Study Design
This study was conducted at the KGK Science Inc. clinic site, London, ON, Canada, from February 28, 2020, to August 29, 2020. The trial received research ethics board approval on February 14, 2020, from the Institutional Review Board (IRB) Services, Aurora, Ontario (Pro00041616). All participants provided written informed consent to participate in the study prior to initiation of study procedures. All participants with identifying information included in the manuscript provided written consent to the use of their photograph. The study was conducted in compliance with the International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH) Guideline for Good Clinical Practice (GCP) and in accordance with the Declaration of Helsinki guidelines and its subsequent amendments.
Participants met the following inclusion criteria: were male or female between the ages of 19–65; provided two adult contacts to ensure transportation to and from the clinic; were willing to comply with all study requirements, questionnaires, records and assessments. Cannabis-specific inclusion criteria included: abstinence from cannabis for 3 days prior to visit, brought a cannabis product purchased from a legal medical or recreational source, in its sealed, original packaging, for use via an inhalation route of administration during the visit and had the required cannabis use history. Cannabis use history was assessed with a seven-item Cannabis Use Questionnaire (CUQ). Participants reported using cannabis at least a few times/month and no more than 2–3 times/week, usually via an inhalation route of administration (vape, smoke); that they do not struggle to control their high or get dizzy, vomit or become paranoid and can handle a self-reported 7/10 level of high; and that they had not used recreational drugs other than alcohol or cannabis within the last few weeks.
Individuals were excluded if they had participated in any previous Zentrela Inc. EEG data collection; had an allergy, sensitivity or stated that they had religious practices that precluded the application of the EEG headset; had a history of a clinically significant adverse event (AE) associated with cannabis, brain injury, epilepsy, seizures, concussion, stroke, anxiety, depression, dementia, mood disorder, schizophrenia, psychotic disorder or familial history of psychosis or current migraine; had a history of alcohol or drug abuse in the last 12 months; self-reported use of cannabinoid products within 3 days of their study visit; impairment from illicit drugs or alcohol during the study visit as assessed by the medical director; had unstable chronic disease; had current or history of kidney or liver disease; had unstable hypertension; had a significant cardiovascular event within the last 6 months; had a self-reported HIV, hepatitis B or hepatitis C diagnosis. Participants on stable medications for at least 3 months, those with a history of surgery within the last 3 months as well as any other condition which may have adversely affected the participants’ ability to complete the study or its measures, or which may have posed significant risk to them, were considered case by case by the medical director.
Investigational Device
The Cognalyzer® is a new technology to detect the presence of cannabis’ psychoactive effects based on brain signal analysis and machine learning. It utilizes a proprietary battery-powered data collection device that streams data via Bluetooth to a recording system on a nearby laptop. Once data were collected, raw continuous data files were retained and sent with blinded filenames to Zentrela Inc. for evaluation. The algorithm could alternatively be operated on the laptop for near real-time evaluations. The EEG device was an 8-channel system with a 250-Hz sampling rate and 24-bit resolution. Conductive gel was applied to reduce electrode impedance below 35 kΩ. Electrodes were placed on the left and right side of a participant’s head to coincide with their frontal, temporal, occipital and parietal lobe positions. Ground and driven right leg (DRL) electrodes were placed on the forehead, for a total of ten electrodes. Electrodes were held in position using a proprietary ten-electrode headband. Once the collection began, data were segmented into 10-s segments with 5-s overlap. Unreliable segments such as signals with high peak voltage, very high variance, very high or low root of mean square (RMS) voltage, and unusual spectral properties caused by eye or body movements were rejected in real time, and data collection continued until a total of 30 artifact-free segments were collected (Fig. 1).
Algorithm Description
The patent-pending Cognalyzer® algorithm works directly with the segmented EEG data collected from the 8-channel EEG device. During data collection, each segment is analyzed for artifact as described above, and a set of 30 artifact-free segments is submitted to the algorithm. Each segment is independently analysed for features including power spectral density, cross power spectral density, coherence and RMS power. Based on these data, the algorithm produces a classification for each segment, either ‘normal’ or ‘abnormal.’ With this set of classifications, the strength of the psychoactive effect for a given data file is determined by calculating the percentage of total segments (0–100%) that are classified as ‘abnormal.’ Two such recordings were made for each participant’s data collection episode, one immediately after the other. The psychoactive effect of an episode is the average of the percentage of total ‘abnormal’ segments of the two EEG files in that episode. An EEG episode was predicted to be post-inhalation when the percentage psychoactive effect was higher than 38%, or otherwise predicted as being pre-inhalation. Two candidate algorithms using slightly different parameter weightings were evaluated, Cognalyzer® version 1 (V1) and version 2 (V2). Cognalyzer® V2 was designed to be more sensitive than V1 in detecting a cannabis effect and V1 to be more balanced in sensitivity vs. specificity.
