Examination of preliminary behavioral and effective connectivity findings from treatment response to citalopram in cocaine use disorder: A dynamic causal modeling study

Abstract

We sought effective (directional) connectivity parameters associated with response to citalopram in cocaine use disorder (CUD) by conducting a functional magnetic resonance imaging (fMRI) experiment with participants diagnosed with CUD (n = 13) and matched healthy controls (HC; n = 17). CUD participants showed a positive correlation between bilateral DLPFC-to-putamen effective connectivity and treatment effectiveness score. These preliminary results support further investigation of prefrontal-striatal interactions in response to treatment in CUD.

1. Introduction

In CUD, Moeller et al. (2007) found that treatment with citalopram, combined with behavioral therapy, decreased cocaine use. Expanding on this finding, Green et al. (2009) showed that higher baseline Iowa Gambling Task (IGT) scores were associated with a higher number of negative urine drug screens for participants treated with citalopram relative to placebo. Functional connectivity involved in the response to citalopram may be characterized by a circuit spanning the prefrontal cortex, somatosensory cortex, and striatum [Kjome et al. (2010)]. Additionally, involvement of the ventromedial and right dorsolateral prefrontal cortex [Bechara et al. (2001)] and theorized prefrontal top-down control of the striatum may be associated with decision-making and control of impulsivity. [Ma et al. (2015)]

Changes in fronto-striatal circuit connectivity were found in other disorders including mood disorders [Clark et al. (2009)], suggesting hypothesized top-down control of serotonergic systems [Challis & Berton (2015)]. Recent neuroimaging evidence indicates involvement of serotonin in the modulation of these circuits [Robinson et al. (2013); Vai et al. (2016)]. The present study sought to test whether top-down, prefrontal-to-striatal effective (directional) connectivity (EC) is associated with CUD improvement from citalopram.

2. Methods

2.1. Subjects

Participants were drawn from a subset of a larger clinical trial conducted at the University of Texas Health Science Center at Houston and Virginia Commonwealth University; CUD participants (n = 13) carried lifetime CUD diagnoses. Normal controls (NC) (n = 17) were matched for age, sex, handedness and education level (Supplemental Information).

2.2. Procedures

Identical to the Ma et al. (2015) procedure, participants completed pretreatment testing, including Barratt Impulsivity Scale and IGT. During fMRI, participants completed a Go/NoGo task. Over the subsequent six weeks, CUD participants were treated with 20–40 mg citalopram, and seen three times weekly for urine drug screens, and participated in once weekly individual cognitive behavioral therapy (CBT).

2.3. fMRI data acquisition

fMRI acquisition methodology and scanning parameters were identical to Ma et al. (2015).

2.4. fMRI bold analysis

The pre-processing and analysis protocol was very similar to that used by Ma et al. (2015). Functional images were spatially smoothed with a Gaussian filter of 6 mm isotropic full width at half-maximum. First- and second-level BOLD analyses were completed using SPM12 (fil.ion.ucl.ac.uk/spm) for pooled data including both CUD and HC.

2.5. DCM analysis

Similar to Ma et al. (2015), we utilized the Dynamic Causal Modeling (DCM) module in SPM12 for effective connectivity (EC) analysis. DCM nodes were constructed from activation-associated clusters constrained by atlas-derived a priori regions of interest (ROIs) (Ma et al., 2015); except here the height threshold was T = 1.25 (two-tail p = 0.10) to discover individual participant activation within each ROI while filtering out much of the noise (Zaghlool and Wyatt, 2014).

Correlations between treatment effectiveness score (TES) and individual EC for all top-down connections were tested, and prefrontal to insula, putamen, and thalamus ECs were also tested. Exploratory correlations with TES were also conducted on all remaining ECs.

3. Results

3.1. Clinical outcomes

One participant demonstrated enough negative urine drug screens to merit designation as a “responder” with a TES of 0.51. Two participants reached scores of approximately 0.10, three were in the 0.05–0.07 range, three in the 0.02–0.03 range, and the remaining four had TES = zero.

3.2. Behavioral

CUD participants performed significantly worse than NC on the BIS with significantly higher non-planning and total scores (p < 0.001); motor impulsivity also had a trending difference. CUD participants also had lower average IGT scores than NC. The responder participant was one of four with a positive IGT score. More behavioral data information is included in the Supplemental Information.

3.3. Effective connectivity correlations with treatment outcome

Easy NoGo modulation of EC from the left dorsolateral prefrontal cortex to left putamen demonstrated a positive correlation with TES score for the full treatment group (r = 0.6086, p_corrected = 0.0208, FDR-corrected for L DLPFC-originating connections tested). Easy NoGo modulation of right dorsolateral prefrontal cortex to left putamen effective connectivity was also significantly correlated with treatment effectiveness score (r = 0.5664, p_corrected = 0.0436, FDR-corrected for all R DLPFC-originating connections tested).

3.4. Between-group effective connectivity differences

No top-down (i.e., prefrontal cortex to deep brain structures) modulatory effects survived FDR correction. On exploratory analyses, a single bottom-up, negative modulatory effect for the easy NoGo task from left putamen to left dorsolateral prefrontal cortex EC remained for HC after FDR correction with EC = −0.0281 Hz (two-tail, p_corr = 0.0191).

4. Discussion

Poorer BIS and IGT performance indicated that this treatment group would respond relatively poorly to citalopram [Green et al. (2009)]. Regardless, correlation analysis suggests there is a relationship between treatment response and modulation of connectivity from both left and right dorsolateral prefrontal cortex to left putamen. The sole between-group difference in modulatory effect (greater bottom-up inhibitory modulation of left dorsolateral prefrontal cortex by left putamen) for NC appears to point to the same circuit. Though limited by poor treatment response, these findings suggest there may exist a prefrontal-striatal effective connectivity relationship to be elucidated with further investigation.

This is a preliminary study. Though we acknowledge the limitations of this study, there are relatively few studies examining treatment options for CUD in the context of associated brain connectivity changes with treatment. It is also noteworthy that all FDR-corrected findings pointed to DLPFC-to-striatum (via putamen) connectivity for correlation with treatment response and for inter-group differences in modulation of EC.

Fig. 1.

Inter-VOI effective connectivity which correlated with response to treatment with citalopram (cyan arrows) as well as modulatory effects (pink arrow) which correlated with response to treatment with citalopram.