Introduction
The ability to effectively inhibit a craving is crucial to prevent relapse. Impairments in response inhibition − a form of impulsivity characterized by the inability to suppress suboptimal or inappropriate responses − are prevalent in neuropsychiatric disorders including addiction (Domínguez-Salas, Díaz-Batanero, Lozano-Rojas, & Verdejo-García, Reference Domínguez-Salas, Díaz-Batanero, Lozano-Rojas and Verdejo-García2016; Verdejo-García, Lawrence, & Clark, Reference Verdejo-García, Lawrence and Clark2008). Deficits in this domain can be assessed using experimental paradigms such as the Stop Signal Task (SST) (Logan, Zandt, Verbruggen, & Wagenmakers, Reference Logan, Zandt, Verbruggen and Wagenmakers2014; Verbruggen & Logan, Reference Verbruggen and Logan2009) and Go/No-Go Task (Drewe, Reference Drewe1975; Garavan, Ross, & Stein, Reference Garavan, Ross and Stein1999), whereby individuals are required to withhold a prepotent motor action after the presentation of an external stopping cue (for a review, see Chambers, Garavan, & Bellgrove (Reference Chambers, Garavan and Bellgrove2009)). In alcohol use disorder (AUD), impaired performance on these measures can be observed prior to alcohol use onset in at-risk individuals (Nigg et al., Reference Nigg, Wong, Martel, Jester, Puttler, Glass and Zucker2006; Squeglia, Jacobus, Nguyen-Louie, & Tapert, Reference Squeglia, Jacobus, Nguyen-Louie and Tapert2014), concurrently with chronic use (Lipszyc & Schachar, Reference Lipszyc and Schachar2010), and in those prone to relapse within 1 year of treatment (Czapla et al., Reference Czapla, Simon, Richter, Kluge, Friederich, Herpertz and Loeber2016; Rupp et al., Reference Rupp, Beck, Heinz, Kemmler, Manz, Tempel and Fleischhacker2016); suggesting that response inhibition contributes to multiple key alcohol-related outcomes (Groman, James, & Jentsch, Reference Groman, James and Jentsch2009; Wilcox, Dekonenko, Mayer, Bogenschutz, & Turner, Reference Wilcox, Dekonenko, Mayer, Bogenschutz and Turner2014).
Inhibitory control relies on the structural and functional integrity of the prefrontal cortex (PFC), which exerts ‘top-down’ influence via its abundant connections to a wide range of cortical and subcortical brain regions (Dalley, Everitt, & Robbins, Reference Dalley, Everitt and Robbins2011). Specific to reactive stopping, rodent and human research has implicated fronto-striatal circuitry comprised of two PFC subregions − the right inferior frontal cortex (rIFC) and the dorsomedial frontal cortex (particularly, the presupplementary motor area [pre-SMA]) − which send hyperdirect projections to the subthalamic nucleus (STN) − a primary inhibitory hub of the basal ganglia (Aron, Reference Aron2007, Reference Aron2011). Information processing within this network begins with orientation toward the stopping cue by the PFC structures (Cai, Chen, Ide, Li, & Menon, Reference Cai, Chen, Ide, Li and Menon2017) which, in tandem, innervate the STN to suppress competing striatal output to the primary motor cortex (M1) − thus canceling the initiated action (Aron, Behrens, Smith, Frank, & Poldrack, Reference Aron, Behrens, Smith, Frank and Poldrack2007; Jahfari et al., Reference Jahfari, Waldorp, van den Wildenberg, Scholte, Ridderinkhof and Forstmann2011).
