Undervaluing non-drug rewards or overvaluing cocaine? Cocaine demand relates to cocaine use severity more strongly than anhedonia in individuals with cocaine use disorder

Abstract

Cocaine use disorder (CUD) is a major public health issue, and greater cocaine use severity has been associated with worse treatment retention and outcomes. Therefore, greater understanding of processes that influence cocaine use is needed. Both anhedonia (i.e., undervaluation of non-drug rewards) and cocaine demand (i.e., cocaine valuation) are related to cocaine use severity and thematically related to each other at face value, but no studies have directly compared these outcomes to our knowledge. The current study represents a secondary analysis from a two-phase sequential, multiple assignment, randomized trial (SMART) aimed at developing adaptive interventions for CUD. We examined the relationship between anhedonia and cocaine demand and how these measures were related to cocaine use severity. Participants (N = 116) were treatment-seeking adults with CUD. All measures were taken at baseline prior to treatment initiation. Analyses revealed 1) “moderate” and “very strong evidence” of relationships between cocaine demand factors (i.e., persistence, amplitude) and anhedonia (PP values ≥77.8%); 2) positive association between cocaine demand (both persistence and amplitude) and measures of cocaine use severity, with the exception of one relationship, which was in the opposite direction; and 3) demand amplitude continued to be positively related to cocaine use severity, even when considering anhedonia. Overall, findings from this study indicate cocaine demand relates to cocaine use severity more strongly than anhedonia.

Introduction

Cocaine use disorder (CUD) is a major public health issue affecting over one million individuals in the United States (Substance Abuse and Mental Health Services Administration, 2021). CUD is associated with substantial health problems (Butler et al., 2017; Schwartz et al., 2010), comorbid psychiatric diagnoses (Narvaez et al., 2014), and poor quality of life (Narvaez et al., 2015). Greater CUD severity, captured by factors such as frequent cocaine use and more cocaine withdrawal symptoms, has further been predictive of poorer treatment retention and outcomes (Poling et al., 2007). Given CUD’s potential adverse effects and lack of FDA approved pharmacotherapies, greater understanding of the factors influencing cocaine reward value and use severity is needed. Two paradigms that provide greater understanding of behavioral factors associated with cocaine valuation and use severity are cocaine demand and anhedonia.

Drug demand is a behavioral economic measure of drug valuation that assesses drug purchase and consumption as a function of increasing cost of drug under conditions of constraint (e.g., limited income, availability of other drugs, window of time to use drug, etc.). Hypothetical drug purchasing tasks are commonly used to assess drug demand in clinical settings where providing access to drugs would be either unethical or impractical (Zvorsky et al., 2019). Meta-analyses have shown drug demand to be positively related to various aspects of drug use severity across numerous drugs of abuse (Strickland et al., 2020; Zvorsky et al., 2019), including cocaine (Bruner & Johnson, 2014; Strickland et al., 2016; Webber et al., 2022; Yoon, de Dios, et al., 2021, 2021; Yoon et al., 2020; Yoon, Suchting, et al., 2021). For example, our previous research revealed that 1) baseline cocaine demand predicted treatment outcomes for CUD (Yoon et al., 2020); and 2) treatment for CUD decreased cocaine demand, with greater changes in demand seen in treatment responders (defined as reaching two weeks of consecutive cocaine abstinence; Yoon, Suchting, de Dios, et al., 2021). These findings support the utility of cocaine demand as a measure of cocaine valuation.

Anhedonia, diminished interest or pleasure in non-drug rewards, is a common feature of CUD and several other psychiatric disorders (Franken et al., 2007; Garfield et al., 2013; Gorwood, 2008). This difficulty with experiencing pleasure from “natural” rewards often makes individuals vulnerable to more potent sources of reward such as drugs (cocaine). Individuals who use cocaine demonstrate significantly higher levels of anhedonia relative to controls, especially during initial abstinence (Garfield et al., 2013; Leventhal et al., 2008, 2010). These elevated levels have been captured by reported measures of state, trait, physical, and social anhedonia (Morie et al., 2014). Higher levels of anhedonia also relate to greater cocaine use severity (Morie et al., 2014; Wardle et al., 2023). Finally, greater self-reported anhedonia at baseline has been associated with worse treatment outcomes in individuals with CUD (i.e., there is a decrease in treatment effectiveness as self-reported anhedonia increases; Crits-Christoph et al., 2018; Wardle et al., 2017).

