As non-prescription CBD is used as a self-prescribed anxiety treatment as an alternative to traditional antianxiety drugs (Corroon and Phillips, 2018; Moltke and Hindocha, 2021), it is important to assess this type of CBD in comparison with anxiolytic drugs like benzodiazepines. A step toward clarifying whether CBD is an effective substitute for benzodiazepines is to compare the subjective stimulus properties of the drugs—in simple terms, to determine how much they ‘feel’ like one another. The present experiment was designed to evaluate shared stimulus properties of chlordiazepoxide (CDP, trade name Librium) and over-the-counter CBD oil from two different suppliers. In addition, the two over-the-counter CBD samples used were analyzed using mass spectrometry to independently assess label information.

Method

Subjects

Adult male experimentally and drug-naive Sprague–Dawley rats (n = 8) were used as subjects. Rats were pair-housed throughout the experiment, had free access to water in the home cage, and were fed approximately 15 g of food 30 min after the termination of each experimental session, resulting in a food-restriction cycle of about 22 h before the start of the next experimental session. Vivarium temperature and humidity were controlled, and a reverse 12-h light-dark cycle was maintained, with all experimental sessions conducted at approximately the same time each day during the dark phase.

Apparatus

Experimental sessions were conducted in eight standard operant-conditioning chambers for rats, each enclosed in a melamine, sound-attenuating cubicle (Med Associates, VT). Each chamber contained a working area of 30.5 cm by 24.5 cm by 21.0 cm, a grid floor, and a 45-mg pellet dispenser with a pellet receptacle centered between two retractable response levers. The levers were 11.5 cm apart from each other and required at least a force of 0.25 N for a response to be recorded. The levers were 4.8 cm wide, protruded 1.9 cm into the chamber, and were elevated 8 cm from the grid floor. Two 28-V stimulus lights that were 2.5 cm in diameter were approximately 7 cm above each lever. Each chamber contained a 28-V houselight on the wall opposite the wall containing the operanda. A ventilation fan circulated air and served to mask extraneous noise. Equipment was interfaced to a computer and experimental sessions and data collection were programmed and conducted with MedPC-IV (Med Associates, VT).

Procedure

Preliminary training

Preliminary training sessions were 30 min in duration and were conducted 5 days a week (Monday–Friday). In preliminary training, rats were placed into the operant-conditioning chamber, the ventilation fan was turned on, and the session began following a 15-min blackout. At the onset of the session, the houselight and stimulus lights above both levers were illuminated. Food pellets were made available on a fixed-ratio (FR) 1 schedule. After five pellets had been delivered for responding on one lever, the FR 1 schedule alternated to the opposite lever. Schedule alternation between levers continued after the delivery of every fifth pellet to ensure contact with food-pellet delivery following lever presses on both levers. The response requirement gradually increased from FR 1 to FR 10 across sessions. Sessions terminated once 40 food pellets were delivered (20 pellets via each lever). Training on the FR 10 schedule continued until all available food pellets were earned for two consecutive sessions. A variable-interval (VI) 15” component (Fleshler and Hoffman, 1962) was then introduced, resulting in a TAND VI 15” FR 10 schedule of reinforcement. Once responding was stable under this schedule, the VI component of the schedule was increased to 30 s to result in a tandem (TAND) VI 30” FR 10 schedule (Anderson and van Haaren, 1999; van Haaren et al., 1999). Preliminary training sessions continued until subjects had earned all available food pellets on the TAND VI 30” FR 10 schedule for two consecutive sessions, with the active lever alternating sides across sessions.

Discrimination training

Subjects were trained to discriminate CDP from saline in training sessions that were conducted Monday through Friday. The training drug (D) or saline vehicle (V) was administered (i.p.) prior to the experimental session in the following order: VDVDD DVDVV, which repeated every 2 weeks (Monday–Friday). CDP was first administered at a dose of 10.0 mg/kg, but was lowered to 5.6 mg/kg following either a) five consecutive sessions at minimum 80% accuracy before the delivery of the first food pellet or b) 30 sessions at the 10 mg/kg dose.

Immediately after injection, subjects were placed into the chamber, the ventilation fan turned on, and a 15-min blackout began. Once the blackout had elapsed, house-lights and stimulus lights were illuminated, and both levers were extended into the chamber and remained available throughout the session. The placement of the CDP and vehicle levers was counterbalanced across subjects: for four subjects, the CDP-associated lever was on the right side and the vehicle-associated lever was on the left side, and vice versa for the remaining four subjects. All responses on each lever were recorded, but only responses on the injection-appropriate lever were reinforced with the delivery of one food pellet according to the TAND VI 30” FR 10 schedule. Extinction was in effect for presses on the other lever. Training sessions terminated once 40 food pellets were delivered or 30 min elapsed, whichever occurred first. Discrimination training continued until 80% of responses that occurred before the delivery of the first food pellet were emitted on the correct lever for five consecutive sessions.

