Measuring Within-Individual Cannabis Reduction in Clinical Trials: A Review of the Methodological Challenges

Abstract

Purpose:

Cannabis abstinence traditionally is the primary outcome in cannabis use disorder (CUD) treatment trials. Due to the changing legality of cannabis, patient goals, and preliminary evidence that suggests individuals who reduce their cannabis use may show functional improvements, cannabis reduction is a desirable alternative outcome in CUD trials. We review challenges in measuring cannabis reduction and the evidence to support various definitions of reduction.

Findings:

Reduction in number of cannabis use days was associated with improvements in functioning across several studies. Reductions in quantity of cannabis used was inconsistently associated with improvements in functioning, though definitions of quantity varied across studies. Different biomarkers may be used depending on the reduction outcome.

Conclusions:

Biologically-confirmed reductions in frequency of cannabis use days may represent a viable endpoint in clinical trials for cannabis use disorder. Additional research is needed to better quantify reduction in cannabis amounts.

1. Introduction

Up to one-third of individuals with past year cannabis use report symptoms consistent with Cannabis Use Disorder (CUD; [1]), as defined by the Diagnostic and Statistical Manual of Mental Health Disorders (DSM IV-TR; [2]). Current treatment options are inadequate, as psychosocial interventions have limited efficacy and no medications are Food and Drug Administration (FDA) approved for the treatment of CUD [3-5]. The number of cannabis users requiring treatment is likely to remain constant or may even increase with increasing legalization of cannabis [6]. Thus, treatment development for CUD is a priority.

Individuals seeking treatment for CUD are often more interested in reducing their cannabis use than abstaining completely [7, 8]. However, most clinical trials use cannabis abstinence as their primary outcome variable [9]. Matching the individual’s cannabis treatment goals (abstinence vs. reduction) to the outcomes assessed is an important factor when selecting a primary outcome. Individuals with a reduction goal are likely to achieve reduction, but less likely to achieve abstinence [10]. With legalization, reduction goals may become more plausible and fewer individuals will be motivated to abstain due to the illegality of cannabis.

An expert workgroup (Analgesic, Anesthetic, and Addiction Clinical Trial Translations, Innovations, Opportunities and Networks; ACTTION) recently convened to discuss measurement of CUD outcomes in clinical trials and, consistent with the goals of most treatment-seekers, they concluded that reduction in cannabis use is a valid potential outcome [11]. However, research is hindered by lack of clear guidelines for measuring cannabis reduction. As discussed more later, cannabis reduction is challenging to measure because individuals lack sophisticated knowledge about their cannabis products (i.e., concentration of various constituents), people often share cannabis making their specific intake difficult to estimate, and biomarkers provide little information about quantity and frequency when assessed infrequently. In this review, we summarize definitions of reduction previously adopted in CUD trials, examine parallels to research on other substances (i.e., alcohol), and review the evidence base for potential cannabis reduction self-report and biomarker metrics.

2. Operationalizing Cannabis Reduction in Clinical Trials

Cannabis reduction may be accomplished in several ways. It may involve a reduction in the frequency of use or amount of cannabis used (e.g., fewer use sessions per day, fewer grams used per day/week, days of no use). Cannabis reduction was examined as an outcome in many randomized controlled trials testing treatments for CUD. A recent review found that 71% of CUD trials (n=41) examined reduced frequency of cannabis use (e.g., number of days used, number of times per day) and 47% (n=27) examined reduced quantity of cannabis use as a primary or secondary outcome [9]. Though the most common definition of frequency was number of days or weeks in which cannabis was used [9], a recent study also conceptualized frequency as amount of time spent “high” [12]. Reduction in quantity of cannabis was diversely measured across CUD trials and was operationalized as reduction in grams, number of joints or cones, number of standard cannabis units, number of inhalations per day, money spent on cannabis, or change in quantitative cannabinoid concentrations [9].

