Anticipatory reward processing among cocaine-dependent individuals with and without concurrent methadone-maintenance treatment: Relationship to treatment response^*

1.1. Cocaine dependence among methadone-maintained individuals

Cocaine abuse or dependence is common among opiate-dependent individuals receiving methadone-maintenance treatment (Kolar et al., 1990; Dobler-Mikola et al., 2005); e.g., approximately 58% of patients admitted for methadone-maintenance treatment also meet criteria for cocaine dependence; reviewed in (Kosten et al., 2003). Cocaine use in this population is associated with numerous negative factors including high-risk sexual behaviors (Tross et al., 2009), illegal activities (Hunt et al., 1986), poorer treatment outcomes (Schottenfeld et al., 2005; Williamson et al., 2006) and greater societal costs (Hunt et al., 1986; Carroll et al., 2014). While clinically recognized for a number of years (Kolar et al., 1990), very little is known about the neurobiology of this treatment refractory population. In addition, despite significant progress in the development of effective intervention strategies, the treatment of cocaine dependence among methadone-maintained individuals remains particularly challenging (Carroll et al., 2014).

1.2. Neural function and treatment response

Studies conducted in non-methadone-maintained populations suggest that individual variability in response to treatment interventions for cocaine dependence is related to variation in neural functional responses in regions including the striatum, cingulate and the prefrontal cortex (PFC; e.g., Brewer et al., 2008; Marhe et al., 2013; Clark et al., 2014). Alterations in both reward-processing-related neurocircuitry (e.g., striatum, insula) and in fronto-parietal regions (e.g., dorsolateral prefrontal and orbitofrontal cortices, posterior cingulate cortex (PCC), cuneus and precuneus) have been reported both among individuals with cocaine dependence (Brewer et al., 2008; Potenza et al., 2012; Patel et al., 2013; Clark et al., 2014) and among non-cocaine-dependent, methadone-maintained individuals (Langleben et al., 2008; Jiang et al., 2011; Gradin et al., 2014). However, given differences in treatment responses between individuals with cocaine dependence with and without methadone-maintenance treatment – in addition to the relatively diffuse action of methadone on multiple neurotransmitter systems (e.g., µ-opioid, dopamine and acetylcholine systems; Merali et al., 1974; Koob, 1992) - it is possible that the functional neurobiology underlying treatment responses may differ between these two groups. Elucidation of the functional neurobiology of this difficult-to-treat population may provide important mechanistic knowledge (Balodis and Potenza, 2015) that may eventually aid in the development of novel therapies or in the improvement of current interventions and may represent an important first-step toward a priori assignment of patients with appropriate interventions on an individual level (e.g., Feldstein Ewing and Chung, 2013).

1.3. Anticipatory reward and loss processing in addiction

Monetary incentive delay (MID) tasks (Knutson, 2000; Andrews et al., 2011) allow for assessment of anticipatory reward and loss processing and have been widely used to study neural responses among individuals with a range of addictions (for a recent review, see Balodis and Potenza, 2015). Findings from studies employing MID tasks generally suggest alterations in reward-related neurocircuitry during anticipatory processing among individuals with addictions. However, multiple inconsistencies exist and findings have not been the same for all drugs-of-abuse (Balodis and Potenza, 2015). While some MID studies of cocaine dependence have include methadone-maintained individuals (e.g., Patel et al., 2013), these and other possible drug effects have not always been controlled for (reviewed in Balodis and Potenza, 2015). Direct comparison of subgroups of addicted individuals (e.g., cocaine-dependent individuals with and without methadone maintenance treatment) can provide needed insight into seemingly inconsistent findings from some previous studies (e.g., Patel et al., 2013; Bustamante et al., 2014), and inform current understanding of an understudied clinical population – methadone-maintained, cocaine-dependent individuals.