Previous research suggests drug demand and anhedonia measures are associated with cocaine valuation and use severity; however, cocaine use severity could be characterized by either of these processes (drug demand, anhedonia). Individuals could be overvaluing drug reward (i.e., drug demand), or they could also be undervaluing alternative non-drug rewards (i.e., anhedonia). Either or both of these processes could result in a shifted allocation of value from non-drug to drug rewards (Bickel et al., 2014), but they might require different treatment approaches – e.g., blunting drug reward vs. enhancing natural reinforcers. Yet, no previous studies have investigated the relationship between drug demand and anhedonia to determine if they are related or independent processes, or how they uniquely contribute to cocaine use severity. Thus, in the present study we examined the role of possible undervaluation of non-drug rewards in more depth by testing whether anhedonia is related to the valuation of cocaine in cocaine use disorder. First, we examined the relationship between anhedonia and cocaine demand. We hypothesized that anhedonia and demand would be related, suggesting that undervaluation of “natural” rewards contributes to valuation of cocaine. However, our second hypothesis was that the ultimate effect of anhedonia on cocaine use severity would extend beyond just a contribution to increased demand, such that both anhedonia and demand would be significantly and independently related to cocaine use severity. We reached this hypothesis, because it seemed likely that anhedonia might contribute to cocaine use in ways aside from just increasing the value of cocaine, such as by decreasing engagement in other activities, thus giving more unfilled time that could be claimed by drug use (Bertz et al., 2022).

Methods

Overall Design

The current study represents a secondary analysis from a two-phase sequential, multiple assignment, randomized trial (SMART) aimed at developing adaptive interventions for CUD, which has been described thoroughly elsewhere (NCT02896712; Schmitz et al., 2018). Briefly, in phase 1, participants received contingency management (CM) and were randomly assigned to receive either Acceptance and Commitment Therapy (ACT) or Drug Counseling. In phase 2, treatment was augmented with pharmacotherapy (modafinil or placebo) for participants who did not respond to initial treatment. The data used for the current study were based on the clinical trial’s intake visit and baseline session. Deidentified data from this study will be available from the corresponding author on reasonable request 1 year after all aims of the project are completed. Requestors of data will be asked to complete a data-sharing agreement that provides for (a) a commitment to using the data only for research purposes and not to identify any individual participant; (b) a commitment to securing the data using appropriate computer technology; and (c) a commitment to destroying or returning the data after analyses are completed. The parent trial from which this data was derived was pre-registered at https://clinicaltrials.gov/study/ NCT02896712; however, this paper represents a secondary analysis that was not pre-registered.

Participants

One hundred sixteen treatment-seeking adults (18 – 60 years old) with CUD of at least moderate severity were included in this analysis. Given that this was a secondary analysis, the sample size was determined by the sample size calculations for the parent trial (NCT02896712). Criteria for inclusion were: between 18 and 60 years of age; current CUD of at least moderate severity (≥ 4 symptoms), per the Diagnostic and Statistical Manual of Mental Disorders, 5^th edition (DSM-5; American Psychological Association, 2013); have at least one positive toxicology screen for cocaine metabolite, benzoylecgonine (BE level ≥ 150 ng/mL) during intake; acceptable health on the basis of interview, medical history, and physical exam; if female, agreed to use acceptable birth control method during study; be able to understand and provide written informed consent; and able to provide the names of at least 2 locators. Exclusion criteria were: current DSM-5 diagnosis for substance use disorder other than cocaine, marijuana, or nicotine; DSM-5 axis I psychiatric disorder or neurological disease; significant current suicidal or homicidal ideation, medical conditions contraindicating modafinil pharmacotherapy; taking medications that could adversely interact with modafinil; having conditions of probation or parole; impending incarceration; pregnant or nursing for female patients; and inability to read, write, or speak English. The University of Texas Health Science Center at Houston Committee for the Protection of Human Subjects (IRB) approved this study, and all participants provided written informed consent in accordance with the Declaration of Helsinki.

Measures

Cocaine purchasing task (CocPT).

