Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Oct 28.
Published in final edited form as: J Pharmacol Exp Ther. 2025 Jun 6;392(7):103622. doi: 10.1016/j.jpet.2025.103622

Subjective drug-effect ratings predict cannabis self-administration in people who use cannabis daily

Thomas P Shellenberg a,b, Justin C Strickland d, Sean D Regnier a, Preston T Tolbert a,e, Stephanie Lake f, Michael J Wesley a,b, William W Stoops a,b,c, Ziva D Cooper f, Margaret Haney g, Joshua A Lile a,b,c
PMCID: PMC12556786  NIHMSID: NIHMS2116024  PMID: 40570551

Abstract

As interest in the potential therapeutic benefits of cannabis-derived products grows, accurate predictors of abuse potential will be vital for informing regulatory decisions. Currently, the Food and Drug Administration recommends using subjective effect ratings of Drug Liking as the primary measure in human abuse potential studies. However, dissociations between subjective ratings and drug-taking behavior have been previously reported. This retrospective analysis determined if any subjective effects questionnaire items uniquely predicted cannabis self-administration in people who use cannabis daily (N=89; 71 male and 18 female). Data from four previous studies across two research sites that included cannabis self-administration procedures were combined. Cannabis concentrations were similar across studies (5.5%, 5.6%, or 5.9% THC), though mg THC dose varied based on administration procedures. Ratings of Good Effect, High, Sedated/Tired, Drug Liking, Stimulated, and Willingness to Take Again were harmonized across studies and used as predictors. In each study, a money option ($0.50 or $1.00) was scheduled as an alternative to cannabis puffs in self-administration procedures. Proportion of choices for cannabis puffs over total choice trials (3 or 8 trials) was used as the primary outcome. Generalized linear models revealed that higher ratings of Willingness to Take Again (OR = 1.04) were associated with increased odds of self-administering active THC cannabis, while higher ratings of Stimulated were associated with decreased odds of self-administering placebo cannabis (OR = 0.94). These results suggest that subjective ratings of Willingness to Take Again should be considered as a primary outcome when assessing abuse potential for novel cannabinoid products.

Keywords: cannabis self-administration, subjective drug-effect ratings, human abuse potential studies

Introduction

In May 2024, the US Drug Enforcement Agency recommended the reclassification of cannabis from a schedule I to a schedule III substance under the Controlled Substances Act (CSA) following a comprehensive review by the United States (US) Department of Health and Human Services, signaling a potential historic shift in the regulatory landscape for cannabis in the US. This change is accompanied by growing interest in the potential therapeutic effects of cannabis products and cannabinoid drugs (Legare et al., 2022). For example, the synthetic cannabinoid agonist drugs dronabinol and nabilone, which produce their therapeutic effects largely via stimulation of CB1 receptors, are currently approved as Schedule II substances under the CSA by the US Food and Drug Administration (FDA) for the treatment of anorexia, weight loss in AIDS patients, and chemotherapy-induced nausea and vomiting. These and other cannabinoid compounds have been evaluated for additional indications like chronic pain, sleep disorders, and various mental health conditions (see reviews in AminiLari et al., 2022; Aviram and Samuelly-Leichtag, 2017; Walsh et al., 2017). Given the growth of research and development of cannabinoid therapeutics, evidence-based assessments of abuse potential of new cannabinoid products will be critical for determining appropriate regulatory control to mitigate risks to public health.

The current FDA guidance defines drug abuse as the “intentional, non-therapeutic use of a drug product or substance, even once, to achieve a desired psychological or physiological effect,” with abuse potential being the likelihood of drug abuse occuring (FDA, 2017; Schuster and Henningfield, 2003; although “abuse” can be considered stigmatizing language, it is used here for consistency with the FDA technical definition). Traditionally, human laboratory studies have been influential in informing policy decisions for drugs with abuse potential. Self-administration procedures are widely considered the most valid means of assessing the abuse liability of drugs because they directly measure drug-taking behavior (Comer et al., 2008; Haney and Spealman, 2008). However, these procedures are more time consuming, require additional drug exposure, and can be difficult to implement, limiting their feasibility. As an alternative, researchers have used subjective drug-effect ratings to predict whether a drug is likely to be abused (Fischman, 1989; Fischman and Foltin, 1991). Typically, these measures instruct participants to indicate the extent to which they are experiencing interoceptive drug effects by rating a series of words or short phrases. For example, administration of delta9-tetrahydrocannabinol (hereafter “THC”) reliably increases subjective ratings of Positive Mood, Good Effect, High, Drug Liking, and Willingness to Take Again (e.g., Chait and Zacny, 1992; Haney et al., 2016; Zacny and de Wit, 1991). Subjective drug effects questionnaires often use visual analog scales (VAS) which contain response options ranging from 0–100 to capture the intensity of the subjective experience, but other scales (e.g., Likert, true/false) have also been used.

Highlighting the importance of subjective effects questionnaires in regulatory decisions, the FDA guidance for industry on the assessment of the abuse potential of drugs recommends a Drug Liking VAS as the primary measure for human abuse potential (HAP) studies (FDA, 2017). Specifically, the guidance document recommends first using the Drug Liking measure to screen people who use drugs recreationally for participation by requiring a meaningfully different self-reported response (i.e., 15-point difference) between double-blind administration of a positive control drug and placebo (Qualification Phase). Those who qualify proceed onto a Treatment Phase in which a range of at least three doses of the test drug are administered, along with one or two doses of the positive control drug and placebo. The FDA further recommends that abuse liability be determined by using linear mixed-effect models to test differences between means of peak Drug Liking effects produced by the test drug, the positive control drug, and placebo. Other VAS ratings such as High, Stoned, and Take Drug Again may also be included as secondary measures.