Generalization testing

Generalization tests were conducted on Tuesdays and Fridays. Discrimination training sessions continued during the three remaining sessions each week on Monday, Wednesday, and Thursday. Generalization tests were conducted under extinction conditions in which no food pellets were delivered for responses on either lever so that contingencies were identical across alternatives (Stolerman, 1993). Due to concerns about suitability of CBD oil for i.p. injection, CBD was administered via gastric gavage (i.g.) prior to generalization tests. Prior research indicates that a variety of drugs will substitute in generalization tests when administered via a different route of administration than what was used in discrimination (Bruner and Anderson, 2009; Fujiwara et al., 2018; Wiley et al., 2021). These results suggest that the interoceptive stimuli and time course of some drugs, including CDP (Bruner and Anderson, 2009) are similar following both i.g. and i.p. administration. In the present experiment, both i.p. and i.g. routes of CDP administration were tested in generalization sessions to further support generalization across different routes of administration.

Generalization tests were conducted at two different time points in an attempt to capture pharmacokinetic effects of CBD following administration (Anderson and van Haaren, 1999). Deiana et al., (2012) and Hložek et al., (2017) found that plasma and brain CBD levels and behavioral effects peaked around 120 min following intragastric (i.g.) administration. Accordingly, time 1 tests were conducted 90 min post-CBD administration to capture drug effects on the ascending limb of the time-course function, and time 2 tests were conducted 120 min after CBD administration to capture the most robust effects at the peak of drug action. Rats were transported in their home cages to the laboratory, administered with CBD intragastrically, and placed back into their cages in the laboratory for 75 min. Once 75 min had elapsed, subjects were placed into the operant-conditioning chambers, where a 15-min blackout occurred prior to the start of the session. Following the blackout, the houselight and stimulus lights were illuminated and the two levers extended into the chamber, beginning the test at 90 min (time 1) or 120 min (time 2) post-CBD administration. The test terminated after the response requirement of the TAND VI 30” FR 10 schedule was fulfilled on either lever but was not followed by food delivery, or 5 minutes elapsed, whichever occurred first (Anderson and van Haaren, 1999; van Haaren et al., 1999).

Subjects were administered with CDP (i.g.) 60 min prior (followed by 45 min in their home cages and a 15-min blackout period in the operant-conditioning chamber) to the test session to account for onset of behavioral effects as a result of intragastric administration. CDP (i.p.) and d-amphetamine (d-amp) (i.p.) were administered 30 min prior (followed by 15 min in their home cages and a 15-min blackout in the operant-conditioning chamber) to the test session.

Drugs

CBD oil was purchased from two sources: a local brick-and-mortar CBD store and an online CBD store. The bottles from the local and online store were labeled as containing 167 mg/ml and 333 mg/ml CBD, respectively. Each supplier provided copies of its own mass-spectrometry analysis that corresponded to the concentrations specified on the labels of the purchased CBD. Test doses were initially selected based on the purported CBD content of each bottle. CBD tests were conducted at a range of doses initially believed to be 56–333 mg/kg CBD, according to the CBD-bottle labels. Following the conclusion of the experiment, however, independent mass-spectrometry analyses were performed on both CBD samples. This analysis indicated that both bottle labels underestimated the CBD content of each bottle, with the oil from the local supplier containing 1064 mg/ml CBD and the oil from the online supplier containing 1569 mg/ml CBD. For the sake of accuracy, the doses reported in this paper are those obtained from our independent mass-spectrometry analysis conducted after the conclusion of the experiment, rather than the doses from the bottle labels.

CDP (Sigma-Aldrich, St. Louis, MO) was dissolved in 0.9% saline vehicle and administered 15 min (i.p.) prior to training, and 30 min (i.p.) and 60 min (i.g.) prior to testing. CBD oil (357–1064 mg/kg, local supplier; 1569 mg/kg, online supplier) was suspended in medium-chain triglyceride (MCT) liquid coconut oil vehicle (Nestlé Health Science, Vevey, Switzerland) and administered for 90 min (i.g., time 1 sessions) and 120 min (i.g., time 2 sessions) prior to testing. d-Amp (Sigma-Aldrich, St. Louis, MO) was dissolved in 0.9% saline vehicle and administered 30 min (i.p.) prior to testing.

Generalization tests were conducted for CDP (i.p. and i.g.) (1.0–10.0 mg/kg), CBD (i.g.) (357–1569 mg/kg), d-amp (i.p.) (0.3–1.0 mg/kg), saline (i.p.), and coconut oil (i.g.). Behaviorally active doses of d-amp were tested as a negative control, as it was expected to produce no substitution for CDP due to distinct behavioral effects as a result of distinct pharmacological mechanisms, and no research to date indicates that CBD would share stimulant-like properties with d-amp. The purpose of a negative control in this context was to ensure that responding on the CDP-associated lever during a test session was indicative of shared specific discriminative-stimulus properties of the training drug and a given test compound, rather than the presence of any drug. d-Amp was tested up to a dose of 1.0 mg/kg due to suppression of responding in some subjects at this dose. Most subjects received each dose twice at each test time, but three subjects received some doses only once or not at all due to health problems and, subsequently, euthanasia.

Data analysis

Discrimination training

In discrimination training, the primary dependent variable was percent injection-appropriate lever presses before the delivery of the first food pellet, calculated by dividing the number of correct responses made before delivery of the first food pellet by the total number of responses made before delivery of the first food pellet, multiplied by 100.

