Behavioral Economic Demand as a Unifying Language for Addiction Science: Promoting Collaboration and Integration of Animal and Human Models

Abstract

The intersection of pharmacological, psychological, and economic theory within behavioral economics has helped advance an understanding of substance use disorder. A notable contribution of this approach is the conceptualization of reinforcement from a behavioral economic demand perspective. Demand analyses provide a multidimensional view of reinforcement in which distinct behavioral mechanisms are measured that impact decision-making and drug consumption. This review describes the state of research on behavioral economic demand as a common language for addiction science researchers across varied model systems and stages of a translational continuum. We first provide an overview of the theoretical concepts and procedures used to evaluate demand in animal and human models. The potential for demand to serve as a common language for diverse research groups in psychopharmacology and addiction science (e.g. those evaluating neurobehavioral outcomes, medications development, clinical practice) is then described. An overview is also provided of existing empirical studies that, while small in number, suggest good linguistic and conceptual overlap between animal and human demand models when studying biological, environmental, and pharmacological individual difference vulnerabilities underlying drug-taking behavior. Refinement of methodological procedures and incorporation of more nuanced environmental features should help improve correspondence between animal and human demand studies as well as clinical translation of such findings. Our hope is that this review and commentary ultimately serves as inspiration for new collaborative efforts involving behavioral economic demand between animal and human researchers who share a common goal of improving substance use treatment outcomes and broader psychological well-being.

Introduction

The marriage of pharmacological, psychological, and economic sciences over the past five decades has advanced a theoretical and practical understanding of choice, broadly, and substance use disorder, specifically (Bickel, Yi, Mueller, Jones, & Christensen, 2010; Chung & Herrnstein, 1967; Rachlin, 2006; Rachlin & Green, 1972). This interdisciplinary approach seeks to characterize systematic decision-making mechanisms to inform basic and applied research on substance use. For example, decades of work on delayed reward discounting has demonstrated that the devaluation of future consequences may function as a key behavioral mechanism underlying substance use disorder (Amlung, Vedelago, Acker, Balodis, & MacKillop, 2017; Bickel, Koffarnus, Moody, & Wilson, 2014; MacKillop, 2013) as well as a transdiagnostic variable relevant to other clinical conditions such as borderline personality and bipolar disorders (Amlung, Marsden, et al., 2019; Bickel, Jarmolowicz, Mueller, Koffarnus, & Gatchalian, 2012).

Another popular topic in this tradition is behavioral economic demand (hereafter often referred to as demand for simplicity). Demand may be operationally defined as the consumption of a commodity at a given cost with demand curves characterizing this association across a range of prices. Such quantitative analyses of demand are central to understanding consumer choice in economics. Psychologists have also long appreciated the benefits of applying demand and other microeconomic principles within an experimental analysis of behavior framework. Howard Rachlin, for example, was one of the first to argue that concepts such as substitutability from consumer demand theory could be used to help understand behavioral allocation and decisions made in response to environmental fluctuations in response cost (e.g., Rachlin, Green, Kagel, & Battalio, 1976). The application of similar microeconomic frameworks to psychological theories of choice were later extended to drug self-administration and abuse liability testing (Bickel, DeGrandpre, Higgins, & Hughes, 1990; Bickel, DeGrandpre, Hughes, & Higgins, 1991; Carroll, 1993; Hursh, 1991).

Today, demand analysis has witnessed a renaissance and rapid growth in the addiction sciences. This resurgence can be partially traced to the refinement of easy-to-implement human laboratory tasks (e.g., the purchase task methodology) as well as the development of accessible mathematical models for evaluating drug demand. A growing body of work has also demonstrated the relevance of demand measures for both contemporaneously and prospectively collected substance use outcomes. For example, in non-human animals, less price sensitive cocaine demand predicts greater cocaine self-administration despite aversive consequences, more responding during extinction, and greater cue-induced and cocaine-primed reinstatement (Bentzley, Jhou, & Aston-Jones, 2014). Similarly, self-reported demand for cocaine and other illicit substances has been associated with both quantity-frequency and severity of use measures in human participants (Strickland, Campbell, Lile, & Stoops, 2019).

Several comprehensive reviews have explored these applications of demand to substance use disorder. These reviews have described topics including the basic principles of behavioral economics (such as demand) as they relate to drug-taking behavior and the experimental analysis of behavior (Bickel, Johnson, Koffarnus, MacKillop, & Murphy, 2014; Cassidy & Kurti, 2018), animal procedures and quantitative methods used to examine demand (Bentzley, Fender, & Aston-Jones, 2013; Hursh & Silberberg, 2008), human procedures for measuring demand and decision-making (Aston & Cassidy, 2019; Kaplan et al., 2018; Koffarnus & Kaplan, 2018; Reed et al., 2019), the construct validity of human demand methods for understanding substance use disorder (González-Roz, Jackson, Murphy, Rohsenow, & MacKillop, 2019; Kiselica, Webber, & Bornovalova, 2016; MacKillop, 2016; Strickland, Campbell, et al., 2019), and the relationship between public policy and demand curve outcomes (Hursh & Roma, 2013).