Although subjective drug experiences are clearly related to subsequent use, human laboratory studies that include self-administration procedures have reported dissociations between subjective responses and drug-taking behavior (Fischman, 1989; Haney et al., 1999; Haney et al., 2015; Hart et al., 2004; Lamb et al., 1991; Sevak et al., 2010; Stoops et al., 2012; Wachtel and de Wit, 2000). For example, across several early experiments, self-reported cannabis effects were not predictive of, or concordant with, smoked cannabis choices (Chait & Zacny, 1992; Kelly et al., 1994, 1997; Zacny & de Wit, 1991). Zacny and de Wit (1991) found that only subjective ratings were sensitive to different THC concentrations (0.0%, 0.8%, 3.6%) whereas cannabis self-administration was not. On the other hand, Kelly and colleagues (1997) showed that only cannabis self-administration was sensitive across active THC concentrations (2.0% and 3.5%), whereas subjective responses were similar. The reasons for these inconsistencies are unclear, but could be due to testing small ranges of relatively low concentration cannabis and/or small sample sizes (i.e., 5–10 participants).

The present report describes a retrospective analysis that sought to address some of the limitations of the prior research and determine whether specific subjective effects questionnaire items uniquely predict choices to self-administer cannabis under double-blind conditions which would inform future refinements of HAP studies to improve prediction of abuse potential. We combined data from four previous studies conducted at two US research facilities that administered higher strength cannabis than prior work, and administered subjective effects questionnaires and cannabis self-administration choice procedures in people who reported daily cannabis use (Cooper et al., 2018; Haney et al., 2015; Lile et al., 2022; Wesley et al., 2018).

Methods

Data Sources

Table 1 provides a description of the procedures used in each of the four studies that contributed data to this analysis. Only data that were harmonizable across all 4 studies were used for the analyses and reported in the results (Table 2). Study 1 (n = 37) determined the influence of oral naltrexone maintenance on the reinforcing and subjective effects of smoked cannabis (0.0 and 5.5% THC; Haney et al., 2015). Study 2 (n = 16) measured the effects of concurrently administered smoked cannabis (0.0 and 5.6% THC) and oral oxycodone on measures of analgesia, subjective response and self-administration (Cooper et al., 2018). The last two studies determined the influence of maintenance on tiagabine (Study 3; n = 12; Wesley et al., 2018) or pregabalin (Study 4; n = 24; Lile et al., 2022) on the reinforcing and subjective response to smoked cannabis (0.0 and 5.9% THC). For the current analysis, we only included data from sessions in which placebo cannabis or cannabis containing active THC was administered in the absence of other active study drugs (e.g., we excluded data in which cannabis was combined with naltrexone in Study 1). The Institutional Review Boards at New York State Psychiatric Institute (Studies 1 & 2) and the University of Kentucky (Studies 3 & 4) approved the conduct of the respective protocols. All participants provided their written, informed consent before completing any study procedures, which were carried out in compliance with relevant laws and instituitional guidelines.

Table 1.

Study Information

# Study N (total = 89) Active THC Concentration VAS timepoints (minutes after sampling) Self-Administration Timepoint Self-Administration Procedure Number of Choice Trials Alternative Reinforcer Magnitude
1 Haney et al., 2015 37 5.5% Post-dose, 15, 30, 60, 90, 120 165 minutes 3-Choice 3 $1
2 Cooper et al., 2017 16 5.6% 15, 30, 60, 90, 120, 180 195 minutes 3-Choice 3 $1
3 Wesley et al., 2018 12 5.9% Post-dose, 15, 30, 45, 60, 90, 120, 180, 240 Approximately 24 hours Independent, concurrent choice (Fixed-Ratio 1000 schedule) 8 $0.50
4 Lile et al., 2021 24 5.9% Post-dose, 15, 30, 45, 60, 90, 120, 180, 240 Approximately 24 hours Independent, concurrent choice (Progressive Ratio schedule) 8 $0.50
Open in a new tab

Note. All cannabis was administered via smoking using the uniform, cued-puff procedure (Foltin et al., 1987).

Table 2.

Harmonized Variables Across Included Studies

Variable Type Variables
Demographic Age
Sex
Race
Years of Education
Substance Use Drinking Days Per Week (Alcohol)
Cigarettes Per Day
Monthly Cannabis Use
Age First Used Cannabis
Self-Administration Number of Drug Choices Made
Number of Money Choices Made
Subjective Effects Good Effect
High
Drug Liking
Willingness to Take Again
Sedated/Tired
Stimulated
Open in a new tab

Participants

Across the four studies, 89 unique healthy participants (71 male and 18 female), aged 18–48 years old, who reported daily (or near daily) cannabis use as defined by at least 25 days per month (Budney et al., 2007) were included in the final analysis. This subject sample aligns with HAP study guidance that participants should be “experienced recreational drug users who have a recent history of using drugs in the same general pharmacological class as the test drug” (FDA, 2017). To determine health status, participants across all studies were screened for their medical and substance use history via physical examinations, psychiatric assessments, electrocardiogram, urinalysis and blood chemistry. Participants with a serious physical disease or Axis I disorder according to DSM-IV criteria, other than nicotine and/or cannabis dependence, were excluded. All participants were also required to not be regularly using substances other than cannabis, alcohol, nicotine or caffeine and not be seeking treatment for their cannabis use. Recent substance use was biologically verified using urine toxicology screening. The screening procedures and additional inclusion/exclusion criteria are described in greater detail in the original manuscripts. If participants participated in more than one study, only the data from their first participation were included.