Generalization tests

The primary dependent variable for generalization tests was percent CDP-associated responding, calculated as responses on the CDP-associated lever divided by total responses emitted during the test multiplied by 100, as no food pellets were delivered and tests terminated after the response requirement for the non-resetting TAND VI 30” FR 10 schedule was fulfilled on one lever or 5 min elapsed from the start of the test. Data for a test were excluded from analysis if fewer than 11 total responses occurred on one lever before the test timed out, as a minimum of 11 responses was required to fulfill the TAND VI 30” FR 10 schedule. Overall response rates were calculated by dividing the total number of presses on both levers emitted during each test by the duration of the test in minutes.

Mean effective dose (ED₅₀) and 95% confidence intervals were generated by log-linear interpolation of the dose-response curve for any compound that produced more than 50% responding on the CDP-associated lever during a generalization test (Anderson and van Haaren, 1999; Bruner and Anderson, 2009).

Statistical analysis

Percent selection of the CDP-associated lever was first analyzed using linear mixed-effects regression; however, this approach violated assumptions of normality and linearity of residuals. Violations were not attenuated by applying a log or arcsine square-root transformation to the percent-selection variable. Thus, as recommended by Young (2018), responding on the CDP-associated lever was analyzed at the proportion level using a generalized linear mixed model with a logit link function (i.e. repeated-measures binomial logistic regression) using the glmer function in the lme4 package in R (Bates, 2015). A mixed (i.e. multilevel, random effects) model can account for individual-subject variability across nested data by varying regression parameters across subjects.

For the current study, the predictors were dose (treated as continuous and scaled to provide results in units of SD) and test time. Dose was also included as a random slope due to visually apparent differences across subjects. Models were analyzed separately for CBD, CDP, and d-amp because the doses were different for each drug, and route of administration was included as a covariate for CDP only, as only CDP was tested via two routes of administration (i.p. and i.g.). The outcome variable was count of responses on the CDP-associated lever weighted by total responses. The R code for these models was formulated as the following:

$$glmer(choiceDrug∕choiceTot\sim scale(dose)\ast testTime+(1+scale(dose)\mid Subject),family=binomial)$$

Exponentiated odds ratios and 95% confidence intervals are reported in tables along with bootstrapped P-values with 200 resamples.

Results

Discrimination training

Discrimination training began at a dose of 10.0 mg/kg CDP. Training at this dose continued until the training criteria were met or 30 days of administration at this dose occurred, at which point the training dose was lowered to 5.6 mg/kg CDP. Discrimination at 5.6 mg/kg CDP required an average of 40 training sessions (range = 30–67 sessions) before the training criteria were met and generalization testing could begin. For the last five training sessions of each type (i.e. CDP or saline; ten sessions total) before generalization testing began, mean percent correct responding was 90.50% (range = 82.03–100%).

Vehicle generalization testing

Percent CDP-associated responding during vehicle tests is presented in Fig. 1. Statistics for CDP-associated responding during vehicle tests are summarized in Tables 1 and 2. Response rates on both levers were calculated during vehicle tests; these data are presented in Fig. 2 and summarized in Tables 3 and 4.

Figure 1.

Note. Mean percent and SEM of total responses on the CDP-associated lever for each test dose of CDP (i.p.), CDP (i.g.), d-amp (i.p.), and CBD (i.g.). Data points for saline and MCT coconut oil are presented at VEH. Time One generalization tests were conducted 30 min, 60 min, 90 min, and 30 min following administration of CDP (i.p.), CDP (i.g.), CBD (i.g.), and d-amp (i.p.), respectively. Time Two generalization tests were conducted 60 min, 90 min, 120 min, and 60 min following administration of CDP (i.p.), CDP (i.g.), CBD (i.g.), and d-amp (i.p.), respectively. Dotted lines at 20% and 80% indicate criteria for partial and full substitution, respectively. The open square at 1569 mg/kg CBD represents CBD purchased from the online retailer.

Table 1.

Summary statistics for CDP-associated responding at Time One

Drug and dose in mg/kg (route)	Mean % CDP-associated responding (SEM)	Range of % CDP-associated responding	N of subjects tested	N of determinationsa
CDP (i.p.)
0 (Saline)	6.66 (2.06)	0–24.49	8	16
1.0	23.77 (7.14)	0–96.88	8	16
3.0	60.73 (8.86)	0–100	8	16
5.6	96.74 (1.09)	86.54–100	8	16
10.0	96.42 (1.16)	83.33–100	8	16
CDP (i.g.)
1.0	1.41 (0.64)	0–5.77	6	12
3.0	52.42 (14.22)	0–100	6	12
5.6	88.13 (7.89)	3.57–100	6	12
10.0	87.13 (8.55)	16.67–100	6	12
CBD (i.g.)
0 (MCT oil)	1.97 (0.89)	0–11.54	8	15
357.3	5.93 (2.72)	0–39.58	8	16
638.0	8.75 (6.16)	0–100	8	16
1064.0	5.34 (4.35)	0–65.62	8	15
1569.0	12.98 (9.05)	0–100	6	11
d-Amp (i.p.)
0.3	1.44 (0.80)	0–7.69	6	11
1.0	6.84 (3.34)	0–16.67	3	6

Note. Time One generalization tests were conducted 30 min, 60 min, 90 min, and 30 min following administration of CDP (i.p.), CDP (i.g.), CBD (i.g.), and d-amp (i.p.), respectively. SEM = standard error of the mean.

