Study design and procedures
This double-blind, within-subjects crossover study included three experimental sessions that were scheduled at least 7 days apart to avoid carryover effects. Participants received the three treatments (one per session) in a randomized and counterbalanced order. The randomization schedule was created by an independent researcher, and only the study pharmacist had access to it. Between 1 and 4 weeks prior to the first session, participants attended an orientation session in which they practiced the driving simulation and cognitive tasks. Practice was continued until participants demonstrated competence in each task. Participants were instructed to abstain from illicit drugs for the duration of the study (i.e., from the time of study enrollment until the final session) and from alcohol on the night before research sessions, and to maintain any use of regular medications. Participants were also instructed to consume no more than their regular caffeine intake on the morning of research sessions.
Participants were healthy adults with a history of infrequent cannabis use. Inclusion criteria were aged 18–65 years, self-reported cannabis consumption ≤ 2 times/week in the previous 3 months and ≥ 10 lifetime exposures, and possession of and minimum 1-year driving on an unrestricted Australian license (i.e., > 4 years total driving experience). Exclusion criteria included current mood disorder, lifetime major psychiatric illness, history of clinically significant adverse response to previous cannabis exposure, any moderate or severe substance use disorder as assessed by an addiction medicine specialist, pregnant/nursing, interest in treatment to reduce cannabis use, current use of medications known to affect driving, active hypertension, cardiovascular disease, or chronic pulmonary disease. Volunteers were recruited through online advertisement, social media (e.g., Facebook), and word of mouth. After an initial phone screen, participants meeting inclusion/exclusion criteria were invited to attend a medical screen which involved a detailed medical and psychiatric evaluation. All participants gave written informed consent prior to study enrollment. All procedures were approved by the Sydney Local Health District (RPAH Zone) Human Research Ethics Committee and were in accordance with the Declaration of Helsinki. The trial was listed on the Australia New Zealand Clinical Trials Registry (No. 12616000414415).
The order of events during research sessions is presented in Fig. 1. Participants arrived at the clinical research unit at 09:00 on the morning of research sessions. Nil breath alcohol concentration (BrAC) was confirmed via breathalyzer (Alcotest 5510, Draeger, Lübeck, Germany), and oral fluid was screened (DrugWipe 5s, Securetec, Neubiberg, Germany) to rule out recent drug use. Participants testing positive for any drug (cannabis, cocaine, opiates, or amphetamines/MDMA/methamphetamines) were sent home and the session was rescheduled. Participants completed a baseline questionnaire at the start of each session which asked about recent use of drugs, alcohol and caffeine, adverse events since the previous session, and perceived sleep quality during the previous night. Baseline cognitive task performance and subjective drug effects were assessed 30 min prior to dosing.
Participants inhaled 125 mg THC (11% THC, < 1% CBD), THC/CBD (11% THC, 11% CBD), or placebo (< 1% THC, < 1% CBD) cannabis (Tilray, BC, Canada) via vaporization at 200 °C (Mighty Medic, Storz & Bickel, Tuttlingen, Germany) over 5 min according to a standardized procedure (inhale 3 s, hold 3 s, exhale, and rest 30 s). If vapor was still visible in exhaled breath at 5 min, then this procedure was continued until vapor was no longer visible.
Blood collection and plasma cannabinoid levels
Blood was collected via indwelling peripheral venous catheter into purple-top (EDTA) Vacutainer® tubes (Becton, Dickinson and Company, Franklin Lakes, NJ) 20 min prior to and 10-, 60-, 120-, and 180-min post-inhalation. The blood was centrifuged at 1228×g for 10 min and the supernatant plasma was decanted and stored in 3.6-mL Nunc® cryotubes (Thomas Scientific, Swedesboro, NJ) at − 80 °C until analysis. Plasma was subsequently thawed for analysis via liquid chromatography-tandem mass spectrometry (LC-MS/MS) according to previously published methods (Kevin et al. 2017; Schwope et al. 2011). Duplicate 200-μL plasma aliquots were fortified with an internal standard mixture containing d3-THC, d3-THC-COOH, and d3-11-OH-THC. Duplicate calibrator samples were prepared with cannabinoid-free plasma (obtained from the Red Cross), spiked with appropriate amounts of a standard mixture of THC, 11-OH-THC, THC-COOH, CBD, and internal standards to generate a standard curve and quality control samples for each analyte. All samples were diluted 1:1 in 0.1% formic acid in water, and analytes were extracted using 400 μL capacity ISOLUTE SLE+ supported liquid extraction columns (Biotage, Sydney, Australia). The analytes were eluted with 500 μL dichloromethane, 300 μL ethyl acetate, and 1.2 mL methyl tert-butyl ether. The eluate was evaporated under a gentle stream of nitrogen without heating, and analytes were reconstituted in 100 μL of 40:60 acetonitrile and 0.1% formic acid in water, transferred to 2-mL autosampler vials fitted with 100-μL glass inserts, and placed in the LC-MS/MS autosampler held at 4 °C. Chromatographic separation was achieved using an Eclipse XDB-C18 column (50 mm × 2.1 mm i.d., particle size 3.5 μm; Agilent Technologies, Singapore) using gradient elution with mobile phases 0.1% formic acid in water and acetonitrile, at a flow rate of 0.3 mL/min. This was coupled to a Shimadzu LCMS-8030 mass spectrometer for analyte identification and quantification.