General Procedures

All studies were conducted on an outpatient basis. Participants were required to abstain from illicit substances besides cannabis, which was biologically verified via urine screening throughout their participation. Urine samples from female participants were also tested for pregnancy throughout participation. Participants completed a similar pre-session process across the studies which included sobriety testing and consumption of a standardized breakfast prior to completing experimental procedures. In Studies 1 & 2, participants were allowed a maximum of three tobacco cigarettes at predetermined intervals throughout the session and participants in Studies 3 & 4 were allowed one tobacco cigarette before beginning the session and were not permitted to smoke during the session.

Study Drug

Active (5.5, 5.6, or 5.9% THC) and placebo (0.0% THC) cannabis cigarettes (800 mg total weight) were provided by the National Institute of Drug Abuse for all studies. Cannabis cigarettes were stored frozen in an airtight container and humidified at room temperature for 24 h prior to the session.

Experimental Procedures

Cannabis Administration

In each study, experimenter- and self-administered cannabis was delivered by the smoked route using a uniform, cued-puff procedure (Foltin et al., 1987). Participants were first instructed to light the cannabis cigarette to prepare for the procedure. Then participants were instructed to inhale (5 s), hold breath (10 s), then exhale and rest (40–45 s) before taking the next puff. This puff procedure was repeated until the number of puffs required to complete the sampling or self-administration procedure requirement had been met (further described below).

Prior to completing self-administration procedures, participants sampled smoked cannabis cigarettes containing either placebo (0.0% THC) or active cannabis (5.5, 5.6, or 5.9% THC) to orient them to the dose effects and provide a fixed dose for subjective effects and other outcome assessments. In Study 1, participants repeated the puff procedure until 50% of the cannabis cigarette has been smoked. In Study 2, the puff procedure was repeated until 70% of the cigarette had been smoked. In Studies 3 & 4, the puff procedure was repeated for 4 puffs from two identical cannabis cigarettes for a total of 8 puffs (approximately 50–70% of each cigarette). Thus, although the same approximate THC concentration was used across studies, the individual study sampling procedures yielded a range of THC doses (Study 1 = 22 mg; Study 2 = 31 mg; Studies 3 & 4 = 47–66 mg; note that these calculations do not account for drug loss to side-stream smoke).

Subjective Effects Questionnaires

Immediately following cannabis sampling and at predetermined timepoints throughout the sessions (see Table 1), participants completed subjective effects questionnaires that used a 0–100 unipolar VAS in which 0 was anchored by the descriptor Not at All and 100 was anchored by the descriptor Extremely. For Studies 1 & 2, these measures were collected from a locally developed Subject Effects Visual Analog Scale questionnaire and a Cannabis Rating Form (originally reported in Haney et al., 1999). In Studies 3 & 4, a locally developed VAS Subject-Rated Drug-Effect Questionnaire (Lile et al., 2012) was used to collect these measures. Ratings of Good Effect, High, Sedated/Tired, Drug Liking, Stimulated, and Willingness to Take Again were harmonizable across all studies and selected for the present analysis.

Self-Administration Procedures

Self-administration choice procedures were used to measure the reinforcing effects of the previously sampled cannabis. In Study 1, sampling occurred 165 minutes prior, and in Study 2 sampling occurred 195 prior, to the self-administration choice procedure. In Studies 3 & 4, sampling occurred approximately 24 hours prior to the self-administration choice procedure. Studies 1 & 2 used a 3-puff choice procedure in which participants could purchase up to 3 puffs of the same cannabis they sampled earlier in the same day for $1 each. Given these parameters, a maximum of 3 additional puffs or $3.00 was available during these procedures. Participants self-administered puffs of cannabis they earned using the cued-puff smoking procedure immediately after making their choices.

In Study 3 and 4, participants completed response requirements via computer mouse clicks to earn individual puffs of cannabis they sampled from the previous day, or an alternative monetary reinforcer ($0.50) over 8 discrete trials according to concurrent, independent fixed-ratio (Study 3) or progressive-ratio (Study 4) reinforcement schedules. A maximum of 8 puffs or $4.00 was available during these procedures. The fixed-ratio schedule required 1000 clicks in each trial to earn the chosen reinforcer. For the progressive-ratio schedule, the initial response requirement for either reinforcer was 400 clicks. The completion of a response requirement for a given reinforcer increased the response requirement for that reinforcer in the next trial by 200 clicks. After completing all trials, participants self-administered puffs of cannabis they earned using the puff procedure as described above.

Data Analysis

Descriptive statistics were used to describe the demographic and substance use history of the combined sample and are reported as means or percentages. Subjective effects data were analyzed as peak response during sampling (i.e., the rating with largest magnitude at any timepoint) for each questionnaire item. Self-administration data were analyzed as the proportion of cannabis choices, which was calculated by dividing the number of cannabis choices made by the total number of choice trials available in each respective self-administration session.

A Wilcoxon Signed-Rank Test was used to determine if the proportion of choices to self-administer placebo and active THC cannabis differed. Differences in mean peak subjective effects produced by placebo and active THC cannabis were assessed using paired-sample t-tests.

Generalized linear models (GLM) determined whether the response to experimenter-administered cannabis, measured by each subjective effects questionnaire item, uniquely predicted subsequent choices to self-administer cannabis. Separate GLMs were fit for the active THC and placebo data using a binomial distribution with logit link function. GLMs are appropriate for handling the non-normal response distribution of the proportional outcome variable used in the analysis (Dobson and Barnett, 2018). Peak effect ratings for each subjective effects questionnaire item were used as the primary predictors in all models and proportion of cannabis choices as the primary outcome. To adjust model estimates for potential confounding effects, age and sex at birth were also included in all models as covariates. Additionally, the study origin was included as a categorical covariate in the model to adjust for differences due to study procedures (i.e., sampling doses, reinforcement schedules, study site, etc.).

