The State of Clinical Outcome Assessments for Cannabis Use Disorder Clinical Trials: A Review and Research Agenda

Abstract

There is considerable variability in the use of outcome measures in clinical trials for cannabis use disorder (CUD), and a lack of consensus regarding optimal outcomes may have hindered development and approval of new pharmacotherapies. The goal of this paper is to summarize an evaluation of assessment measures and clinical endpoints for CUD clinical trials, and propose a research agenda and priorities to improve CUD clinical outcome assessments. The primary recommendation is that sustained abstinence from cannabis should not be considered the primary outcome for all CUD clinical trials as it has multiple limitations. However, there are multiple challenges to the development of a reliable and valid indicator of cannabis reduction, including the lack of a standard unit of measure for the various forms of cannabis and products and the limitations of currently available biological and self-report assessments. Development of a core toolkit of assessments is needed to both allow flexibility for study design, while facilitating interpretation of outcomes across trials. Four primary agenda items for future research are identified to expedite development of improved clinical outcome assessments for this toolkit: (1) determine whether minimally invasive biologic assays could identify an acute level of cannabis use associated with psychomotor impairment or other cannabis-related harms; (2) create an indicator of quantity of cannabis use that is consistent across product types; (3) examine the presence of cannabis-specific functional outcomes; and (4) identify an optimal duration to assess changes in CUD diagnostic criteria.

1.0. INTRODUCTION

There is a common set of challenges encountered in trial design for nearly all substance use disorders (SUDs), in particular whether the primary outcome to evaluate the intervention’s efficacy should be based on abstinence or reduction in use. The selection of primary outcome indices, including the ideal methods of assessment, has long been debated in SUD treatment literature (Donovan et al., 2012; Van den Brink et al., 2006; Wells et al., 1988). However, relative to other SUDs, clinical trials for cannabis use disorder (CUD) carry unique design considerations due to cannabis’ chemical complexity, challenges in measuring exposure, and its changing legal status. Historically, there has been considerable variability across treatment trials in the use of outcome measures for cannabis use. A recent systematic review of CUD clinical trials reported by Lee et al. (2019) identified over 60 unique outcome measures that were used within 58 identified randomized controlled trials (RCTs). This lack of consistency in outcome measures makes comparison across trial results challenging and complicates the selection of an optimal primary endpoint for future clinical trials.

The choice of outcome measures and associated endpoints in published CUD trials has varied depending on whether the tested intervention was behavioral versus pharmacological (Lee et al., 2019). These differences could be attributable to the federal regulatory and evidentiary oversight of clinical trials that test a drug product, which is not required for behavioral intervention trials. Designs for clinical trials that test a pharmacotherapy for commercial development are informed by guidance from the United States Food and Drug Administration (FDA) Center for Drug Evaluation and Research (CDER; FDA, 2014). CDER publications emphasize choosing measures with strong psychometric properties, and measures that most directly indicate treatment response (for more detail, see FDA, 2017; FDA, 2009). Clinical trials for behavioral interventions, while also emphasizing strong psychometrics and indicators of treatment response, typically include more outcome measures than trials for pharmacotherapies and are less likely to report a single defined primary endpoint (Lee et al., 2019). The major challenge in selecting appropriate outcome assessments, regardless of the type of intervention, is the lack of consensus on how best to quantify cannabis use, how to define problematic use, and how to determine what constitutes a meaningful therapeutic change. Therefore, pharmacological intervention trials typically have primary endpoints similar to those one would find in clinical trials for other illicit drugs of abuse (Donovan et al., 2012; Lee et al., 2019; Marshall et al., 2014), which are based on periods of sustained abstinence that are not unique or specific to cannabis, and that may not be sensitive to detecting meaningful change.

In March of 2018, a working group of experts from across the U.S. was assembled and met with the goals of: 1) reviewing and evaluating current assessment tools and endpoints included in CUD clinical trials, 2) defining the types of research needed to develop more refined clinical outcome assessments, and 3) developing recommendations for outcome assessment in CUD trials. The purpose of this review paper is to provide an overview of the issues related to measuring cannabis use, review the strengths and limitations of current assessment tools, report on best practices for current clinical outcome assessments and endpoints for CUD trials, and provide recommendations for improving CUD assessments in the future.