Frequency and quantity of use are common outcomes assessed in alcohol use disorder (AUD) clinical trials. In fact, the FDA communicated in a draft guidance that absence of heavy alcohol drinking, defined as any day with 4 or more drinks for females and 5 or more drinks for males, is a viable endpoint to evaluate efficacy in clinical trials for AUD pharmacotherapies [13]. A number of other alcohol reduction outcomes were considered, including drinks per day, drinks per drinking day, percent days abstinent, percent heavy drinking days, and reduction in grams of ethanol consumed per day [14-16]. Researchers also demonstrated that certain measures of alcohol reduction are more strongly associated with AUD symptom reduction than others. For example, in a nationally representative sample from the United States, reductions in ethanol grams per day were more strongly associated with reduced risk of AUD than reductions in ethanol grams per drinking day [17]. The latter definition ignores frequency of drinking. Thus, although many definitions of alcohol reduction exist, researchers are beginning to examine which of these definitions best reflect clinically meaningful endpoints. Few studies compared various definitions of cannabis reduction to other important clinical outcomes, such as problems due to use or general functioning. Here we review the evidence to-date supporting use of specific cannabis reduction metrics.

3. Choosing a Reduction Endpoint: Achieving Clinical Significance

The clinical importance of transitioning from using a problematic substance to abstaining from that substance is obvious. If researchers are to consider cannabis reduction as an alternative to abstinence, it is necessary to validate the clinical significance of cannabis reduction endpoints. While it is possible that improvements in functioning are linearly associated with decreases in cannabis use, it is also possible that a specific amount of reduction must occur before an individual experiences benefit.

For example, it is unclear if a reduction in grams used per week (quantity) is associated with improvements in functioning, even if the individual spends the same amount of time using cannabis (frequency). Furthermore, it is unclear how much of a reduction is needed to achieve benefit (e.g., 50% reduction in days used; no more than 1 gram per week).

A number of studies examined the relationship between cannabis use/reduction and clinically relevant outcomes (e.g., risk for CUD, overall functioning). Some research shows that frequency metrics are more highly related to CUD symptoms than quantity metrics [18]. However, relying solely on frequency may miss important information about total cannabis exposure [19-22]. One research group proposed a cannabis use quantity threshold that represents high risk for CUD [23]. This group proposed a standard joint unit that is roughly equivalent to 0.25 grams of cannabis or 7 mg Δ⁹-tetrahydrocannabinol (THC; [24]). This recommended standard joint unit was derived from the median grams and THC content found in 315 donated joints in Barcelona, Spain. In further research, they showed that individuals who report an average of at least 1.2 standard joint units per cannabis use day were at greater risk for CUD [23]. Given that the majority of their sample were daily users, it is unclear whether this threshold would apply to non-daily users. Another study of individuals who used cannabis at least once in the past month found that use of more than 0.50 grams of cannabis per day placed individuals at risk for experiencing cannabis-related problems [25]. Studies of this nature are critical for the screening of problematic use in clinical settings and educating the public regarding higher- versus lower-risk use. However, they have limitations. For instance, the standard joint unit does not readily translate to cannabis products that are in wax or vaporizer form [11] and does not account for varying content (e.g., THC vs. other constituents).

Definitions of high- versus low-risk cannabis use may be helpful in determining ideal clinical targets for individuals with CUD. However, few studies examined associations between frequency and/or quantity of cannabis use and clinical outcomes specifically among individuals who attempted to reduce cannabis use. Studies demonstrating that cannabis reduction metrics are associated with improved functioning among individuals with CUD are particularly important, as these studies help to disaggregate risk factors for problematic cannabis use from improvements resulting from reduction. Thus, we review the convergent validity (and lack thereof) for various definitions of frequency/quantity of cannabis use among problematic users (see Table 1). Frequency of cannabis use (i.e., number of days used) was consistently associated with negative outcomes while quantity inconsistently showed associations with problematic outcomes above and beyond frequency. Additionally, some research showed that number of days used produced larger effect sizes, and was therefore more sensitive to change over the course of a clinical trial, than cannabis quantity metrics [31-33]. As described in the next section, it is possible that difficulties in estimating cannabis quantity limit its utility as an outcome measure in clinical trials.