Chronic alcohol use promotes a shift from PFC to striatal dominance over responding (Everitt & Robbins, Reference Everitt and Robbins2005), leading to a loss of control over alcohol-seeking and consumption (Goldstein & Volkow, Reference Goldstein and Volkow2002). These large-scale alterations in PFC function and their behavioral sequelae in AUD are attributable, in part, to the impact of alcohol on both inhibitory (i.e. gamma-aminobutyric acid [GABA]-ergic) and excitatory (i.e. glutamatergic) synaptic transmission; the latter being a core regulator of experience-dependent neuroplasticity (Koob & Volkow, Reference Koob and Volkow2016). Extensive preclinical research in rodents indicates that acute alcohol administration produces an overall reduction in cortical excitability through the potentiation of selective GABA transmission and an accompanied suppression of glutamate release (Abrahao, Salinas, & Lovinger, Reference Abrahao, Salinas and Lovinger2017). Prolonged exposure, however, results in an enduring state of cortical hyperexcitability − via decreased GABAergic and increased glutamatergic output countering the inhibitory effect of acute consumption (Kalivas, Reference Kalivas2009) − which facilitates the development and perseveration of alcohol dependence (Littleton, Reference Littleton2001).
In humans, transcranial magnetic stimulation (TMS) techniques have been applied to identify acute and chronic alcohol-related changes in intracortical inhibitory and excitatory processes in vivo. In line with neuromolecular evidence, healthy volunteers administered acute alcohol in single doses show enhanced inhibition in M1 (Conte et al., Reference Conte, Attilia, Gilio, Iacovelli, Frasca, Bettolo and Inghilleri2008; Ziemann, Lönnecker, & Paulus, Reference Ziemann, Lönnecker and Paulus1995), and dampened excitability of the PFC (Kähkönen, Wilenius, Nikulin, Ollikainen, & Ilmoniemi, Reference Kähkönen, Wilenius, Nikulin, Ollikainen and Ilmoniemi2003); with a corresponding decrease in functional connectivity between these areas (Kähkönen et al., Reference Kähkönen, Kesäniemi, Nikouline, Karhu, Ollikainen, Holi and Ilmoniemi2001). Conversely, clinical populations with AUD − including those with alcohol withdrawal syndrome (Nardone et al., Reference Nardone, Bergmann, Kronbichler, Caleri, Lochner, Tezzon and Golaszewski2010) − have shown reduced M1 (Conte et al., Reference Conte, Attilia, Gilio, Iacovelli, Frasca, Bettolo and Inghilleri2008; Quoilin, Wilhelm, Maurage, de Timary, & Duque, Reference Quoilin, Wilhelm, Maurage, de Timary and Duque2018) and PFC (Naim-Feil et al., Reference Naim-Feil, Bradshaw, Rogasch, Daskalakis, Sheppard, Lubman and Fitzgerald2016) inhibition, with more pronounced reductions linked to greater behavioral disinhibition at time of testing and likelihood of relapse at reassessment after 1 year (Quoilin et al., Reference Quoilin, Wilhelm, Maurage, de Timary and Duque2018). These observations are contrary to the global cortical hypoexcitability demonstrated in chronic nicotine (Lang, Hasan, Sueske, Paulus, & Nitsche, Reference Lang, Hasan, Sueske, Paulus and Nitsche2008) and cocaine (Boutros et al., Reference Boutros, Lisanby, Tokuno, Torello, Campbell, Berman and Kosten2001) use − likely due to the distinct receptor profiles of these drugs (Barr et al., Reference Barr, Farzan, Wing, George, Fitzgerald and Daskalakis2011). Thus, translational findings indicate that chronic alcohol promotes a pathophysiology characterized by widespread neuroadaptations in both GABAergic (Hanlon, Dowdle, & Henderson, Reference Hanlon, Dowdle and Henderson2018; Quoilin et al., Reference Quoilin, Wilhelm, Maurage, de Timary and Duque2018; Zhou, Zhan, He, & Luo, Reference Zhou, Zhan, He and Luo2019) and glutamatergic (Conte et al., Reference Conte, Attilia, Gilio, Iacovelli, Frasca, Bettolo and Inghilleri2008; Nardone, Trinka, Sebastianelli, Versace, & Saltuari, Reference Nardone, Trinka, Sebastianelli, Versace and Saltuari2019) mediated systems implicated in cortical excitability, which may, in turn, adversely affect neuroplasticity reliant on these receptor activities (Aroniadou & Keller, Reference Aroniadou and Keller1995; Koob & Volkow, Reference Koob and Volkow2016).