2.0. MEASURING CANNABIS USE

Historically, both behavioral and pharmacological clinical trials for CUD focused on abstinence as the most meaningful indicator of therapeutic change in a treatment for CUD. Abstinence has traditionally been considered necessary for recovery from addiction (National Council on Alcoholism, 1976) and a required treatment goal in many behavioral therapies provided in community treatment facilities, such as 12-step approaches (Miller et al., 1989). Sustained periods of abstinence have also been an accepted endpoint by regulatory agencies for evaluating and approving pharmacotherapies for illicit drug use (Food and Drug Administration Draft Guidance, 2018; Food and Drug Administration: Psychopharmacologic Drugs Advisory Committee, 2013). Measuring use of and abstinence from cannabis, however, is challenging because of the multitude of endpoints that can be selected. For example, the number of days since last use, longest duration of continuous abstinence, point prevalence abstinence (i.e., no use on the day before the assessment), proportion of individuals with continuous abstinence during treatment or between follow-up visits, and time to relapse are all examples of endpoints for various measures of abstinence in substance-related clinical trials (Hughes, 2001). Moreover, these endpoints could be assessed through multiple potential measurement tools, including biological assessments and participant self-report.

2.1. Biological Assessments

The potential advantage of choosing a biological assessment for measuring cannabis use in clinical trials depends in part on the specific compound being measured. Cannabis contains hundreds of phytocannabinoids and non-cannabinoid chemical constituents (ElSohly et al., 2017); these substances or their metabolites are potential targets to determine cannabis exposure via biological assessments. Delta-9-tetrahydrocannabidiol (THC) is the primary cannabinoid responsible for the central nervous system effects of cannabis, including intoxication (Grotenhermen, 2003), and THC (or its metabolite, 11-Nor-9-carboxy-delta9-tetrocannabinol: ‘THCCOOH’) can be measured in a variety of biological matrices using multiple commercial technologies (Huestis and Smith, 2018).

When cannabis is consumed, THC is widely distributed to highly perfused tissues including the brain, liver, and spleen, and later stored in adipose tissue. Chronic, frequent cannabis users have a large body burden of THC that is released from body stores into the blood over time, where it is metabolized and excreted (Sharma et al., 2012). Therefore, chronic, frequent cannabis users have measurable amounts of THC and its metabolites in their blood for a much longer period of time (up to several weeks) than occasional cannabis users (6 to 8 hours) (Bergamaschi et al., 2013; Sharma et al., 2012). The lipophilic nature of THC makes differentiating new cannabis use (i.e., relapse) from residual cannabinoids that remain from prior use difficult. Available biological matrices for detecting THC and other cannabinoids and metabolites include: blood, breath, hair, sweat, urine, and oral fluid, which differ in their sensitivity and window of drug detection. The detection windows of these matrices are summarized by primary analyte measured (Table 1). Each assessment is briefly considered in relation to its utility in clinical trials below.

Table 1.

Detection of Cannabis Analytes in Available Biological Matrices

Biological Matrices, Analytes	Detection Window Utility	Considerations
Blood
THCVCOOH	Presence indicates chronic, frequent use	Cannot detect recent use
THCV	Short window of detection	Low detectability
THC-glucuronide	Short window of detection	Low detectability Indicates recent use May not be detectable following oral administration
CBG	Short window of detection	Indicates recent use May not be detectable following oral administration
CBN	Short window of detection	May not be detectable following oral administration May not be present in some cannabis preparations
Breath
THC	Shortest window of detection (4 hours)	Still in development No data on sensitivity and specificity of new cannabis breathalyzers
Hair
THCCOOH	Largest window of detection (5 days - 3 months is typical) Useful for long-term follow-up	Expensive with few studies to guide interpretation
Sweat
THC	Provides cumulative record of cannabis exposure during time path is worn (typically 7 days)	Only ‘ne laboratory in the US to analyze samples
Urine
THCCOOH	Detection dependent on dose cutoff and chronicity of exposure (up to 2 days in infrequent users; more than 30 days in chronic)	With models can predict abstinence and relapse Easiest to adulterate

Note. THCCOOH = 11-nor-9-carboxy-Δ9-tetrahydrocannnabinol; THCV = tetrahydrocannabivarin; CBG = cannabigerol; CBN = cannabinol.