Coefficient estimates for fixed effects were converted to and are reported as odds ratios (ORs) with 95% confidence intervals (CIs). Full models were fit that included all predictors to calculate ORs that accounted for the effect of each subjective effect rating to determine if any item uniquely predicted proportion of cannabis choice. Before interpreting results, we a priori decided to exclude predictors that showed high multicollinearity based on a Variance Inflation Factor (VIF) > 10 to avoid unstable coefficient estimates and inflated standard errors (Gareth et al., 2013). To facilitate interpretation, the OR estimates from the model were raised to the power of ten to estimate the effect of a ten-unit increase in VAS rating on cannabis choice, respectively. All ORs can be interpreted as the change in odds of making a cannabis choice during the self-administration procedure for every one unit and ten-unit increase in the peak effects scores from each subjective effects questionnaire item during sampling.

All data procedures and analyses were conducted in R Statistical Language using publicly available package stats (R Core Team, 2023). Figures were generated in R Studio using the publicly available packages ggplot2 (Wickham, 2016) and sjPlot (Ludecke, 2023), and GraphPad Prism 10. Statistical outcomes were considered significant when p < 0.05. The R code used to conduct the analysis is available upon request.

Results

Demographics and Substance Use History

Table 3 shows the demographic and substance use characteristics of the combined sample.

Table 3.

Demographic and Substance Use Characterisitics

Mean (SD)/%
Demographics
 Age 27.7 (6.7)
 Female 20.2%
 White 72.2%
 Years of Education 12.9 (2.0)
Cannabis Use
 Monthly Cannabis Use (Days) 29.1 (2.0)
 Age of First Cannabis Use 14.6 (2.8)
Other Substance Use
 Weekly Alcohol Use (Days) 1.4 (1.8)
 Any Past-Month Tobacco Use 60.0%
 Cigarettes Per Day 4.8 (5.9)
Open in a new tab

Cannabis Choice

Participants made a significantly higher proportion of choices to self-administer active THC cannabis relative to placebo cannabis (V = 174.5, p < .001; Figure 1).

Figure 1. Cannabis Choice.

Figure 1.

Open in a new tab

Note. Plotted are the mean proportion of choices for active cannabis compared to placebo cannabis. The proportion of choices to self-administer active cannabis was significantly greater than placebo THC. ***p < .001.

Subjective Effects Ratings

Active cannabis produced significantly greater ratings of Good Effect, High, Drug Liking, Stimulated, and Willingness to Take Again when compared to placebo (t values > 7.1, p values <.001, Cohen’s ds = 0.63 – 1.54; Figure 2). Ratings of Sedated/Tired did not differ between the two doses and were therefore excluded from subsequent modeling analyses.

Figure 2. Peak Effects of THC on Subjective Effects Ratings.

Figure 2.

Open in a new tab

Note. Plotted are the mean peak effect ratings for each subjective effect item following placebo (white) and active cannabis (gray) administration. Active cannabis significantly increased ratings of Good Effect, High, Drug Liking, Stimulated, and Willingness to Take Again. Sedated/Tired did not differ as function of active THC. Standardized mean differences are reported as Cohen’s d above comparisons that reached statistical significance. ***p < .001.

Relationship Between Subjective Effects Ratings and Cannabis Choice

Active THC Cannabis

Each subjective effects questionnaire item had VIFs less than 10 in the full model, so all were included in the GLM for the active cannabis data. Table 4 shows the model-derived ORs with 95% CI for questionnaire items in terms of 1-unit and 10-unit increases in VAS ratings. According to the model, there was a significant relationship between ratings of Willingness to Take Again during sampling and subsequent cannabis self-administration (p = 0.027). Specifically, the OR estimate indicated that a ten-unit increase on Willingness to Take Again was associated with a 52% increase in the odds of making a cannabis choice (Figure 3). Ratings of Good Effect, High, Drug Liking, and Stimulated had no statistically significant effect on the likelihood self-administering active cannabis.

Table 4.

Results from Generalized Linear Models

Predictor OR1 (95% CI) OR10 (95% CI) p
Active THC
(Intercept) 0.45 (0.03–5.71) 0.45 (0.03–5.71) 0.537
Willingness to Take Again 1.04 (1.01–1.09) 1.52 (1.08–2.28) 0.027
Drug Liking 0.98 (0.94–1.02) 0.84 (0.56–1.22) 0.375
High 0.99 (0.96–1.02) 0.91 (0.67–1.23) 0.547
Good Effect 1.00 (0.98–1.03) 1.04 (0.80–1.34) 0.785
Stimulated 0.99 (0.97–1.01) 0.92 (0.77–1.07) 0.284
Covariates
Age 1.00 (0.92–1.07) 1.00 (0.92–1.07) 0.909
Female (REF= Male) 0.85 (0.26–2.74) 0.85 (0.26–2.74) 0.790
Study 2 (REF = Study 1) 0.53 (0.07–3.53) 0.53 (0.07–3.53) 0.524
Study 3 2.23 (0.62–8.77) 2.23 (0.62–8.77) 0.231
Study 4 1.78 (0.41–8.71) 1.78 (0.41–8.71) 0.453
Placebo THC
(Intercept) 0.51 (0.02–13.40) 0.51 (0.02–13.40) 0.687
Stimulated 0.94 (0.88–0.98) 0.52 (0.27–0.81) 0.017
Drug Liking 1.00 (0.95–1.05) 0.99 (0.61–1.63) 0.984
High 1.02 (0.97–1.08) 1.22 (0.74–2.07) 0.439
Good Effect 1.03 (0.99–1.07) 1.34 (0.91–2.02) 0.134
Willingness to Take Again 1.03 (0.99–1.07) 1.32 (0.93–1.98) 0.144
Covariates
Age 0.98 (0.88–1.08) 0.98 (0.88–1.08) 0.656
Female (REF= Male) 1.00 (0.16–4.82) 1.00 (0.16–4.82) 0.995
Study 2 (REF = Study 1) 0.36 (0.03–2.70) 0.36 (0.03–2.70) 0.348
Study 3 1.39 (0.28–7.75) 1.39 (0.28–7.75) 0.690
Study 4 0.69 (0.09–4.69) 0.69 (0.09–4.69) 0.704
Open in a new tab

Note. All Odds Ratios are adjusted for the effect of age, sex, and study of origin. Odds ratios were also raised to 10th power to describe the effect of an 10-unit increase in the predictor on the odds of making a cannabis choice. Bold formatted text indicates statistical significance (p < .05). REF = reference group for categorical variables in the model. OR = Odds Ratio.