Paired associative stimulation (PAS) is a TMS protocol by which the repeated delivery of low-frequency paired-pulses from two differing sources (e.g. cortical, median nerve, or deep brain stimulation) can modify excitability, and thus, functional activity between brain regions via spike-timing-dependent plasticity mechanisms (Stefan, Kunesch, Cohen, Benecke, & Classen, Reference Stefan, Kunesch, Cohen, Benecke and Classen2000). Cortical PAS (cPAS) involves the paired stimulation of two cortical sites, producing effects which seemingly extend to distant, yet interconnected subcortical regions which subserve more rudimentary behaviors (Burt, Lisanby, & Sackeim, Reference Burt, Lisanby and Sackeim2002). In these protocols, the relative order of stimulation site and duration between (i.e. interstimulus interval [ISI]) paired-pulses can induce either excitation or inhibition; putatively reflecting long-term potentiation (LTP)-like or long-term depression (LTD)-like effects (Stefan et al., Reference Stefan, Kunesch, Cohen, Benecke and Classen2000). In line with the presumed involvement of plasticity mechanisms, the effects of PAS develop rapidly, endure beyond acute stimulation in a reversible manner (Stefan et al., Reference Stefan, Kunesch, Cohen, Benecke and Classen2000), and can be blocked by administration of drugs which interact with glutamate subtype N-methyl-D-aspartic (NMDA)-receptors (Wolters et al., Reference Wolters, Sandbrink, Schlottmann, Kunesch, Stefan, Cohen and Classen2003). Previous research aimed to induce short-term plasticity with PAS during acute alcohol administration in healthy volunteers has shown augmentation of LTD-like (Fuhl, Müller-Dahlhaus, Lücke, Toennes, & Ziemann, Reference Fuhl, Müller-Dahlhaus, Lücke, Toennes and Ziemann2015) but disruption of LTP-like mechanisms in M1 (Loheswaran et al., Reference Loheswaran, Barr, Rajji, Blumberger, Le Foll and Daskalakis2016; Lücke et al., Reference Lücke, Heidegger, Röhner, Toennes, Krivanekova, Müller-Dahlhaus and Ziemann2014) and the PFC (Loheswaran et al., Reference Loheswaran, Barr, Zomorrodi, Rajji, Blumberger, Foll and Daskalakis2017). These more standard PAS protocols used M1- or PFC-TMS preceded 25 milliseconds (ms) earlier by median nerve stimulation with either motor-evoked potential (MEP) or electroencephalograph (EEG) output as outcome measures. cPAS protocols also appear to affect MEP when the conditioning stimulus is applied to regions integrated with M1, such as contralateral M1 (Rizzo et al., Reference Rizzo, Siebner, Morgante, Mastroeni, Girlanda and Quartarone2009), supplementary motor area (Arai et al., Reference Arai, Müller-Dahlhaus, Murakami, Bliem, Lu, Ugawa and Ziemann2011), and ventral premotor (Buch, Johnen, Nelissen, O'Shea, & Rushworth, Reference Buch, Johnen, Nelissen, O'Shea and Rushworth2011) and posterior parietal (Veniero, Ponzo, & Koch, Reference Veniero, Ponzo and Koch2013) cortices. However, cPAS has not been used in individuals with chronic exposure to alcohol.
