Human Laboratory Models of Reward in Substance Use Disorder

Abstract

Human laboratory models in substance use disorder provide a key intermediary step between highly controlled and mechanistically informative non-human preclinical methods and clinical trials conducted in human populations. Much like preclinical models, the variety of human laboratory methods provide insights into specific features of substance use disorder rather than modelling the diverse causes and consequences simultaneously in a single model. This narrative review provides a discussion of popular models of reward used in human laboratory research on substance use disorder with a focus on the specific contributions that each model has towards informing clinical outcomes (forward translation) and analogs within preclinical models (backward translation). Four core areas of human laboratory research are discussed: drug self-administration, subjective effects, behavioral economics, and cognitive and executive function. Discussion of common measures and models used, the features of substance use disorder that these methods are purported to evaluate, unique issues for measure validity and application, and translational links to preclinical models and special considerations for studies wishing to evaluate homology across species is provided.

1. Human Laboratory Models of Reward in Substance Use Disorder

Human laboratory models are critical for evaluating the mechanisms of substance use disorder to inform effective intervention development. These methods occupy an intermediary step between highly controlled and mechanistically informative non-human preclinical methods and clinical trials conducted in human populations. Human laboratory studies allow for much of the control afforded by precise experimental manipulations in laboratory settings while also offering an opportunity to translate mechanistic discoveries to clinical application. When effectively conducted these studies can provide an important translation from preclinical signals to clinical trials for drug development (Czoty et al., 2016).

There has been extensive discussion establishing that individual preclinical models of substance use often provide insights into particular features of substance use disorder rather than recapitulating the full disorder (e.g., Lamontagne and Olmstead, 2019; Smith, 2020; Vanderschuren and Ahmed, 2021). There has been less dialogue scrutinizing whether human laboratory methods similarly exemplify specific features of substance use disorder rather than modelling the diverse causes and consequences simultaneously in a single model. The purpose of this narrative review is to discuss the advantages and disadvantages of existing models of reward and key neurocognitive processes in substance use disorder from this perspective. We offer a narrative review that emphasizes the specific contributions that each model has towards informing clinical outcomes (forward translation) and analogs within preclinical models (backward translation).

We review four core areas of human laboratory research on reward and related neurocognitive mechanisms in substance use disorder: drug self-administration, subjective effects, behavioral economics, and cognitive and executive function. In each section we first provide an overview of common measures and models used to evaluate the domain of interest. We then describe intervention applications and the features of substance use disorder that these methods are purported to evaluate. Unique issues for measure validity and application are then described. Finally, we discuss the translational links to preclinical models and special considerations for studies wishing to evaluate homology across species. We conclude with a discussion of outstanding gaps between the laboratory and clinical environment and novel methods that may help lessen these gaps for effective translation.

2. Drug Self-Administration

Grounded in operant theory, self-administration procedures evaluate the positive reinforcing effects of drugs. In other words, drugs increase the likelihood of the behaviors leading to their delivery due to their positive consequences for the organism (Comer et al., 2008; Jones and Comer, 2013; Panlilio and Goldberg, 2007). Within this framework, self-administration paradigms serve multiple purposes that include (but are not limited to) (i) modeling patterns of drug-taking behavior in both human and pre-clinical laboratory studies; (ii) determining the abuse potential of pre-existing and novel psychoactive compounds, including pharmacotherapies for substance use disorders; (iii) assessing the effects of candidate medications on drug intake; (iv) exploring the impact of various reinforcement contingencies on drug-seeking and drug-taking behaviors; and (v) elucidating the neural substrates mediating drug reward (Jones and Comer, 2013; Richardson and Roberts, 1996). Multiple reviews have comprehensively summarized human drug self-administration procedures (e.g., Comer et al., 2008; Henningfield et al., 1991; Jones and Comer, 2013); as such, the present paper will briefly describe common contemporary human self-administration procedures, examine their impact on medications development, discuss consideration of these models, and describe the translational appeal of their pre-clinical analogues.

2.1. Review of Non-Operant and Operant Drug Intake

Self-administration procedures in the human laboratory confer many advantages. Notably, they offer strong face validity, as drug intake constitutes a vital component of chronic substance use. Additionally, self-administration assays have proven effective in evaluating candidate medications (Czoty et al., 2016; Haney, 2009; Haney and Spealman, 2008; Panlilio and Goldberg, 2007; Regnier et al., 2022). The below section reviews common procedures used in these methods.

2.1.1. Free Access

In free access procedures, participants consume drug ad libitum; these procedures are often when the risk of transient adverse effects and overdose are low (e.g., drugs like nicotine and caffeine). For instance, Spiga et al. (1998) examined the effects of ad libitum nicotine on methadone self-administration in which participants freely consumed cigarettes that were provided by the experimenters prior to methadone availability. Free access procedures may be especially useful when included in the context of larger experimental studies (e.g., residential medications development studies) to determine secondary effects on drug consumption (e.g., decreases in ad lib tobacco use under a putative pharmacotherapy).

2.1.2. Intermittent Operant Schedules

During basic operant paradigms, participants emit a specific behavior (such as pressing a button or moving a joystick) to receive drug. Pre-clinical work influenced these approaches for human drug self-administration, which were subsequently modified for human subjects (Deneau et al., 1969; Haney and Spealman, 2008; Jones and Comer, 2013; Schuster and Thompson, 1969). Fixed ratio (FR) schedules are commonly used in the human laboratory; completing a set number of operant responses on these schedules results in reward delivery. For instance, on a FR 10 schedule, every tenth lever press may result in a heroin infusion. The primary dependent variables include the total amount of drug delivered, the number of drug deliveries, the pattern of responding on the active lever, and the rate of responding on the active lever. Other simple intermittent schedules include the fixed interval (FI), variable ratio (VR), and variable interval (VI) schedules in which responses may result in drug delivery after some period of time has passed (i.e., interval schedules) or may have a variable number of responses or time required (i.e., variable schedules). Under controlled laboratory conditions, assays typically incorporate limited holds (i.e., refractory periods where scheduled consequences are not available) and establish a maximum number of infusions to ensure that participants do not overdose or experience significant adverse health consequences.

2.1.3. Progressive Ratio Schedules

To measure motivation to obtain a drug, studies can rely on progressive ratio schedules (Richardson and Roberts, 1996). Under these conditions, each reinforcer delivery causes a systematic increase in the ratio value required, such that the amount of effort to obtain the same unit dose of drug increases over the course of the session. For example, Stangl and colleagues (2022) describe a progressive ratio schedule with intravenous alcohol administration that instructed participants to press a button to self-administer alcohol, and each infusion sequentially increased the ratio requirement for the next reward delivery. Ratio values may increase either arithmetically according to defined rule sets or logarithmically. PR sessions conclude when subjects reach the “breakpoint,” the ratio value at which the subject no longer completes the scheduled requirement. This value serves as the primary dependent variable. Broadly speaking, higher breakpoint values denote greater abuse potential. These schedules are conceptually similar to behavioral economic demand approaches discussed later, but differ in their focus on determination of a breakpoint value (as opposed to consumption levels across the price sequence).