Figure 3. Model-Predicted Probability Curves for Significant Predictors.

Figure 3.

Open in a new tab

Note. Plotted are the model-derived marginal effects for Willingness to Take Again (top panel) and Stimulated (bottom panel). Gray ribbons are 95% prediction intervals for proportion of cannabis choices at given Visual Analog Scale ratings.

Placebo Cannabis

Each subjective effects questionnaire item had VIFs less than 10, so all were included in the GLM for placebo cannabis. Table 4 shows the model-derived Odds Ratios with 95% CI for questionnaire items in terms of 1-unit and 10-unit increases in VAS ratings. In the model, only ratings of Stimulated were significantly associated with the odds of making a choice to self-administer placebo cannabis (p = 0.017). Specifically, a ten-unit increase in ratings of Stimulated was associated with a 52% decrease in the odds of making of making a cannabis choice (Figure 3). Ratings of Good Effect, High, Drug Liking, and Willingness to Take Again had no statistically significant effect on the likelihood self-administering placebo cannabis.

Discussion

HAP studies are only one, albeit highly influential, component of what determines a new drug product’s overall abuse potential. According to FDA guidance, a broad range of evidence from studies of chemistry, pharmacology, pharmacokinetics, abuse-related adverse events, and behavior, as well as reports from law enforcement officials and medical professionals should be compiled for the abuse potential assessment that is submitted in a New Drug Application (FDA, 2017). If the test drug is approved, information from the application is used to guide decisions regarding the appropriate labeling and scheduling of the drug under the CSA. As interest in cannabinoid-derived therapeutics continues to grow, strong evidence-based assessments of abuse liability will be increasingly important for determining appropriate labeling and scheduling for mitigating potential harms to public health (Cooper and Abrams, 2019).

The present analysis found that higher ratings of Willingness to Take Again following the experimental administration of active cannabis predicted a greater proportion of choices to self-administer the same type of cannabis later. This result supports the use of this questionnaire item to predict choices to use cannabis again, which aligns with the FDA definitions of drug abuse and abuse potential. On the other hand, ratings of Drug Liking did not serve as a unique predictor of cannabis choices. Because active cannabis administration consistenly increases ratings of Good Effect, High, and Drug Liking, and given the emphasis on Drug Liking in the FDA guidance, it was somewhat surprising that these questionnaire items did not serve as predictors of cannabis self-administration. Instead, these findings suggests that Willingness to Take Again should be included or considered as a replacement for Drug Liking, or at least as a complementary primary measure, in the FDA guidance for HAP studies of novel cannabinoid products.

For placebo cannabis, higher ratings of Stimulated predicted fewer choices, suggesting that this questionnaire item might have captured an aversive experience for individuals who regularly smoke cannabis, akin to feeling anxious, jittery or nervous. The reason why placebo cannabis increased ratings of Stimulated in some participants is unclear, though it is worth noting that the magnitude of this response to placebo was relatively low with average mean peak ratings of approximately 20 out of 100 on the VAS. One possibility is that placebo administration might be eliciting a conditioned opponent response (e.g., McCaul et al., 1989; Newlin, 1986), in which the prediction of positive cannabis effects like euphoria and/or relaxation resulted in a compensatory response of feeling Stimulated that was unpleasant for some individuals. This possibility is highly speculatative and additional research is needed to elucidate why increased ratings of Stimulated were associated with less placebo self-administration.

To our knowledge, this is the first study to determine the correspondence between the subjective and reinforcing effects of cannabis in experimental contexts in people who use cannabis daily. However, several observational studies have examined the initial subjective experience with cannabis as a predictor of developing problematic use later. In general, this literature suggests that positive effects experienced during initial cannabis exposure are indeed associated with increased use and higher risks of developing cannabis use problems. For example, two studies found that those who experienced positive effects from cannabis use (e.g., feeling high, happy, and relaxed) at ages 14–16 years old were more likely to be dependent on cannabis at ages 16–21 years old, even after adjusting for numerous confounding factors (Fergusson et al., 2003; Le Strat et al., 2009). Other studies using latent class and confirmatory factor analysis techniques found diverse classes of subjective responses to cannabis that were associated with varying risk for developing problematic cannabis use. In particular, those with who reported a more positive initial subjective experience generally were shown to have more problematic cannabis use (Scherrer et al., 2009; Zeiger et al., 2010). These findings support the predictive validity of positive subjective effects as a risk factor for subsequent cannabis use. However, comparable observational data are not usually available for novel drug products and instead HAP assessments are needed to inform regulatory policies.