1.4. Study overview and hypotheses

In this study we assessed neural responses associated with anticipatory reward and loss processing during performance of a MID task (Knutson, 2000; Andrews et al., 2011). Neural responses were compared between: (i) cocaine-dependent, methadone-maintained (CD-MM) patients (n=24); (ii) cocaine-dependent (CD) patients not maintained on methadone (n=20); and (iii) healthy comparison (HC) participants (n=21). All patients were scanned prior to behavioral treatment for cocaine dependence. Correlational analyses were conducted to explore associations between BOLD responses and durations of abstinence during treatment.

Based on previous findings (reviewed above) of alterations within cortico-striatal and fronto-parietal regions among individuals with cocaine and opioid dependence, separately (Brewer et al., 2008; Langleben et al., 2008; Jiang et al., 2011; Potenza et al., 2012; Patel et al., 2013; Clark et al., 2014; Gradin et al., 2014), we hypothesized: (i) a significant main effect of diagnostic group on neural responses during anticipatory reward and loss processing within cortico-striatal and fronto-parietal networks. Based on previous reports of decreased BOLD responses during reward (Bustamante et al., 2014) and loss (Patel et al., 2013) anticipation among individuals with cocaine dependence, we further hypothesized that this would involve decreased BOLD responses during anticipatory reward and loss processing among CD-MM and CD participants, in comparison to HC participants. Given methadone’s inhibitory effects on µ-opioid, cholinergic and dopaminergic systems (Merali et al., 1974; Koob, 1992), we further hypothesized that: (ii) reductions in BOLD response would be most pronounced among CD participants also receiving methadone-maintenance therapy; and (iii) there would be a negative association between BOLD responses and daily methadone dose (mg/day) in this patient group.

Several recent studies conducted in non-methadone-maintained populations suggest a negative association between insular functioning and subsequent relapse. Specifically, reduced insular BOLD responses (Clark et al., 2014) and reduced insular-striatal connectivity (McHugh et al., 2013) have been positively associated with subsequent relapse to stimulants. In addition, increased caudate BOLD response has been found to be positively associated with treatment retention among individuals with cocaine dependence (Bustamante et al., 2014). Based on these data, a final exploratory hypothesis (iv) was that neural responses within the caudate and insula during anticipatory reward and loss processing would be positively associated with better treatment outcomes (maximum days of consecutive abstinence) among both CD-MM and CD patients.

2.1. Participants and recruitment

Treatment-seeking CD-MM participants (n=24), treatment-seeking CD patients (n=20) and demographically similar HC participants (n=21) were included in this study. All patients were recruited from randomized clinical trials (RCTs) of behavioral interventions for cocaine dependence (details in Supplemental Materials). Inclusion criteria for all RCTs included a diagnosis of DSM-IV cocaine dependence, as assessed using structured clinical interview (SCID; First, 1995) and willingness to commit to eight weeks of treatment. Data from the CD patient group have been published previously and were collected as part of a separate study (Jia et al., 2011); however, this patient group is included here as a comparison group in order to allow for assessment of neural responses specifically related to methadone-maintenance treatment among individuals with cocaine dependence. Importantly, this earlier publication (Jia et al., 2011) did not include data relating to the anticipatory phase (A2 phase) of the MID task that is focus of the current manuscript. In order to ensure consistency across publications, the same treatment outcome variable (maximum days of self-reported consecutive abstinence; Jia et al., 2011) is used here to quantify treatment responses in both patient groups. Notably, this measure has been found to be a significant predictor of longer-term abstinence (Carroll et al., 2014).

HC participants were recruited from the community via advertisement. Exclusion criteria for HC participants included current or past use of any psychotropic medication, and any current or past Axis-I disorder (including lifetime alcohol- or other substance-use disorder with the exception of nicotine dependence) as assessed using a SCID (First, 1995).

2.2. Demographic and clinical characteristics

Demographic, clinical and cocaine-use characteristics are shown in Table 1. In comparison to both CD-MM and CD participants, HC participants had significantly more years of education (p<.001), but did not differ in age or gender (Table 1). CD-MM and CD participants did not differ in years of education, age or gender.