2.1.1. Blood.

Blood testing can provide useful data to demonstrate recent acute exposure to cannabis (Bergamaschi et al., 2013). For example, identifying cannabigerol (CBG), cannabinol (CBN), THC-glucuronide, or tetrahydrocannabivarin (THCV) in a blood sample suggests recent use (Newmeyer et al., 2016). Likewise, similar to 11-nor-9-carboxy-Δ9-tetrahydrocannnabinol (THCCOOH), THCVOOH can often be detected in blood for an extended period after a single use. Blood testing, therefore, could be used within clinical trials to indicate relapse. However, blood tests for cannabinoids do not have perfect sensitivity; while identifying CBG, CBN, THC-glucuronide, or THCV in a blood sample does indicate recent use, their absence does not preclude recent use. Frequent blood collection is also difficult and burdensome for both staff and participants, and is typically conducted in specialized laboratories so results are not immediate. Blood collection and testing is expensive, more invasive, and carries a higher risk of collection-related harm (albeit low) compared with other biological sample collections.

2.1.2. Breath.

The window of detection for THC in breath after controlled cannabis inhalation is short (less than 4 hours in chronic frequent users, less than 2 hours in occasional users), making it a potentially good indicator of recent acute dosing (Himes et al., 2013). Investigators are attempting to develop a breathalyzer that can detect acute cannabis intoxication, but to date the sensitivity, specificity, and efficiency of these devices have not been validated. Moreover, a 2 to 4-hour window of detection might not prove useful within the context of an outpatient clinical trial for CUD.

2.1.3. Hair.

Hair sample tests are easy to collect and offer the longest window of drug detection (Musshoff and Madea, 2006). Though hair tests may be ideal for documenting any exposure over an extended period of time, they would not be suitable for confirming short-term periods of abstinence within a clinical trial. Hair testing can only provide categorical information, i.e., that the person was exposed to cannabinoids, not precisely when or the magnitude of exposure. Although hair sample tests are susceptible to producing false positive tests from secondhand environmental drug exposure, targeting THCCOOH in hair avoids any issue of environmental drug exposure because it is not a constituent of cannabis smoke (Huestis et al., 2006; Romano et al., 2001).

2.1.4. Sweat.

Sweat tests are convenient assessments for measuring the cumulative record of THC exposure during the period in which an individual wears a sweat patch, and provides longer detection windows than urine (Huestis et al., 2008). Sweat tests’ cumulative means of monitoring drug use over time makes it an attractive biological assessment. However, individuals produce sweat inconsistently over time, leading to weak dose-concentration relationships (Barnes et al., 2008). Sweat tests can also easily be contaminated from secondhand or thirdhand environmental exposure.

2.1.5. Urine.

Urinalysis for cannabis typically assesses for the presence of the THC metabolite THCCOOH. Urinalysis testing can easily be incorporated into clinical trials because it is the most common method of detecting substances of abuse and is often used in treatment settings as part of standard care. Qualitative results of urinalysis (positive or negative) are also available immediately. The limitation of this form of testing is that it can often take a month or longer for a chronic frequent cannabis user during abstinence to have a “negative” urine test (Lowe et al., 2009). There are also sex differences in concentration of THCCOOH in urine; female chronic frequent cannabis users have a significantly longer window of detection than male chronic frequent users (Karschner et al., 2009). Despite its limitations, urinalysis testing could still hold utility as an outcome assessment in clinical trials if used in conjunction with available models for calculating predicted metabolite levels (see Huestis and Smith, 2018; Schwilke et al., 2011; Smith et al., 2009). These models can be used to predict trajectories of THCCOOH concentrations in urine during periods of sustained abstinence. When patients’ THCOOH concentrations fall outside of the ranges predicted by the models (one model for occasional users and another model for chronic frequent cannabis users), it indicates recent use (relapse) occurred between two samples. These models are utilized in outpatient studies to corroborate participant self-report of abstinence (Schuster et al., 2016), but need additional validation.