2.1.4. Second-order Schedules

To examine the effects of drug-related stimuli on operant responding, investigators can use second-order schedules, which merge two intermittent schedules of reinforcement to form a more complex contingency. Consider the following example of participants on a FR 10 (FR 5). Participants must complete five lever presses (FR 5), which results in a cue (e.g., a green light). The participant must complete that schedule 10 times (FR 10) to obtain the drug reward. In this example, the drug is the primary reinforcer, while the light (or cue stimulus) becomes a secondary reinforcer due to its association with the terminal drug delivery. Both the primary and the secondary reinforcers play an important role in maintaining responding in these procedures. Given that abuse potential is contextual, these schedules can isolate contextual factors that impact use. Chronic substance use can also stem from a lack of stimulus control (Lamb and Ginsburg, 2018). Thus, second-order schedules allow for the evaluation of stimulus control over drug-seeking behaviors. In addition, these schedules also reduce the impact of locomotor impairment of operant responding that can limit the evaluation of rate-dependent outcomes.

2.1.5. Drug Choice Procedures

Self-administration may also present participants with a discrete choice between a drug reward and some alternative, such as money. Such procedures are thought to more closely emulate conditions in the natural environment and recognize the heterogeneity of real-world drug taking, given that choices are often made between drugs and at least one other outcome. During self-administration sessions, participants in drug choice procedures can complete an operant response (e.g., clicks on a computer mouse) to earn portions of drug or money on separate schedules (Jones and Comer, 2013). Cumulative amounts of money and drug are typically delivered at the end of the session as a bolus drug dose, although for shorter-acting drugs (e.g., intranasal cocaine) doses can be delivered throughout a session after completion of each scheduled requirement. Examples include evaluating cocaine choice in the context of varying alternatives (Stoops et al., 2010), assessing the reinforcing efficacy of the atypical analgesic tramadol (Babalonis et al., 2013), and examining the effects of sublingual naloxone-buprenorphine maintenance on oxycodone choice (Jones et al., 2011).

2.2. Application of Clinical Models of Drug Self-Administration to Varying Drug Classes

The literature details a wide variety of schedules of reinforcement governing opioid intake (e.g., FR, PR, drug-money choice assays, and other combinations thereof). Self-administration studies using opioids often use intravenous self-administration through an intravenous catheter placed at the start of experimental sessions. Just as people tend to use opioids with high periodicity (partly to avoid the negative consequences of opioid withdrawal), opioid self-administration may be available 3-4 times at evenly spaced intervals during the course of a human laboratory study. Mello and colleagues (1981), for example, modeled this phenomenon when they tested the impact of naltrexone pre-treatment on heroin intake. During sessions (held four times daily), participants chose between heroin and money by pressing a button; reinforcer delivery was governed by high-ratio value FR second-order schedule. To ensure participants did not overdose or experience serious adverse events, an experimentally-imposed ceiling was placed on the amount of heroin participants could earn each day. A more recent study also used drug self-administration studies to demonstrate the safety and feasibility of using the rapid-acting opioid agonist remifentanil (Lile et al., 2024). Participants received intravenous infusions of remifentanil every minute for 40 minutes with safe and effective doses observed in this rapid, repeated procedure. These procedures document the ability of self-administration methods to adapt to the drug use pattern and pharmacological effects of the drug and class of interest.

In the clinical laboratory setting, alcohol may be self-administered orally or intravenously (Chukwueke and Le Foll, 2019). For the former, sessions can begin with a priming drink of alcohol to achieve a specific blood or breath alcohol concentration (BAC) followed by a self-administration phase. On occasion, participants may receive monetary compensation for drinks declined, similar to drug versus money choice procedures. Participants may also self-administer ethanol via an intravenous catheter. Emitting an operant response (e.g., pressing a button) results in one contingent infusion of ethanol. Similar to oral alcohol self-administration, intravenous alcohol procedures consist of both a priming phase and a self-administration phase. Intravenous approaches can achieve steady state BAC values and avoid stimulus impacts related to alcohol consumption (e.g., taste) allowing for the precise evaluation of pharmacological effects.

Cannabis self-administration experimental designs can follow several procedures, including operant self-administration (e.g., pulling lever to receive a puff of smoke), controlled-smoking procedure (where experimenter cues precede inhalation), choice procedures (such as drug versus money tasks), and free-access (ad libitum) studies (Spindle et al., 2018; Xiao et al., 2023). Combusted smoke is a common route of administration, though studies can also use vaporized or oral routes of administration. Use of inhalation procedures necessitates additional control over dosing and consideration of the administration process (e.g., to account for sidestream smoke loss).

2.3. Interventions Development and Drug Self-Administration

Drug self-administration procedures play a salient role in medications development; they elucidate the efficacy and abuse potential of candidate medications for SUDs. These procedures provide direct evaluation of changes in the positively reinforcing effects of drugs and reductions that may occur under varying experimental settings. An extensive literature exists describing medications development for substance use disorders (e.g., Comer et al., 2008; Hart, 2005; Regnier et al., 2022). The current review underscores examples of clinical translation to illustrate the relevance of self-administration in screening novel pharmacotherapies across common drug classes.

Self-administration procedures have been used to evaluate novel or existing medications for opioid use disorder (MOUDs). These treatments translated from the human laboratory to the clinic with varying degrees of success. For example, methadone effectively decreased opioid intake in self-administration procedures (e.g., Donny et al., 2005). Naltrexone similarly successfully reduced opioid use among non-treatment seekers (e.g., Mello et al., 1981). Currently, there are three FDA-approved treatments for alcohol use disorder: disulfiram, naltrexone, and acamprosate (Chukwueke and Le Foll, 2019). Naltrexone has been most extensively evaluated in human laboratory contexts for its effects on alcohol self-administration. One meta-analysis of human laboratory studies demonstrated that naltrexone reduced alcohol consumption with a small-to-medium effect size (Hedges’ g = −.277) (Hendershot et al., 2017). Currently, no FDA-approved treatment for cocaine use disorder exists. Both pre-clinical and clinical trials have suggested that d-amphetamine may be an effective treatment for cocaine use disorder (Chiodo et al., 2008; Czoty et al., 2011; Czoty et al., 2010; Negus and Mello, 2003; Rush et al., 2009; Rush et al., 2010). Human laboratory studies, specifically, have demonstrated efficacy with oral d-amphetamine treatment decreasing intranasal cocaine choice (Rush et al., 2009; Rush et al., 2010) consistent with data demonstrating that individuals maintained on sustained release d-amphetamine and methadone had a significantly greater proportion of cocaine negative urine samples relative to controls (Grabowski et al., 2004). More broadly, synthesis of the preclinical, human laboratory, and clinical literature on cocaine trials showed that when concordant self-administration procedures were used between preclinical and human laboratory studies that concordant outcomes were observed with clinical trials (Czoty et al., 2016).

2.4. Validity and Considerations of Human Laboratory Drug Self-Administration Procedures

2.4.1. Schedules of Reinforcement

Schedule of reinforcement can have significant impacts on the interpretation of self-administration data and clinical translation. FR schedules indicate whether a specific unit dose of drug maintains self-administration, thereby suggesting that the drug is a reinforcer (Richardson and Roberts, 1996). Nonetheless, FR data can present several theoretical problems. Of interest, reinforcement alone does not elucidate the specific mechanism(s) that underlie drug-seeking or drug-taking behavior, but simply describes a relationship between environmental contingencies and the relevant behavior (Arnold and Roberts, 1997). Drug self-administration on an FR schedule is also not indicative of the level of (or changes in) reinforcing efficacy (Arnold and Roberts, 1997; Richardson and Roberts, 1996). In contrast, PR tests measure motivation to obtain drug reward and, by proxy, are thought to help isolate reinforcing efficacy (Richardson and Roberts, 1996). These procedures can be limited, however, by the specific sequence selected and the sensitivity of response to this sequence. Collectively, these data emphasize the importance of considering schedule of reinforcement when designing and interpreting data from human laboratory designs.