Recently, our research group demonstrated that a novel cPAS protocol targeting the rIFC and pre-SMA putatively influencing the STN hyperdirect pathway improved response inhibition − particularly for older (⩾30 years) adults − in two separate healthy samples. Specifically, when rIFC was stimulated 4 ms prior to pre-SMA, it was presumed this repeated pairing concomitantly strengthened pre-SMA to STN input; consequently facilitating faster reactive stopping (Kohl et al., Reference Kohl, Hannah, Rocchi, Nord, Rothwell and Voon2018; Mandali, Tsurumi, Popa, & Voon, Reference Mandali, Tsurumi, Popa and Voon2021). (For further discussion of the logic behind our cPAS protocol, see Kohl et al. (Reference Kohl, Hannah, Rocchi, Nord, Rothwell and Voon2018).) Importantly, cPAS delivered to this circuit influenced response inhibition with no effect on delay discounting − a dissociable form of impulsivity wherein choice preference for small, immediate rewards outweighs that for delayed, yet larger rewards (Voon & Dalley, Reference Voon and Dalley2016) − thus indicating target specificity of the cPAS intervention (Kohl et al., Reference Kohl, Hannah, Rocchi, Nord, Rothwell and Voon2018). To our knowledge, no prior research has examined whether cPAS can modulate this disorder-relevant cognitive process as a function of chronic alcohol use. Thus, in the current study, we assessed plasticity of the inhibitory network in AUD v. healthy controls (HC) with a cPAS protocol targeting the rIFC and pre-SMA, and measured response inhibition using the SST.
Materials and methods
Participants
We contacted 55 potential participants from the greater Cambridgeshire area, United Kingdom (UK). Twenty patients meeting the criteria for the Diagnostic and Statistical Manual of Mental Disorders Fifth-Edition (DSM-V-TR; American Psychiatric Association, 2013) AUD (26–74 years, 6 [30%] females) were recruited from the outpatient hepatology clinic at the University of Cambridge Addenbrooke's medical site, with a majority diagnosed with moderate to severe alcohol-related liver disease (ARLD) at time of testing (patient somatic health status is reported in online Supplemental Material Section 1); of these patients, 15 reported they were fully abstinent from alcohol, while five reported their condition as ongoing. Thirty-five HCs (23–73 years, 11 [31.4%] females) were enrolled via SONA online research recruitment system; within the sample of 35 HCs, 20 (26–73 years, 6 [30%] females) were matched for age, gender, and years of education with AUD patients.
Safety screening for all participants undergoing TMS was undertaken via either phone or in-person interview by a trained research assistant. Exclusion criteria included TMS contraindications, past or current major neurological or psychiatric disorders, and undergoing pharmacotherapy programs that could influence task performance or neurological activity (including benzodiazepine or withdrawal medications). Further exclusion criteria for HCs included any past or current Substance Use Disorder (SUD), while AUD patients were excluded if they reported co-morbid SUDs or chronic polysubstance (with the exception of nicotine) use. All participants reported a right-handed predominance.
Upon arrival at the testing center, participants were briefed about the experimental design and gave written informed consent, and were reimbursed £7.50 per hour for study participation. All experimental procedures contributing to this work were approved by the University of Cambridge Research Ethics Committee and performed according to the ethical standards of the Declaration of Helsinki, as revised in 2008.
Response inhibition measure
Response inhibition was assessed with the SST (Cambridge Cognition, Cambridge, UK; Figure 1A); all task audio and visual stimuli were produced by and participant responses recorded on a portable computer monitor. Participants were instructed to respond as quickly as possible to an arrow pointing in either a right or left direction (go signal) by pressing one of two buttons on a button box connected to the monitor with the right or left index finger coinciding with the direction of the arrow. If an audio tone (stop signal) was presented, the participants were required to withhold the response. The stop signal onset time was step-wise modified by the stopping success of the previous response.