A notable gap in these reviews and in the empirical literature more generally is a synthesis of research conducted in animal subjects and human participants. To date, both animal and human research has effectively leveraged behavioral economic demand to demonstrate important basic and applied information about substance use disorder. However, these research interests have grown in a parallel, but largely independent fashion. This is unfortunate because behavioral economic demand, as we will argue, represents a possible mechanistic approach that is “horizontally translatable” across animal and human model systems in addition to its oft-described benefits of vertical translation from neuropharmacology to policy implementation.

The primary purpose of this review is to describe the state of research on behavioral economic demand as a common language for addiction science researchers across varied model systems and stages of a translational continuum. To this end, we first provide an overview of the theoretical concepts and procedures currently used to evaluate demand in animal and human models. We then describe the possibilities for demand as a translational language for addiction science research. Preliminary examples of findings from animal and human research that demonstrate a conceptual overlap are then described. Finally, we provide recommendations and directions for refining methodology and improving collaborative efforts between animal and human research programs utilizing demand approaches. Our hope is that this review and commentary ultimately serves as inspiration for new collaborative efforts between animal and human researchers who share a common goal of improving substance use outcomes and broader psychological well-being.¹

Overview of Behavioral Economic Demand and Multidimensionality

Behavioral economics is broadly concerned with applying ideas from psychological theory to understand mechanisms contributing to economic decision-making (Kahneman, 2003; Mullainathan & Thaler, 2000). Behavioral economic demand specifically has been understood to help describe individual and environmental factors that may influence the relationship between the price of a commodity and its consumption. These concepts can align with traditional psychopharmacology research when a commodity is considered a reinforcer (e.g., food, drugs) and cost is defined as an operant requirement on a specific schedule of reinforcement (Bickel, Johnson, et al., 2014; Hursh, 1984; Hursh & Roma, 2013; Rachlin et al., 1976). Accordingly, the unit price of a particular drug commodity may be defined using the operant requirement needed to obtain that drug and its dose (i.e., unit price = responses required/dose). Manipulation of either the work necessary or the dose delivered alters the unit price.² Subsequent observation of responses across a range of unit prices can then provide a means for generating demand functions effectively and efficiently.

The analysis of reinforcement from a demand curve perspective is one of the major strengths of a behavioral economic demand approach when applied to drug-taking behavior. Reinforcement evaluated in this way emphasizes that independent behavioral mechanisms contribute to overall rates of reinforcement and allocation of behavior under conditions of constraint (Figure 1). That is, a demand approach is predicated on the idea that reinforcement is a multidimensional construct for which distinct behavioral mechanisms independently alter the shape and position of a demand curve to impact decision-making and drug consumption. Demand curve analyses therefore represent a distinctive view that reinforcing value is not a property that is inherent to a particular substance, but instead defined by individual and environmental factors that influence choice and decision-making.

Figure 1. Prototypic Behavioral Economic Demand Curve.

Plotted are prototypic demand (top) and demand and expenditure/effort (bottom) curves. Also plotted are typical demand curve outcomes, including derived demand intensity and elasticity (top curve) and curve-observed O_max, P_max, and Breakpoint (bottom).

Theoretical and empirical accounts support the notion that demand curves are likely to capture two distinct behavioral mechanisms: 1) demand intensity and 2) demand elasticity (Bidwell, MacKillop, Murphy, Tidey, & Colby, 2012; Hursh & Silberberg, 2008; Mackillop et al., 2009). Demand intensity reflects the consumption of a commodity at a theoretical unit price of zero or when that commodity is free. This value has sometimes been considered to represent a hedonic set point of maximal consumption against which an organism wishes to defend consumption in the face of increasing constraints. Elasticity, in contrast, reflects the sensitivity of consumption to changes in price along the demand curve. Elastic portions of the demand curve are those where consumption is highly sensitive to price, whereas inelastic portions are those where consumption is generally insensitive to price (see Figure 1). Demand curve analyses typically focus on elasticity quantified as a global measure along the demand curve, which is referred to as demand elasticity or a. This a value indexes the rate of change in elasticity as price increases along the demand curve. Higher demand elasticity values indicate more price sensitive consumption (i.e., a greater increase in elasticity along the demand curve reflected by greater reductions in consumption with increases in price). Therefore, higher demand intensity and lower demand elasticity are considered indicative of more problematic patterns of substance use. As illustrated in Figure 1, other measures of demand may also be observed directly from the demand curve including breakpoint (i.e., price point at which consumption drops to zero), O_max (i.e., maximum expenditure), and P_max (i.e., price point at maximum expenditure).³

One of the most popular methods for evaluating demand curve data is by using the exponential demand equation (Hursh & Silberberg, 2008). This equation plots consumption as a nonlinear function of price, demand intensity, and demand elasticity:

$${\log }_{10}Q={\log }_{10}\left(Q_{\mathit{0}}\right)+k\left({\text{e}}^{\left(-\alpha *Q\mathit{0}*C\right)}-1\right)$$

Where Q = consumption at a given price; Q₀ = derived demand intensity (consumption at a hypothetical zero price); k = a constant that denotes the range of consumption values in log units; C = commodity price; and α = derived demand elasticity. More recently, the exponentiated equation has been proposed as an alternative to account for zero consumption values (Koffarnus, Franck, Stein, & Bickel, 2015). Zero values (i.e., prices at which no commodity is purchased, commonly observed at high prices) present quantitative challenges when applying the exponential equation because the exponential model requires a logarithmic transformation of consumption. One common solution is replacing zeros with small, non-zero values (e.g., 0.1, 0.01, or 0.001), but the selection of replacement values can have a considerable impact on derived values and model fit (Koffarnus, Franck, et al., 2015; Strickland, Lile, Rush, & Stoops, 2016). Koffarnus and colleagues (2015) developed the modified exponentiated demand equation in which both sides of exponential model are raised to the power of 10:

$$Q=Q_{\mathit{0}}*{10}^{k\left(\text{e}\left(-\alpha *Q\mathit{0}*C\right)-1\right)}$$

This modification improved fits in both the initial demonstration using cigarette demand data (Koffarnus, Franck, et al., 2015) as well as for alcohol, cigarette, and cocaine demand in an independent follow-up evaluation (Strickland et al., 2016).