2.4.2. Drug-Specific Considerations

Route of administration differences and impacts on drug delivery can also impact the interpretations across drugs and classes. For example, oral alcohol self-administration provides improved homology with administration in real-world environments, but intravenous delivery circumvents individual pharmacokinetic variability, orosensitivity, and calorie-related confounds that can impact controlled alcohol delivery (Stangl et al., 2022). Similarly, the human laboratory cannabis literature presents significant methodological variability, particularly related to ad libitum procedures. Lack of standardization of the primary dependent variables (number of puffs, puff volume, duration of smoke inhalation, number of cannabis cigarettes smoked, or pre-post weight comparison of cigarettes) can hinder comparisons between studies (Xiao et al., 2023). Titration poses another threat to translation when using inhalation methods. Participants can regulate breath depth and duration to titrate cannabis or other inhaled drug intake. In sum, self-administration assays should be tailored to the drug of interest.

2.4.3. Ethical Considerations and Recruitment

Due to ethical concerns, human laboratory self-administration studies primarily recruit non-treatment seeking individuals who use drugs. This population is readily responsive to drugs’ reinforcing effects and drug and contextual manipulations in laboratory settings (Haney, 2009; Jones and Comer, 2013; Moeller and Stoops, 2015). However, systematic exclusion of populations in or seeking treatment may limit generalizability and translation of findings. Recent research in alcohol self-administration has recognized this limitation and proposed the use of self-administration procedures in populations that are seeking treatment, but for which reduced drinking is a treatment goal (Roberts et al., 2021). This design is made possible by the recognition that alcohol use reduction is a clinically viable treatment goal (as well as an FDA-approved endpoint). Alternative paradigms including some of those discussed in this review (e.g., simulated behavioral economic demand tasks) that do not necessarily entail drug intake may also allow researchers to ethically examine reinforcing effects among individuals in recovery or seeking treatment (Moeller and Stoops, 2015).

2.4.4. Pre-Clinical Self-Administration Analogues: Translational Implications

Pre-clinical behavioral studies constitute a critical screen in the drug development pipeline, allowing investigators to gauge the abuse potential of novel pharmacotherapies. Importantly, drug self-administration across species (from rodent to humans) is largely conserved. In other words, animal models of drug self-administration offer strong face validity and translational homology. Consistent with the human laboratory literature, pre-clinical drug self-administration may occur under free access conditions or may rely on operant contingencies (e.g., FR and PR schedules of reinforcement). Rodent models of psychostimulant and opioid self-administration are robust, demonstrating predictable dose-dependent levels of responding.

In contrast, alcohol and cannabis self-administration in preclinical models has proven challenging. Although rodent models of oral alcohol self-administration exist, the aversive taste of ethanol slows acquisition and may require special acquisition procedures. Studies often use sucrose fading (e.g., sweetening alcohol with sugar) to facilitate alcohol intake. However, sucrose can act as a confound, as the reinforcing effects of sugar may drive alcohol consumption. As alcohol self-administration is predicated on the assumption that responding is driven by alcohol’s reinforcing effects, the aforementioned issues may pose a threat to translation (Vena et al., 2020).

Pre-clinical experimental design may also impact predicted treatment efficacy. Many studies rely on acute models of substance use, where drug is available under limited access conditions. In contrast, chronic models of use with extended drug access may more accurately mirror human substance use patterns (Panlilio and Goldberg, 2007). Short catheter lifespans often prevent chronic drug self-administration which can produce tension in achieving concordance between preclinical and human laboratory contexts. These design considerations should be considered when translating preclinical findings to the laboratory and clinical setting.

3. Subjective Effects

Subjective effect measures are considered a complement to drug self-administration procedures that provide insight into the interoceptive effects of a drug as well as other internal states that may be altered by drug administration. Other methods may evaluate subjective states related to the absence of drug use (e.g., in states of withdrawal) that may contribute to the likelihood of relapse and return to substance use. Subjective effects have a long history in the evaluation of the abuse liability of drugs as well as general cultural depictions of drugs and the human experience (see historical review in NIDA Research Monogram Jaffe and Jaffe, 1989)

3.1. Common Clinical Models Using Subjective Reporting

3.1.1. Subjective Drug Effects

Subjective evaluation of drug effects remains one of the primary methods for evaluating drug reward and abuse liability. This history can be traced back to at least the use of subjective drug reporting at the Addiction Research Center (ARC) also known as the Lexington Narcotic Farm (the progenitor of the NIDA intramural research program in the 1970s; Campbell et al., 2021). Research at the ARC contributed to the widely used Addiction Research Center Inventory (ARCI) which contains several scales thought sensitive to specific drugs and drug classes (Haertzen, 1966; Haertzen et al., 1963). Today numerous subjective scales exist for measuring subjective drug experience with items such as “How much do you feel Good Effects from the drug right now” or “How much do you feel Anxious from the drug right now”. Measurement can occur on a variety of scales such as visual analog scales (e.g., place a line on a 100 mm line), Likert items, or dichotomous choice (e.g., yes/no, true/false). Items may index effects directly attributed to drug administration or may evaluate more general changes in mood (e.g., how much did the drug make you feel anxious versus how anxious do you feel at this moment). Subjective measures are typically taken prior to drug dosing and at periodic intervals following dosing (e.g., hourly post drug administration). Responses can be analyzed for the timecourse of effects or summarized into a single value as peak effect or area under the curve.

3.1.2. Drug Craving

Another common subjective experience queried in human laboratory settings is drug craving. Patients commonly report drug craving as a significant barrier to long-term recovery and predictor of relapse during a treatment attempt (Panlilio et al., 2019). Importantly, craving can also serve as detriment to quality of life even when a return to substance use does not occur and is one of the core diagnostic features that defines DSM-5 substance use disorder. Drug craving is commonly defined as an intrusive and overwhelming drive to use drugs (Kakko et al., 2019). Research has also emphasized, however, that craving is a heterogenous concept that likely consists of multiple dimensions (e.g., anticipation of negative reinforcement, intrusive thoughts, and desire to use) (Bergeria et al., 2020; Sayette et al., 2000). This has likely contributed to the heterogeneity of drug craving measures available and varied approaches to studying drug craving in human laboratory and clinical settings (Goodyear and Haass-Koffler, 2020; Kleykamp et al., 2019).

Craving measures can be broadly categorized by their measurement of molar levels of craving aggregated over certain periods (i.e., tonic craving) versus momentary craving in response to acute changes (i.e., phasic craving). Tonic craving levels are typically measured using questionnaires indexing certain aspects of craving (e.g., how much have you desired cocaine over the past week) or using the direct query of craving itself (e.g., how much have you craved cocaine in the past week). Phasic craving is often used in tandem with experimental manipulations such as cue presentation (i.e., cue-induced craving). Cue-induction procedures can use visual (e.g., substance images) or tactile (e.g., touching drug paraphernalia or product without using) exposure to elicit craving. Phasic craving can then be measured via physiological markers like heart rate or brain response and/or patient report (e.g., Ekhtiari et al., 2016; Ray and Roche, 2018).