Response inhibition measure and cortical paired associative stimulation (cPAS) coil location target and orientation. (A) Stop Signal Task (SST) schematic. (B) Stimulation coil location and orientation. Coil 1 was placed over the right IFC (MNI coordinates [in mm]: x = 48, y = 16, z = 16) at a 20° angle to the coronal plane (shown here in a sagittal view), while coil 2 was placed over the right pre-SMA (MNI coordinates [in mm]: x = 10, y = 10, z = 60) parallel to the midline (shown here in an axial view).
SST performance can be influenced by factors related to execution of the motor response including signal discrimination (measured in amount of direction errors), reaction time (measured as the mean reaction time on go trials), and stopping accuracy (measured by the proportion of successful stops during stop trials), as well as interacting control mechanisms involved in performance monitoring and adjustment (for a review, see Verbruggen & Logan, Reference Verbruggen and Logan2008). Taken together, the primary outcome measure of interest is stop signal reaction time (SSRT); defined as the median reaction time on trials correctly performed with a button press response subtracted from the stop signal delay (Logan et al., Reference Logan, Zandt, Verbruggen and Wagenmakers2014). The lower the SSRT − or the less time necessary to cancel a motor response prompted by the stopping cue − the greater capacity for response inhibition.
Self-report measures
A series of self-report psychiatric measures were employed at baseline. The Alcohol Use Disorders Identification Test (AUDIT; Saunders, Aasland, Babor, de la Fuente, & Grant, Reference Saunders, Aasland, Babor, Fuente and Grant1993) and the Fagerström Test for Nicotine Dependence (Heatherton, Kozlowski, Frecker, & Fagerström, Reference Heatherton, Kozlowski, Frecker and Fagerström1991) assessed the severity of alcohol and nicotine use, respectively. Current clinical status was assessed using Beck's Depression Inventory (BDI-II; Beck, Ward, Mendelson, Mock, & Erbaugh, Reference Beck, Ward, Mendelson, Mock and Erbaugh1961) and the State-Trait Anxiety Inventory (STAI; Julian, Reference Julian2011). Two forms of impulsivity separable from response inhibition were assessed with: (1) the Impulsive-Behaviors Scale (UPPS-P; Whiteside & Lynam, Reference Whiteside and Lynam2001) which contains five trait impulsivity subscales: lack of perseveration, lack of premeditation, sensation-seeking, negative urgency, and positive urgency; and (2) the Monetary Choice Questionnaire (MCQ; Kirby, Petry, & Bickel, Reference Kirby, Petry and Bickel1999) which measures delay discounting; the more quickly the reward loses value as a function of delay (represented by an increase in K value), the more impulsive the individual is considered.
Stimulation protocol
We delivered two off-line cPAS protocols. In both protocols, we first used neuronavigation for precise targeting registered to Montreal Neurological Institute (MNI) space (Brainsight; Rogue Research Inc., Montreal, Quebec, Canada). Resting motor threshold (RMT) was assessed via single TMS pulses to right M1, defined as the lowest intensity stimulation to effectively elicit responses from the first dorsal interosseous muscle of the participant's non-dominant hand (i.e. MEPs) as monitored by electromyograph (EMG) acquired through Signal software (Cambridge Electronics design, Cambridge, UK).
For cPAS, stimulation pulses were administered using two Magstim 2002 machines (Magstim Company Limited, Whitland, UK) via two 70-mm figure-of-eight air-film coils. Coil targets in the form of MNI coordinates were derived from a functional imaging meta-analysis of response inhibition (Cieslik, Mueller, Eickhoff, Langner, & Eickhoff, Reference Cieslik, Mueller, Eickhoff, Langner and Eickhoff2015). Coil 1 was situated over the rIFC (MNI target site [in mm]: x = 48, y = 16, z = 16) at a 20° angle to the coronal plane, and coil 2 was situated over the right pre-SMA (MNI target site [in mm]: x = 10, y = 10, z = 60) parallel to the midline when viewed axially (Fig. 1B). Pulse intensity was set to 120% of RMT. Both cPAS sessions comprised 100 pairs of stimuli at 0.2 Hz to achieve an 8.3-minute duration.