Table 1.

Studies Examining Correlates of Reduction Among Individuals Experiencing Problems with Cannabis Use

Citation	Convergent Validity	Level of Support
Frequency: Number of cannabis use days
26. Buu et al. 2017a	Symptoms of CUD	Among individuals with CUD at baseline in a national sample, frequency of use significantly predicted CUD persistence at 3 year follow-up
27. Hser et al. 2017b	Hospital Anxiety & Depression Scale; Pittsburgh Sleep Quality Index; Quality of life assessment from Phenx Toolkit	Treatment-seeking adults with CUD who showed a negative trajectory in number of cannabis use days during a CUD trial showed improvement in functioning in numerous areas except quality of life
28. Brezig et al. 2018	Quality of Life Enjoyment & Satisfaction Questionnaire-Short Form	Reduction in number of cannabis use days was significantly associated with improvement in quality of life at end of 12 weeks treatment in a clinical CUD sample
29. Mooney et al. 2018c	PROMIS Global Health Scale- Physical Health; Patient Health Questionniare-9 (PHQ-9)- Depression	Some evidence that individuals using cannabis 3 times/week or fewer have better physical health and lower depression than individuals using more frequently
21. Tomko et al. 2018b	Urinary THCCOOH; Marijuana Problems Scale	Number of days used explained most variance in outcomes within a CUD clinical trial
30. Liao et al. 2019c	Short-Form Health Survey (SF-12)-mental and physical health;Simplified Nutritional Appetite Questionnaire; St. George’s Respiratory Questionnaire; PROMIS Pain Intensity & Sleep Disturbance instruments; Perceived Stress Scale; Specific Psychotic Experiences Questionnaire; Hospital Anxiety & Depression Scale; Marijuana Problems Scale	Number of cannabis use days showed small, significant correlations with cannabis-related problems, overall mental health functioning, sleep disturbance, appetite problems, and physical pain intensity (rs=0.20-0.40)
Frequency: Number of use episodes per cannabis use day
30. Liao et al. 2019c	Short-Form Health Survey (SF-12)-mental and physical health;Simplified Nutritional Appetite Questionnaire; St. George’s Respiratory Questionnaire; PROMIS Pain Intensity & Sleep Disturbance instruments; Perceived Stress Scale; Specific Psychotic Experiences Questionnaire; Hospital Anxiety & Depression Scale; Marijuana Problems Scale	Number of episodes per cannabis use day showed small, significant correlations with cannabis-related problems, anxiety, depression, stress, sleep disturbance, & poor physical health (rs=0.20-0.28)
Quantity: Joints/cones used per day
26. Buu et al. 2017	Symptoms of CUD	Among individuals with CUD at baseline in a national sample, number of joints per day did not predict CUD persistence at 3 year follow-up
21. Tomko et al. 2018b	Urinary THCCOOH; Marijuana Problems Scale	No evidence that joints/cones per day explained more variance in outcomes in a CUD trial than number of cannabis use days alone
Quantity: Number of grams used
28. Brezig et al. 2018	Quality of Life Enjoyment & Satisfaction Questionnaire-Short Form	Reduction in grams as weighed by a surrogate substance not associated with improvement in quality of life at end of 12 weeks of treatment in a clinical CUD sample.
21. Tomko et al. 2018b	Urinary THCCOOH; Marijuana Problems Scale	Grams as weighed by a surrogate substance explained a small, but significant increase in variance in outcomes in a clinical CUD sample, even after accounting for number of days used and joints/cones per day