3.2. Interventions Development and Subjective Effects

Subjective effects are extensively used in interventions development research due to the ease of their implementation and their perceived ease of interpretation (Fischman and Foltin, 1991). Subjective effects of a drug are most closely tied to the interoceptive state that a participant may experience when using that substance, which has implications for its abuse liability as well as interventions to reduce use. Craving, in contrast, may be independent of substance use, but remains a significant barrier to treatment and quality of life in recovery as noted above.

One of the key areas in which subjective effects are applied in human laboratory contexts is in human abuse potential (HAP) studies as a part of the Food and Drug Administration (FDA) review of novel medications. Existing FDA guidance on the assessment of the abuse potential of drugs specifies that human HAP studies should test novel medications when abuse-related signals are observed in preclinical studies and/or abuse-related adverse events observed in clinical trials (Food and Drug Administration, 2017). These studies evaluate a variety of outcome measures including the subjective effects of drugs as a primary endpoint of abuse liability. The FDA recommends that subjective effect scales include key metrics like “Drug Liking” as a primary outcome collected throughout the drug timecourse as well as secondary measures such as “Take Drug Again” or “Drug Similarity” as end of day metrics. Secondary measures may also include effects that are specific to the tested drug class (e.g., “stoned” for cannabinoids or “stimulated” for stimulants). Data from a new medication or drug formulation are then compared to drugs from similar drug class as a positive control and comparator to determine potential abuse liability (e.g., a novel opioid medication might be compared to the opioid agonist hydromorphone; a novel sleep medication might be compared to the sedative z-drug zolpidem). This practical use of subjective effects in FDA guidance combined with the ability to implement in tandem with more complex procedures makes subjective effect data a common feature in laboratory studies of novel interventions and medications for substance use.

3.3. Validity and Considerations for Subjective Effects Paradigms

3.3.1. Interoception and Objectivity

A common criticism of subjective effect data is the inability to “objectively” verify the internal state experienced. This critique is attributable to at least two historical features of addiction science – the skepticism towards patient insight or honesty in substance use disorder and the historical roots of behavioral pharmacology in behaviorism. Debate over the validity of patient self-report of substance use has existed for decades (e.g., Brown et al., 1992; Darke, 1998; Del Boca and Noll, 2000; Midanik, 1988; Zanis et al., 1994). At the same time, behavioral pharmacological methods are borne out of research from the experimental analysis of behavior and behaviorism that emphasized observed behavioral processes and rejection of their contemporaries’ use of internal states and cognitive processes. While resolution of these concerns is beyond the scope of this review, an important extension of this is debate is questions concerning the extent to which patients with substance use disorder may accurately assess internal states in the context of human laboratory studies that remains today. Several studies have documented potential limitations in interoception that occur among people with substance use disorder (Paulus and Stewart, 2014; Wiśniewski et al., 2021), which may impact the ability to accurately attend to the internal states elicited by drug administration. Nonetheless, numerous studies have shown the sensitivity of subjective drug effects to manipulations such as dose and route of administration supporting the accuracy of people who use drugs to attend to these experimental changes.

An important additional consideration is that while behavioral measures may provide objective measurement of clinical phenomenon, patient report should remain an important and prioritized clinical outcome. The emergence of biomarkers in psychiatric health has shifted emphasis toward development of diagnostic markers for substance use disorder. These measures may enhance delineation of risk and avoid biases that can occur in self-report or patient-rated symptom measurement. However, patient experience and quality of life remains a critically significant outcome, and patients may report clinical distress even in the context of a discrepant objective behavioral measure signifying resolution of symptoms. This clinical relevance emphasizes the continued importance of measuring subjective state while also taking steps to improve fidelity of measurement in the human laboratory.

3.3.1. Association with Drug Intake and Use

Another consideration of subjective measures is the non-isomorphic association of these measures with those of drug intake. Consistent with the goals of this review, data has demonstrated that the subjective effects of a drug and drug craving are not necessarily directly related to or predictive of drug self-administration. For example, several promising pharmacotherapies have been shown to decrease subjective response to cocaine or cocaine craving without changing cocaine self-administration in the human laboratory (see review in Haney, 2009). In broader clinical settings, however, a recent meta-analysis found that measures of drug craving were associated with a moderate effect sized increased odds of drug use and/or relapse, emphasizing a positive association that was not deterministic (Vafaie and Kober, 2022). This should be, in some ways, not surprising given that subjective effects often quantify internal states generated by acute drug intake while the decision to use a drug under a specific context can be altered by a myriad of factors including alternatives available, the environment, and aspects of drug cost and benefit that are not related to direct pharmacological effects (see discussion of incorporating this multifaceted nature of drug intake in Section 6 Clinical Gaps and Future Directions).

3.4. Pre-Clinical Subjective Effects Analogues: Translational Implications

Caution should be taken when considering translation of data from human laboratory subjective effects and craving paradigms to preclinical models. This difficulty relates to the features of language that are explicit in the measurement of subjective effects in human participants and the difficulty in recapitulating these features in animal models (de Wit et al., 2018; Field and Kersbergen, 2020). For instance, numerous studies have evaluated “incubation of craving” as an animal model of craving (Venniro et al., 2021). These procedures rely on the observation that subjects trained on an operant response (e.g., lever pressing for opioid self-administration) will show a time-dependent increase in non-contingent responding after periods of deprivation (i.e., greater non-reinforced lever pressing after 14 days than 1 day of restricted access to the operant context) (Grimm et al., 2001). While these data have been extensively interpreted as incubation of craving, data from clinical populations has failed to show a similar time-dependent increases in cue-induced opioid craving (see review in Bergeria et al., 2024). Whether similar difficulties in translation are observed for other drug classes such as psychomotor stimulants have not been systematically reviewed. Yet, these data emphasize the need for consideration of the clinical relevance of preclinical models particularly when considering interoceptive states like subjective responding.

Instead, practical advances in the evaluation of measured internal states in human participants within animal models will likely come from refinement of efforts to evaluate the behavioral mechanisms that underly these subjective processes. As an example, recent work has documented how craving for drug and non-drug rewards can be described by computational models of multiplicative gain that generalizes based on similarity to a cue (Biernacki et al., 2022; Konova et al., 2018). Preclinical models are well equipped to explore these kinds of investigations that are not reliant on interrogation of internal state and may advance our understanding of computational models of subjective states.

