Exploring the role of reward functioning in the overlap of post-traumatic stress and cocaine use disorder

Abstract

Background:

Posttraumatic stress disorder (PTSD) symptoms are common in people with cocaine use disorder (CUD), and even sub-threshold PSTD symptoms result in worse treatment outcomes (Hadad et al., 2020; Najavits et al., 2007). Difficulties with reward functioning may drive this comorbidity. Impairments in reward functioning are prominent in both PTSD and CUD and contribute to development of substance use problems after trauma (Fani et al., 2020; Vujanovic et al., 2017). There are three distinct reward processes that may be involved in the PTSD/CUD overlap: consummatory reward (ability to experience pleasure), motivational reward (willingness to exert effort for rewards), and reward learning (adapting behavior based on reward history) (Berridge et al., 2009; Treadway & Zald, 2011). Here we test whether impairments in these reward functions account for the relationship between PTSD and CUD symptoms.

Methods:

This is a secondary analysis of data from a clinical trial (NCT02773212) that measured of PTSD symptoms, CUD severity, consummatory reward, motivational reward, and reward learning in 53 treatment-seeking people with CUD.

Results:

Greater PTSD symptoms related to (1) more severe CUD and (2) less ability to learn from reward; however, impaired reward learning did not significantly account for the overlap in PTSD and CUD symptom severity.

Conclusions:

The observed relationship between PTSD and CUD symptoms was not accounted for by reduced ability to experience pleasure from rewards, reduced motivation for rewards, or reduced ability to learn from rewards. Thus, treatments that attempt to enhance reward functioning seem unlikely to address this complex comorbidity.

Introduction

Up to 40% of people with cocaine use disorder (CUD) have comorbid post-traumatic stress disorder (PTSD; Back et al., 2000; Hadad et al., 2020). People with PTSD symptoms have worse substance use treatment outcomes, even when the symptoms are sub-threshold for a PTSD diagnosis (Brancu et al., 2016; Kim et al., 2020; Pietrzak et al., 2021). To improve treatment outcomes, we must understand how PTSD and CUD relate (María-Ríos & Morrow, 2020). One possibility is that people use cocaine to cope with negative emotions arising from PTSD, including fear, anxiety and depression (Hawn et al., 2020; Waldrop et al., 2007). The Research Domain Criteria model of psychiatric disorders suggests that negative mood symptoms in PTSD, such as fear and anxiety, stem from the “negative valence system”, a set of brain circuits which respond to aversive stimuli and contexts (Schmidt & Vermetten, 2017). However, PTSD can also involve dysfunction in the “positive valence system”, a set of circuits which respond to motivational or rewarding stimuli and contexts (Schmidt & Vermetten, 2017). Dysfunction in the positive valence system produces anhedonia, i.e. an absence of pleasure or motivation, which people may also try to combat with cocaine (Jaffe & Kilbey, 1994). Consistent with the idea that negative mood and anhedonia arise from distinct brain systems, they respond differentially to different treatments (Serretti, 2023). Thus, to select appropriate treatments, we need to understand whether CUD/PTSD comorbidity is driven by coping with negative emotions (negative valence system), lack of pleasure/motivation (positive valence system) or both. The hypothesis that people use substances to cope with negative emotions has been thoroughly tested, with a recent metanalysis concluding that coping with negative emotions explains some but not all of the relationship between substance use and PTSD symptoms (Luciano et al., 2022). However, the role of anhedonia has been less examined (Vujanovic et al., 2017).

There is reason to believe anhedonia is important to the PTSD-CUD relationship. Anhedonia is elevated in both disorders (Acheson et al., 2022; American Psychological Association, 2013; Crits-Christoph et al., 2018; Fani et al., 2020; Garfield et al., 2013; Nawijn et al., 2015; Wardle et al., 2017), and post-traumatic anhedonia predicts development of substance use problems (Fani et al., 2020). Note, anhedonia may be just one facet of the “emotional numbing”, i.e. overall diminished emotional experience, seen in PTSD (Wisco et al., 2020). Blunting of emotions after trauma may be adaptive, protecting patients from intense affect that could overwhelm them or cause detrimental behavior like self-harm (Putica et al., 2021). If, as we hypothesize, PTSD and CUD are linked via anhedonia, clinicians may want to be cautious in treating anhedonia, so as not to remove a self-protective mechanism without providing alternate coping strategies. Nevertheless, to direct treatment appropriately we need to understand whether PTSD and CUD are linked primarily by negative emotion, or if anhedonia also plays a role.