The cocaine purchasing task is a measure of cocaine demand that was administered at baseline, week 2, and week 5. Participants were asked how many rocks of crack cocaine they would purchase across a range of prices per rock (i.e., 0, 1, 2, 3, 4, 5, 10, 15, 20, 25, 30, 40, 50, 100, 200, 500, 1000, 2000, 5000, and 10,000 dollars). The majority of participants reported primary crack cocaine use and all reported a history of crack cocaine use. Participants were asked to assume the following: their income and savings are what they typically are; the quality of cocaine is the type they normally purchase; there are no other sources of cocaine; it must be consumed within 24 hours; and their craving and desire for cocaine are similar to how they felt when completing the task. This task’s prices and instructions were similar to those presented in previous cocaine demand studies (Bruner & Johnson, 2014; Webber et al., 2022; Yoon, de Dios, et al., 2021; Yoon et al., 2020; Yoon, Suchting, et al., 2021). Compared to previous cocaine purchasing tasks in the literature, prices in the current task were chosen to 1) provide relatively more price points at lower values to potentially capture responding from a treatment-seeking population; and 2) present a high enough price ($10,000) to ensure that consumption would eventually reach 0.

Snaith-Hamilton Pleasure Scale (SHAPS).

The SHAPS (Snaith et al., 1995) is a 14-item self-report questionnaire used to measure hedonic capacity. It has been found to be a reliable and valid questionnaire to assess hedonic tone in substance using populations (Franken et al., 2007). Participants rate statements (e.g., “I would be able to enjoy a beautiful landscape or view”) on a 4-point Likert scale ranging from 0 (strongly disagree) to 3 (strongly agree). These responses were reverse coded and summed, with higher total scores suggesting greater levels of anhedonia.

Addiction Severity Index-Lite (ASI-Lite).

The ASI-Lite (Cacciola et al., 2007), an abbreviated version of the original ASI (McLellan et al., 1980), is a standardized interview that assesses problems in seven domains: medical, employment/support status, alcohol, drug, legal, family/social, and psychological. In the alcohol/drugs domain, participants were asked to report their use for two time periods, past 30 days and lifetime. From the ASI-Lite, we obtained the number of days (in the past 30 days) and number of years (in their lifetime) that participants used cocaine, with more days and years indicating greater cocaine use severity.

Kreek-McHugh-Schluger-Kellogg Scale (KMSK).

The KMSK (Kellogg et al., 2003) is a brief standardized interview that was administered at baseline to quantify substance use. This instrument assesses the frequency, duration, and amount used for several substances during the participant’s heaviest period of use, in addition to current/past use, mode of use, and substance of choice. The KMSK is effective in performing rapid dimensional analyses for several substances, including cocaine (Butelman et al., 2018), and it has demonstrated good construct validity, sensitivity and specificity for cocaine (Kellogg et al., 2003; Tang et al., 2011). From the KMSK, we obtained total scores for lifetime and past month cocaine use with higher scores indicating greater severity of cocaine use, as well as the most money spent on cocaine (lifetime) in a single episode of use and the average amount spent on cocaine use in the last 30 days.

The Structured Clinical Interview for DSM-V – Research Version (SCID-5-RV).

The SCID-5-RV (First et al., 2015) is a semi-structured interview that is used to make DSM-5 diagnoses. From the SCID-5-RV, we obtained number of CUD symptoms, with higher number of symptoms indicating greater cocaine use severity.

Procedure

All data were collected at The University of Texas Health Science Center, Houston. The measures used in this secondary analysis are from baseline and intake assessments prior to any treatment for CUD. During the intake visit, participants completed the ASI-Lite, KMSK, SHAPS, SCID-5-RV, a medical screening, and provided a urine sample to determine eligibility for the clinical trial. If eligible, participants provided informed consent and were enrolled in the clinical trial. In the baseline sessions, participants completed the CocPT. Participants were compensated for completing study procedures.

Data Analytic Plan

Data Quality.

Out of 120 individuals with data from the SHAPS, 117 had data from the CocPT. Individual data from the CocPT from these 117 individuals were initially assessed for systematicity. Demand data were identified as non-systematic if 1) units of drug consumed at a given price were at least 20% greater than at the preceding price, or 2) units of drug consumed at the final price were not less than the first price by at least 10% (Bruner & Johnson, 2014). An exception to the latter was made for data from 7 “zero-responders”, individuals that chose not to consume any cocaine at any prices. The presence of zero-responders was also the justification for utilizing the Bruner and Johnson rules for systematicity rather those developed by Stein and colleagues, as the latter utilizes the log of consumption values (Stein et al., 2015). Data from 4 individuals showed non-systematic data at 5 price points. These specific values were removed for subsequent curve-fits described below. One subject’s demand data were removed from analysis for being an outlier, showing relatively high consumption of cocaine (i.e., up to 1,000 rocks of cocaine).