4. Behavioral Economics

Behavioral economics merges theoretical and analytical approaches from microeconomics and operant behavior to describe drug consumption patterns and specific features of drug use motivation (Bickel et al., 2014; Hursh, 1991; Hursh and Silberberg, 2008; MacKillop, 2016; Strickland and Lacy, 2020). Theoretical models in behavioral economics, such as reinforcer pathology, are focused on how interactions between drug-related, within-person, and environmental factors function as determinants of behavior. For example, the recently described contextualized reinforcer pathology approach posits that disproportionate behavioral allocation to substance use reward in substance use disorder may be understood as relative constraint on access to alternative non-drug activities and rewards resulting in an overvaluation of smaller immediate rewards and higher reinforcement for drug (Acuff et al., 2023b). These conceptual features described by behavioral economic models can be empirically measured in the laboratory using delay discounting, behavioral economic demand, and non-drug reinforcement procedures, respectively. Behavioral economic methods share some features with drug self-administration ones insofar as they can involve drug consumption under varying schedules of reinforcement or differing environmental constraints. However, while drug self-administration considers drug intake as the outcome of interest, behavioral economic assessments consider sensitivity to the constraint of interest as the primary endpoint (e.g., sensitivity to delay, sensitivity to cost). Effective therapeutics may ultimately result in similar decreases in drug intake, but do so through differing behavioral mechanisms of constraint. Behavioral economic procedures allow for the identification of the specific mechanisms underlying change thereby affording prediction of where and when therapeutics may be most effective.

4.1. Common Clinical Models Using Behavioral Economics

Although behavioral economic procedures broadly measure the effect of a particular constraint on reward valuation and decision-making, we focus here on the most common methods including delay discounting, probability discounting, behavioral economic demand, and non-drug reward surveys (see additional review in Koffarnus and Kaplan, 2018). These methods have historically or presently been applied to understand reward in the context of substance use disorder either for the drug itself or for behaviors relevant to drug use (e.g., drug use alternatives, sexual health decision-making).

4.1.1. Delay and Probability Discounting

Discounting procedures broadly index the sensitivity of reward to the delay (i.e., delay discounting) or probability of receipt (i.e., probability discounting) (Ainslie, 1975; Chung and Herrnstein, 1967; Rachlin and Green, 1972). Both human and non-human animals show a highly replicable and consistent devaluation of rewards when receipt is constrained by delay or probability. This devaluation can be measured and mathematically quantified using discounting procedures. From a theoretical perspective, discounting is thought to represent a transdiagnostic process relating to negative or maladaptive health behaviors including substance use (Bickel et al., 2019; Bickel et al., 2012). For example, cigarette use involves the selection of the immediate reinforcement provided by cigarettes via nicotine delivery instead of the long-term benefits of health and avoidance of complications like lung cancer that come with cigarette abstinence.

Numerous discounting procedures are used in the human laboratory, and most require participants to answer a series of choices between a smaller commodity available immediately (e.g., $50 now) versus a larger amount available after a delay (e.g., $100 in 3 months). In the case of probability discounting, this choice is made between a smaller amount available with 100% likelihood versus a larger amount available with some uncertainty. Choices can then be used to generate atheoretical (e.g., area under the curve [AUC]) or model-guided (e.g., hyperbolic model) outcomes quantifying the sensitivity of value to the constraint of interest (Gilroy et al., 2017; Myerson et al., 2001; Rachlin, 2006). Model outputs can subsequently be used to evaluate effects of a specific manipulation, characterize populations based on clinical variables of interest, or as a predictor of some future behavior.

Discounting methods offer an opportunity to understand how devaluation may differ between- or within-person across varied contexts and commodities. Models of discounting recognize both a “trait”-like aspect to discounting representing a combination of biological predisposition and molar learning history and a “state”-like aspect that is sensitive to environmental and contextual factors (Odum et al., 2020; Odum, 2011). Meta-analyses of discounting have shown that delay discounting tends to be greater in populations who smoke cigarettes or use illicit drugs compared to non-clinical populations supporting a trait-like distinction of risk (Amlung et al., 2017; MacKillop et al., 2011; however, see Strickland et al., 2021a for smaller effect size differences for cannabis use). On the other hand, Rung and Madden (2018) provided a meta-analysis demonstrating that discounting rates are also sensitive to aspects of the decision-making context such as choice framing and priming emphasizing a parallel state-like quality based on context. State-like fluctuations are also documented in data showing increased discounting under conditions of environmental deprivation or other scarcity contexts (e.g., Bickel et al., 2016; Craft et al., 2022; Stein et al., 2021).

Although discounting procedures may be used to measure any commodity of interest, the vast majority of research has focused on monetary rewards. This focus is justified by the idea that money represents a universal commodity for which all participants would have significant behavioral history. Recent research has shown, however, that incorporation of non-monetary outcomes may improve the utility of discounting procedures (e.g., Johnson and Bruner, 2012; Johnson et al., 2017; Johnson et al., 2016; Rasmussen et al., 2010; Strickland et al., 2017, 2019c; Tsukayama and Duckworth, 2010). For instance, the sexual discounting task is a variation of delay discounting procedures in which participants are asked their likelihood of engaging in unprotected sex immediately or delaying sex until a condom is available (see reviews in Gebru et al., 2022; Johnson et al., 2021). This procedure has ecological and clinical validity for sexual health decision-making and the transmission of sexually transmitted infections (for which condoms act as a primary method of prevention). One study using these methods also demonstrated that while acute administration of alcohol increased delay discounting of condom use, it did not alter delay discounting for monetary outcomes supporting a specific effect on this non-monetary outcome (Johnson et al., 2016). These examples demonstrate the clinical utility that may occur when non-drug rewards are evaluated in discounting studies.

4.1.2. Behavioral Economic Demand

Behavioral economic demand (referred to as demand in the remainder of this review for simplicity) quantifies the sensitivity of consumption of a commodity to its cost (Hursh, 1984, 1991; Hursh and Silberberg, 2008). Demand procedures typically generate demand curves in which the consumption of a specific good (e.g., cocaine) is measured at a series of escalating prices. Conceptually, demand is thought to provide a means for measuring the relative overvaluation of drug rewards that occurs with the development of a substance use disorder as well as map on to diagnostic features like insensitivity to consequences by indexing broader sensitivity to cost (see discussion in Acuff and Murphy, 2021)

Demand curve analyses isolate unique features of a demand curve that measure distinct aspects of reward and reinforcement (e.g., Hursh and Silberberg, 2008; Johnson and Bickel, 2006; Koffarnus et al., 2015). These indices include consumption of the good at free or unconstrained cost (i.e., demand intensity or “hedonic setpoint”) and the relative sensitivity of consumption to changes in unit price (i.e., demand elasticity). Other observed aspects of a demand curve can also be directly observed from the demand curve including the price at which maximum expenditure/effort occurs (Pmax), maximum expenditure/effort (Omax), and the price at which consumption is suppressed to zero (breakpoint).

Traditional human laboratory measures of demand involved procedures adapted or directly related to common self-administration methods such as operant responses (e.g., lever pulls) under varying schedules of reinforcement (e.g., FR10, FR100, FR1000) (Johnson and Bickel, 2006). Responses on these procedures can be used to generate demand curves and relevant indices as described above. Although providing the benefit of directly observed behavior and consequences, these procedures are also limited by the reliance on specialized equipment, need for a large number of sessions to generate full demand curves, and some specific confounds that may impact the values generated (e.g., time constraints).