^aTotal number of times each dose was administered. Odd numbers indicate that a subject received the dose only once.

Table 2.

Summary statistics for CDP-associated responding at Time Two

Drug and dose in mg/kg (route)	Mean % CDP-associated responding (SEM)	Range of % CDP-associated responding	N of subjects tested	N of determinationsa
CDP (i.p.)
0 (Saline)	4.38 (1.27)	0–12.50	8	15
1.0	5.15 (1.68)	0–25.0	8	16
3.0	75.89 (8.71)	4.88–100	8	16
5.6	96.29 (1.51)	83.13–100	8	16
10.0	99.11 (0.40)	95.11–100	8	15
CDP (i.g.)
1.0	21.64 (11.03)	0–98.15	5	9
3.0	33.44 (14.16)	0–100	5	9
5.6	79.80 (12.83)	2.70–100	5	10
10.0	93.77 (4.49)	55.91–100	5	10
CBD (i.g.)
0 (MCT oil)	4.43 (1.24)	0–14.29	7	13
357.3	5.71 (3.29)	0–40.91	7	14
638.0	9.65 (8.24)	0–100	6	12
1064.0	25.03 (8.53)	0–100	6	12
1569.0	23.50 (10.30)	0–100	6	12
d-Amp (i.p.)
0.3	1.18 (0.49)	0–3.33	5	9
1.0	4.16 (1.52)	0–8.33	3	6

Note. Time One generalization tests were conducted 30 min, 60 min, 90 min, and 30 min following administration of CDP (i.p.), CDP (i.g.), CBD (i.g.), and d-amp (i.p.), respectively. SEM = standard error of the mean.

^aTotal number of times each dose was administered. Odd numbers indicate that a subject received the dose only once.

Figure 2.

Note. Mean response rate on both levers and SEM for each test dose of CDP (i.p.), CDP (i.g.), d-amp (i.p.), and CBD (i.g.). Data points for saline and MCT coconut oil are presented at VEH. Time One generalization tests were conducted 30 min, 60 min, 90 min, and 30 min following administration of CDP (i.p.), CDP (i.g.), CBD (i.g.), and d-amp (i.p.), respectively. Time Two generalization tests were conducted 60 min, 90 min, 120 min, and 60 min following administration of CDP (i.p.), CDP (i.g.), CBD (i.g.), and d-amp (i.p.), respectively. The open square at 1569 mg/kg CBD represents CBD purchased from the online retailer.

Table 3.

Summary statistics for overall response rates at time 1

Drug and dose in mg/kg (route)	Mean resp./ min (SEM)	Range of resp./ min	N of subjects tested	N of determinationsa
CDP (i.p.)
0 (Saline)	83.18 (5.51)	36.59–121.82	8	16
1.0	101.23 (5.98)	63.16–141.82	8	16
3.0	112.53 (7.12)	72.97–181.77	8	16
5.6	123.94 (11.71)	22.32–183.75	8	16
10.0	51.87 (12.55)	4.29–128.57	8	16
CDP (i.g.)
1.0	79.38 (9.79)	18.46–148.13	6	12
3.0	68.45 (8.17)	9.47–121.82	6	12
5.6	85.84 (8.82)	30.85–133.71	6	12
10.0	102.97 (11.89)	29.23–161.82	6	12
CBD (i.g.)
0 (MCT oil)	55.31 (5.35)	27.77–93.63	8	15
357.3	60.44 (6.40)	14.72–95.0	8	16
638.0	57.46 (5.17)	24.70–89.19	8	16
1064.0	59.51 (7.49)	8.08–110.0	8	15
1569.0	58.94 (5.66)	19.46–82.70	6	11
d-Amp (i.p.)
0.3	48.58 (775)	5.67–89.14	6	11
1.0	21.87 (12.51)	4.00–82.29	3	6

Time 1 generalization tests were conducted 30 min, 60 min, 90 min, and 30 min following administration of CDP (i.p.), CDP (i.g.), CBD (i.g.), and d-amp (i.p.), respectively.

^aTotal number of times each dose was administered. Odd numbers indicate that a subject received the dose only once.

Table 4.