Demand data were fit to Equation 1 (Koffarnus et al., 2015) below using GraphPad Prism 6.0f (GraphPad Software Inc.; La Jolla, CA).

$$Q=Q_0\mathrm{*}{10}^{k(e^{-\alpha Q_{0}C}-1)}$$ (1)

Equation 1 represents a relatively recent, modified version of the traditional exponential demand model (Hursh & Silberberg, 2008). Advantages of the exponentiated equation include: 1) inclusion of 0 consumption values; and 2) improved fits compared to the exponential model based on recent studies (Koffarnus, Franck, Stein, & Bickel, 2015; Strickland, Lile, et al., 2016; Yu, Liu, Collins, Vincent, & Epstein, 2014). In the equation, Q represents the amount of consumption at commodity price C. Q₀ represents consumption levels at or near 0 price and is also referred to as intensity of demand. The parameter k represents the range of consumption in log units and was set to 2.5 in the current study based on the highest value observed for Q₀. The term α represents the rate of change in elasticity and is inversely proportional to essential value (EV) according to Equation 2 below (Hursh & Roma, 2016):

$$EssentialValue=\frac{1}{(100·\alpha {·k}^{1.5})}$$ (2)

Advantages of EV over α include easier interpretability as well as ability to deal with zero responders. In previous studies, EV for zero responders has been defined as 0, the lowest possible value, as excluding these values may artificially overestimate group demand values (Heckman et al., 2019; Stein et al., 2017). Additional indices of demand were derived using an Excel template provided by the KU Laboratory in Applied Behavioral Economics (http://www.behavioraleconlab.com/resources---tools.html) (Kaplan & Reed, 2014). These demand indices included O_max, the maximum response output or money spent at a given price of drug, otherwise referred to as P_max. Finally, breakpoint represents the first price at which 0 consumption is observed and was derived from the raw data. In summary, the demand indices assessed in the current study consist of Q₀, EV, O_max, P_max, and breakpoint. For demand data from the 7 zero responders, since Equation 1 cannot be fit to their data, all demand indices were defined as 0.

Factor analysis of demand indices.

A factor analysis was conducted to reduce the five demand indices into two latent factors, persistence (i.e., reflecting insensitivity to escalating price) and amplitude (i.e., reflecting the absolute levels of consumption and price; Cassidy et al., 2020; O’Connor et al., 2016). This was undertaken to avoid issues of multicollinearity in the primary analyses, as well as eliminate the need for arbitrary choices of the most relevant demand index to be entered into the primary regressions. First, exploratory factor analysis was conducted to confirm viability of the two-factor solution proposed by previous studies (Cassidy et al., 2020; O’Connor et al., 2016) via inspection of the scree plot (Goldberg & Velicer, 2006). The entered variables were Q₀, EV, O_max, P_max, and breakpoint. Each variable was increased by .01 and log₁₀-transformed to meet normality assumptions prior to entry. Next, PCA with oblique (oblimin) rotation constrained to two factors was performed. Demand indices were permitted to load on multiple factors, and a threshold of 0.40 was considered evidence that a demand index significantly loaded on to a given factor (Tabachnick & Fidell, 2012). Factor scores were derived using the regression method, so that scores had a mean of 0 and standard deviation of 1. PCA was performed using the psych (Revelle, 2023) and GPArotation packages (Bernaards & Jennrich, 2005) in the R statistical environment (R Core Team, 2023).

Covariates.

Age, gender, race, education, day 1 status, and number of cigarettes per day were assessed empirically as potential confounders using univariate models against demand factors, SHAPS, and cocaine severity variables. For a given regression model in the primary analysis, a characteristic was considered confounding if it was significantly associated with both outcome and predictor based on guidelines (Assmann et al., 2000; Pocock et al., 2002), and was included as a covariate in adjusted models. If the statistical inferences in the covariate-adjusted models did not change from the non-adjusted models, the non-adjusted models were reported.

Bayesian analyses.