These limitations have led to the development of the commodity purchase task as a complementary procedure for evaluating demand in human laboratory and clinical contexts (Jacobs and Bickel, 1999; Murphy and MacKillop, 2006). The purchase task procedure involves presentation of participants with a hypothetical vignette involving various verbal behavior manipulations to control the response environment. Participants respond with the number of a specified commodity unit they would purchase across varied costs allowing for similar generation of demand curves as in the lengthy effortful responding procedures. Benefits of the purchase task include rapid data collection (typically less than 5 minutes) and ability to measure across varied contexts and populations. Meta-analyses also demonstrate the construct validity of the procedure by showing association between demand indices from purchase task assessments and measures of substance use quantity-frequency and severity (Gonzalez-Roz et al., 2019; Strickland et al., 2020).

4.1.3. Non-Drug Reward Assessments

Recent research within behavioral economics has emphasized the importance of measuring non-drug reinforcers when considering substance use reward. These data are based on extensive research demonstrating that substance use is inversely associated with engagement with and the availability of non-drug alternatives (e.g., Acuff et al., 2019; Herrnstein, 1974; Murphy et al., 2006; Nader and Woolverton, 1992; Rachlin, 1997). Self-administration procedures have also considered the role of non-drug alternatives in changing drug consumption (e.g., see Drug versus Money Choice section above). However, these drug versus money choice procedures typically emphasize changes in drug reward that are dependent on changes in non-drug reward rather than changes in non-drug reward that may occur independently of those drug-related changes (but see discussion of a more holistic approach within the drug self-administration literature in Banks and Negus, 2017). Non-drug reward assessments in the human laboratory provide an opportunity for understanding mechanistic changes in non-drug reward that can occur independent of their association with drug rewards.

Non-drug reward from this view has most commonly been evaluated by measuring the frequency and/or enjoyment of activities while abstaining from drugs and while using drugs (Acuff et al., 2019; Correia et al., 1998; Murphy et al., 2005). These methods are typically used to calculate indices of both non-drug reward alone as well as relative to total. Other approaches have indexed the amount of budgetary expenditure towards non-drug versus drug activities as a relative index of substance-free reinforcement (Tucker et al., 2016; Tucker et al., 2008; Tucker et al., 2009). These methods have collectively demonstrated that lower non-drug reinforcement is associated with more harmful substance use and greater difficulty in sustaining alcohol reduction following a reduction attempt (e.g., Meshesha et al., 2017; Tucker et al., 2016; Tucker et al., 2008; Tucker et al., 2009).

These measures provide a more molar account of non-drug reward which sums over an aggregate period of time (e.g., the past month, the past 90 days). While such approaches are useful for indexing population-based differences in non-drug reward or differences following extensive periods of intervention (e.g., at a clinical trial endpoint), they may not capture acute changes or those occurring after experimental manipulations in laboratory contexts. One recent variation that may provide a step towards addressing these concerns uses brief measures that evaluate access and enjoyability of non-drug activities (Acuff et al., 2024). An initial measure development study found that single item measures of non-drug reward access and enjoyability were closely associated with the likelihood of returning to substance use in the month following treatment (e.g., 58.3% versus 12.4% return to use for those with the least and most non-drug reinforcer availability, respectively). Additional follow up in laboratory contexts and refinement is needed, nonetheless, these brief approaches may prove more flexible for dynamic assessment given their use of general response rather than those that rely on engagement with specific types of non-drug reward.

4.2. Interventions Development and Behavioral Economics

As noted above, the strength of behavioral economic manipulations lies in the ability to isolate the specific behavioral mechanisms underlying change in drug (or non-drug) reward. Meta-analyses of demand and discounting procedures effectively show that these procedures can be sensitive to behavioral and pharmacological manipulations (Acuff et al., 2020; Rung and Madden, 2018) emphasizing utility in intervention and medications development. For example, varenicline reduced cigarette demand relative to placebo in a 5-week outpatient study involving a 4-week quit attempt consistent with varenicline’s clinical effects on cigarette reward and abstinence (McClure et al., 2013).

One method that has been introduced to help improve the fidelity of behavioral economic procedures, broadly, and demand procedures, specifically, is the use of blinded purchase tasks (Berry et al., 2023; MacKillop et al., 2019; Strickland et al., 2023). These procedures involve querying based on a double-blindly administered drug or other commodity experienced during an experimental session, which can help to isolate pharmacological effects from responding based on drug expectations. One study found that maintenance on the dual orexin receptor antagonist suvorexant produced dose-related increased demand for blinded doses of cocaine (Strickland et al., 2023), which was consistent with drug versus money choice data from that study suggesting homology in the conclusions from these assays (Stoops et al., 2022).

4.3. Validity and Considerations of Behavioral Economic Models

As noted above, several meta-analyses have demonstrated the construct validity of behavioral economic methods by showing associations with measures of drug use frequency and severity (e.g., Amlung et al., 2017; Gonzalez-Roz et al., 2019; MacKillop et al., 2011; Strickland et al., 2020). Here we focus on two issues particularly relevant to the use of behavioral economic methods in laboratory settings – their temporal reliability/stability and the role of hypothetical report.

4.3.1. Temporal Reliability and Stability

Establishing the temporal reliability and stability of behavioral economic procedures ensures that changes observed in response to acute laboratory challenge are isolated to the manipulation of interest. Temporal reliability typically reflects the stable rank-order or association of measures over time in a population (e.g., people who have high values at timepoint 1 also have high values at timepoint 2). Temporal stability typically reflects the absolute difference in values over time with high stability signified by minimal change in the absolute value of a score. Importantly, values may be reliable (e.g., showing a consistent test-retest correlation in a sample), but not necessarily stable (e.g., showing a consistent increase or decrease across the population) indicating the importance of evaluating both outcomes.

Temporal reliability and stability have been well established for discounting procedures with some variations across procedural manipulations and contexts. A recent meta-analysis of studies evaluating the reliability of delay and probability discounting procedures found evidence for moderate test-retest reliability (r = .67) with greater reliability for delay compared to probability discounting as well as monetary compared to non-monetary discounting (Gelino et al., in press). Stability was also established with small effect magnitude changes in discounting values over time (d = 0.05) that were not statistically significant. Similarly, demand measures have shown good reliability over time. Several studies have established acceptable to good test-retest reliability for demand of alcohol (Acuff and Murphy, 2017; Murphy et al., 2009), tobacco products (Few et al., 2012; Strickland et al., 2021b), cannabis (Aston et al., 2023; Bush et al., 2023), opioids (Strickland et al., 2019c), and non-drug reinforcers (Strickland et al., 2019a). These studies have similarly shown that the reliability and stability of measures is partly dependent on the stability of consumption (Acuff and Murphy, 2017; Strickland et al., 2021b), emphasizing the association between demand measures and measures of naturalistic use.

Assessments of reliability for non-drug reinforcement measures is notably more limited. One study found that the test-retest reliability of reinforcement ratio values was high for general factor values (r = .86) when assessed in college student drinkers over 2-3 day intervals (Hallgren et al., 2016). In contrast, a study conducted in crowdsourced participants found low test-retest reliability of reinforcement ratio measure (r = .29) over an 18-week period (Strickland et al., 2019a). However, stability was high in that study suggesting that the low reliability may be attributable to a restricted range of responses. Test-retest reliability of brief measures have also not been established emphasizing another area for additional assessment needs.