Note. PubMed Database was searched on 8/27/19 using the following search terms: "marijuana OR cannabis" AND "reduction" and reference sections of eligible articles were cross-referenced. Articles were restricted to human subjects’ research. Original research that 1) focused on cannabis persistence or reduction in a cannabis treatment-seeking and/or CUD sample and 2) examined associations between at least one frequency or quantity metric and a clinical outcome (i.e., functioning in one or more domains, CUD symptoms, cannabis biomarkers) were retained, resulting in 6 publications. ^aParticipants reported frequency of use over past year on a 0-10 Likert scale ranging from 0 = never, 5 = once a month, to 10 = every day. ^bHser et al. (2017) and Tomko et al. (2018) used the same sample of adults with CUD; ^cLiao et al. (2019) and Mooney et al. (2018) used the same sample of adults with a history of CUD treatment or heavy use.

4. Choosing a Reduction Endpoint: Maximizing Statistical Power

In addition to being consistent with patient preferences, reduction outcomes offer some practical advantages over abstinence endpoints. One practical advantages is that continuous endpoints (e.g., days used, grams used, number of inhalations) offer more precision, and therefore more statistical power, than binary abstinence endpoints, thereby requiring a smaller sample size [34]. When variables are dichotomized, precision is lost from the dataset (e.g., an individual using cannabis three times per day is coded the same as an individual who had a single episode of use over the designated time frame). Specifically, even under best circumstances (50% abstinence), up to 36.3% of the information available in the continuous variable may be lost due to dichotomization [34]. This loss of information would inflate the necessary sample size by up to 57%. When abstinence rates are significantly different from 50%, this sample size inflation becomes even more dramatic.

Even reduction outcomes could be dichotomized such as the “absence of heavy drinking” endpoint proposed by the FDA for AUD trials [13]. However, if the reduction outcome is more realistic and consistent with patient goals, even a dichotomized reduction endpoint may have greater statistical power than a dichotomized abstinence outcome, depending on the prevalence of success. For example, to have sufficient power to find an odds ratio of abstinence between two treatment groups equal to 3.0 in a sample with a 10% overall abstinence rate, one would need to equally randomize 322 participants (5.4% vs. 14.6% abstinence in the two groups). To find an odds ratio of meeting a reduction threshold of 3.0 in a sample in which 50% of individuals achieve a specific reduction threshold, you would need to randomize only 102 participants (63% vs. 37% meeting reduction thresholds in the two groups). Thus, reduction endpoints are not only consistent with the goals of cannabis users, but are also likely to require fewer resources to develop and evaluate novel CUD treatments, potentially expediting the treatment development process.

Statistical power in clinical trials is also dependent on the precision of the measures used to assess cannabis reduction. Low reliability of measures and increased measurement error results in attenuated effect sizes [35, 36]. Thus, when considering reduction endpoints, it is important to minimize measurement error. This is particularly challenging due to limitations in both self-report and biomarker assessment of cannabis use.

5. Measuring Cannabis Reduction: Self-Report and Biomarkers

Self-report of any behavior, including substance use behavior, is subject to a number of biases, such as recall errors, digit bias, limited awareness, difficulty with estimation, intentional underreporting, and social desirability [37-43]. Measuring quantity of cannabis use via self-report presents several unique challenges [44]. First, there are several methods of administration and a wide variety of available cannabis products, including high concentrate THC preparations (i.e., wax, dabs), that make quantification challenging. Second, cannabis users may lack awareness regarding aspects of their use, such as the percentage of THC typically used or the amount of plant material used. Even regular cannabis users have difficulty estimating grams of cannabis used [45]. Though daily users are better at estimating THC or potency [45], potency across cannabis strains is variable [46]. Researchers implemented a number of strategies, such as gram estimation through the weighing of surrogate plant material [47] and the use of images to aid in quantity reporting [18]. Assessment of frequency of use, though still limited by general self-report biases, circumvents many of the challenges specific to cannabis quantification.