Previous analyses of data from different drug classes have also systematically examined the correspondence between subjective ratings and drug-taking behavior (Bolin et al., 2013; Li et al., 2020; Murray et al., 2021; Strickland et al., 2014). Two prior reports found that choices to self-administer oral d-amphetamine were associated with subject ratings of Positive Mood, Friendly, Drug Liking, and other positively-valenced subjective effect measures based on bivariate correlations and repeated-measures ANOVA (Bolin et al., 2013; Murray et al., 2021). In addition, Strickland and colleagues (2014) found that choices to use intranasal cocaine were positively correlated with 11 subject ratings and best predicted by ratings of Like Drug, Performance Improved, and Rush using an model selection technique based on Akaike Information Criteria. Moreover, Li and colleagues (2020) demonstrated that feeling Sedated was associated with decreased odds of choosing to drink alcohol. Although each of these studies found that specific subjective effects questionnaire item ratings were associated with drug-taking, they did not use consistent methods or analytical approaches, which makes comparisons difficult. Further, these studies did not all include a Willingness to Take Again item. Given that subjective effects tend to increase simultaneously (i.e., collinear), we took steps to ensure that predictors in the GLMs were contributing enough unique variance to cannabis choices based on VIFs. This approach was determined to be the most conservative and informative for addressing which subjective items are optimal for abuse potential assessments based on their capability to uniquely predict choices to self-administer cannabis above and beyond other subjective effect items. Future studies could use more consistent subjective effect measures, self-administration procedures, and similar statistical models that were used in this analysis to provide a clearer account of the relationship between subjective effect ratings and drug-taking across a range of drug classes and study samples (e.g., naïve individuals, occasional users, etc.).

The present study had some limitations that should be noted. First, data from only one approximate strength of active THC cannabis was included in the analysis, which precluded a within-subject dose-effect analysis of the relationship between subjective effects ratings and cannabis self-administration. HAP study guidance indicates that a dose of the positive control drug (e.g., active cannabis) that reliably produces higher responses on prototypical measures when compared to placebo should be selected. The cannabis concentration (approximately 6% THC) from the present analysis clearly produced greater subjective and reinforcing effects than placebo. However, cannabis containing other THC concentrations (e.g., 3% THC) have also consistently been shown to produce reliable increases in positive subjective effect ratings and function as a reinforcer (Haney et al., 1997; Kelly et al., 1997; Mendelson and Mello, 1984; Ward et al., 1997), and inclusion of data from lower (or higher) strength cannabis might have yielded different results. Although within-subject dose-effect analyses could not be conducted, differences in the sampling procedures yielded a range of estimated administered THC doses (22–66 mg) across studies, which was included as a model covariate. This lack of harmonization across studies is a limitation, but the results suggest that the relationship between subjective and reinforcing effects applies across a range of doses.

Another possible limitation is that the current study only included smoked cannabis using a standardized puff procedure which might limit the utility of these results for informing HAP studies of novel cannabinoid products and other modes of administration. Although some therapeutic products are administered by the intrapulmonary route (e.g., inhaler), most existing and novel drug products are administered orally because of its advantages, including superior patient compliance, non-invasiveness, and convenience (Alqahtani et al., 2021). Past studies have shown that although oral and smoked cannabis produce a subjective experience, smoked cannabis has been shown to produce higher peak effects over a shorter time course when compared to oral THC (Chait and Zacny, 1992; Hart et al., 2002; Wachtel et al., 2002). Given these differences, it unclear whether the present findings translate to orally admininstered cannabis and/or other THC and THC-like drugs (e.g., dronabinol, nabilone) which opens a possible direction for future research.

Conclusions

The present study found that higher ratings of Willingness to Take Again predicted more self-administration of active cannabis and higher ratings of Stimulated predicted less self-administration of placebo cannabis in people who used cannabis daily. Although ratings of Drug Liking provide useful descriptive information about the abuse liability of various drugs, it does not appear to be the most predictive measure for the abuse of cannabinoid products among people who use cannabis daily. Instead, these results suggest that Willingness to Take Again might serve as a more predictive one-item assessment for HAP studies of cannabinoid products. Future studies should adopt a similar statistical approach to determine if these findings generalize to other drug classes and study samples.

Significance Statement.

This retrospective analysis found that subjective ratings of Willingness to Take Again was a more predictive measure of cannabis self-administration than Drug Liking in people who use cannabis daily. Refining human abuse potential assessments to prioritize measures that better align with drug-taking behavior could improve regulatory evaluations of novel cannabinoid products.

Financial Support

This work was supported by funding from the National Institutes of Health National Institute on Drug Abuse (R01 DA019239; P50 DA009236; R01 DA047296; R01 DA057252; R01 DA036550; R01 DA025605; T32 DA035200). SL is supported through a Canadian Institutes of Health Research Banting Postdoctoral Fellowship.

Abbreviations:

CI

confidence interval

CSA

Controlled Substances Act

FDA

Food and Drug Administration

GLM

generalized linear model

HAP

human abuse potential

OR

odds ratio

THC

delta9-tetrahydrocannabinol

VAS

visual analog scale

Footnotes

Conflicts of interest

The authors declare no conflicts of interest.

Data Availability Statement

The authors declare that all the data supporting the findings of this study are contained within the paper.