2.1.6. Oral Fluid.

Both THC and THCCOOH are detectable in oral fluid following recent cannabis use. Although it would be predicted that the window of detection for cannabinoids in oral fluid is shorter than the detection of THCCOOH in urine, Quest Diagnostics Annual Report on Workplace Drug Testing shows equivalent detection rates in these matrices. Oral fluid offers an observed, gender neutral collection that is less invasive, and more difficult to adulterate than urine testing (Desrosiers et al., 2014). In addition, oral fluid is greatly preferred by donors and collectors. Cannabinoids do not persist in oral fluid from chronic frequent cannabis users for as long as they persist in urine. However, there is some disagreement on whether oral fluid assessed in early abstinence can consistently and reliably differentiate new cannabis use from residual cannabinoid excretion.

2.2. Self-Report Assessments

Concordance between biological assessments and self-report measures of abstinence are fairly high (kappa range: 0.6 – 0.9; Barrowclough et al., 2014; Copeland et al., 2001; Dennis et al., 2004; Hoch et al., 2014). The most common instruments for measuring self-reported cannabis use in clinical trials are recall diaries, including the time line follow-back (TLFB; Norberg et al., 2012), which uses a retrospective calendar of events to prime a participant’s memory of their use. The TLFB was developed to measure alcohol exposure, but has been adapted to assess other substance use, including cannabis use (Robinson et al., 2014). TLFB instruments are the most frequently used self-report assessments in trials for CUD (Lee et al., 2019), though due to their subjective nature, historically were not recommended as valid outcome measures. Ecological Momentary Assessment (EMA) is another diary method that may improve accuracy of self-reported cannabis use through collection of real- or near-time assessment of use, thereby reducing retrospective recall bias (Shiffman et al., 2008). It also has the potential to assess various cannabis metrics and can provide a fine-grained and ecologically valid evaluation of the association with predictors of use, such as withdrawal, craving, and affect (Buckner et al., 2015; Wycoff et al., 2018).

The challenge with using TLFB, EMA, or any other diary measures to assess cannabis use lies in the ambiguity for the optimal unit of cannabis measurement to be assessed. TLFB and EMA instruments can be used to collect data on how often (frequency, such as number of days out of some period and/or how many occasions per day they use), for how long (duration, such as number of hours or proportion of waking hours acutely intoxicated), and in what situations (context) individuals use cannabis. Assessing quantity of use in a use episode, however, is quite challenging because of the significant variability in cannabis potency, preparation, and method of administration.

The number of joint/bowls/cones or “puffs” used within a given window of time is the most frequently used metric for assessing quantity of cannabis used. However, there is no standard size of a single “joint”, “bowl,” “cone,” or “puff,” and users’ estimate of the typical amount they roll or pack is largely inaccurate (Hindocha et al., 2017; Prince et al., 2018). Indeed, one experimental study that asked participants to roll their own joints found large variability in the amount of plant material used across participants (Mariani et al., 2011). Showing participants images of “standard sizes,” thereby creating a metric for participants to report their average use, does little to improve the accuracy of self-reports (Van der Pol et al., 2013), but could be used to measure within-subject change. Furthermore, cannabis users might share a unit of consumption (e.g., joint) with others, making precise measurement of individual exposure challenging.

Some researchers have begun assessing quantity of smoked/vaporized flower cannabis by asking participants to report on the average grams they consume over time (e.g., Tossmann et al., 2011; Van Dam and Earleywine, 2010). However, units in grams do not map on to other forms of administration, such as oils/tinctures, patches/gels, and edibles, which are all gaining in popularity (Russell et al., 2018). Assessing quantity by number of joints/bowls/cones or grams also does not account for differences in THC potency.