Summary statistics for overall response rates at Time Two

Drug and dose in mg/kg (route)	Mean resp./min (SEM)	Range of resp./min	N of subjects tested	N of determinationsa
CDP (i.p.)
0 (Saline)	58.28 (5.70)	22.32–102.35	8	15
1.0	74.88 (8.44)	27.43–138.18	8	16
3.0	96.64 (9.02)	26.47–167.65	8	16
5.6	110.78 (6.20)	76.36–159.38	8	16
10.0	115.78 (8.55)	65.29–165.88	8	15
CDP (i.g.)
1.0	49.45 (9.44)	18.46–87.13	5	9
3.0	55.88 (8.37)	28.57–96.0	5	9
5.6	76.24 (7.16)	49.23–120.0	5	10
10.0	71.83 (8.49)	35.29-100.58	5	10
CBD (i.g.)
0 (MCT oil)	45.83 (5.96)	2.20-81.82	7	14
357.3	44.87 (5.99)	6.47-80.63	7	14
638.0	38.26 (6.59)	14.40-88.33	6	12
1064.0	35.49 (5.89)	11.13-79.41	6	12
1569.0	31.56 (8.27)	2.0-75.0	6	12
d-Amp (i.p.)
0.3	47.85 (8.13)	8.68-90.85	5	9
1.0	27.41 (9.58)	2.20-61.71	3	6

Note. Time One generalization tests were conducted 30 min, 60 min, 90 min, and 30 min following administration of CDP (i.p.), CDP (i.g.), CBD (i.g.), and d-amp (i.p.), respectively. SEM = standard error of the mean.

^aTotal number of times each dose was administered. Odd numbers indicate that a subject received the dose only once.

Saline

Saline (i.p.) was the vehicle for CDP and d-amp. No substitution was observed at either test time, as mean percent CDP-associated responding remained below 20% when saline was tested (Fig. 1, open triangle at VEH; see Tables 1 and 2 for a summary). Response rate data for saline at both test times are presented in Fig. 2 (open triangle at VEH) and summarized in Tables 3 and 4.

MCT coconut oil

Liquid MCT coconut oil (i.g.) was the vehicle for CBD. No substitution was observed at time 1 or time 2, as mean percent CDP-associated responding remained below 20% for all subjects in both tests (Fig. 1, filled square at VEH). Response rate data for MCT coconut oil at both test times are presented in Fig. 2 (filled square at VEH) and summarized in Tables 3 and 4.

CDP (i.p. and i.g.) generalization testing

Response rates on both levers were calculated during both CDP (i.p. and i.g.) generalization tests and are presented in Fig. 2 (open and filled triangles, respectively) and summarized in Tables 3 and 4.

At time 1, full substitution (>80% responding on the CDP-associated lever) of CDP (i.p. and i.g.) was observed at 5.6 mg/kg and 10.0 mg/kg CDP (Fig. 1, left panel, open and filled triangles). Mean percent CDP-associated responding decreased along with decreasing CDP dose, with partial substitution occurring at 1.0 mg/kg and 3.0 mg/kg. Statistics for CDP-associated responding at time 1 are summarized in Table 1.

At time 2, full substitution of CDP (i.p.) was observed at 5.6 mg/kg and 10.0 mg/kg (Fig. 1, right panel, open triangles). Mean percent CDP-associated responding decreased along with decreasing CDP dose, with partial substitution observed at 3.0 mg/kg and no substitution observed at 1.0 mg/kg CDP (i.p.). CDP (i.g.) produced full substitution at 10.0 mg/kg at time 2 (Fig. 1, right panel, filled triangles). Partial, near-full substitution was observed at 5.6 mg/kg, and partial substitution was also observed at 3.0 and 1.0 mg/kg CDP (i.g.). Statistics for CDP-associated responding at time 2 are summarized in Table 2.

Regression coefficients are provided for the model that allowed effects of CDP dose to vary by subject (Table 3), which outperformed an intercept-only model (χ²(2) = 815.42, P < 0.001). As CDP dose increased, the odds of responding on the CDP-associated lever also increased (odds ratio [OR] = 10.70, P < 0.001). There was no effect of test time (OR = 0.92, P = NS). There was an effect of route of administration (OR = 0.14, P < 0.05), with the i.p. route (ED₅₀: 2.12 mg/kg, 95% confidence interval [CI]: 1.91–2.36 mg/kg) more likely to produce CDP-associated responding than the i.g. route (ED₅₀: 2.96 mg/kg, 95% CI: 2.44–3.59 mg/kg). Both routes, however, produced similar degrees of substitution at all doses except 1.0 mg/kg CDP (Figs. 1 and 2, open and filled triangles). There was no dose-by-test-time interaction (OR = 0.94, P = NS).

CBD (i.g.) generalization

Response rates on both levers were calculated during CBD generalization tests. These data are presented in Fig. 2 (filled and open squares) and summarized in Tables 3 and 4.

At time 1, mean percent responding on the CDP-associated lever remained below 20% for all doses, indicating no substitution of CBD at any dose at that test time. Statistics for CDP-associated responding during CBD-generalization tests at time 1 are summarized in Table 1.

At time 2, a dose-dependent increase in CDP-associated responding was observed (Fig. 1, right panel, filled and open squares). Mean percent responding on the CDP-associated lever remained below 20% for 357 and 638 mg/kg, indicating no substitution at these doses (Fig. 1, right panel, filled and open squares). Partial substitution, however, was observed at the higher doses tested: at 1064 and 1569 mg/kg, mean percent responding on the CDP lever was 25.03% (SEM = 8.53%, n = 12) and 23.50% (SEM = 10.30%, n = 12), respectively. Statistics for CDP-associated responding during CBD-generalization tests at time 2 are summarized in Table 2.