Study Procedures and Assessments
Screening was conducted via telephone. Prior to enrollment of participants, ten mock participant visits were completed to ensure accurate administration of the Cognalyzer® and maintain the scientific rigor and reproducibility of results. At the 1-day, in-clinic visit, participants signed an informed consent form, and medical history and eligibility criteria were reviewed. Seventy-five participants were enrolled in the study. To evaluate primary and secondary objectives, study assessments were conducted pre- and post-cannabis inhalation. For all enrolled participants, pre-cannabis inhalation, seated blood pressure (BP) and heart rate (HR) were measured, oral fluid was collected (Quantisal™ collection device, Immunalysis Corp., limit of detection: 0.5 ng/ml) to measure THC concentration, and two 2.5-min baseline EEG recordings were completed with the Cognalyzer® investigational device. Oral fluid was collected from under the participant's tongue and in accordance with the collection device instructions. For each participant, photographs of the EEG headset applied on their head were captured to document that the ten EEG electrodes were placed in the correct locations. Photos captured the frontal, posterior and lateral left and right views of the electrode placements, and all identifying participant features were removed. Participants were allowed to smoke or vape the cannabis products they brought to the clinic ad libitum. They were required to do so outside of the clinic building, while monitored by clinic staff. Participants were instructed to inhale cannabis until a self-reported 7/10 feeling of high was reached. Immediately after cannabis inhalation, a 5-item Drug Effects Questionnaire (DEQ-5) was administered, and two 2.5-min post-inhalation Cognalyzer® EEG readings were collected [5, 31]. The five questions on the DEQ-5 included: (1) Do you feel the drug effect right now? (2) Are you high right now: (3) Do you dislike any of the effects you are feeling right now? (4) Do you like any of the effects you are feeling right now? (5) Would you like more of the drug you took right now? The last 50 participants enrolled in the study remained in the clinic for 4 h post-inhalation, and the Cognalyzer® and DEQ were repeated at 30 min, 1, 1.5, 2, 2.5, 3, 3.5 and 4 h. These data will be presented elsewhere [32]. For these participants, a second oral fluid sample was collected 4 h post-inhalation. Participants’ vital signs were measured, and they were assessed for safety prior to leaving the clinic at the end of the study visit.
Prior to sending the participants’ blinded EEG data to Zentrela Inc. for analysis, each participant’s EEG data were saved as both blinded and unblinded files for a total of eight files (four blinded and four unblinded) per participant. Files were blinded by replacing the filename with a random nonrepeating integer (http://www.randomization.com), and within each participant and time point (pre and post), the two files collected during one episode were grouped together. To further ensure blinding, each file was saved as a new document to remove the original timestamp and date. A password-protected master list with the identified random number assigned to each of the raw files was generated and retained at KGK Science Inc. The files were clear text CSV data that contained only the eight channels of raw EEG data with no other identifying information. This structure was verified by KGK Science Inc. personnel before submission to Zentrela Inc. for analyses. Zentrela Inc. personnel were not present during data collection, and Zentrela Inc. personnel who conducted the initial (primary outcome) data analyses were only unblinded after results were returned to KGK Science Inc. for tabulation and the database locked.
During the study visit, AEs were recorded and documented in the study record. All AEs were classified as per the description, duration, intensity, frequency and outcome. The medical director determined the causality and intensity of all reported AEs, as appropriate. AEs were coded according to the Medical Dictionary for Regulatory Activities (MEDRA) terminology System Organ Class, version 23.0.
Data and Statistical Analysis
The planned sample size for this study was 150 participants. A previous pilot study estimated a sensitivity of 83.3% for the Cognalyzer®. With a sample size of 150 participants, it was possible to obtain sensitivity of between 75.2% and 90% with at least 80% power and 5% significance level, given 20% attrition. Due to the impact of the global coronavirus disease 2019 (COVID-19) pandemic on timelines and project funding, the number of participants was reduced [33]. A sample size of 75 participants was enrolled in the study. The sensitivity obtained from the study was within the planned range on the sample size calculation.
Two per protocol (PP1 and PP2) populations were assessed in this study. The PP populations consisted of all participants who completed the study visit and procedures connected with the Cognalyzer® EEG measurement. The PP1 population consisted of participants with two EEG files for both the pre-inhalation and the first post-inhalation time points. The PP2 population consisted of participants with two EEG files for pre-inhalation and all post-inhalation time points.
This study used a comparison standard of pre- vs. post-inhalation status. Self-reported, subjective impairment was assessed with the DEQ-5 and was used to classify subjective impairment at 4 h post-inhalation.
The pre- vs. post-inhalation comparison status characterized by Cognalyzer® was summarized for the following parameters: sensitivity, specificity, accuracy, percent false positive, percent false negative, positive predictive value (PPV) and negative predictive value (NPV) of the instrument. The PPV is the proportion of positive test results which are true positives. The NPV is the proportion of negative test results which are true negatives. The diagnostic performance of Quantisal™, Immunalysis oral fluid testing relative to pre- vs. post-inhalation was evaluated with THC concentration cut-offs of 0.5 ng/ml. Since there are different guidelines for determining the cut-off and this study was primarily designed to evaluate the Cognalyzer® and its potential application in synergy with oral fluid, it was decided to include all concentrations > 0.5 ng/ml as positive samples. To calculate a false-positive rate at 4 h post-inhalation, it was arbitrarily defined that an answer to Q1 (“Do you feel the drug effect right now?”) of the DEQ-5 of < 4 indicated not being subjectively impaired; if participants answered with a level ≥ 4 they were subjectively considered impaired. This level was chosen arbitrarily to align with the threshold of 38% used for the Cognalyzer® test. By using the DEQ-5, the participant's own subjective self-reported intoxication level was considered as the comparison standard of intoxication. Participants were encouraged to answer the DEQ-5 questions honestly and were not incentivized to report any specific DEQ-5 values.
The ability to improve current drug testing procedures was evaluated by combining the Cognalyzer® results pre-inhalation and 4 h post-inhalation with oral fluid THC concentration assays.
The null hypotheses were tested using McNemar’s test. p values ≤ 0.05 were considered statistically significant. All statistical analyses were completed using R Statistical Software Package version 3.6.3 or newer for Microsoft Windows [34].