Cannabinoid biomarkers provide an objective measure of cannabis use that can offset some of the limitations of self-report. Self-report and biomarkers provide complementary information and the gold standard for determining abstinence is to use both in SUD clinical trials [48]. In the context of AUD trials, the FDA acknowledged that, at the very least, collection of biomarkers may improve accuracy of self-report [13]. Recent CUD ACTTION workgroup recommendations also support the use of self-report with biological confirmation for CUD trials [11]. However, the workgroup noted that biomarkers are limited in their ability to measure reduced cannabis use [11]. Here, we review how cannabis biomarkers may be used to confirm self-report of cannabis reduction.

Cannabinoids and/or their metabolites can be measured in blood, plasma, urine, oral fluid, sweat, breath, and hair [11, 49]. Because THC is the primary psychoactive constituent in cannabis, THC or its metabolite, 11-nor-9-carboxy-THC are the cannabinoids most often assessed in CUD trials. Cannabinoid biomarkers in some biological matrices have detection windows typically too short to provide a comprehensive picture of use between weekly office visits (e.g., breath, oral fluid) or too long to capture change in use over short-duration clinical trials (e.g., urine, hair). Thus, for a biomarker to have utility in clinical trials, investigators must consider the half-life when planning the timing and frequency of biomarker assessments.

Short detection windows.

Biomarkers with short detection windows (≤ 24 hours) may be administered one or more times per day to confirm participants’ reports of recent cannabis abstinence. This may be useful if the outcome of interest is reduction in frequency of cannabis use days. In this scenario, a cut–off value is selected to classify abstinent versus non–abstinent samples. Plasma THC was previously used to estimate time since last cannabis use with relatively high accuracy [50]; however, it was not used to quantify the magnitude of reduction in use over a full-length clinical trial and repeated plasma collection places a high burden on participants and medical staff. Measuring cannabis in oral fluid or breath is much less invasive.

THC (THCCOOH) and several minor cannabinoids (e.g., cannabinol, cannabidiol, cannabigerol) are measurable in oral fluid. Specific cannabinoid detection windows vary; thus, examining multiple cannabinoids may provide more information regarding timing of cannabis use [51]. The detection window for oral fluid cannabinoids is 24 hours or less depending upon the target analyte and/or the cut-off selected, particularly when requiring a cut-off value for both THC and a minor cannabinoid or metabolite [49, 51-54]. Relative to urinary and oral fluid THCCOOH, oral fluid THC is less impacted by the individual’s chronicity of cannabis use [53, 55-57]. Oral fluid THC decreases at a relatively linear rate except immediately following cannabis use [52, 53]. Oral fluid THCCOOH peaks later than THC, is detectable longer than THC, and is not subject to elevation after passive smoke exposure [49, 51, 58]. However THCCOOH levels do not necessarily show linear declines and, after controlled cannabis consumption, oral fluid samples may be positive for THCCOOH following several negative samples [58].

Breathalyzers to detect recent cannabis use are in development [59] and breath THC may provide an even shorter detection window than oral fluid [60, 61]. Both oral fluid and breath are easily collected without medical oversight and are possible to observe remotely via video calls or recordings. Two pilot studies [62, 63] demonstrated the feasibility of verifying once or twice daily qualitative oral fluid THC tests remotely via video. In these studies, participants’ identities and test results were confirmed 1) by a video uploaded to a mobile app and asynchronously observed by research staff [62] or 2) by live video (i.e., Skype/Facetime calls; [63]). The potential for remote collection is key since the frequency with which oral fluid or breath assessments need to occur to capture continuous use are generally too frequent to conduct in-person during an outpatient clinical trial.

In sum, current evidence suggests that oral fluid THC testing may be feasible for confirming self-reported reductions in cannabis use days when samples are taken at least daily. However, different cut-offs may be needed for THCCOOH and other cannabinoids, especially among more frequent users who experience a build-up of THC in the body. Further work is needed to determine the optimal combination of analytes tested for maximizing accuracy. Breath THC is an alternative option on the horizon.