The two cPAS conditions varied in the ISI of the paired pulses: (1) rIFC stimulation preceded right pre-SMA stimulation by 4 ms (i.e. IFC + 4), and (2) rIFC stimulation preceded right pre-SMA stimulation by 100 ms (i.e. IFC + 100). The former served as the experimental condition based on evidence of modulatory effects on the cortico-subcortical response inhibition network in our previous studies (Kohl et al., Reference Kohl, Hannah, Rocchi, Nord, Rothwell and Voon2018; Mandali et al., Reference Mandali, Tsurumi, Popa and Voon2021). The latter served as a control condition, as the duration between paired pulses is presumed to be too protracted a period to facilitate either cortico-cortical (Buch et al., Reference Buch, Johnen, Nelissen, O'Shea and Rushworth2011) or cortico-subcortical conduction (Lu, Tsai, & Ziemann, Reference Lu, Tsai and Ziemann2012).
Experimental design
We used a single-blind between-subjects design to investigate the effects of cPAS targeting the response inhibition network in AUD and HC adults, consisting of two cPAS sessions delivered in randomized order at least 7-days apart (Fig. 2A). Prior to the cPAS intervention of the first session, participants completed the baseline self-report measures. Post-cPAS in both sessions, participants undertook the SST within the half-an-hour duration by which the cPAS intervention is purportedly active (Stefan et al., Reference Stefan, Kunesch, Cohen, Benecke and Classen2000).
Experimental design and results from cortical paired associative stimulation (cPAS) intervention on Stop Signal Task (SST) performance. (A) Experimental design. (B) Boxplot of mean difference of SST performance during the control (IFC + 100) condition and experimental (IFC + 4) condition in alcohol use disorder (AUD) and matched (N = 20) healthy control (HC) groups. The matched HC group, but not the AUD group, significantly improved SST performance in the experimental condition controlling for the variance of the control condition. (C) Boxplot of mean difference of SST performance during the control (IFC + 100) condition and experimental (IFC + 4) condition in alcohol use disorder (AUD) and larger (N = 35) healthy control (HC) groups. Performance of the larger HC group adhered to that of the matched HC group. p < 0.05*, p < 0.01**. Error bars denote standard error.
Statistical analysis
Statistical analyses were conducted in JASP (Version 0.16.0). First, all data were assessed for normality (Shapiro-Wilk test p > 0.05), homogeneity of variance (Levene's test p > 0.05), and outliers (>3 standard deviations from the group mean) to employ the appropriate t test for the variable type(s) according to statistical assumptions met.
Our primary variable of interest was mean SSRT (in milliseconds). Differences in within-group SSRT were assessed using paired samples t tests. Between-group differences were assessed by removing the variance of the control SSRT (IFC + 100) from experimental SSRT (IFC + 4) to perform independent samples t tests. We then used a one-way ANCOVA model to confirm between-group differences while controlling for variables informed by self-report measures. Concurrent nicotine use was also controlled for in this model given its observed effects on PAS LTP-like plasticity induction (Grundey et al., Reference Grundey, Thirugnanasambandam, Kaminsky, Drees, Skwirba, Lang and Nitsche2012; Thirugnanasambandam et al., Reference Thirugnanasambandam, Grundey, Adam, Drees, Skwirba, Lang and Nitsche2011). Next, we examined the proportion of those who improved SSRT in the IFC + 4 condition compared between groups using a stepwise logistic regression to compute odds ratio. Finally, Spearman correlations were performed to assess whether SST performance was related to independent measures of impulsivity (i.e. UPPS-P and MCQ).