Anhedonia itself is a clinical symptom that may be produced by dysfunction in several sub-functions of the positive valence system (Treadway & Zald, 2011). These neurobehaviorally distinct sub-functions include: 1. “Consummatory reward” - hedonic response when reward is received (Positing cake as a reward, if given a slice of cake, how much do you enjoy it?) 2. “Motivational reward” - exertion of effort for rewards (How hard will you work for a slice of cake?) 3. “Reward learning” - adapting behavior based on reward history (If an action results in you getting a slice of cake, do you learn to repeat that action?) (Dexter et al., in press). We need to understand which of these functions is most important in the PTSD/CUD relationship because they may also be differentially influenced by different treatments (Wardle et al., 2024).

To address these questions, we conducted secondary analysis of a CUD clinical trial in which participants completed measures of CUD symptom severity, PTSD symptom severity, consummatory reward, motivational reward and reward learning (Wardle et al., 2023). We hypothesized that impaired reward functioning would account for some of the relationship between PTSD symptoms and CUD severity. Given prior findings in this same sample showing consummatory reward is most strongly linked to CUD severity (Wardle et al., 2023), we hypothesized that consummatory reward would contribute most to this relationship.

Methods

This is a secondary analysis of a clinical trial for CUD that examined reward functioning and outcomes in CUD treatment, registered at https://clinicaltrials.gov/study/NCT02773212. Refer to our previously published article for methodological details (Wardle et al., 2023).

Participants

Participants were N = 53 adults age 18-60 seeking treatment for moderate to severe CUD, who had no other substance use disorders aside from alcohol, cannabis or nicotine use disorder, provided at least ≥1 cocaine positive urine sample during screening, and were in acceptable mental and physical health for an outpatient medication study (Wardle et al. 2023 for details). Participants were recruited from Houston and Chicago through ads and clinic referrals. The University of Illinois Chicago and the University of Texas Health Science Center at Houston IRBs approved the study.

Measures

PTSD symptoms were assessed using total scores on the PTSD Checklist for DSM-5 (PCL-5), with measurement of trauma exposure and identification of an “index” trauma using the accompanying Life Events Checklist (LEC) (Blevins et al., 2015). We collected seven reward functioning measures. See Wardle et al. (2023) for details, but briefly, measures of consummatory reward were: 1. Total Snaith-Hamilton Pleasure Scale score (SHAPS; Snaith et al., 1995); 2. Valence ratings for positive emotional pictures minus neutral emotional pictures (Marchewka et al., 2014); 3. Arousal ratings for positive emotional pictures neutral emotional pictures; 4. Corrugator (CR; “frown”) electromyography (EMG) during positive emotional pictures minus neutral emotional pictures (Larsen et al. 2003). Measures of motivational reward were: 1. Keypresses to extend viewing of positive emotional pictures minus neutral emotional pictures; 2. Percent of high-effort trials chosen on the Effort Expenditure for Rewards Task (EEfRT; Treadway et al., 2009). Our measure of reward learning was response bias on the Probabilistic Reward Task (PRT; Pizzagalli et al., 2005). Consistent with our prior publication (Wardle et al., 2023), CUD severity was a composite of number of CUD symptoms on the SCID-5, total dollar amount of cocaine used in the past 30 days, and number of days the participant used cocaine in the past 30 days, all z-scored and summed.

We also evaluated the following potential covariates: diagnosis of past year alcohol and cannabis use disorder using the Structured Clinical Interview for DSM-5 (SCID-5; First et al., 2015) no/mild nicotine dependence vs. moderate to severe nicotine dependence using the Fagerstrom Test for Nicotine Dependence (FTND; 50), a binary variable classifying participants as having a past year mood or anxiety disorder or not (the most common comorbid psychiatric diagnoses) from the SCID-5, and cocaine withdrawal using the total sum of the Cocaine Selective Severity Assessment for (Kampman et al., 1998).