References

  1. Alqahtani MS, Kazi M, Alsenaidy MA, and Ahmad MZ (2021) Advances in Oral Drug Delivery. Front Pharmacol doi: 10.3389/fphar.2021.618411. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. AminiLari M, Wang L, Neumark S, Adli T, Couban RJ, Giangregorio A, Carney CE, and Busse JW (2022) Medical cannabis and cannabinoids for impaired sleep: a systematic review and meta-analysis of randomized clinical trials. Sleep doi: 10.1093/sleep/zsab234. [DOI] [PubMed] [Google Scholar]
  3. Aviram J and Samuelly-Leichtag G (2017) Efficacy of cannabis-based medicines for pain management: a systematic review and meta-analysis of randomized controlled trials. Pain Physician 20: E755–E796. [PubMed] [Google Scholar]
  4. Bolin BL, Reynolds AR, Stoops WW, and Rush CR (2013) Relationship between oral D-amphetamine self-administration and ratings of subjective effects: do subjective-effects ratings correspond with a progressive-ratio measure of drug-taking behavior? Behav Pharmacol 24: 533–542. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Budney AJ, Vandrey RG, Hughes JR, Moore BA, and Bahrenburg B (2007) Oral delta-9-tetrahydrocannabinol suppresses cannabis withdrawal symptoms. Drug Alcohol Depend 86: 22–29. [DOI] [PubMed] [Google Scholar]
  6. Chait LD and Zacny JP (1992) Reinforcing and subjective effects of oral Δ9-THC and smoked marijuana in humans. Psychopharmacology (Berl) 107: 255–262. [DOI] [PubMed] [Google Scholar]
  7. Comer SD, Ashworth JB, Foltin RW, Johanson CE, Zacny JP, and Walsh SL (2008) The role of human drug self-administration procedures in the development of medications. Drug Alcohol Depend 96: 1–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Cooper ZD and Abrams DI (2019) Considering abuse liability and neurocognitive effects of cannabis and cannabis-derived products when assessing analgesic efficacy: A comprehensive review of randomized-controlled studies. Am J Drug Alcohol Abuse 45: 580–595. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Cooper ZD, Bedi G, Ramesh D, Balter R, Comer SD, and Haney M (2018) Impact of co-administration of oxycodone and smoked cannabis on analgesia and abuse liability. Neuropsychopharmacology 43: 2046–2055. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Dobson AJ and Barnett AG (2018) An introduction to generalized linear models, 4th ed. Chapman and Hall/CRC, New York. [Google Scholar]
  11. Fergusson DM, Horwood LJ, Lynskey MT, and Madden PAF (2003) Early reactions to cannabis predict later dependence. Arch of Gen Psychiatry 60: 1033–1039. [DOI] [PubMed] [Google Scholar]
  12. Fischman MW (1989) Relationship between self-reported drug effects and their reinforcing effects: Studies with stimulant drugs. NIDA Res Monogr 92: 211–230. [PubMed] [Google Scholar]
  13. Fischman MW and Foltin RW (1991) Utility of subjective-effects measurements in assessing abuse liability of drugs in humans. Br J Addict 86: 1563–1570. [DOI] [PubMed] [Google Scholar]
  14. Foltin RW, Brady JV, Fischman MW, Emurian CS, and Dominitz J (1987) Effects of smoked marijuana on social interaction in small groups. Drug Alcohol Depend 20: 87–93. [DOI] [PubMed] [Google Scholar]
  15. Food and Drug Administration (2017) Assessment of abuse potential of drugs: guidance for industry. Center for Drug Evaluation and Research. [Google Scholar]
  16. Gareth J, Daniela W, Trevor H, and Robert T (2013) An introduction to statistical learning: With applications in R, 1st ed. Springer, New York Heidelberg Dordrecht London. [Google Scholar]
  17. Haney M, Collins ED, Ward AS, Foltin RW and Fischman MW (1999) Effect of a selective dopamine D1 agonist (ABT-431) on smoked cocaine self-administration in humans. Psychopharmacology (Berl) 143: 102–110. [DOI] [PubMed] [Google Scholar]
  18. Haney M, Comer SD, Ward AS, Foltin RW, and Fischman MW (1997) Factors influencing marijuana self-administration by humans. Behav Pharmacol 8: 101–112. [PubMed] [Google Scholar]
  19. Haney M, Malcolm RJ, Babalonis S, Nuzzo PA, Cooper ZD, Bedi G, Gray KM, McRae-Clark A, Lofwall MR, Sparenborg S, and Walsh SL (2016) Oral cannabidiol does not alter the subjective, reinforcing or cardiovascular effects of smoked cannabis. Neuropsychopharmacology 41: 1974–1982. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Haney M, Ramesh D, Glass A, Pavlicova M, Bedi G, and Cooper ZD (2015) Naltrexone maintenance decreases cannabis self-administration and subjective effects in daily cannabis smokers. Neuropsychopharmacology 40: 2489–2498. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Haney M and Spealman R (2008) Controversies in translational research: Drug self-administration. Psychopharmacology (Berl) 199: 403–419. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Haney M, Ward AS, Comer SD, Foltin RW, and Fischman MW (1999) Abstinence symptoms following smoked marijuana in humans. Psychopharmacology (Berl) 141: 395–404. [DOI] [PubMed] [Google Scholar]
  23. Hart CL, Ward AS, Collins ED, Haney M, and Foltin RW (2004) Gabapentin maintenance decreases smoked cocaine-related subjective effects, but not self-administration by humans. Drug Alcohol Depend 73: 279–287. [DOI] [PubMed] [Google Scholar]
  24. Hart CL, Ward AS, Haney M, Comer SD, Foltin RW, and Fischman MW (2002) Comparison of smoked marijuana and oral Δ9-tetrahydrocannabinol in humans. Psychopharmacology (Berl) 164: 407–415. [DOI] [PubMed] [Google Scholar]
  25. Kelly TH, Foltin RW, Emurian CS, and Fischman MW (1997) Are choice and self-administration of marijuana related to Δ–9-THC content? Exp Clin Psychopharmacol 5: 74–82. [DOI] [PubMed] [Google Scholar]