4.3.2. Hypothetical versus Incentivized Behavior

Another relevant consideration in the use of behavioral economic procedures is that many rely on hypothetical report rather than direct drug administration or incentivized behavior. Several studies have demonstrated the correspondence between incentivized and hypothetical behavior for discounting (Johnson and Bickel, 2002; Madden et al., 2003) and demand (Amlung et al., 2012; Amlung and MacKillop, 2015). These studies show that these hypothetical tasks correspond well to incentivized versions when they are measured simultaneously. Other studies show that measures of drug demand in populations with well-established drug repertoires can predict future consumption (Acuff et al., 2023a; Bird et al., 2024; Heckman et al., 2019; Strickland et al., 2019a) suggesting a predictive validity of hypothetical procedures for future, real-world behavior. Similar research has shown that these arrangements can also forecast behavior for behaviors that a participant has more limited experience (Heckman et al., 2019; Strickland et al., 2022). For example, one study found that demand for novel tobacco products (e.g., e-cigarettes) predicted use patterns 15 months later (Heckman et al., 2019).

An important distinction in these procedures is between prediction by hypothetical tasks that is associational in nature versus absolute. The majority of demonstrations of correspondence between hypothetical and incentivized methods show that responding is strongly correlated across procedures, but not the same absolute value. In other words, participants who respond highly on one procedure tend to respond highly on the other, but the absolute value of response is not identical. This distinction suggests that while these procedures are useful for predicting the likely response to manipulations or medications that they may not predict the absolute value or magnitude of decrease. Thus, the strengths of simulated procedures as predictive measures, especially when used in the context of experimental manipulations, likely lies in the ability to forecast the general impact of medications, environmental changes, or policy conditions on substance use reward and the behavioral mechanisms underlying reward.

4.4. Pre-Clinical Behavioral Economic Analogues: Translational Implications

Either direct or indirect homology is observed between preclinical models using behavioral economic procedures and human laboratory models. For example, preclinical models of delay discounting can evaluate either fixed delays across experimental sessions or within-session adjusting delay or amount procedures to determine indifference points and derive delay discounting metrics (Evenden and Ryan, 1996; Mar and Robbins, 2007; Reynolds et al., 2002; Wahab et al., 2018). Similarly, manipulations of drug dose or “cost” via schedule manipulations can be conducted across or within-session to evaluate behavioral economic demand and demand curve analyses (Bentzley et al., 2013; Bentzley et al., 2014; Lacy et al., 2020; Oleson et al., 2011). A benefit of these approaches is that non-human and human laboratory models have shared features not only mechanistically, but also linguistically (i.e., the language used to describe these mechanisms is similar) across models and to broader contexts (Strickland and Lacy, 2020). Demand approaches, for instance, use mechanistic and linguistic terms that align with pre-existing meaning and are conceptually related to public health oriented problems (e.g., markets and demand within them).

Some aspects of human laboratory methods should be considered when comparing for direct translation to preclinical assays. Contemporary methods in human laboratory methods primarily use within-session determinations of discounting or demand outcomes due to their efficiency while preclinical methods often use multiple session determinations to generate these data (e.g., manipulations across sessions of response cost). Some data suggest that these procedural variations may influence the output metrics, for instance, within-session procedures resulting in more elastic nicotine demand than between-session procedures in one study (Powell et al., 2019). The impact for translatability or sensitivity of between- versus within-session determinations for indexing pharmacotherapeutic impacts is not known.

5. Cognitive Mechanisms and Executive Function

Recent targets for substance use intervention include cognitive-behavioral mechanisms and executive control functions that may underlie substance use and substance use disorder. These aspects of reward processing include diverse targets that broadly capture mechanisms that mediate approach and/or avoidance activities or other aspects of cognitive control. These targets are premised on the idea that interventions that compensate for cognitive deficits observed in substance use disorder may improve daily functioning or directly decrease substance use by addressing mechanisms underlying drug reward itself (Nixon and Lewis, 2019; Verdejo-Garcia, 2016). Laboratory tasks designed to evaluate these mechanisms such as attentional bias, working memory, and response inhibition may thus be used to evaluate population-based differences, predict treatment outcomes, or evaluate the efficacy of a novel intervention.

A comprehensive description of all tasks and mechanisms targeted within substance use is beyond the scope of this review. Instead, we below describe some considerations in the implementation of these procedures and their interpretation (note that some of these considerations may also be applied to other measures of reward). We also describe applications in substance use treatment and preclinical translation.

5.1. Interventions Development and Cognitive Mechanisms

Cognitive and executive function mechanisms as a target in substance use disorder has most commonly been evaluated in the context of cognitive training and remediation interventions. These interventions such as working memory training or attentional bias modification are designed to address these purported deficits through training (Anderson et al., 2021; Nardo et al., 2022). Pharmacological interventions may also be tested for their effects on these cognitive processes independent of their effects on the primary reinforcing effects of drugs (e.g., use of modafinil or galantamine as cognitive enhancers Brady et al., 2011). Studies on the impact of these training interventions has generally showed robust improvements for the trained or similar task (e.g., Houben et al., 2011a; Houben et al., 2011b; Snider et al., 2018; Strickland et al., 2019b), but more limited when considering improvements on different cognitive domains (i.e., far-transfer) or substance use (e.g., Rass et al., 2015; Schulte et al., 2018).

Recent efforts have been made to identify effective means to improve interventions. For example, a recent Delphi consensus study emphasized core areas of interest in cognitive remediation that could be targeted and mechanisms for improving current interventions (Verdejo-Garcia et al., 2023). That expert panel reached consensus that implicit biases, positive affect, arousal, executive functions, and social processes were the key cognitive areas for intervention. Complementary interventions including cognitive bias modification, contingency management, emotion regulation training, and cognitive remediation (e.g., goal management training) to improve these areas. This panel also emphasized that cognitive training and remediation should occur as an adjunct to treatment rather than stand-alone intervention. This point is methodologically relevant insofar as laboratory investigations often focus on the isolated effects of cognitive interventions, but their clinical deployment will likely require an integrated approach.

5.2. Validity and Considerations of Cognitive Mechanisms

5.2.1. Reliability Paradox and Between versus Within-Person Variability

Cognitive tasks may be used in a variety of ways including as individual difference predictors of outcomes like treatment response or substance use status (i.e., between-subject differences) as well as to evaluate the effects of a specific manipulation (i.e., within-subject differences). An important consideration here is what Hedge and colleagues (2018) refer to as the reliability paradox. The reliability paradox emphasizes how laboratory behavioral tasks are typically designed to maximize within-person variability at the expense of minimizing between-person variability in order to produce the maximal effect for experimental manipulations. The Stroop task exemplifies this in that task design is maximized to produce the within-person difference between congruent and incongruent trials while minimizing variability between people in the expression of this effect (Dang et al., 2020). A consequence of this minimization of between-person variability is that it can constrain reliability of these measures.

This has implications for the evaluation of cognitive tasks in substance use disorder. Low reliability due to minimized between-person variability can constrain the ability to detect between-person differences on task performance and use these tasks as a measure of individual differences. This may also help to explain the diversity of hierarchical structures used to represent executive and cognitive function that often fail to replicate across studies (Dang et al., 2020). The implication in the context of substance use reward is that these measures are likely best incorporated to understand the effect of manipulations on changes in these processes rather than as prognostic predictors of risk or diagnostic features.