Other alternatives for analyzing demand data have been explored such as mixed-effect modeling in traditional nonlinear forms (Yu, Liu, Collins, Vincent, & Epstein, 2014) or using left-censored (Liao et al., 2013), two-part (Zhao et al., 2016), or Bayesian approaches (Ho, Nhu Vo, Chu, Luo, & Le, 2018). These methods are designed to address many of the same problems identified above (e.g., zero consumption values) and have proved beneficial in initial demonstrations. Further research, however, is needed to explore the utility of these models for better or alternatively describing drug demand data. For example, some of these models make specific and diverging assumptions about the nature of zero consumption responding (e.g., that zero consumption is undetected consumption versus desired abstinence) that should be theoretically and empirically explored for their suitability. The lack of independent follow-up assessments similarly makes conclusions about the utility and generalizability of these models unclear.

Procedures for Measuring Demand

Several procedures have been developed to assess demand in animal subjects and human participants. These procedures generally manipulate unit price by changing the dose delivered and/or ratio requirement on an active schedule of reinforcement, as described above. For example, if unit price is functionally defined as lever presses needed to obtain 1.0 mg of cocaine then increasing the dose delivered while maintaining the fixed-ratio (FR) requirement will decrease unit price (i.e., “more bang for the buck”). On the other hand, increasing the FR requirement while holding the dose constant will increase the unit price (i.e., “pay more for the same product”).

Between-Session Methods

Between-session methods for measuring demand involve manipulations of cost across experimental sessions. These manipulations of cost may occur through changes in the FR requirement or dose delivered to systematically study unit price. Such manipulations can occur in standard self-administration chambers for animal subjects research. Similarly, human laboratory research may use effortful responding approaches to index demand through variations in the response cost (analogous to varying lever response requirements in an operant chamber). For example, the number of lever plunges needed to receive a cigarette puff can be evaluated under conditions for which the FR response is increased across sessions as a method for increasing cost (M. W. Johnson & Bickel, 2003; M. W. Johnson, Bickel, & Kirshenbaum, 2004; Shahan, Odum, & Bickel, 2000). Simply put, in these procedures, unit price is manipulated across sessions thereby allowing for evaluation of a full cost-consumption range and subsequent demand curve analyses, over multiple days/sessions.

Between-session procedures are appealing because they allow for evaluation of demand under conditions involving tight experimental control and with more limited possibilities of direct carryover effects. In the case of human participants research, these procedures also facilitate direct experience with the cost-consumption contingency by using actual rather than hypothetical behavior. However, between-session approaches are also liable to extraneous variables that can cause daily fluctuations in behavior (e.g., mood or hormonal changes). Similarly, the resource intensive nature of the demand curve generation via between-session methods means they are more cost and resource prohibitive as well as can result in significant attrition (e.g., participant dropout in humans, loss of catheter patency in animals).

Within-Session Procedures

Within-session procedures also manipulate unit price, but do so by varying that price within a single experimental session. In the animal laboratory, within-session manipulations are typically accomplished using the threshold procedure in which subjects complete successive 10-minute components involving progressively decreasing doses (Bentzley et al., 2013; Oleson, Richardson, & Roberts, 2011). An analog of the threshold procedure in human participants research is the commodity purchase task (Jacobs & Bickel, 1999; Kaplan et al., 2018). The commodity purchase task is a questionnaire in which individuals are asked to report consumption of specific commodities (e.g., cigarettes) across changes in price per unit (e.g., $0.01, $0.10, $1.00 per cigarette). Although some parameters can differ across different task implementations (e.g., price range or time of purchase; see reviews by Kaplan et al., 2018; Reed et al., 2019), most purchase tasks utilize a similar instructional set involving hypothetical consumption that cannot be stockpiled and is purchased under typical income availability. Proliferation of the commodity purchase task from its initial demonstration with heroin and cigarette demand (Jacobs & Bickel, 1999) has resulted in a growing list of substances successfully studied using the procedure, including alcohol, nicotine (e.g., cigarettes, e-cigarettes), cannabis, cocaine, prescription drugs, and synthetic cathinones (e.g., Aston, Metrik, & MacKillop, 2015; Bruner & Johnson, 2014; Cassidy, Tidey, Colby, Long, & Higgins, 2017; P. S. Johnson & Johnson, 2014; MacKillop et al., 2008; J. G. Murphy, MacKillop, Skidmore, & Pederson, 2009; Strickland, Lile, & Stoops, 2019).