Procedure

Measures were collected during an in-person screening over 1-3 visits totaling 4-6 hours and a 2.5 hour baseline laboratory session. All measures were collected prior to treatment. See Wardle et al. (2023) for details.

Data Analytic Plan

Analyses were conducted using R v4.2.5 (RStudio Team, 2015). We imputed missing data for EMG for 3 participants, the PRT for 9 participants and the EEfRT for 3 participants (see Wardle et al. 2023 for details) using predictive mean matching in the mice package in R, and used pooled results from five imputed data sets for all analyses (Buuren & Groothuis-Oudshoorn, 2011). We examined all continuous variables for normality and transformed any non-normal variables. Next, we assessed all characteristics in Table 1 (aside from probable PTSD diagnosis, trauma exposures and SHAPS cut point) as possible covariates. Per published guidelines (Pocock et al., 2002), only variables that significantly related to both PTSD symptoms and either CUD severity or any reward functioning measure were used as covariates. Only location (Houston/Chicago) met these criteria, with Chicago participants showing higher severity across self-report, interview and behavioral measures of distress. Thus, we report results both with and without location as a covariate, as it may simply proxy greater PTSD and CUD severity. We then completed primary analyses: Step 1. Linear regression of CUD severity on PTSD symptoms; Step 2. Linear regressions of each reward functioning measure on PTSD symptoms, applying Benjamini-Hochberg corrections for False Discovery Rate; Step 3. Any reward functioning measure that significantly related to PTSD severity in Step 2 was included as a covariate in the PTSD/CUD regression, and bias-corrected and accelerated bootstrapping was applied to establish whether that reward function significantly accounted for overlap between PTSD symptoms and CUD severity (BCa, mediate package in R; Tingley et al., 2014). Note, these steps are used to test mediation, but our analysis is cross-sectional, so we do not refer to this as mediation.

Table 1.

Sample characteristics

Characteristic	Houston (n = 30) M (SD) or n (%)	Chicago (n = 23) M (SD) or n (%)	Total (n = 53) M (SD) or n (%)
Male gender	26	18	44 (83%)
Race
Native American	1	1	2 (4%)
Black	25	19	44 (83%)
White	1	1	2 (4%)
Not reported	3	3	6 (11%)
Latinx ethnicity	4	4	8 (15%)
Age	49.73 (6.95)	52.35 (5.56)	50.87 (6.46)
Years of education	13.23 (2.17)	13.52 (1.70)	13.36 (1.97)
Monthly income	$1750 (1965)	$1519 (1260)	$1650 (1685)
Location Houston			30 (57%)
Participated pre-COVID*	30	13	43 (81%)
Alcohol use disorder	7	5	12 (23%)
Cannabis use disorder	3	2	5 (9%)
FTND Nicotine dependence	17	8	25 (47%)
Mood or anxiety disorder past year*	1	7	8 (15%)
CSSA Cocaine withdrawal score	17.15 (14.83)	18.82 (14.10)	17.87 (14.40)
PCL-5 Probable PTSD*	1	8	9 (17%)
Trauma exposures (from LEC)
Natural disaster*	23	6	29
Fire or explosion	12	9	21
Transportation accident	22	15	37
Serious other accident	7	9	16
Toxic substance exposure	3	2	5
Physical assault*	15	19	34
Sexual assault	3	8	11
Unwanted sex	5	8	13
Combat	2	1	3
Captivity	2	1	3
Life-threatening illness or injury	6	10	16
Severe human suffering	5	4	9
Sudden violent death	6	11	17
Sudden accidental death	5	5	10
Causing injury, harm or death to another	3	7	10
Other	5	7	12
Total # of types of trauma reported			5.17 (3.52)
SHAPS Clinical level of anhedonia	5	5	10 (19%)

FTND = Fagerstrom Test of Nicotine Dependence; LEC = Life Events Checklist; PCL-5 = Posttraumatic Stress Disorder Checklist for DSM-5; SHAPS = Snaith-Hamilton Pleasure Scale; CSSA = Cocaine Selective Severity Assessment. * = p < 0.05 difference between locations on t-test or chi-square test. Note, high levels of exposure to natural disaster in Houston were due to study being run across Hurricane Harvey; however this was not a common index trauma selected for reporting on the PCL-5 (chosen by only one person).