Long detection windows.

If the outcome of interest is reduction in quantity of use, a quantitative biomarker with a strong dose-concentration relationship would be useful. In particular, the ideal biomarker is one with THC levels that rise and fall depending on the level of THC exposure. The most frequently used biomarker in clinical trials is urinary THCCOOH concentration [9]. However, decreases in urinary THCCOOH levels are not always linear and rates of metabolism vary across individuals, making it difficult to directly interpret quantitative levels [64]. In a recent trial examining Fatty Acid Amide Hydrolasea (FAAH)-inhibitor PF-04457845 as a treatment for CUD, reduction in joints per day and THCCOOH concentrations were both assessed as independent outcomes [65]. PF-04457845 significantly reduced cannabis use as indicated by both joints per day and urinary THCCOOH concentrations. However, group differences emerged after two weeks of treatment when joints per day were used as the outcome, and after four weeks when urinary THCCOOH was the outcome [65]. Sophisticated mathematical models exist that compare two consecutive urinary THCCOOH values to determine if new use occurred within the interval between assessments [66, 67], but were not designed to discern when a participant initiated new abstinence [43] or reduced their use. Mathematical models that take into account creatinine and factors associated with metabolism (e.g., sex, genetics, body fat) may eventually allow researchers to translate quantitative THCCOOH concentrations into a value that directly relates to quantity of THC consumed; however, further work is necessary to determine whether such models are possible.

Sweat can be measured via a patch worn over a specific period of time, generally one week. Sweat patches detect cannabis use a few days prior to application and any new use that occurs while the patch is worn [49, 68]. Huestis and colleagues [68] demonstrated that THC is likely to be detected in sweat for at least one week following abstinence. A smaller portion of chronic frequent cannabis users may show elevations for up to at least 4 weeks following abstinence onset. Research on cannabis quantification via sweat patches is limited and further work is necessary to determine how well THC concentrations in ng/patch map on to actual cannabis consumption and how this may be affected by method of use. At least one study showed that daily oral ingestion of THC in capsule form (up to 15 mg/day) was not detected via sweat [68].

THC/THCCOOH concentrations in hair could serve as a potential indicator of significant cannabis reduction since lighter cannabis use (less than daily) often goes undetected [69]. Unfortunately, even some daily cannabis users may test negative, indicating low sensitivity [69]. Additionally, because the detection window in hair is long (up to 90 days; [69]), this is only an option for longer duration clinical trials.

6. Conclusions

Cannabis reduction may be a viable and desirable alternative to abstinence for individuals with CUD. Most CUD trials with a reduction outcome to-date have defined reduction as change in frequency (i.e., number of days) of cannabis use, rather than reductions in quantity [9]. Number of cannabis use days is consistently associated with clinical outcomes, while assessment of quantity inconsistently demonstrates convergent validity. This may be partially due to difficulty in accurately reporting and estimating cannabis quantity. Further, number of cannabis use days can currently be confirmed by frequent, remote biomarker assessment while biomarker confirmation of cannabis quantity requires additional research. Though future improvements in our ability to measure cannabis quantity will likely increase precision of clinical trial outcomes, evidence to-date supports use of frequency of cannabis use days as a measure of cannabis reduction in current trials.

The ideal definition is one that is sensitive to clinically meaningful change as evidenced by improvement in functioning, maximizes statistical power, can be accurately reported, and has a corroborating biomarker. Lack of consensus regarding how to define cannabis reduction limits consistency across clinical trials. However, incorporation of multiple definitions of reduction in future clinical research will allow for head-to-head comparisons until a standard definition is established.

Key points.

1. Reduction in cannabis use is an alternative to abstinence in cannabis use disorder treatment

2. Reduction in cannabis use is challenging to assess and biologically confirm, and definitions vary across trials

3. Preliminary evidence suggests that reduced number of days of cannabis use is associated with improvements in functioning