Within-session tasks provide several benefits that complement those limitations of between-session approaches. The rapid nature of data collection allows a high-throughput and high-resolution evaluation of changes in demand as a function of individual subject characteristics (e.g., hormonal fluctuations; acute drug administration). The use of hypothetical choice in human participants research also affords the opportunity to work with populations for whom other measures of drug use, such as drug self-administration, are impractical or not ethically feasible. This more widespread application is an important strength considering that treatment-seeking populations or those with medical contraindications are often those for whom interventions development efforts would ideally generalize. This is not to say that within-session tasks are not without limitations. These limitations, in most cases, reflect the within-session nature of the task and are addressed by those strengths of between-session approaches. For instance, the close proximity of measurements in within-session tasks may increase the likelihood of carryover effects that influence self-administration or purchasing behavior, a limitation that may be addressed by switching to a between-session design should these carryover effects prove problematic.

Demand as a “Horizontally Translatable” Concept

As reviewed above, behavioral economic demand has contributed an important theoretical lens through which to understand addiction science and drug-taking behavior. Applications of demand have emphasized the multidimensionality of reinforcement and, in doing so, helped provide a more nuanced view of choice and decision-making as it relates to drug and non-drug reinforcers. A growing diversity of tasks have also allowed for the widespread study of demand in diverse animal and human research contexts.

Such theoretical and methodological advances have allowed for broader utilization of behavioral economic demand and the potential for it to serve as a common language among addiction science researchers. Harnessing demand as a translational method for facilitating integration of animal and human models is conceptually similar to those advances made in the application of neuroeconomics in the broader, decision-making sciences. For instance, Glimcher and colleagues (2011; 2004) have argued that the ideas of expected utility and subjective value generalize across neuroscience, psychology, and economic research. These generalizations allow for a connection between levels of investigation through measurable and quantifiable functions. These functional links can then be empirically tested for describing connections from neural representations of value to behavioral measures of choice to broader market dynamics. This hypothesizing of measurable and falsifiable connections means that neuroeconomics has allowed for a mechanistic because approach to understanding economic decision-making as opposed to a neoclassic economic as if approach based on purely predictive understanding (Glimcher, 2011). For instance, neuroeconomic approaches can and have allowed for hypotheses that behavioral observations of expected utility theory occur because there is an underlying neural representation of utility-a testable and falsifiable question (Glimcher, 2011).

Research at the intersection of behavioral economic demand and addiction science has not advanced to the same extent neuroeconomics has with regard to vertical reductions of behavioral and neural measurements in decision-making (but see examples of this in addiction science from Bedi, Lindquist, & Haney, 2015; Gray et al., 2017; MacKillop et al., 2014). However, what is becoming increasingly clear from existing research is that a demand approach can also facilitate integration along a horizontal continuum spanning different model systems at a particular level of analysis or investigation. In other words, demand as applied in varied animal and human models shows excellent linguistic and conceptual overlap which, when taken together, simplifies cross-species efforts and advances both basic and applied research.

Linguistically, the terminology used for demand research has proved easy to apply across contexts and has occurred in a largely consistent manner. Although this benefit may seem trivial in nature, the impact is notable in practice. Numerous problems in psychological science may be traced to translational problems related to language and the (mis)-interpretation of that language (e.g., Lilienfeld et al., 2015). For example, issues may arise when one invokes a single term for two or more things that are quite different (i.e., the jingle fallacy) or uses different language for two or more things that are quite the same (i.e., the jangle fallacy; Block, 1995; Thorndike, 1904). For instance, the term “impulsivity” as used in the psychological and addiction science likely suffers from the jingle fallacy insomuch as it is often used to define distinct behavioral concepts (e.g., response inhibition, delay discounting) that while being quite different in many respects are labeled using this same language.

Similar complications occur when the language surrounding an idea conceptually departs from its lay or general usage. Behavioral scientists are typically forced to work with language for which the surrounding cultural and linguistic structure places pre-existing meaning for particular terms. This co-existence can prove problematic when the scientific meaning of a psychological principle departs from its meaning in a public venue. The experimental analysis of behavior, for instance, has long struggled with the differences between a reinforcer or punisher as applied in behavior analysis (i.e., outcomes that increase or decrease the probability of a behavior, respectively) compared to as used in English common (i.e., “good” or “bad” outcomes; Deitz & Arrington, 1983; Foxx, 1990, 1996). Demand, in contrast, obviates many of these concerns by using a language structure that largely aligns with pre-existing meaning and is conceptually-related. This similarity then facilitates cross-talk between researchers using different species or contexts as well as dissemination to and interpretation by other non-expert researchers or lay audiences.

Demand has also proved similar across animal and human empirical findings, at least in the preliminary studies conducted without the goal of congruence. Although still limited in number, existing studies conducted with animal and human models on similar neurobehavioral and environmental variables have shown good overlap. These studies have demonstrated that demand can be applied in such a way that theoretical meaning is not lost as one moves from model systems in a translational spectrum. Namely, demand intensity or demand elasticity retain their proposed theoretical mechanism whether studied in a rodent self-administration procedure, a human participant completing an FR protocol in a laboratory study, or a patient completing a commodity purchase task in a clinical trial.

Below we provide a narrative review of existing research that has demonstrated this preliminary conceptual overlap between animal and human demand models. These studies span a range of topics involving individual difference vulnerabilities relevant to substance use disorder including biological, environmental, and pharmacological variables.