Results

Descriptive Statistics

Per Table 1, our sample was majority male and Black. Seventeen percent of participants met PCL-5 criteria for probable current PTSD (Blevins et al., 2015). Nineteen percent exceeded an established cutpoint for clinically-significant anhedonia on the SHAPS (using the typical "last few days" timeframe; Snaith et al., 1995).

Primary Analyses

Higher PCL-5 Totals significantly related to greater CUD severity, as expected (see Figure 1). This relationship was no longer significant when controlling for study location (see Table 2). As location may have proxied overall greater severity, we continued with subsequent steps. After FDR correction, greater PTSD symptoms related to lower reward bias on the PRT (see Figure 2). After controlling for location, this relationship was marginal (see Table 3). We tested whether reward bias on the PRT significantly accounted for the relationship between PTSD symptoms and CUD severity. There is no established way to pool BCa from multiple imputed datasets, so we ran the analysis in each imputed data set separately and saw no significant indirect effect (indicating no significant overlap) in any imputed data set.

Figure 1.

Higher scores on the PTSD Checklist for DSM-5 (PCL-5), indicating more PTSD symptoms, relate to greater cocaine use severity (composite score includes frequency, amount and life impact of cocaine use). Location of recruitment (Chicago vs. Houston) also shown because this relationship was no longer significant after controlling for location, possibly because recruitment from Chicago was confounded with an overall increase in severity across many measures of distress. This graph shows predicted scores and 95% confidence intervals.

Table 2.

Regression Results for PTSD Symptoms and Cocaine Use Severity

Effect	Estimate	SE	Statistic	p-value
Model No Covariates	Adj. R2 = 0.06	--	F(1, 51) = 4.57	0.04
Intercept	−0.61	0.44	t(51) = −1.40	0.17
PCL-5 Total (sqrt)	0.26	0.13	t(51) = 2.14	0.04
Model With Covariates	Adj. R2 = 0.16	--	F(2, 50) = 5.95	0.005
Intercept	−0.98	0.43	t(50) = −2.23	0.03
Study location (Houston/Chicago)	1.68	0.64	t(50) = 2.61	0.02
PCL-5 Total (sqrt)	0.12	0.13	t(50) = 0.96	0.34

PCL-5 = PTSD Checklist for DSM-5; sqrt = square root transformed

Figure 2.

Lower response bias on the Probabilistic Reward Task (PRT), indicating less effective learning from reward, relates to higher scores on the PTSD Checklist for DSM-5 (PCL-5), indicating more PTSD symptoms. Location of recruitment (Chicago vs. Houston) also shown because this relationship was no longer significant after controlling for location, possibly because recruitment from Chicago was confounded with an overall increase in severity across many measures of distress. This graph shows predicted scores and 95% confidence intervals, drawn from one randomly selected imputed data set of five.

Table 3.

Results of linear regressions testing relationships between PTSD Symptoms and reward functioning measures, without and with adjustment for study location

Variable	M (SD) or n (%)	Standardized beta for PCL-5 Total (sqrt)	Standardized beta for PCL-5 Total (sqrt), adjusted for location
PCL-5 Total (raw)	12.24 (15.23)	--	--
PCL-5 Total (sqrt – used in all analyses)	2.51 (0.34)
Consummatory Reward
SHAPS	8.66 (0.72)	β = −0.02, p = 0.89	β = 0.04, p = 0.82
Valence Ratings	1.76 (0.12)	β = −0.19, p = 0.17	β = −0.12, p = 0.41
Arousal Ratings	0.98 (0.15)	β = −0.05, p = 0.73	β = −0.01, p = 0.96
Corrugator EMG +	−0.52 (0.17)	β = 0.11, p = 0.46	β = 0.15, p = 0.35
Motivational Reward
Keypresses to Extend (sqrt)	4.47 (0.15)	β = 0.01 p = 0.96	β = 0.24, p = 0.09
EEfRT % Hard Task +	31% (0.03)	β = −0.04, p = 0.77	β =0.06, p = 0.71
Reward Learning
PRT Response Bias +	0.12 (0.02)	β = −0.44, p = 0.001*	β = −0.31, p = 0.03