Some researchers have attempted to measure quantity of cannabis use by assessing the dollars paid for the amount used. Price would theoretically incorporate information on quantity and potency, and be less dependent on differences in method of administration. However, not all cannabis users purchase their product (Walsh et al., 2013). Moreover, price can change dramatically depending on the legal status of cannabis in the individual’s municipality, as well as fluctuations in the licit and illicit cannabis economies (Sifaneck et al., 2007). Further, this metric cannot account for individual quantity when cannabis is shared.

Assessing the psychoactive potency of cannabis is challenged by the complexity of the cannabis plant and the variability in the content of the principal active constituents THC and CBD; preparations that contain high concentrations of CBD and very low concentrations of THC are associated with far less risk of problematic use compared to preparations containing high concentrations of THC (Babalonis et al., 2017). Changing from a cannabis product that is higher in THC content to a cannabis product that primarily contains non-intoxicating CBD during a clinical trial might not be identified through assessment of price paid. To avoid some of these issues, some researchers have innovatively used assessment of typical and maximum intoxication as a proxy for quantity, potency, and cannabinoid ratio (Walden and Earleywine, 2008). Intoxication, however, varies across individuals, is dependent on history of use, and does not account for the development of tolerance (Kirk and De Wit, 1999).

3.0. MEASURING PROBLEMATIC USE

Clinical endpoints should reflect the desired effect of treatment. FDA resources for drug development focus on patient reported outcomes (PROs) as endpoints that capture how a patient feels, functions or survives, rather than merely quantifying some reduction in exposure to a drug (FDA Center for Drug Evaluation and Research). Treatment of many non-substance-related psychiatric syndromes, such as anxiety, PTSD, and depression, conceptualize the desired clinical benefit as a reduction in symptoms or remission of the disorder itself. A diagnosis of CUD is typically the primary inclusion criterion for CUD RCTs. However, a reduction in the severity of the CUD (number of diagnostic criteria met pre/post intervention), time-to-treatment response (average time for patients to see a clinically significant reduction), and diagnostic rate (number of patients who still meet CUD criteria at end of treatment) are rarely included as primary endpoints in clinical trials. Instead, abstinence or days of use are emphasized in most CUD trials (Sherman and McRae-Clark, 2016).

Face-to-face interviews and self-administered measures could be used in CUD trials to directly assess changes in diagnostic status. The most commonly used diagnostic measures include the clinician-administered Structured Clinical Interview for DSM-IV or 5 (SCID) assessment for CUD (First et al., 2015), the Mini-International Neuropsychiatric Interview (M.I.N.I.; Sheehan et al., 1998), and the Psychiatric Research Interview for Substance and Mental Disorders (PRISM; Hasin et al., 1996; Hasin et al., 2006). These semi-structured assessments, however, can be lengthy and require specialized training when administered in person. Brief, self-administered assessments, such as the Cannabis Use Disorders Identification Test (CUDIT-R), have demonstrated high concordance between identifying DSM criteria and a CUD diagnosis (Adamson et al., 2010), and would be quick and easy to administer in a clinical context. However, brief measures, such as the CUDIT, are not intended to be diagnostic and cannot be used to identify psychiatric comorbidities.

Relying on a measure that is anchored to DSM has the limitation that it may require updating as revisions to the DSM are published. An alternative approach could focus only on assessment of pre-/post-treatment change in DSM-5 severity modifiers (e.g., a shift from severe (6+ symptoms) to mild (2–3 symptoms)), though, given the potential for floor effects, this might limit enrollment to those with only moderate or severe forms of CUD in clinical trials.