Empirical Examples of Cross-Model Correspondence

Sex Differences

Several studies have evaluated sex differences in drug demand. These studies are predated by a large body of work that emphasizes a central role for gonadal hormones in influencing drug use behaviors. Specifically, estrogen signaling is thought to augment several drug use behaviors including subjective drug effects (Evans & Foltin, 2006; Sofuoglu, Dudish-Poulsen, Nelson, Pentel, & Hatsukami, 1999) and drug self-administration (Lynch, 2006), thereby being a mechanism by which women may be more vulnerable to the reinforcing effects of drugs than men (see reviews by Anker & Carroll, 2011; Becker, McClellan, & Reed, 2017)

A recent study in male and female rats evaluated cocaine and remifentanil demand using a threshold procedure (Lacy, Austin, & Strickland, 2020). Freely cycling female subjects showed higher demand intensity for both cocaine and remifentanil during testing sessions in which estrogen was high relative to progesterone (e.g., the estrus phase). Notably, overall sex differences were not observed, emphasizing the importance of these specific deviations related to estrous cycling and changes in gonadal hormone circulation. Similar results have been reported in other preclinical studies for which no overall sex differences were found in cocaine demand intensity or elasticity (Kawa & Robinson, 2019) or nicotine demand intensity or elasticity (Grebenstein, Burroughs, Zhang, & LeSage, 2013; Powell et al., 2019) between male and female rats during pre-manipulation assessments. Important to note differing from these findings is a recent between-sessions study in which greater demand elasticity was observed for fentanyl and a liquid food reinforcer in male compared to female rats (Townsend, Negus, Caine, Thomsen, & Banks, 2019). It is possible that differences between within- and between-session approaches contributed to these divergent findings, although systematic tests are needed (see Future Directions section).

Human laboratory assessments have similarly failed to identify significant overall gender differences in tobacco cigarette demand across varied populations (Green & Ray, 2018; O’Connor et al., 2016; Strickland et al., 2016), cocaine demand in persons with cocaine use disorder (Bruner & Johnson, 2014), or prescription opioid demand in persons reporting non-medical prescription opioid use (Strickland, Lile, et al., 2019). Importantly, there are other studies that have reported more inelastic cocaine demand for men with cocaine use disorder (Strickland et al., 2016) and higher opioid demand intensity for men within a sample of college students (Pickover, Messina, Correia, Garza, & Murphy, 2016). These potentially conflicting findings could be due to differences in the contextual or social factors that differentially influence these varied populations, emphasizing the relevance of recognizing non-biological influencers of sex and gender differences in human settings (Becker et al., 2017). Notably, however, a recent small laboratory study found evidence that women in the follicular phase (i.e., the human estrous phase in which estrogens are high) compared to luteal phase had higher cigarette demand intensity and expenditure (effect sizes d = 0.71 to 0.74; Farris, Abrantes, & Zvolensky, 2019). Although this finding requires replication in a larger sample designed for significance testing, these biological outcomes align with the elevated demand observed under conditions of high levels of circulating gonadal hormones in female animal models.

Taken together, existing studies on sex and gender differences in drug demand have shown some correspondence, particularly for the limited research evaluating the influence of gonadal hormones on drug demand measures. Unfortunately, many of these studies with human participants have failed to evaluate those effects of circulating gonadal hormones as a moderating factor in demand outcomes. Sex differences in human drug demand are also likely informed by a convergence of factors that include such hormonal influences in addition to sociocultural and environmental factors, emphasizing factors that animal models may also seek to model to improve conceptual overlap and animal model design.

Environmental Influences on Drug Demand

A rich body of preclinical work and a growing body of human laboratory work has evaluated ways in which manipulations of the environment may influence demand measures (for reviews of the broader impacts of enviornmental manipualtions in substance use see Neisewander, Peartree, & Pentkowski, 2012; Stairs & Bardo, 2009; Strickland & Smith, 2014). These manipulations include both those that result in more impoverished environmental conditions (e.g., acute or chronic stress exposure) and those that improve environmental circumstance (e.g., environmental enrichment).

The effects of stress exposure on drug demand as well as individual difference vulnerabilities to stress exposure have been studied using varied animal models. Several of these studies have used forced swim stress as a model of stress exposure. One such study evaluated cocaine self-administration using a threshold procedure following four days of repeated forced swim stress (Groblewski, Zietz, Willuhn, Phillips, & Chavkin, 2015). Increases in P_max (a measure that is conceptually and empirically related to elasticity) were found for cocaine demand in Wistar rats, but not Wistar-Kyoto rats, following repeated stress thereby indicating a strain-dependent effect. These effects of stress in the Wistar subjects were also blocked by pre-treatment with the kappa opioid receptor antagonist norbinaltorphimine (norBNI) suggesting a role for kappa opioid signaling in the stress-demand relationship.

Another preclinical study used a similar intermittent swim stress procedure to evaluate effects of stress exposure on heroin demand (Stafford et al., 2019). Elevated heroin demand essential value (an inverse transformation of the demand elasticity coefficient) was associated with both heightened behavioral (i.e., climbing behavior during the forced swim test) and biological (i.e., corticosterone release) responses to stress again emphasizing individual vulnerabilities relevant to the stress response. These effects were also consistent with another study that found that intra-VTA microinjections of corticotropin releasing factor (CRF) resulted in more inelastic cocaine demand in male rats tested using a threshold procedure (Leonard, DeBold, & Miczek, 2017). Interestingly, another study that modeled early-life adversity using a limited bedding and nesting procedure found this stress manipulation resulted in increased cocaine demand intensity, but no change in demand elasticity (Bolton et al., 2018). These findings suggest that different stress procedures that model distinct qualitative and quantitative features of stress may impact distinct stress pathways and produce functionally different effects on demand mechanisms underlying drug-taking behavior (e.g., impacting demand intensity versus elasticity or P_max).