CD-MM and CD participants did not differ in years of pretreatment cocaine use or days of cocaine use in the month prior to treatment (p’s>.05; Table 1). However, consistent with the general literature, CD-MM participants had significantly higher pretreatment Addiction Severity Index (ASI) scores (p=.005) and achieved significantly less days of abstinence from cocaine during treatment than did non-methadone-maintained CD participants (p<.001). CD-MM and CD participants did not differ in rates of tobacco-smoking, alcohol-use, mood, anxiety or personality disorders (Table 1).

2.3. Monetary incentive delay task

The modified MID task used in this study has been previously described (Andrews et al., 2011); Supplemental Figure 1¹. At the start of each trial participants were presented with a cue for 1000msec, indicating the amount of money to be won or lost on that trial (e.g., ‘WIN $1’, ‘LOSE $1’), followed by a fixation cross (prospect of reward; A1 phase; variable duration). Participants were then presented with a target stimulus for a variable 3–5 sec duration. In order to win (on ‘Win’ trials), or avoid losing (on ‘Loss’ trials) money on each trial, participants responded with a button press while the target was on the screen. Following the target stimulus, a fixation cross was again presented (anticipation of reward or loss phase; A2; variable 3–5 second duration). Finally, participants were given feedback on the outcome of the trial based on whether they successfully hit the target (e.g., ‘WON $1’; ‘DID NOT WIN $1’; ‘LOST $1’; ‘DID NOT LOSE $1’). Further details (duration, subject payment) are provided in the Supplemental Materials².

2.4. Image acquisition, spatial processing and subject-level statistics

CD participants were recruited and scanned between 2006 and 2007. CD-MM participants were recruited and scanned between 2008 and 2009. All imaging data were acquired using the same Siemens Trio 3T scanner (Siemens AG, Erlangen, Germany). Further details of image acquisition, spatial processing and comparison of subject motion are presented in the Supplemental Materials. Onsets of task events were convolved with the hemodynamic response function in SPM12. Models were high-pass filtered at 128sec and included additional motion regressors from realignment. All events were modelled with temporal derivatives. Given debate regarding the significance of a neutral comparison (e.g., ‘WIN $0) condition (Balodis and Potenza, 2015), events of interest were compared to implicit (non-modelled) baseline activity, consistent with previous studies (e.g., Andrews et al., 2011).

2.5. Group-level statistics

A single multi-level voxel-wise ANOVA including the between-subjects factor of group (CD-MM, CD, HC) and the within-subjects factor of trial type (win/loss) was used to test the primary study hypothesis of a significant main effect of diagnostic group on neural responses during anticipatory processing. In order to determine the direction of the identified main effects, post-hoc group-wise comparisons were conducted using voxel-wise t-tests. All statistical maps were voxel-level thresholded at p<.001 prior to undergoing cluster-based family-wise-error (FWE) correction (pFWE<.05). Associations between daily methadone dose and BOLD responses within the identified clusters were assessed using Pearson’s r.

2.6. Exploratory correlational analyses related to abstinence during treatment

As described in the introduction, the insula and caudate were selected as a priori ROIs for correlational analyses with days of abstinence during treatment (details below), based on findings from previous studies of cocaine dependence ( e.g., Marhe et al., 2013; McHugh et al., 2013; Clark et al., 2014), and methadone-maintained opioid-dependence (e.g., Langleben et al., 2008; Gradin et al., 2014). Previous MID studies of cocaine dependence have differed with respect to the functional coordinates used to define the caudate (reviewed in Balodis and Potenza, 2015) and both the anterior and posterior aspects of the insula are implicated in cocaine dependence (e.g., Kilts, 2001; McHugh et al., 2013). Given this lack of specificity, both ROIs were defined anatomically (see Figure S2³). Bilateral anatomical ROIs were defined using the automatic anatomical labeling (AAL) atlas (Tzourio-Mazoyer et al., 2002). Comparisons of BOLD signal within these ROIs were conducted using small volume correction (SVC) as implemented in SPM12 (pFWE<.05). Average BOLD signal values for ROI were extracted from individual participant contrast images and entered into SPSS for correlational analyses with methadone dose (CD-MM group only) and percentage of cocaine-negative urines during treatment (all patients) and were conducted using Pearson’s r, two-tailed α=.05.