Of the six rats that received 1064 mg/kg CBD, three rats (R4, R5, and R7) emitted at least 20% of responses on average on the CDP-associated lever, with a mean of 45.02% (SEM = 10.51, range = 25–100%). Between determinations, rats R4 and R5 displayed largely consistent selection of the CDP-associated lever (25–33.33%), whereas percent selection of the CDP-associated lever for rat R7 was 100% in the first determination and 61.76% in the second.

Of the six rats that received 1569 mg/kg CBD, three subjects (R3, R4, and R7) emitted at least 20% of responses on average on the CDP-associated lever, with a mean of 48.31% (SEM = 14.00, range = 4.35–100%). Between determinations, there was a higher degree of intrasubject variability compared to 1064 mg/kg CBD. In the first determination, percent selection of the CDP-associated lever was 100%, 20%, and 94.18% for rats R3, R4, and R7, respectively. In the second determination, percent selection of the CDP-associated lever was 21.43%, 4.35%, and 50% for rats R3, R4, and R7, respectively. Two subjects (R2 and R8) showed no substitution of CBD for CDP at either dose of CBD, with a maximum of 7.47% of responses occurring on the CDP-associated lever at 1064 mg/kg CBD and 2.78% of responses at 1569 mg/kg CBD.

Regression coefficients are provided for the model that allowed effects of dose to vary by subject (Table 5), which outperformed an intercept-only model (χ²(2) = 22.03, P < 0.001). As CBD dose increased, the odds of responding on the CDP-associated lever also increased (OR = 1.88, P < 0.001). There was an effect of test time (OR = 2.53, P < 0.001), with CDP-associated responding more likely to occur at time 2. There was no dose-by-test time interaction (OR = 0.86, P = NS).

Table 5.

Results of binomial logistic regression of CDP-associated responding during generalization tests

Test Drug	Variable	Odds ratio (exponentiated)	95% confidence interval	p-value
CDP	Dose	2.37 (10.70)	1.32, 3.81	<0.001
	Test time	−0.08 (0.92)	−0.19, −0.01	0.12
	Route of administration	−1.95 (0.14)	−2.08, −1.85	0.01
	Dose*Test time	−0.06 (0.94)	−0.19, 0.04	0.25
CBD	Dose	0.63 (1.88)	0.27, 1.10	<0.001
	Test time	0.93 (2.53)	0.69, 1.15	<0.001
	Dose*Test time	−0.15 (0.86)	−0.41, 0.12	0.25
d-Amp	Dose	0.66 (1.93)	−1.34, 6.74	0.23
	Test time	−0.14 (0.87)	−6.23, 9.67	0.77
	Dose*Test time	0.04 (1.04)	−5.19, 9.76	0.95

Note. Route of administration was included as a covariate for CDP only, as CDP was the only drug tested via two different routes of administration (i.p. and i.g.).

Response rates on both levers were calculated during both CBD generalization tests and are presented in Fig. 2 (filled and open squares). Average response rate remained near 60 responses/min at all doses of CBD at time 1 (Fig. 2, left panel). At time 2, average response rate was near 40 responses/min, and there was a slight decrease in response rate as doses increased (Fig. 2, right panel). Response rate data for CBD-generalization tests are summarized in Tables 3 and 4.

Negative control (d-amp) generalization

No substitution was observed for d-amp (0.3 and 1.0 mg/kg) at either test time. Mean percent responding on the CDP-associated lever remained well below 20% at 0.90% (SEM = 0.60%, n = 9) and 6.84% (SEM = 3.34%, n = 6) for 0.3 and 1.0 mg/kg d-amp, respectively, at time 1. At time 2, mean percent responding on the CDP-associated lever was 1.32% (SEM = 0.53%, n = 8) and 4.99% (SEM = 1.52%, n = 5) for 0.3 and 1.0 d-amp, respectively.

Regression coefficients are provided for the model that allowed effects of d-amp dose only to vary by subject (Table 5), which outperformed an intercept-only model (χ²(2) = 15.971, P < 0.001). There was no effect of d-amp dose (OR = 1.93, P = NS), test time (OR = 0.87, P = NS), or dose by test time interaction (OR = 1.04, P = NS).

Overall response rates on both levers during both d-amp generalization tests were calculated and are summarized in Tables 3 and 4 and presented in Fig. 2 (open circles). At time 1, average response rate on both levers was 48.58 resp/min (SEM = 7.75 resp/min, n = 11) and 21.87 resp/min (SEM = 12.51 resp/min, n = 6) for 0.3 and 1.0 mg/kg d-amp, respectively. At time 2, average response rate on both levers was 47.85 resp/min (SEM = 8.13 resp/min, n = 9) and 27.41 resp/min (SEM = 9.58 resp/min, n = 6) for 0.3 and 1.0 mg/kg, respectively.

Discussion

In the present experiment, partial substitution of CBD (1064 and 1569 mg/kg) was found at time 2 (120 min post-administration) in rats trained to discriminate 5.6 mg/kg CDP from saline. Dose-dependent substitution of CDP was observed at both test times, and neither vehicle (saline and MCT coconut oil) nor negative-control (d-amphetamine) compounds were found to substitute for 5.6 mg/kg CDP in either test.