4.0. FUNCTIONING AND OTHER OUTCOMES

All SUDs are defined by a pattern of continuing to use the substance despite its interference in major domains of functioning. Similar to assessing disordered use, functioning is rarely assessed as a primary or even secondary outcome for most CUD trials. This may be attributable to the difficulty in demonstrating functional change within the temporal constraints of most RCTs, which are brief. However, the changing landscape related to cannabis suggests that assessment of secondary effects of the CUD and/or functioning may be equally if not more important than assessment of use or an individual CUD criterion. Indeed, relying on change scores for CUD assessments that map on to DSM-5 criteria come with their own set of limitations. Given the currently defined criteria for CUD, many of these assessments include items that correlate with frequency and quantity of use, such as items that assess time or money spent seeking or using cannabis. Reliance on these items runs the risk of over- or under-identifying CUD in populations that use cannabis frequently, but may or may not be experiencing problems, such as medicinal cannabis users (Loflin et al., 2017). Nevertheless, even “medicinal use” can become problematic if it is worsening or not improving daily functioning.

Assessing functioning and other consequences of use as primary or secondary endpoints of an intervention might be appropriate for populations where the benefit of some moderate amount of use might outweigh risk, and where reduction in use (and improved functioning), rather than abstinence, is the goal of treatment. A meaningful reduction in substance use is typically defined by a level of change that is predictive of improvements in long-term functioning and reduced consequences (Kiluk et al., 2019; Winchell et al., 2012). For example, “no heavy drinking days” has been validated as a clinically meaningful endpoint for treatment of Alcohol Use Disorder (Falk et al., 2010). One secondary analysis of a medication trial for CUD found that reduced frequency of cannabis use was associated with improvements in overall quality of life (Brezing et al., 2018). However, there are no published CUD clinical trials with reduced cannabis use as a primary endpoint, and there is no currently accepted “safe” level of use. This lack of consensus on a clinically meaningful reduction endpoint for cannabis could be attributable to the weak association between cannabis use and broad areas of psychosocial functioning (Kiluk et al., 2019).

Even in instances where abstinence is the explicit goal of treatment, assessing functioning as a secondary outcome would be beneficial. For example, most cannabis users in treatment for CUD cite abstinence as their primary goal (Lozano et al., 2006). However, for many of these individuals the reason for seeking abstinence is to be able to pass a drug screen for employment or because they are legally mandated to do so through a drug court (Stephens et al., 1993). Likewise, adolescents rarely self-refer for treatment (Tims et al., 2002), and a goal of abstinence may be imposed by family or social agencies. In these examples, the motivation for reduced use or abstinence is external and/or functional, suggesting that a measurement of functioning or other indicators of pathology might be a more appropriate indicator of change.

5.0. RECOMMENDATIONS

Despite historical reliance on total abstinence as a typical outcome measure, abstinence does not need to be the primary outcome for all CUD clinical trials. The multiple challenges in measuring abstinence from cannabis, as well as acknowledgement that abstinence is a high bar that may be difficult to achieve and sustain (Volkow et al., 2018), suggests other outcomes should also be considered. Recent FDA draft guidance for industry regarding endpoints for demonstrating effectiveness of medications for opioid use disorder indicate potential acceptance of drug use patterns other than abstinence as thresholds to define response to treatment (FDA, 2018). Also, there is precedent for regulatory agencies accepting non-abstinence endpoints (e.g., no heavy drinking days) that have demonstrated associations with reduced harm from substance use (European Medicines Agency, 2010; Food and Drug Administration Draft Guidance, 2015). Additional outcomes reflecting the view of patients with regard to how they “feel, function, or survive” may also be appropriate (FDA, 2014). Moreover, there should be flexibility in what outcome assessments are selected for any one trial based on the population and type of intervention. Therefore, to be able to make cross-study comparisons and to be able to perform meta-analyses to assess treatment effects, most primary and secondary outcome measures for CUD RCTs should be selected from a core standard toolkit of CUD assessment measures, which urgently needs to be developed.

While development of a core assessment toolkit for CUD is critically needed, it will likely require validation of new outcome measures. Given the pursuit of clinical trial endpoints that are cannabis-specific and sensitive to detecting meaningful change, there are four primary research agenda items for new CUD measure development that are needed:

(1). Determine whether minimally invasive biologic assays could be used to identify an acute level of cannabis use associated with psychomotor impairment or other cannabis-related harms.