3.1. Task performance

The modified MID task utilizes an individual calibration method (based on participants’ out-of-scanner reaction times) to ensure relatively equal numbers of ‘hits’ (successful wins on win trials and successful avoidance of losses on loss trials) across participants (Andrews et al., 2011). Consistent with this calibration method, neither hit rates nor total amounts of money won differed significantly across participant groups (Table 1). In addition, groups did not differ on overall reaction times for either win or loss trials (Table 1).

3.2. Whole-brain effects of group

Whole-brain ANOVAs indicated significant main effects of group on neural responses during anticipatory reward and loss processing (A2 phase of the MID task) within nine clusters (Table 2; Figure 1). The first cluster included regions of the left posterior cingulate, cuneus, precuneus, visual cortex and parahippocampal gyrus. The second cluster included regions of the left dorsolateral prefrontal cortex (DLPFC) and left inferior and middle frontal gyri. Together, clusters three, four, five and six included regions of the right lingual, fusiform and inferior temporal gyri, of the left medial frontal gyrus and of bilateral middle and superior temporal gyri. The seventh cluster was located in the right putamen. The eighth cluster was located in the right preuneus and the ninth cluster in the right cuneus.

3.3. Groupwise comparisons

Findings from whole-brain-corrected post-hoc groupwise comparisons related to effects of group on neural responses during anticipatory reward and loss processing are described below and are shown in Figure 1 and Table 3.

3.4. Exploratory correlational analyses related to abstinence during treatment

Among CD-MM patients, there was a significant negative association between treatment response (percentage of cocaine-negative urines during treatment) and pretreatment BOLD responses during win anticipation (A2 phase) within the right caudate (r_(df=22)=−.43, p=.036; Figure 2). This relationship remained significant after controlling for current methadone dose (r_(df=21)=−.41, p=.0496) and was not observed among non-methadone-maintained CD patients (r_(df=18)=.13, p=.573). Comparison of r-values using Fisher’s r-to-z transformation indicated significant differences in the strength of the correlations between days of consecutive abstinence and BOLD responses within the caudate among CD-MM versus CD patients (Z=−1.82, one-tailed p=.034). Follow-up comparisons controlling for treatment group confirmed these findings and are presented in Table S2⁹.

4.1. Whole-brain between-group differences

During anticipatory reward and loss processing, CD-MM individuals exhibited significantly decreased activity within fronto-parietal regions implicated in cognitive processes including aspects of working memory, cognitive control and sustained attention (Owen et al., 2005; Swick et al., 2008), but not within reward-related neurocircuitry (e.g., caudate), as has been previously reported in some studies of cocaine dependence (e.g., Bustamante et al., 2014). All previous MID studies conducted in cocaine dependence have employed ROI-based approaches focusing primarily on reward-related regions (e.g., none included the IFG or PCC; Jia et al., 2011; Patel et al., 2013; Bustamante et al., 2014). Thus, the extent to which MID task performance is associated with alterations within the identified regions among individuals with cocaine dependence has not been assessed previously. These data suggest that augmentation of current treatments with interventions specifically targeting fronto-parietal attentional networks (e.g., neurocognitive training, cholinergic enhancement; Sofuoglu et al., 2013) may be beneficial among cocaine-dependent individuals who are maintained on methadone.

The PCC and precuneus regions identified in our main effects analyses are central hubs of the default mode network, which is typically deactivated during cognitive task performance (Raichle et al., 2001). Thus, greater deactivations within these regions during reward task performance may relate to default mode functional connectivity alterations previously reported in both opioid- and cocaine-dependent populations (e.g., Upadhyay et al., 2010; Konova et al., 2015). Similarly, decreased IFG activity may relate to alterations within cognitive control and/or salience networks (Sharp et al., 2010). However, further research using network-based analysis approaches (e.g., independent component analysis; ICA) is needed to test this hypothesis.