We conducted Bayesian models to test our hypotheses. First, generalized linear models (GLM) predicted anhedonia as a function of amplitude and persistence in separate models. Next, GLM predicted 1) each measure of cocaine use severity as a function of amplitude and of persistence, and 2) anhedonia as a function of each measure of cocaine use severity. Multiple GLM then predicted each measure of cocaine use severity as a function of anhedonia, amplitude, and persistence. Bayesian statistics permit the quantification of the relative strength of the evidence for the alternative hypothesis, i.e., the likelihood that there is a relationship among demand, cocaine use severity, and anhedonia. Bayesian models provided a posterior distribution with credible range of values for each effect (95% credible interval (CrI), and quantified the posterior probability (PP) that an effect greater (PP(b>0) or less than 0 (PP(b<) exists. Models used weakly informative priors (b=~Normal [μ=0, σ²=100]) to maximize influence of the data on PP. Guidelines broadly characterize PP thresholds (Jeffreys 1961; Lee & Wagenmakers, 2013), as anecdotal (PP=50–74%), moderate (PP=75%-90%), strong (PP=91%-96%), very strong (97%-99%), and extreme (>99%) support for an effect. A PP≥75% was chosen as a minimum level of support, and relationships with at least this level of support were tested in post-hoc analyses. Estimation used four Markov chains of 4000 total iterations each (with 2000 of these discarded as burn-in). Convergence diagnostics via scale reduction factors (R-hat), effective sample size, and posterior predictive distributions were examined to ensure satisfaction of Bayesian modeling assumptions. Assumptions for each model were satisfied, including R-hat < 1.01, sufficiently large effective sample size, and graphical confirmation that the observed distribution of each outcome fell within the range of distributions produced by 1000 replications drawn from the posterior distributions of the outcome. All outcomes were modeled using the gaussian distribution with the following exceptions for non-normal data: the student distribution was used to model continuous outcomes with few outliers (SHAPS), and the lognormal distribution was used to model right-skewed monetary outcomes (KMSK average money spent in past 30 days, KMSK most money spent in lifetime). Analyses were performed via the brms package (Bürkner, 2017) in the R statistical environment (R Core Team, 2023).

Results

Socio-demographic and cocaine use characteristics.

Participants were predominately male (80%), Black/African American (78%), and non-Hispanic (91%), with mean age of 49.93 and 12.82 years of education (Table 1). Participants reported an average of 17.67 (SD = 9.61) days of cocaine use in the past 30 days and an average of 17.47 (SD = 9.84) years of use in their lifetime.

Table 1.

Sample characteristics

Characteristic	M (SD) or n (%)
Age	49.93 (7.99)
Male gender	93 (80%)
Race
Black/African American	90 (78%)
White	17 (15%)
Asian	3 (3%)
More than one race	3 (3%)
Not Reported	3 (3%)
Ethnicity
Non-Hispanic	106 (91%)
Years of education	12.82 (1.54)
SCID CUD Symptoms	7.2 (1.95)
SHAPS Total
Drug Use
Cocaine Positive on 1st Day of Treatment	19 (16%)
Days using cocaine this month	17.67 (9.61)
Years using cocaine	17.47 (9.84)
Cigarettes per day	8.12 (8.87)
KMSK 30 Days Total Score	9.34 (2.17)
KMSK Lifetime Total Score	14.36 (2.07)
KMSK 30 Days Average USD spent	73.35 (62.94)
KMSK Lifetime Most USD spent	178.16 (187.49)

Note. SHAPS = Snaith-Hamilton Pleasure Scale; KMSK = Kreek-McHugh-Schluger-Kellogg Scale; SCID = Structured Clinical Interview for DSM-V.

Factor analysis of demand indices.

Principal component analysis (PCA) suggested a 2-factor model, accounting for 95% of the variance (Table 2). The first factor, persistence (68% variance), loaded heavily on EV, O_max, P_max, and Breakpoint, reflecting insensitivity to increases in price of cocaine. The second factor, amplitude (27% variance), loaded heavily on Q₀, reflecting the level of demand in unrestricted conditions. The two factors were significantly correlated (r=0.69, p<.001). Both factors were used to represent demand in subsequent analyses.

Table 2.

Demand factor loadings

Index	Factor 1 Persistence	Factor 2 Amplitude
Q0	0.10	0.93
Essential Value	0.77	0.25
Omax	0.77	0.29
Pmax	1.10	−0.18
Breakpoint	0.83	0.16
% Variance	68.0	27.0
Eigenvalue	3.39	1.37

Covariate testing.

Covariate testing revealed that age met criteria for confounding, being significantly related to amplitude (b=1.97, p=.008), ASI lifetime number of years (b=0.32, p<.001), and KMSK 30 days score (b=0.70, p=.042). Age was entered as a covariate in models predicting ASI lifetime number of years or KMSK 30 days score as a function of amplitude.