  26. Kelly TH, Foltin RW, Mayr MT, and Fischman MW (1994) Effects of Δ9-tetrahydrocannabinol and social context on marijuana self-administration by humans. Pharmacol Biochem Behav 49: 763–768. [DOI] [PubMed] [Google Scholar]
  27. Lamb RJ, Preston KL, Schindler CW, Meisch RA, Davis F, Katz JL, Henningfield JE, and Goldberg SR (1991) The reinforcing and subjective effects of morphine in post-addicts: A dose-response study. J Pharmacol Exp Ther 259: 1165–1173. [PubMed] [Google Scholar]
  28. Le Strat Y, Ramoz N, Horwood J, Falissard B, Hassler C, Romo L, Choquet M, Fergusson D, and Gorwood P (2009) First positive reactions to cannabis constitute a priority risk factor for cannabis dependence. Addiction 104: 1710–1717. [DOI] [PubMed] [Google Scholar]
  29. Legare CA, Raup-Konsavage WM, and Vrana KE (2022) Therapeutic potential of cannabis, cannabidiol, and cannabinoid-based pharmaceuticals. Pharmacology 107: 131–149. [DOI] [PubMed] [Google Scholar]
  30. Li J, Murray CH, Weafer J, and de Wit H (2020) Subjective effects of alcohol predict alcohol choice in social drinkers. Alcohol Clin and Exp Res 44: 2579–2587. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Lile JA, Alcorn JL, Hays LR, Kelly TH, Stoops WW, Wesley MJ, and Westgate PM (2022) Influence of pregabalin maintenance on cannabis effects and related behaviors in daily cannabis users. Exp Clin Psychopharmacol 30: 560–574. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Lile JA, Kelly TH, and Hays LR (2012) Separate and combined effects of the GABA reuptake inhibitor tiagabine and Δ9-THC in humans discriminating Δ9-THC. Drug Alcohol Depend 122: 61–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Ludecke D (2023) sjPlot: Data Visualization for Statistics in Social Science. https://CRAN.R-project.org/package=sjPlot
  34. McCaul ME, Turkkan JS, and Stitzer ML (1989) Conditioned opponent responses: effects of placebo challenge in alcoholic subjects. Alcohol Clin and Exp Res 13: 631–635. [DOI] [PubMed] [Google Scholar]
  35. Mendelson JH and Mello NK (1984) Reinforcing properties of oral Δ9-tetrahydrocannabinol, smoked marijuana, and nabilone: influence of previous marijuana use. Psychopharmacology (Berl) 83: 351–356. [DOI] [PubMed] [Google Scholar]
  36. Murray CH, Li J, Weafer J, and de Wit H (2021) Subjective responses predict d-amphetamine choice in healthy volunteers. Pharmacol Biochem Behav 10.1016/j.pbb.2021.173158. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Newlin DB (1986) Conditioned compensatory response to alcohol placebo in humans. Psychopharmacology (Berl) 88: 247–251. [DOI] [PubMed] [Google Scholar]
  38. R Core Team (2023) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. [Google Scholar]
  39. Scherrer JF, Grant JD, Duncan AE, Sartor CE, Haber JR, Jacob T, and Bucholz KK (2009) Subjective effects to cannabis are associated with use, abuse and dependence after adjusting for genetic and environmental influences. Drug Alcohol Depend 105: 76–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Schuster CR and Henningfield J (2003) Conference on abuse liability assessment of CNS drugs. Drug Alcohol Depend 70: S1–4. [DOI] [PubMed] [Google Scholar]
  41. Sevak RJ, Stoops WW, Glaser PEA, Hays LR, and Rush CR (2010) Reinforcing effects of d-amphetamine: Influence of novel ratios on a progressive-ratio schedule. Behav Pharmacol 21: 745–753. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Stoops WW, Lile JA, Glaser PEA, Hays LR, and Rush CR (2012) Influence of acute bupropion pre-treatment on the effects of intranasal cocaine. Addiction 107: 1140–1147. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Strickland JC, Lile JA, Rush CR, and Stoops WW (2014) Relationship between intranasal cocaine self-administration and subject-rated effects: predictors of cocaine taking on progressive-ratio schedules. Hum Psychopharmacol 29: 342–350. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Wachtel S, ElSohly M, Ross S, Ambre J, and de Wit H (2002) Comparison of the subjective effects of Δ9-tetrahydrocannabinol and marijuana in humans. Psychopharmacology (Berl) 161: 331–339. [DOI] [PubMed] [Google Scholar]
  45. Wachtel SR and de Wit H (2000) Naltrexone does not block the subjective effects of oral Δ9-tetrahydrocannabinol in humans. Drug Alcohol Depend 59: 251–260. [DOI] [PubMed] [Google Scholar]
  46. Walsh Z, Gonzalez R, Crosby KS, Thiessen M, Carroll C, and Bonn-Miller MO (2017) Medical cannabis and mental health: A guided systematic review. Clinical Psychol Rev 51: 15–29. [DOI] [PubMed] [Google Scholar]
  47. Ward AS, Comer SD, Haney M, Foltin RW, and Fischman MW (1997) The effects of a monetary alternative on marijuana self-administration. Behav Pharmacol 8: 275–286. [DOI] [PubMed] [Google Scholar]
  48. Wesley MJ, Westgate PM, Stoops WW, Kelly TH, Hays LR, and Lile JA (2018) Influence of tiagabine maintenance on cannabis effects and related behaviors in daily cannabis users. Exp Clin Psychopharmacol 26: 310–319. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Wickham H (2016) ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag, New York. [Google Scholar]
  50. Zacny JP and de Wit H (1991) Effects of food deprivation on subjective effects and self-administration of marijuana in humans. Psychol Rep 68: 1263–1274. [DOI] [PubMed] [Google Scholar]
  51. Zeiger JS, Haberstick BC, Corley RP, Ehringer MA, Crowley TJ, Hewitt JK, Hopfer CJ, Stallings MC, Young SE, and Rhee SH (2010) Subjective effects to marijuana associated with marijuana use in community and clinical subjects. Drug Alcohol Depend 109: 161–166. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The authors declare that all the data supporting the findings of this study are contained within the paper.

RESOURCES