That partial substitution was observed at time 2 but not time 1 suggests that the 90-min time point was too early in the time-course effect of CBD to capture any similar discriminative-stimulus effects between CDP and CBD. In three subjects for which partial substitution of CBD occurred at 1569 mg/kg, probe tests were conducted for this dose at longer time points (150 and 180 min post-administration) in an attempt to better capture substitution in the event that 120 min was not long enough, but no differences in substitution were observed between these more distal time points and 120 min (data not shown). These results are consistent with previous findings that plasma and brain levels and behavioral effects of intragastric CBD peak around 120 min post-administration (Deiana et al., 2012; Hložek et al., 2017).

The results of the present experiment indicate that at very high doses, over-the-counter CBD may share some stimulus properties with 5.6 mg/kg CDP These findings build on those of Gray et al., (2022), who found that doses of CBD up to 150 mg/kg (i.g.) did not increase responding on the benzodiazepine lever in rats trained to discriminate 0.5 mg/kg (i.p.) midazolam from saline. Aside from the differences in doses tested (the present experiment tested doses over 10 times higher than those tested in Gray et al., 2022), there are three primary differences in procedure that may have contributed to the contrasting findings: the training drug, the reinforcement schedule used in discrimination training, and the sources of the CBD.

Midazolam and CDP, though both members of the benzodiazepine class with similar pharmacodynamic profiles, have similar yet distinct discriminative-stimulus properties that may be related to their pharmacokinetic profiles (Woudenberg and Slangen, 1989). These pharmacokinetic differences make them appropriate for distinct clinical purposes: CDP is a longer-acting drug (half-life: 24–48 h; Ahwazi and Abdijadid, 2022) that tends to be prescribed used for the treatment of anxiety disorders (Woudenberg and Slangen, 1989; Ahwazi and Abdijadid, 2022), whereas midazolam is shorter-acting (half-life: 1.5–2.5 h; Lingamchetty et al., 2022) and is more commonly used to induce sleepiness and memory impairment before medical procedures (Woudenberg and Slangen, 1989; Lingamchetty et al., 2022). It is possible that the difference in stimulus properties and time-course effects between these two training drugs is responsible to a degree for the partial substitution of CBD for CDP in the present experiment, particularly given CBD’s long time-course effects when administered i.g. (Deiana et al., 2012; Hložek et al., 2017). Failure of CBD to substitute for midazolam in contrast to partial substitution for CDP suggests that CBD may share stimulus properties to a higher degree with CDP; a definitive conclusion cannot be drawn, however, as equivalent CBD doses were not tested in rats trained to discriminate midazolam from saline in Gray et al. (2022).

Another variable that can influence substitution of a test drug is the reinforcement schedule used when training drug discrimination. Gray et al., (2022) used an FR schedule, which has been shown to produce strong stimulus control but tends to engender quantal, or all-or-none, responding during the generalization test: all (or nearly all) of the responses occur on one lever (Stolerman, 1993). A TAND VI FR schedule, like the one used in the present experiment and others (Anderson and van Haaren, 1999; van Haaren et al., 1999), produces somewhat weaker stimulus control, though still sufficient to maintain correct responding across training sessions, and is more likely to produce graded responding in the generalization test, with responding distributed across the two levers at test doses that do not fully substitute for the training drug (Stolerman, 1993).

The third major distinction between the present study and Gray et al., (2022) is the provenance of the CBD used in testing. Gray et al., (2022) used the purified Epidiolex active ingredient, which consists of ≥98% CBD and ≤0.1% THC and has been approved by the FDA. The CBD oils used in this experiment were purchased over the counter from a local brick-and-mortar shop and an online retailer. Although mass-spectrometry paperwork obtained from the retailers, along with independent mass-spectrometry analyses conducted post hoc, indicated that the CBD oil used in the present experiment did not contain any unexpected adulterants, it remains that over-the-counter CBD products are far less regulated than Epidiolex and may contain additional compounds, like flavorants, terpenes, and carrier oils. Future research may attempt to replicate the present findings with similar doses of Epidiolex.

Relatedly, an interesting secondary finding of the present experiment was the under-labeling of the two CBD oils purchased. The oils from the local and the online retailer were labeled as containing 167 and 333 mg/ml, respectively, but independent mass spectrometry analysis (data not shown) suggested that each bottle contained significantly more CBD than was indicated on the label. This finding is consistent with several studies that have found under-labeling is a common phenomenon in the over-the-counter CBD market (e.g. Bonn-Miller et al., 2017; Gardener et al., 2022; Johnson et al., 2022; Miller et al., 2022), with estimates of under-labeled products ranging from 18% (Gardener et al., 2022) to nearly 43% (Bonn-Miller et al., 2017).