Bayesian regressions.

Bayesian results are displayed in Tables 3 and 4; relationships with PP≥75% are emphasized and summarized as follows. Univariate regressions showed that SHAPS was positively related with persistence but negatively related to amplitude. Unexpectedly, SHAPS was negatively related with all measures of cocaine severity except number of symptoms on the SCID. On relationships between demand factors and cocaine severity measures, analyses showed persistence was positively associated with ASI lifetime number of years, ASI 30 days number of days, KMSK 30 days score, and KMSK average money spent in past 30 days. Persistence was negatively associated with KMSK lifetime score. Amplitude was positively associated with all cocaine severity measures except ASI lifetime number of years (Table 3).

Table 3.

Bayesian simple regressions among demand factors, anhedonia, and cocaine severity measures

	SHAPS		Persistence		Amplitude
	b [95% CrI]	PP	b [95% CrI]	PP	b [95% CrI]	PP
Persistence	4.20 [−2.61, 10.71]	77.8%
Amplitude	−2.49 [−5.67, 0.02]	97.4%
ASI lifetime number of years	−0.09 [−0.46, 0.19]	75.6%	0.67 [−1.15, 2.52]	76.6%	0.56 [−1.26, 2.40]	73.5%
ASI 30 days number of days	−0.10 [−0.44, 0.18]	77.6%	0.75 [−1.03, 2.48]	79.0%	1.01 [−0.81, 2.77]	86.9%
KMSK lifetime score	−0.51 [−1.85, 0.71]	81.6%	−0.14 [−0.53, 0.23]	76.9%	0.25 [−0.15, 0.63]	89.8%
KMSK 30 days score	−0.70 [−2.05, 0.46]	89.0%	0.29 [−0.11, 0.67]	91.7%	0.34 [−0.06, 0.73]	95.4%
KMSK lifetime $	0.00 [−0.01, 0.00]	78.7%	0.05 [−0.11, 0.21]	71.2%	0.19 [0.03, 0.35]	99.0%
KMSK 30 days $	−0.05 [−0.12, 0.00]	98.2%	0.16 [0.01, 0.31]	98.4%	0.15 [0.01, 0.30]	97.8%
SCID number of symptoms	0.22 [−1.37, 1.79]	62.1%	0.11 [−0.26, 0.48]	71.7%	0.21 [−0.15, 0.57]	87.1%

Note. b, estimate; CrI = credible interval; PP = posterior probability. PP≥75% emphasized.

Table 4.

Bayesian multiple regressions predicting cocaine severity measures

	Outcome, b [95% CrI], PP
Effect	ASI lifetime number of years	ASI lifetime number of years (adjusted)	ASI 30 days number of days	KMSK lifetime score	KMSK 30 days score	KMSK 30 days score (adjusted)	KMSK lifetime $	KMSK 30 days $	SCID number of symptoms
Intercept	18.28 [14.78, 21.63], >99.9%	−6.81 [−18.03, 4.63], 88.2%	18.04 [14.67, 21.46], >99.9%	14.44 [13.74, 15.13], >99.9%	9.64 [8.89, 10.41], >99.9%	7.46 [4.77, 10.11], >99.9%	4.81 [4.52, 5.10], >99.9%	4.08 [3.81, 4.35], >99.9%	7.06 [6.39, 7.74], >99.9%
SHAPS	−0.08 [−0.32, 0.17], 73.8%	−0.06 [−0.29, 0.17], 69.5%	−0.04 [−0.28, 0.20], 63.1%	−0.01 [−0.06, 0.04], 61.5%	−0.02 [−0.08, 0.03], 82.2%	−0.02 [−0.07, 0.03], 81.0%	0.00 [−0.02, 0.02], 51.7%	−0.01 [−0.03, 0.01], 78.7%	0.01 [−0.04, 0.06], 67.1%
Persistence	0.64 [−1.91, 3.09], 69.5%	1.02 [−1.23, 3.39], 80.8%	0.12 [−2.47, 2.60], 53.7%	−0.58 [−1.13, −0.06], 98.7%	0.12 [−0.45, 0.67], 66.7%	0.16 [−0.38, 0.70], 71.1%	−0.16 [−0.38, 0.05], 93.3%	0.11 [−0.10, 0.32], 85.8%	−0.07 [−0.58, 0.43], 60.8%
Amplitude	−0.03 [−2.54, 2.55], 50.8%	−1.22 [−3.71, 1.17], 83.7%	0.85 [−1.70, 3.47], 74.7%	0.63 [0.08, 1.17], 98.8%	0.21 [−0.35, 0.80], 76.5%	0.12 [−0.46, 0.66], 65.5%	0.30 [0.09, 0.53], 99.8%	0.06 [−0.15, 0.27], 71.7%	0.27 [−0.24, 0.78], 85.2%
Age (adjusted models only)		0.50 [0.28, 0.71], >99.9%				0.04 [−0.01, 0.09], 95.0%