The precuneus, PCC and IFG are implicated in multiple addiction-related processes including drug-cue reactivity (Wilcox et al., 2011; Courtney et al., 2014), craving (Garavan, 2000; Brody et al., 2007) and vulnerabilities to relapse (Clark et al., 2014; Stewart et al., 2014). Alterations within the primary visual cortex (also identified in the main effects and post-hoc analyses) have been previously reported among individuals with a range of addictions, including opioid addiction (Gradin et al., 2014), and may relate to impairments in visual memory noted in this population (Darke et al., 2000). Our findings are further consistent with recent meta-analytic data indicating robust effects of both pharmacological and behavioral treatment interventions for addictions on neurofunctional responses within the precuneus, PCC, IFG and primary visual cortex (Konova et al., 2013). As such, our findings of methadone-related reductions in neural responses within these regions add to an emerging body of literature highlighting the role of fronto-parietal neurocircuitry in the development, maintenance and treatment of addictions (Courtney et al., 2014). Our correlational analyses suggest that BOLD signal reductions within the IFG and precuneus may be related to methadone in a dose dependent fashion. These data are consistent with methadone’s indirect effects on multiple neurotransmitter systems including (in addition to the µ-opioid system) those involved in attentional control and reward-related processes; e.g., acetyl-cholinergic, dopaminergic (Merali et al., 1974; Koob, 1992).

4.2. ROI analyses

Although not detected in our whole-brain analyses, a priori ROI-based analyses further indicated a significant effect of diagnostic group on neural responses within the insula and caudate during anticipatory processing. For the insula, groupwise comparisons indicated that this involved decreased responses among CD-MM patients in comparison to both CD and HC participants. This finding is consistent with a recent report of decreased insula activation during reward anticipation among opiate-dependent individuals maintained on methadone (Gradin et al., 2014), as well as with structural MRI findings suggesting differences in insular cortex activity between cocaine- and opiate-dependent individuals (Lyoo et al., 2004). Given the central role of the insula in a range of interoceptive and reward-related processes (e.g., approach behaviors, drug cue-reactivity; Paulus and Stewart, 2014) one possible interpretation of these findings is that methadone’s efficacy in reducing opioid use might involve a reduction in drug cue-reactivity and associated approach behaviors via attenuation of insular activity during anticipatory processing. While further research is needed to support this hypothesis, further research into treatments specifically targeting insular function in this and other addicted populations appears warranted (Paulus and Stewart, 2014).

The main effect of diagnostic group within the caudate was primarily driven by relatively decreased activity during loss anticipation among the CD group when compared to the HC group. Previous studies using variations of the MID task to study anticipatory reward and loss processing have yielded somewhat mixed findings, with both increases, decreases and no alterations in caudate activity reported among individuals with cocaine dependence (Jia et al., 2011; Patel et al., 2013; Bustamante et al., 2014). Seemingly discrepant findings may relate to between-study differences in the clinical characteristics of patients or to within-study heterogeneity, as previous studies have not consistently assessed the effects of current medication status (e.g., methadone) on neural responses among individuals with cocaine dependence (Patel et al., 2013; Balodis and Potenza, 2015).