Preclinical research on stress exposure is complemented by other research evaluating protective factors relevant to environmental enrichment. One of these studies evaluated the effects of environmental enrichment on cocaine and methylphenidate demand in male rats using both threshold and between-session procedures (Yates, Bardo, & Beckmann, 2019). Across these varied experimental approaches, increased demand elasticity was observed among subjects from the enriched compared to isolated environments (suggestive of less drug use vulnerability). The negative effects of isolation on demand elasticity were also attenuated when subjects were moved to an enriched environment emphasizing the context-dependency of the enrichment’s protective effects. Interestingly, a similar study involving environmental enrichment effects on nicotine demand found that female rats showed more elastic demand when reared in enriched environments, whereas male rats showed less elastic demand when reared these enriched environments (Powell et al., 2019). These findings suggest a potential sex difference for the impact of environmental enrichment, however, follow-ups involving replication and systematic study of other substances is needed.

Studies in human subjects have also evaluated the impact of stress induction on drug demand, although these studies are limited to effects on alcohol and cigarette demand (Amlung & MacKillop, 2014; Dahne, Murphy, & MacPherson, 2017; Owens, Ray, & MacKillop, 2015). One of these studies evaluated adult heavy drinkers following acute laboratory stress induction using the Trier Social Stress Test (Amlung & MacKillop, 2014). Increases following stress induction were observed for alcohol intensity, breakpoint, and maximum expenditure (O_max) on an alcohol purchase task. These findings were later replicated in another study that used a personalized stress exposure for which increases in breakpoint and maximum expenditure as well as decreases in elasticity were observed compared to a neutral event control (Owens et al., 2015). The relationship between stress induction and alcohol expenditure in that study was also moderated by a genetic polymorphism in the gene for corticotrophin releasing hormone-binding protein (CRH-BP). Specifically, a subset of individuals defined by their genotype at this locus showed a greater increase in alcohol expenditure following stress induction than those with the alternative genotypes. This finding is relevant given the demonstrated role for HPA-axis signaling in animal models of drug demand (Leonard et al., 2017; Stafford et al., 2019) as well as a role for CRH-BP in modulating the relationship between stress and drug intake in other animal models of reinforcement (Wang et al., 2005; Wang, You, Rice, & Wise, 2007). Although no studies have evaluated specific manipulations of enriching the environment among human participants (a practically and ethically challenging manipulation to conduct), existing research suggests that socio-economic status (SES) shows some negative correlation with demand measures (J. G. Murphy, MacKillop, Tidey, Brazil, & Colby, 2011; Strickland, Lile, et al., 2019). For example, individuals with higher SES have reported lower demand intensity and higher demand elasticity for cigarettes (J. G. Murphy et al., 2011) and lower demand intensity for non-medical prescription opioids (Strickland, Lile, et al., 2019).

Existing research on environmental features influencing drug demand have emphasized the negative impact of stress and protective effects of environmental enrichment (or higher SES in the case of human participants). Despite the fact that no studies have been conducted in direct collaboration, findings from the animal and human laboratory have each demonstrated a role for HPA-axis signaling for influencing drug demand and drug intake. Environmental enrichment studies have similarly demonstrated a protective effect whether directly manipulated in the environment for animal subjects or through observational analyses of human participants.

Tobacco Cessation Pharmacotherapies

A considerable body of work has evaluated putative and approved pharmacotherapies for substance use disorder by applying demand methods in animal and human models (Bentzley & Aston-Jones, 2017; Bentzley et al., 2014; Bujarski, MacKillop, & Ray, 2012; Stoops et al., 2019; Stoops et al., 2016). Despite this diversity of studies, the only compounds that have been evaluated in both animal subjects and human participants for the same indication are FDA-approved pharmacotherapies for tobacco cessation: the monoamine transport inhibitor bupropion and the nicotinic partial agonist varenicline. The only preclinical study to evaluate nicotine demand in this area tested the predictive validity of nicotine demand for pharmacotherapy outcomes in male rats (Kazan & Charntikov, 2019). Demand curves for nicotine were first generated using a between-session approach and the dose-response effects of bupropion and varenicline on progression ratio (PR) nicotine self-administration tested. Later acute treatment with these pharmacotherapies produced more robust decreases in PR responding for subjects that were higher in demand essential value for nicotine at the baseline assessment, suggesting an individual difference variable that may contribute to therapeutic efficacy. Sucrose demand essential value was also not predictive of nicotine demand essential value suggesting a commodity specificity of these demand assessments.