Note. b = estimate; CrI = credible interval; PP = posterior probability. PP≥75% emphasized. All estimates from unadjusted models except where specified.

When controlling for both demand factors, SHAPS was negatively associated only with KMSK 30 days score (adjusting for age or not) and with KMSK average money spent in past 30 days. When controlling for SHAPS and amplitude, persistence was negatively associated with KMSK lifetime score and KMSK most money spent in lifetime, and was positively associated with KMSK average money spent on cocaine in past 30 days. Persistence only showed a relationship with ASI lifetime number of years when additionally adjusting for age. When controlling for SHAPS and persistence, amplitude was positively associated with KMSK lifetime score, KMSK 30 days score, KMSK most money spent in lifetime, and SCID number of symptoms. Amplitude showed a relationship with ASI lifetime number of years when additionally adjusting for age (Table 4).

Discussion

To our knowledge, the present study is the first to examine the relationship between anhedonia (i.e., undervaluation of non-drug rewards) and demand (i.e., cocaine valuation) in cocaine use disorder. Consistent with our hypothesis, analyses did suggest potential relationships between both factors of demand (persistence, amplitude) and anhedonia, but these did not consistently suggest our hypothesized relationship between higher anhedonia and more demand for cocaine. To the contrary, although greater anhedonia was related to more persistence – continuing to purchase cocaine in the face of price increases – it actually related to lower amplitude of demand when cocaine was less costly (with a higher posterior probability, 97.4%, suggesting “very strong evidence”). This presents a more subtle potential picture of anhedonia’s effects, suggesting anhedonia might relate to individuals’ continued cocaine use in the face of discouraging factors (potentially including during treatment; Crits-Christoph et al., 2018; Wardle et al., 2017), but may not contribute to use of greater amounts of cocaine/more baseline severity.

When analyzing demand’s relationship with cocaine use severity, results showed that greater demand for cocaine was associated with higher cocaine use severity, as expected, with the exception of one relationship in the opposite direction. Although not significant, we observed a similar pattern in one of our previous studies. In that study, lifetime use variables (cocaine severity and money spent), were negatively associated with the demand index related to maximum response output or money spent at a given price of drug (Yoon, Suchting, et al., 2021). It is possible that some individuals with greater lifetime severity might have higher levels of motivation or desire to change the financial impact of drug use, which may relate to decreased persistence. It may be helpful for future studies to explore this further. In our study, amplitude of demand, reflective of the total level of demand that would be demonstrated in unrestricted conditions, displayed a stronger relationship with cocaine use severity, relative to persistence of demand. (Notably, these stronger relationships were accompanied with comparatively higher ranges of posterior probabilities overall (87–99% versus 77–98%), suggesting higher degrees of evidence for these associations with amplitude of demand). These results are consistent with previously reported findings regarding baseline demand and self-reported measures of cocaine use, both from our group (Yoon, de Dios, et al., 2021; Yoon et al., 2020) and others (Zvorsky et al., 2019). It should be noted there is participant overlap with the studies referenced from our group; however, those studies examined the relationship between specific demand indices (e.g., Q₀) and cocaine use rather than factors of demand. This highlights another contribution of this study, which reinforces prior findings that there are two latent factors of demand, persistence and amplitude (Bidwell et al., 2012; Cassidy et al., 2020; Mackillop et al., 2009; O’Connor et al., 2016). In summary, these findings support the Reinforcer Pathology theory (Bickel et al., 2011), which posits that individuals who use drugs display an excessive valuation for addictive reinforcers (cocaine), and continue to show the feasibility and utility of hypothetical purchasing tasks in measuring cocaine demand (Bruner & Johnson, 2014; Webber et al., 2022; Yoon et al., 2020; Yoon, Suchting, et al., 2021).