4.3. Associations with abstinence

Partially contrary to our exploratory hypothesis, no significant associations between neural responses within the insula and durations of abstinence were observed within either patient group. Among CD-MM patients, neural responses within the right caudate were negatively associated with abstinence during treatment (percent days of cocaine-negative urines). This relationship was not observed among non-methadone-maintained CD patients. Statistical comparison of r-values confirmed a differential relationship between caudate activity and abstinence between the two patient groups, which remained after controlling for the type of treatment received. Individual variations in pretreatment neural responses relate to differences in response to treatments for cocaine dependence (e.g., Brewer et al., 2008; Marhe et al., 2013). However to our knowledge, this is the first report of a differential relationship between pretreatment neural activity among individuals with and without methadone-maintenance treatment. While preliminary, these data support the hypothesis that individual differences in pretreatment neural function and neurochemistry relate to heterogeneities in treatment responses for addictions (Martinez et al., 2011). They further demonstrate the importance of assessing the relationship between neural function and treatment responses among different sub-populations of disorders (e.g., methadone-maintained, cocaine-dependent individuals).

4.4. Strengths, limitations and conclusions

To our knowledge, this is the first study to compare neural responses between individuals with cocaine dependence with and without methadone-maintenance treatment, and in comparison to HC participants. This study’s strengths include recruitment of a difficult-to-treat population, the use of a well-validated MID task, and the relatively close matching of CD individuals in the outpatient versus the methadone samples. Study findings should nonetheless be considered within the context of some limitations, including the absence of a non-cocaine-dependent, methadone-maintained comparison group. Based on our a priori hypotheses, we constrained our analyses to the anticipatory (A2) phase of the MID task. This phase was selected given the particular relevance of anticipatory processing to action selection and to addictive behaviors more generally (Balodis and Potenza, 2015). However, future studies assessing other aspects of reward-processing (e.g., consummatory processing of reward outcomes) are warranted. In particular, studies employing computational modeling techniques to assess more nuanced aspects of reward and loss processing (e.g., prediction errors) within different subdivisions of the striatum during MID task performance in this and other addicted populations are needed (Robinson et al., 2010; Balodis and Potenza, 2015).

This study did not include assessment of some clinical factors which could have influenced treatment responses; e.g., trauma history (Heffner et al., 2011), neurocognitive factors (Aharonovich et al., 2006). Thus we cannot exclude possible effects of these variables on study findings. Several patients met criteria for past or current depression, which may influence neural reward responses (Hagele et al., 2015). While we were not sufficiently powered to assess between-group differences in neural responses based on depression variables, the CD-MM and CD groups did not differ in rates of past or current depression. Similarly, while our inclusion of only treatment-seeking, cocaine-dependent individuals allowed for correlational analyses with treatment outcomes, the behavioral treatments delivered differed across patients and we were not sufficiently powered to explore group-by-therapy (e.g., CBT4CBT vs. TAU) interactions. While correlational findings remained unchanged after controlling for the type of therapy received, caution is warranted when interpreting these data. Thus, further studies in larger samples are needed to adequately assess relationships with treatment.

Previous research suggests neurostructural differences (as assessed via quantification of white matter hyperintensities) between opiate- and cocaine-dependent individuals within regions including the insular cortex (Lyoo et al., 2004). Neural structural factors within regions including the caudate and hippocapmus are related to treatment outcomes among individuals with addictions (Xu et al., 2014; Yip et al., 2014). In addition, individual variability in neurofunctional responses during MID task performance has been linked to neurostructural variability (Yip et al., 2015). Thus, a final limitation of this study is the absence of accompanying structural (e.g., T1- or diffusion-weighted) data analyses, as would be required to explore the effects of these factors on functional neural responses in this population.

These data nonetheless suggest significant neurofunctional differences between individuals with cocaine dependence who are also maintained on methadone for opioid dependence, in comparison to those who are not. Consistent with previous findings of methadone-associated attenuation of neural responses (Langleben et al., 2008) – and with methadone’s general inhibitory effects on neurotransmitter systems including µ-opioid, cholinergic and dopaminergic systems (Merali et al., 1974; Koob, 1992) - negative associations between daily methadone dose and BOLD responses were observed. These associations remained after controlling for addiction severity measures. Thus, differences in neural responses observed between cocaine-participant groups in this study may relate specifically to methadone dosing. These data highlight the importance of comparing functional responses between different subgroups of addicted populations, particularly as these responses link to treatment-related variables.