⁺Missing data, all estimates involving this variable based on pooled models with multiple imputation

^{* =}p < 0.05 after Benajmini-Hochberg correction for seven reward functioning measures; sqrt = square root transformed; CUD = Cocaine use disorder; SHAPS = Snaith-Hamilton Pleasure Scale; EMG = electromyography; EEfRT = Effort Expenditure for Rewards Task; PRT = Probabilistic Reward Task

Discussion

To our knowledge, this is the first study to evaluate the role of reward functioning in the PTSD/CUD relationship (Vujanovic et al., 2017). As expected, PTSD symptoms were associated with greater CUD severity. Further, people with higher PTSD symptoms were less likely to learn to repeat behaviors that had been rewarded. This is theoretically consistent with the pathology of PTSD, which affects brain networks involved in reward learning (Seidemann et al., 2021), but surprising from an empirical standpoint, as a previous study showed enhanced reward learning in PTSD (Myers et al., 2013). We also expected stronger relationships with other reward functions (Elman et al., 2005; Hopper et al., 2008; Nawijn et al., 2015), particularly given prior results in this same sample showing consummatory reward related to CUD severity. Finally, impaired reward learning did not significantly account for the observed relationship between PTSD and CUD.

Our hope was that these results would support applying treatments that attempt to enhance reward functioning (e.g. Mindfulness-Oriented Recovery Enhancement; Garland, 2023) to address comorbid PTSD and improve CUD outcomes. These null findings suggest that coping with negatively-valenced emotions is still the strongest explanation for CUD/PTSD overlap. Coping with negatively-valenced emotions is already addressed in common treatments for this comorbidity, such as Seeking Safety (Najavits, 2002), so unfortunately these findings do not provide an immediate direction for enhancing existing therapies. They do suggest considering other plausible links, such the joint contribution of impulsivity to PTSD and SUD (María-Ríos & Morrow, 2020), as some overlap in symptoms is unaccounted for by coping motives (Luciano et al., 2022).

Limitations include our cross-sectional design, which cannot establish directionality. The smaller sample size may also have led to under-detection of relationships – although, we did see relationships between CUD and PTSD. We also could not examine trauma characteristics, such as type (e.g., accidental, interpersonal). Childhood trauma in particular elevates risk for anhedonia and substance use (Dillon et al., 2009; Fan et al., 2021; Hanson et al., 2015; Moustafa et al., 2021). We also did not have access to participant treatment histories, and thus do not know if these are untreated or residual trauma symptoms. Additionally, most participants identified as male and Black. Although common in studies of CUD (Ling et al., 2016; Ma et al., 2022; Yoon et al., 2020), this is not representative of the general population of people with CUD (Substance Abuse and Mental Health Services Administration, 2023). This also prevented examination of influences of gender, race, or ethnicity, which are important in the manifestation of both PTSD and substance use (Giarratano et al., 2020; Gradus et al., 2017; Miguel et al., 2019). Finally, not all results were robust to covarying for study location, which may have proxied greater overall severity. Strengths included rigorous multi-modal assessment of multiple reward functions.

In conclusion, in this novel investigation of PTSD symptoms, reward functioning, and CUD, people with higher PTSD symptoms were less likely to repeat behaviors that had produced rewards previously. Over time, this could lead to fewer rewarding experiences, contributing to clinically observed symptoms of anhedonia in PTSD. However, the PTSD/CUD relationship was not strongly explained by this problem with reward functioning. Thus, treatments that attempt to enhance reward functioning seem less likely to address this complex comorbidity.