Currently, none of the biological tests described above can be used to easily and reliably confirm self-reported reductions in cannabis use frequency/severity without relying on complex algorithms. There are potential analytes for differentiating new cannabis use from residual excretion in blood and oral fluid in occasional cannabis users and chronic frequent cannabis users (Newmeyer et al., 2016; Swortwood et al., 2017), but additional research is needed to develop objective biological indicators of early abstinence from cannabis. Furthermore, refinement of biological assessments will need to be able to distinguish critically different patterns of use and types of cannabis products To be meaningful and valid according to FDA standards, the biologic assay should demonstrate association with improvements in how patients feel and function, i.e., reduced cannabis-specific adverse consequences or improved functional outcomes.

(2). Determine whether it is possible to create an indicator of quantity of use that is consistent across cannabis product types.

While the TLFB provides a standardized method for assessing frequency of self-reported use (Robinson et al., 2014), there is no consensus on a valid unit of measurement for quantity that could be consistently obtained across cannabis product types. Indeed, a consensus paper for stimulant use disorder endpoints was that such a standard unit of measure does not exist for stimulants, such as cocaine, and developing such a measure might not be possible (Kiluk et al., 2016). ‘Standard THC units’ have been proposed for quantifying cannabis exposure, with one standard unit fixed at 5mg of THC for all cannabis products and methods of administration (Freeman and Lorenzetti, 2019). However, this approach may be most feasible in legal markets where cannabis products are marketed and sold with product labels and known content. Creating an objective, standardized unit of measurement for quantity of cannabis use that will not prove burdensome in a clinical setting will be challenging, but is critically needed if reduction in cannabis use is chosen as a primary outcome for a clinical trial. This unit will likely need to incorporate/reflect information on: (1) estimated cannabinoid content (ratio and potency), (2) weight/mass of product, and (3) method of administration. Until this measure has been created and validated, clinical trials should consider at a minimum collecting information on days of use and number of times used per day.

(3). Determine whether there are cannabis-specific functional outcomes and how they are best assessed.

The most common secondary measures of CUD pathology and domains of functioning that are assessed in relation to CUD include: withdrawal, craving, mood, sleep, cognitive functioning, psychosocial functioning, problems related to cannabis use, self-efficacy to manage cannabis use, and quality of life. Several trials have used craving and withdrawal as secondary endpoints (e.g., Hill et al., 2017; Killeen et al., 2012; Trigo et al., 2018), and three pharmacotherapy trials have used withdrawal as a primary endpoint (Bhardwaj et al., 2018; D’Souza et al., 2019; Johnston et al., 2014). However, the majority of both pharmacotherapy and behavioral RCTs have not assessed how changes in withdrawal and craving impact other measures of functioning throughout a quit attempt. Moreover, even if it were standard practice to include assessment of functioning and correlates of CUD, the sheer number of potential assessments that could be selected would make drawing comparisons across interventions enormously challenging. For example, in the systematic review by Lee et al. (2019), at least 17 instruments were used to collect data on psychosocial functioning in trials for CUD. Research is needed to determine which of these measures of functioning are most suitable for inclusion in clinical trials. To promote utilization and comparison across clinical trials, it is recommended that selected measures be brief, readily available, and culturally sensitive.

(4). Identify an appropriate period of time for assessing changes in DSM-5 criteria for CUD.

An advantage of using reduction in DSM-5 CUD severity levels is that the outcome is the same condition that is being treated, and it also indirectly permits a non-abstinent reduction in use. However, the major issue with using assessment measures that map on to DSM-5 criteria is that DSM-5 criteria for CUD specify that symptoms must occur within a 12-month period (Hasin et al., 2013), making the use of unmodified DSM-5 criteria potentially incompatible with RCTs that are less than one year in duration. However, DSM-5 does include a specifier of ‘in early remission’, defined by the absence of symptom criteria for at least three months, which could be useful for detecting a meaningful indicator of improvement following treatment. A one-month duration was previously specified in DSM-III for the diagnosis of Substance Abuse or Dependence (Spitzer et al., 1980). While there may be promise in using a 30-day duration for measuring change in DSM-5 substance use disorder diagnostic status or severity (Kiluk et al., 2018), a consensus will need to be reached for the optimal duration of time for absence or reduction in DSM-5 CUD criteria.