The majority of research related to tobacco pharmacotherapies has been conducted in the human laboratory and clinic. One study investigated the effects of bupropion treatment on cigarette demand one-week prior to initiation of a smoking cessation attempt (Madden & Kalman, 2010). Bupropion failed to significantly alter demand measures for cigarettes at that one-week time point, however changes in demand elasticity at one week were predictive of cigarette abstinence at treatment follow-up (10 weeks later). Other studies have examined the effects of varenicline on cigarette demand with mixed findings (Green & Ray, 2018; McClure, Vandrey, Johnson, & Stitzer, 2013; C. M. Murphy et al., 2017; Schlienz, Hawk, Tiffany, O’Connor, & Mahoney, 2014). In the first of these studies, one week of varenicline exposure increased demand elasticity relative to placebo (McClure et al., 2013). Another study observed significant reductions in demand intensity following one week of varenicline or nicotine replacement patch medication use prior to a quit attempt date (C. M. Murphy et al., 2017). Notably, the magnitudes of reduction in demand intensity in that study were again predictive of length of abstinence at 1 and 3 months. Significant reductions in O_max, but no changes in other demand outcomes, were observed in a third study in which participants received 10-days of varenicline treatment (Green & Ray, 2018). A final study found no differences between varenicline and placebo treatment with respect to changes in cigarette demand measures, with both groups showing similar magnitude increases in demand elasticity and decreases in demand intensity over a four-week trial (Schlienz et al., 2014). These discrepant outcomes could be due to differences in analytic strategies, attrition rates, and study setting with further tests directly manipulating these parameters in larger samples needed to test these possibilities. Taken together, however, existing human evidence of the predictive capability of cigarette demand in these studies does align with the one study evaluated these pharmacotherapies in animal models.

Future Directions for Animal, Human, and Translational Demand Research

The research reviewed above documents examples of studies in which biological, environmental, and pharmacological factors have shown some correspondence between studies in animal and human models despite the lack of direct collaborative efforts. Although this work is promising, the literature remains small with respect to instances in which such overlap can be evaluated. The remainder of this review focuses on areas that we believe future research could benefit for further refinement and utilization of demand approaches for understanding substance use and addressing substance use disorder.

Direct Animal-to-Human Collaborations

Our primary recommendation is the need for more direct animal-to-human translational efforts. The empirical examples provided in this review are promising, but rely upon suggestive and coincidental overlap between studies conducted largely independently of one another. This is not surprising given the intense collaborative effort required for bridging translational projects across species. For instance, one review found that among over 100 randomized clinical trials for cocaine medications development involving over 60 different drugs, only 10 compounds had also been examined in both controlled human and animal laboratory studies (Czoty, Stoops, & Rush, 2016). Thus, efforts are needed to promote collaborative projects that further integrate animal and human experiments, with demand as a “common language” to harmonize the interpretation of these efforts.

A review of existing psychopharmacology literature provides some examples of how such integration across animal and human research programs may occur. For instance, successful efforts were recently described for cocaine medications screening and the integration of choice procedures in animal and human laboratories (A. R. Johnson et al., 2016; Lile et al., 2019; Lile et al., 2016). These studies involved the testing of a translational model using homologous cocaine self-administration procedures in non-human primates and human participants. The development and refinement of a drug versus non-drug choice procedure showed similar effects of dose magnitude and alternative reinforcers on drug self-administration (A. R. Johnson et al., 2016; Lile et al., 2016) as well as a consistent impact of treatment with the clinically effective pharmacotherapy d-amphetamine (Lile et al., 2019). These findings provide a good “proof-of-concept” demonstrating how animal and human procedures may methodologically and empirically align. Similar efforts to match demand methods for animals and humans should prove equally useful with the added benefit of providing a theoretical perspective that incorporates the multidimensional view of reinforcement that demand approaches provide.

Method Validation and Refinement

In this regard, method refinement and critical assessment of existing techniques is also needed. Although it is tempting to move research efforts immediately to clinical applications, there remain outstanding questions about the reliability and validity of some commonly used demand techniques in addition to the homology across animal and human procedures. The threshold procedure is widely used to test within-session demand curves, however, parametric evaluations of features including dose randomization or correspondence with between-session methods are generally lacking. Powell and colleagues (2019) recently found, for example, that within-session procedures resulted in more elastic nicotine demand than between-session procedures in male and female rats. Interestingly, these results are similar to those observed with the commodity purchase task data for which more elastic alcohol demand is seen under vignette conditions involving shorter (i.e., 1 hour) compared to longer (i.e., 9 hours) access times (Kaplan et al., 2017). Tests of specific price ordering (i.e., ascending, descending, or randomized) in human participants have found few differences, at least in the case of alcohol purchase task data (Amlung & Mackillop, 2012). Similar tests have not, to our knowledge, been conducted with animal subjects making the influence of using a descending dose order in the threshold procedure unclear. Such studies evaluating the clinical impact of differences in derived demand parameters among within- and between-session methods as well as under varying price/cost sequences are needed.

Similarly, few studies have evaluated the relationship between commodity purchase task data evaluated using hypothetical and incentivized outcomes (Amlung, Acker, Stojek, Murphy, & MacKillop, 2012; Amlung & MacKillop, 2015; Wilson, Franck, Koffarnus, & Bickel, 2016) and none, to our knowledge, have evaluated differences between purchase task and effortful responding procedures in humans. In one study, alcohol demand using hypothetical and incentivized tasks were highly correlated at both individual price points and for overall demand metrics (Amlung et al., 2012). These relationships were later replicated in a similar design conducted with an independent sample of heavy drinkers (Amlung & MacKillop, 2015). Wilson and colleagues (2016), however, reported lower elasticity values when measuring cigarette demand using hypothetical purchase tasks compared to incentivized ones. Additional work is needed to replicate these effects in larger and more heterogeneous samples as well as to determine how these findings of the relationship between hypothetical and incentivized demand may vary across commodity type.