Limitations, future directions, and conclusion

That the CBD oils tested were unexpectedly much more concentrated than intended was both a benefit and a drawback to the present experiment. It provided the benefit of testing a range of doses not previously reported in the literature, and thereby some evidence that very high doses of CBD may share a low degree of stimulus properties with CDP. However, it also resulted in doses exceeding those previously reported to be anxiolytic in both rats and humans, which complicates comparison to prior research into CBD’s anxiolytic effects. Previously, systemic CBD’s effects on anxiety have been described with an inverted U-shaped dose-response function (in rats: Guimarães et al., 1990; in humans: (Linares et al., 2019), with higher doses (in rats: > 10 mg/kg; in humans: > 300 mg/kg) ineffective at reducing anxiety-like behavior in the elevated plus-maze (Guimarães et al., 1990) or symptoms of anxiety during a simulated public-speaking task (Linares et al., 2019). For comparison, the highest dose tested in the present experiment (1569 mg/kg) converts to about 125 g of CBD for an 80-kg human. Although inverted U-shaped functions are relatively uncommon in drug-discrimination experiments, future research may test lower doses of CBD that more closely align with those previously reported as effective for reducing anxiety.

Although there was evidence of some degree of shared stimulus properties between CBD and CDP, it was low under the conditions of this experiment. The present findings, along with those of Gray et al., (2022), suggest that CBD alone is not a suitable substitute for benzodiazepines. However, if CBD was found to be able to supplement benzodiazepines in anxiety treatment, CBD could potentially be used as an adjunctive treatment. Future directions for this research include drug-drug interaction experiments, in which sub-training doses of benzodiazepines that do not engender full substitution alone are combined with high doses of CBD that produce partial substitution to evaluate potential interactive effects of CBD and benzodiazepines. Additive effects occur when two drugs administered together produce an effect that is equivalent to the individual effects exerted by each drug (Niu, et al., 2019). Synergistic effects occur when the combined effect of two drugs supersedes what would be expected based on the individual effect of each drug (Niu et al., 2019). Drug-drug interaction experiments regarding combinations of CBD and clobazam, a benzodiazepine used to treat severe seizure disorders, have produced mixed findings regarding whether the two drugs interact to increase anticonvulsive effects (Anderson et al., 2019; Gaston et al., 2019). Research regarding interaction effects of CBD and benzodiazepines with regard to anxiolytic effects may be warranted.

The results of the present experiment suggest that CBD at lower doses does not ‘feel’ like (i.e. have stimulus properties that generalize to) 5.6 mg/kg CDP. As differences in stimulus properties may suggest differences in pharmacological mechanisms of action (Andrews and Stephens, 1990; Young, 2009; McMahon, 2015), the relatively low degree of substitution observed of CBD for CDP may indicate other mechanisms of action responsible for the purported benzodiazepine-like effects of CBD. Evidence indicates that CBD’s action at the 5-HT_1A receptor may contribute to its anxiolytic effects (de Almeida and Devi, 2020; De Gregorio et al., 2019; García-Gutiérrez et al., 2020; see Britch et al., 2021), which suggests a need for future drug-discrimination experiments with CBD and anxiolytic/antidepressants from other drug classes such as selective serotonin reuptake inhibitors, some of which have previously been used as training drugs in drug-discrimination experiments (e.g. sertraline, Lucki et al., 1994; Marona-Lewicka and Nichols, 1998; citalopram, Marona-Lewicka and Nichols, 1998; fluvoxamine, Olivier et al., 1993).

The statistical approach used in the present experiment was chosen as previous work has demonstrated via simulation that using inappropriate statistical models for choice data can drastically increase the risk of false positives (Frankot et al., 2023). Although meant as a complement to the more traditional analyses used in drug-discrimination research, the inferential statistical analyses are not without some limitations. First, the data have a nested structure, with responses nested within sessions and sessions nested within individual subjects. A multi-level analysis was used to account for nestedness; however, differing approaches to calculating degrees of freedom in a multilevel model can impact P-values, with some methods inflating the likelihood of type I error, particularly at small sample sizes (Luke, 2017). Second, there was a non-linear relation between dose and percent CDP-associated responding, which can limit the use of certain parametric tests. Lastly, the sample size may have limited statistical power to detect differences across doses. Here, a generalized linear mixed-effects model was used to account for non-linearities (Young, 2018). Bootstrapping was chosen to help mitigate bias of parameter estimates, although error rates increase at smaller sample sizes (Luke, 2017). As the statistical analysis is not without limitations, over-interpretation should be avoided—it should be considered complementary (DeHart and Kaplan, 2019) to the traditional drug-discrimination analysis also used in the present experiment, in which <20% responding on the drug-associated lever is interpreted as no substitution, 20–80% is interpreted as partial substitution, and >80% is interpreted as full substitution.

Given the public’s interest in CBD products as alternatives to traditional pharmacological anxiety treatments, the present experiment was designed to evaluate stimulus properties of CBD in comparison to CDP. Although a low degree of partial substitution was observed at doses above 1000 mg/kg, these data do not presently support the use of CBD as a replacement for benzodiazepine medications. The CBD doses that were found to engender partial substitution are several times higher than approved therapeutic doses of Epidiolex and the doses offered by over-the-counter CBD products, which would make for a particularly cost-ineffective anxiety treatment. In summary, the present finding that CBD substitutes to any degree for a benzodiazepine is a relatively novel one that warrants further research.