Potential new measures or modifications of existing measures should attempt to remedy the existing limitations of CUD outcome assessments, while maintaining consistency with the FDA’s CDER best practices guidelines (FDA, 2014). Specifically, new or modified clinical outcome assessments must include representative patient/layperson input, demonstrate strong psychometric properties (e.g., construct validity, internal and external reliability), and be capable of providing direct evidence of a clearly defined treatment response (FDA, 2009).

Once developed, multiple factors should be considered when selecting among these measurement tools and choosing corresponding endpoints for clinical trials, such as the population under study, type of intervention, mechanism of the intervention, and length of treatment. In addition, researchers must be mindful of participant and staff burdens associated with data collection. For those seeking immediate guidance on best choice of CUD clinical outcome assessments, future CUD trials (as well as other psychiatry trials that may include individuals with CUD or using cannabis) should include, at a minimum, each of the elements listed in Table 2. Time and frequency assessments can provide flexibility in data summarizations (i.e., assess daily use over a week, and daily use over a month). If all clinical trials for CUD include at least one assessment from each of the content elements listed in Table 2, researchers will be able to compare results across pharmacological and behavioral treatment trials for CUD while refining development of a standard toolkit of clinical assessment measures.

Table 2.

Recommendations for a Core Standard Assessment Toolkit

Critical Assessment Elements	Recommended Assessments	Future Development
Primary Outcome Measures
Symptom Checklist corresponding to the DSM-5 Criterion for Cannabis Use Disorder, which assesses: diagnostic status, number of symptoms, severity of the disorder.	SCID (APA, 2015) MINI (Sheehan et al., 1998) PRISM (Hasin et al., 2006) CUDIT-R (Adamson et al., 2010)	Assessments must be periodically updated to reflect changes in DSM and ICD-10 criteria for CUD. Consensus needed on optimal duration of time for presence of diagnostic criteria (e.g., 30 days, 90 days).
Self-report assessment of frequency of use.	TLFB (Robinson et al., 2014b)	Consensus needed on time period to be assessed.
Self-report assessment of quantity of use.	N/A	Development of measure assessing quantity of use that incorporates: total amount, cannabinoid potency, and/or method of administration.
Biological assessment to validate self-reported reduced use or abstinence.	Urinalysis (THCCOOH) Oral Fluid (THCV, CBG)	Creatinine-normalized ratio models used to predict relapse from urinalysis results will need to be validated in outpatient samples.
Secondary Outcome Measures
Assessment of Quality of Life and/or Psychosocial Functioning	N/A	Cross-validation is needed between the most commonly used CUD measures and assessments of function and quality of life.

Note. N/A = no consensus was reached for a recommended assessment; SCID = Structured Clinical Interview for DSM-5; MINI = Mini-International Neuropsychiatric Interview, PRISM = Psychiatric Research Interview for Substance and Mental Disorders; CUDIT-R = Cannabis Use Disorders Identification Test, Revised; TLFB = Timeline Follow Back Instrument; THCCOOH = 11-nor-9-carboxy-Δ9-tetrahydrocannnabinol; THCV = tetrahydrocannabivarin; CBG = cannabigerol.

6.0. CONCLUSIONS

Following a thorough review of the literature and discussion of the limitations and challenges inherent in measuring cannabis use and CUD, expert consensus was reached on several core outcome domains that should be included in all CUD trials, but consensus was not reached on any specific set of outcome assessments that should be consistently used. There is a pressing need for development of a standard toolkit of assessments, which could be used flexibly to meet the aims of CUD clinical trials. Four primary research agenda items that could help to move the field closer to reaching consensus on best practices when selecting CUD outcome measures and development of a standard clinical outcome assessment toolkit for CUD trials were identified.

Highlights.

• Provides a review of clinical outcome assessments for CUD trial endpoints.

• Abstinence is not the only valid indicator of CUD treatment success.

• A core toolkit of assessments for CUD treatment studies is needed.