The custom-built driving simulator (Hyperdrive, Adelaide, Australia) consisted of a fixed-base equipped with original vehicle controls (steering wheel, indicators, seat, safety belt), hi-resolution Fanatec® pedals, and a servo motor wheel base (Endor AG, Landshut, Germany) linked to four networked computers running SCANeR™ Studio simulation engine (v1.6, OKTAL, Paris, France). Visual images were displayed on three 32-in. LCD monitors using three channels set to provide a 100° field of view. A digital dashboard displayed speed, rpm, and status of vehicular control systems (e.g., traction control). A complete rear visual scene was displayed on three separate channels (rear vision mirror, left, and right-side mirrors). Graphics refreshed at a rate of 60 Hz, and data were sampled at a rate of 20 Hz. Surround sound provided auditory feedback, and force feedback steering provided haptic feedback. Data collected by the simulator’s software program included measures of lateral control (lateral position, steering wheel angle), longitudinal control (speed, acceleration), and interaction with other vehicles.
The driving simulation started with a 5-min highway car-following task in which participants were required to follow and maintain a constant distance (headway) to a lead vehicle that would accelerate or decelerate every 30 s in a sinusoidal manner (between 90 and 110 km/h). The task occurred on a two-lane, dual-carriageway highway in steady highway traffic. Outcome measures included standard deviation of lateral position (SDLP; a measure of lane weaving (Verster and Roth 2011)), mean headway (i.e., distance to the lead vehicle), and standard deviation of headway.
The remainder of the task (25 min; “secondary driving task”) consisted of highway and rural segments. Participants were instructed to follow the spoken GPS directions and drive as they normally would. Highway segments involved a two-lane, dual-carriageway road with posted 110 km/h speed limits, rural segments involved winding single-lane roads with various posted speed signs (60–100 km/h), and intersections with and without signal-controls. Outcome measures included SDLP, mean speed (MSP), and standard deviation of speed (SDSP). Throughout the task, there were various hazards (e.g., roadworks, aggressive drivers), cyclists, pedestrians, and traffic in variable density. To minimize familiarity, the appearance (i.e., make, model and color) of other vehicles was generated randomly for each drive. The time of day for each drive was set to match the actual time of testing. Tests of simulated driving began 30 min (T1) and 210 min (T2) after dosing.
Cognitive/psychomotor performance was assessed using three computerized tasks which are known to be sensitive to the impairing effects of THC (Vandrey et al. 2017). These included the Digit Symbol Substitution Task (DSST; (Mcleod et al. 1982)), Divided Attention Task (DAT; (Kleykamp et al. 2010)), and Paced Auditory Serial Addition Task (PASAT; (Herrmann et al. 2015)). Performance assessments were completed in this order at baseline and 20 min (T1) and 200 min (T2) after dosing.
In the DSST, participants were presented with a series of geometric patterns labeled from 1 to 9, each consisting of an array of filled and blank squares in a 3 × 3 grid. When a number appeared in the middle of the screen, participants were instructed to replicate the pattern corresponding to that array using the numeric keypad of a computer keyboard. Participants had 90 s to replicate as many patterns as possible. Outcome measures included number of patterns correct and accuracy (number of patterns correct/number of patterns attempted).
In the DAT, participants were required to track a horizontally moving stimulus on the screen using their mouse while simultaneously responding to peripheral visual stimuli by clicking the left mouse button whenever a number in any corner of the screen matched a target number presented at the bottom of the screen. Outcome measures included mean distance of the cursor from the target (tracking error), the number of target numbers correctly identified (/24), and response time.
In the PASAT, participants watched single digits appear on the screen and were instructed to sum each new digit with the preceding one. Participants responded by clicking on the correct answer from a list of numbers (1–10) presented on the screen. Outcome measures included response time on correct trials and the total number of correct trials (/90).
Subjective drug effects
Subjective drug effects were assessed at baseline and 15, 60, 120, 180, and 240 min after dosing using a series of Visual Analog Scales (VAS). Participants rated on a 100- mm line their responses to the statements: “Strength of drug effect”, “Liking of drug effect”, “Stoned,” “Sedated,” “Anxious,” and “Confident to drive”. All scales were unipolar except for “Liking of drug effect” which was bipolar (dislike very much – like very much). The State-Trait Anxiety Inventory (Y-1 – state form only) (Spielberger 1983) was also administered at baseline and 15, 60, 120 and 180 min after dosing. Self-reported driving ability was further assessed by the Adelaide Driving Self-Efficacy Scale (George et al. 2007) (which provides a score from 0 to 120) at baseline and at 60 and 180 min after dosing.
Sample size was determined by a power calculation based on the effect size (ηp2 = 0.14) associated with the main effect of Dronabinol on simulated driving performance reported in a previous study (Veldstra et al. 2015). Data from the driving simulator tasks were reviewed and cleaned to remove recognizable artifacts (e.g., increased SDLP while overtaking other vehicles). All data were analyzed in SPSS v24 (IBM Corp, Armonk, NY) with Linear Mixed Models (LMMs). The restricted maximum likelihood method was selected, and a first-order autoregressive (AR1) covariance structure was specified for repeated factors as it provided the lowest Akaike information criterion (AICC) model fit values. Fixed factors included treatment (3 levels), time (2, 3, and 5 levels for driving, cognitive and pharmacokinetic/subjective drug effects data, respectively), session (3 levels), and treatment by time. For data which included baseline assessment (i.e., cognitive, pharmacokinetic, and subjective drug effect data), baseline scores were included in the model as a covariate. In each model, planned Bonferroni pairwise comparisons were used to compare treatment means at each level of time. For additional pharmacokinetic data (e.g., area under the curve), non-parametric Wilcoxon signed-rank tests were used to assess differences between treatments. The statistical significance level was set at p < 0.05.