The impact of commodity accumulation has also received little attention in the existing literature. We are not aware of systematic research that has evaluated issues related to accumulation of the commodity that most notably would occur during within-session procedures (e.g., the threshold procedure). Human methods typically address commodity accumulation through instructional control for which limits are placed on “stockpiling” (e.g., inform individuals that they must consume all of a given commodity in a given situation). It is possible, however, that the accumulation of a commodity during procedures for which active drug administration occurs may influence decision-making in a systematic or differential manner (see Beckmann, Chow, & Hutsell, 2019 for an example of recent efforts in controlling the influence of within-session drug exposure in drug choice procedures). Tests of these hypotheses are, to date, lacking and may further inform methodological procedures in demand (and in drug administration research more broadly).

Incorporating Cross-Commodity Relationships and Environmental Complexity

Research efforts that move beyond simple schedules of demand assessment are also essential for the future clinical implications of demand methods. Specifically, cross-commodity demand procedures are one way to better incorporate information about alternative reinforcer availability and directly measure behavioral allocation and reinforcer choice that could offer additional translational appeal across species. Cross-commodity demand reflects the responsiveness of quantity demanded for a commodity as a function of the change in price of another commodity. Cross-price elasticity then quantifies this relationship between price-changes with a price-manipulated commodity and demand for an alternative price-fixed commodity. Commodities may act as substitutes meaning that as the price increases for a price-manipulated commodity that consumption increases for its alternative (e.g., Coca-Cola and Pepsi). Commodities may also function as complements meaning that as the price increases for one commodity that consumption decreases for the alternative (e.g., hotdogs and hotdog buns).

Existing research in the human laboratory has begun to leverage cross-commodity purchase task procedures to advance tobacco regulatory science. These studies have found nicotine replacement commodities serve as substitutes for traditional tobacco cigarettes at varying magnitudes of effect (M. W. Johnson, Johnson, Rass, & Pacek, 2017; Snider, Cummings, & Bickel, 2017; Stein, Koffarnus, Stepanov, Hatsukami, & Bickel, 2018). Such findings are consistent with prior human laboratory studies using effortful responding protocols that showed similar cross-price substitutability of nicotine replacement products for tobacco cigarettes (M. W. Johnson & Bickel, 2003; M. W. Johnson et al., 2004; Shahan et al., 2000). A recent study in male rats also indicated consistent outcomes in that e-cigarette liquid containing nicotine substituted for nicotine solution alone (i.e., in saline) (Smethells, Harris, Burroughs, Hursh, & LeSage, 2018). These examples are consistent with a broader observation that fixed price non-drug alternative reinforcers can reduce demand for drug reinforcers with increases in price (Carroll, 1993; but see Comer, Hunt, & Carroll, 1994; Cosgrove & Carroll, 2003; Rodefer & Carroll, 1997).

Other studies have provided an improved understanding of demand by better incorporating the environmental contingencies at play during real-world decision-making. Reinforcer availability, next-day responsibilities, and income availability manipulations have all been shown to influence demand in human participants (Gentile, Librizzi, & Martinetti, 2012; Kaplan, et al., 2017; Koffarnus, Wilson, & Bickel, 2015; J. G. Murphy et al., 2014; Skidmore & Murphy, 2011). Comparable effects have been observed in animal models when providing for analogous environmental constraints (e.g., session length as an analog of income; Carroll, Rodefer, & Rawleigh, 1995). For example, changing from a closed economy (i.e., the experimental setting is the only one where the commodity is available) to an open economy (i.e., the commodity is available outside the experimental context) has generally been shown to result in more elastic demand in both animals and humans (see review by Kearns, 2019; Kim, Gunawan, Tripoli, Silberberg, & Kearns, 2018; Mitchell, de Wit, & Zacny, 1994; Mitchell, De Witt, & Zacny, 1998).

Conclusions

Behavioral economic demand is situated to become a potential unifying language for preclinical, clinical, and community perspectives. Prior research has elegantly illustrated how information from laboratory demand models may be applied to inform policy by predicting expected behavioral decision-making in a real-world setting (Amlung, Reed, et al., 2019; Grace, Kivell, & Laugesen, 2015; Kaplan & Reed, 2018; Salzer, Gelino, & Reed, 2019; see review by Hursh & Roma, 2013). The established benefits of translating demand vertically across levels of analysis is complemented by the flexibility of application across different model systems at these levels of analysis. This recognition of demand’s utility in understanding drug-taking behavior fits well within broader theoretical shifts recognizing and appreciating the role of behavioral models of addiction (Heyman, 2013; Lamb & Ginsburg, 2018). An understanding of these mechanisms underlying the decision to allocate behavior towards drug use over more prosocial alternatives may ultimately lead to improved targets in the prevention and treatment substance use disorder.

Public Significance Statement.

Demand, or the relationship between price and consumption, has helped improve an understanding of reinforcement related to drug-taking behavior. This review describes the potential benefits of utilizing a demand approach for improving communication and correspondence in clinical findings for animal and human researchers in addiction science. Such efforts should help advance clinical translation to address vulnerabilities and treatments for substance use disorder.