Analyses suggested potential relationships between anhedonia and measures of cocaine use severity, but these relationships were in the opposite direction hypothesized. Although this is consistent with the idea proposed above that anhedonia might have more effect on the persistence of cocaine use (and thus treatment outcomes) rather than the amount or severity of cocaine use at baseline, they are in direct contrast with prior studies using very similar measures of both anhedonia and cocaine use severity (Morie et al., 2014; Wardle et al., 2023). Indeed, in our prior study in a substantively similar population we found that this same measure of anhedonia related specifically to a composite measure of cocaine use severity comprised of amount and frequency of use in the last 30 days in both frequentist and Bayesian analyses (Wardle et al., 2023). Yet, in the current study greater levels of anhedonia were related to lower amount and frequency of use. Potential explanations for this difference in results include pre- and post-COVID-19 pandemic effects (e.g., impact of disruption of routines and daily activities, shifts in substance use patterns), seasonal variations, differences in study procedures (e.g., the influence of the experimenter administering the study, in our previous study procedures were adapted due to COVID-19), or cohort differences. Of note, the present study’s participant characteristics are similar to those of our prior work (Wardle et al., 2017, 2023); although this study includes a larger sample size. The current study’s findings also contrast with a larger literature on anhedonia and substance use that suggests anhedonia is related to more severe use and worse treatment outcomes across substances (Garfield et al., 2013; Kiluk et al., 2019). Resolving these contradictory findings may require a more comprehensive systematic review or metanalytic approach. Future studies may also assess how anhedonia relates to cocaine use severity overtime to explore these relationships further.

Finally, when analyzing these constructs’ (anhedonia, demand) shared impact on measures of cocaine use severity, analyses suggested “moderate” and “very strong” evidence for relationships between both factors of demand and cocaine use severity, even when controlling for anhedonia. In these analyses, the same counter-intuitive relationship between greater anhedonia and less cocaine use severity held, even when controlling for demand.

Overall, these results suggest demand (i.e., cocaine valuation) may be playing a more consistent and important contribution to cocaine use severity than anhedonia (i.e., undervaluation of non-drug rewards). This overvaluation for cocaine may motivate individuals to seek and consume the drug. This relates to previous research on cocaine motivational value, in which higher brain reactivity towards cocaine cues compared to nondrug rewards was associated with greater motivation to consume cocaine (Webber et al., 2022). It has also been found that individuals attach high incentive salience to cocaine-related cues relative to non-drug cues, and display higher levels of anhedonia compared to other groups (Webber et al., 2021). These findings suggest that blunting the value of cocaine is an important and independent treatment target. Exploring interventions aimed at blunting the perceived value of cocaine may offer promising avenues for CUD treatment. Although, prior studies relating both anhedonia and demand to actual treatment outcomes show it is still important to consider targeting both demand and anhedonia when providing CUD treatment.

The current study had a number of limitations. First, it is cross-sectional; thus, we are not able to establish causality when examining these relationships. Second, most of our sample identified as Black/African-American, Non-Hispanic males, which might limit the generalizability of these finding. Although this is similar to many previous studies on cocaine use disorder (Bruner & Johnson, 2014; Wardle et al., 2017; Webber et al., 2022; Yoon et al., 2020), it still limits the generalizability of our findings. Further studies should consider targeted recruiting of more diverse samples to explore these relationships and analyze the impact of race, gender, and ethnicity. Third, a potential limitation related to hypothetical purchasing tasks is that participants’ choices on the task may differ from decision making in real world scenarios; although, previous literature has shown that decision making in hypothetical purchasing tasks closely relates to real world consumption (Amlung et al., 2012; Wilson et al., 2016).

In sum, the current study investigated the relationship between anhedonia and cocaine demand and their shared impact on cocaine use severity among treatment-seeking individuals with cocaine use disorder. Our findings indicate cocaine demand relates more strongly to cocaine use severity than anhedonia, which presents important treatment and research implications. Indeed, this highlights the need for the development of interventions that target motivational value of drug rewards. Future research may focus on extending this work by evaluating how cocaine demand, anhedonia, and cocaine use severity change over the course of treatment, which can further inform relapse risk. Additionally, as both anhedonia and demand have been associated with cocaine treatment outcomes, assessing both may provide improved predictive utility and help identify individuals with greater treatment need.

Public Health Significance:

Cocaine use disorder is a significant public health concern; thus, it is important to understand processes influencing cocaine use. Our findings suggest cocaine demand plays a stronger role than anhedonia in cocaine use severity.