5.2.2. Attribution Errors and Epiphenomenon

Evaluation of cognitive-behavioral mechanisms as a pathway for intervention is based on the idea that substance use disorder is characterized by specific deficits in these factors. Several large scale meta-analyses have evaluated executive function dysfunctions in substance use and identified generalized and specific factors dysregulated in substance use disorder (Lovell et al., 2020; Potvin et al., 2018; Stavro et al., 2013). However, relevant to consider here is the extent to which these deficits are causally related to substance use and if they exist outside of normative ranges (Hart et al., 2012).

Specifically, while differences between controls and populations that use drugs may be observed on some of these cognitive mechanisms, the extent to which these measures depart from normative ranges that would characterize “deficit” is not typically considered. It is also important to note that in many cases the broader context in which substance use occurs (e.g., environments with decreased access to alternative, non-drug rewards) is not controlled for in totality within case-control designs. Therefore, the extent to which cognitive deficits may be attributed to substance use is less clear and some of these mechanisms may be more directly related to other environmental determinants that covary with substance use.

These issues are also related to concerns about epiphenomenonality in the measurement of cognitive deficits in substance use. Epiphenomenal processes are secondary phenomenon that occur in parallel to a primary process that may connotate causation, but in which causal pathways may be limited. In the case of cognitive deficits in substance use, cognitive mechanisms that covary with substance use (e.g., increased attentional bias towards drug-related stimuli) may occur concurrently with the development of substance use disorder, but bare no causal effect on their development. In this case, addressing this cognitive target would have limited impact on the resolution of substance use and related problems.

These issues of attribution and epiphenomenon are important to consider when evaluating drug reward for several reasons. First, as described above, the viability of cognitive targets for substance use is based on a causal pathway to decrease substance use reward. Should mechanism have little causal impact on substance use, the clinical impact of these interventions on direct substance use reward would be limited (albeit may have some impacts on broader quality of life nonetheless). Second, attributions of cognitive deficits that are either absent or overemphasized may have an unintended consequence of furthering stigma regarding substance use. Future research in these domains is needed to clarify whether and the extent to which various cognitive mechanisms represent a causal pathway of clinical significance in substance use disorder to address these issues. Large scale longitudinal projects such as the Adolescent Brain and Cognitive Development (ABCD) study will be essential in this regard.

5.3. Pre-Clinical Cognitive Methods: Translational Implications

Preclinical analogs exist for the many of the cognitive tasks used in the human laboratory. The use of preclinical models in which the behavioral and substance use history of an animal is precisely controlled and understood can help reconcile questions regarding the causal nature of substance use and cognitive changes (Melugin et al., 2021). Similar to consideration with the translation of subjective effects data, an important consideration in the translation of preclinical assays is the differences in non-human animals’ executive function and cognitive processes. Consideration of the complexity of the task design and higher-order process evaluated is needed as are the difficulty in evaluating some aspects that may be dependent on features like language or uniquely human social interactions (de Wit et al., 2018).

One method that has proven helpful for improving translational homology across species is touchscreen-based methods (Galbo-Thomma and Czoty, 2023; Kangas and Bergman, 2017). Touchscreens allow for dynamic variations in task design affording the opportunity to flexibly deploy existing and novel tasks for use in behavioral pharmacology research. Recent research has demonstrated close homology in both performance and response to pharmacological manipulations across preclinical models (e.g., rodents, monkeys) and with human analogs using similar touchscreen methods (e.g., Kangas and Bergman, 2017; Luc and Kangas, 2023; Robble et al., 2021). Further technological advances may similarly improve the homology and measurement of cognitive processes in preclinical models.

6. Clinical Gaps and Future Directions

Above we have discussed traditional and emerging human laboratory methods designed to evaluate drug (and non-drug) reward relevant to substance use and substance use disorder. Translational linkages to preclinical and clinical domains have been emphasized. However, important gaps still exist between human laboratory (and preclinical) methods as currently employed and the clinical context of substances use disorder that can limit clinical translation even when conducted under optimal conditions. Some of these gaps exist because of the noted lack of correspondence between the clinical outcomes desired and the specific models evaluated in the human laboratory and preclinical settings (e.g., see discussion in Negus and Banks, 2021). Laboratory measures of self-administration, for instance, focus on reductions in drug-taking behavior yet the current landscape of FDA-approval requires drug abstinence as an outcome.

One key reason for this overall lack of correspondence is the complexity of drug-taking in real-world contexts and the many factors that can increase or decrease the likelihood of substance use. Diverse contributors to substance use occur in real-world environments including (but not limited to) changes in the social context of use, mood state of the individual, environmental features associated with substance use, and sociocultural and political factors that can enhance or diminish drug reward and consequence. Seldom are these factors incorporated into the human laboratory. When incorporated, these factors are often isolated to a single feature implemented in a rigorous, but simplistic manner. An important direction for future research is to improve translation by developing improved balances between the strong internal validity offered by laboratory procedures with the strong external validity of contexts in which substance use commonly occurs.

One approach may incorporate more complex environmental factors into laboratory studies. As an example, research has begun to incorporate social factors into laboratory studies given the well-documented role that social context plays in drug-taking behavior (see reviews in Creswell, 2021; Strickland & Acuff, 2023; Strickland & Smith, 2014). Emerging human laboratory studies using contemporary methods have documented that pharmacodynamic effects can differ under social and isolated contexts and based on the nature of the social group (e.g., substance use behavior of a peer), recapitulating well-documented findings in the epidemiological literature (e.g., Kirkpatrick & de Wit, 2013, 2015). Extensive of these more complex designs not only to understand basic biobehavioral mechanisms of use, but also to screen candidate medications and interventions for substance use stands to improve translation to the diverse clinical environment.

Another approach to improve clinical translation of laboratory methods is leveraging novel technological advances to move the laboratory into the real-world. Advances in methods like ecological momentary assessment (EMA) which allow for measurement of people in natural environment through dense daily sampling combined with advances in passive biosensor and biometric data collection can allow for many of the same pharmacokinetic and pharmacodynamic assessments employed in traditional laboratory research, but with the improved ecological validity of real-world sampling. Similarly, mobile laboratories that evaluate participants in real-world environments can employ many of the rigorous techniques used in laboratory settings with the regulatory and practical flexibility allowed by naturalistic sampling. For example, federal restrictions on the use of new cannabis products available in the market has historically limited advances in understanding cannabis effects in this emerging market. These challenges have been partly addressed through the use of mobile laboratory assessments that allow for evaluation of these products purchased by participants in real-world marketplaces, thus better modelling and measuring those pharmacological factors relevant to real-world use (e.g., Bidwell et al., 2020).

7. Conclusions

The current review provided an overview of common methods used in the evaluation of reward and related neurocognitive processes in substance use disorder. A particular emphasis was placed on considering the degree to which a single model is unlikely to translate into the totality of substance use disorder given the multifaceted nature of the condition and its etiology. Instead, translational research on substance use disorder should strive to include multiple methods and models – concurrently or simultaneously – to provide the most accurate prediction of clinical settings. Perhaps most importantly, venues for continued discussion among preclinical, human laboratory, and clinical researchers is essential for the refinement of models and development of common language.

Highlights.

• This review overviews popular models of reward used in substance use research.

• Unique issues for measure validity and application are discussed.

• Translational links to preclinical models and homology are considered.

• Directions to improve human laboratory models of reward are provided.