We observed 20 participants per group was sufficient to demonstrate, under standard assumptions (80% power, alpha-level = 0.05), effect sizes of greater than or equal to 0.22 (partial eta squared) and 0.8 (Cohen's d) across significant results. Furthermore, our sample size was consistent with or exceeded the sample sizes of previous TMS studies in plasticity induction in alcohol use (Fuhl et al., Reference Fuhl, Müller-Dahlhaus, Lücke, Toennes and Ziemann2015; Loheswaran et al., Reference Loheswaran, Barr, Rajji, Blumberger, Le Foll and Daskalakis2016; Reference Loheswaran, Barr, Zomorrodi, Rajji, Blumberger, Foll and Daskalakis2017; Lücke et al., Reference Lücke, Heidegger, Röhner, Toennes, Krivanekova, Müller-Dahlhaus and Ziemann2014). All tests were two-tailed with significance assigned at p < 0.05, and Bonferroni corrected for multiple comparisons. Confidence intervals are provided for all statistically significant findings.
Results
Demographic and psychiatric factors
The matched HC group showed lower alcohol use frequency and severity as well as lesser severity of psychiatric factors depression, anxiety, and urgency impulsivity than the AUD group. The AUD group also more quickly discounted delayed rewards. Demographic and psychiatric characteristics for the AUD and matched HC (N = 20) groups are summarized in Table 1. Compared to the larger HC group (N = 35), the AUD group was significantly older and had undergone less years of formal education, while showing similar differences in psychiatric factors observed with the matched HCs. Full demographic and psychiatric characteristics for the larger HC group are summarized in online Supplemental Material Section 2.
Demographic and psychiatric factors between alcohol use disorder (AUD) and gender- and age-matched (N = 20) healthy control (HC) groups
Factor = type of demographic or psychiatric factor under evaluation (from top to bottom: age, gender [in proportion of females], years of education, alcohol use [AUDIT], nicotine use [Fagerström], depression [BDI-II], anxiety [STAI], impulsivity [UPPS-P] with subscales sensation-seeking [SS], negative urgency [NU], positive urgency [PU], lack of premeditation [LOPre], and lack of perseveration [LOPer], and delay discounting [MCQ, K value]); AUD mean (s.d.) = the mean and standard deviation in the AUD group by factor; HC mean (s.d.) = the mean and standard deviation in the HC group by factor; Test = type of t test used according to statistical assumptions met; Statistic = test statistic; Df = degrees of freedom; p-Value = significance level of test (asterisks by a p-value indicate statistical significance; p < 0.05*, p < 0.005**); CIs = confidence intervals of average (i.e. mean or median) group difference for each statistically significant finding
RMT in AUD and HC groups
The average RMT of the AUD (IFC + 4: 44.5 ± 6.65, IFC + 100: 45.2 ± 7.12) group did not significantly differ from matched HC (IFC + 4: 42 ± 6.74, IFC + 100: 42.5 ± 5.41) group (p > 0.05).
SST practice effects and cPAS plasticity induction carryover
SST practice effects were indexed by comparing mean SSRT as a function of testing session order. Participants who underwent IFC + 4 cPAS during their first testing session (N = 27; 174 ± 64.4) did not significantly differ in mean IFC + 4 SSRT than those who underwent IFC + 4 cPAS during their second testing session (N = 28; 158.5 ± 45.4) (p > 0.05). Additionally, no significant differences were observed in mean IFC + 100 SSRT between those who underwent IFC + 4 cPAS during their first (157.2 ± 34.7) v. their second (186.3 ± 65.3) testing session (p > 0.05), further indicating adequate wash out of the plasticity enhancing effects of IFC + 4 cPAS on SST performance in the control condition.
cPAS effects on SST performance in AUD and healthy controls
The matched HC group (N = 20) showed faster average SSRT in the IFC + 4 condition (152.2 ± 58.4) compared to the IFC + 100 condition (173.2 ± 48.4), t(19) = −3.01, p = 0.007, CIs:−35.6 to −6.42. Conversely, in the AUD group, there was no significant difference in SST performance between IFC + 4 (184.4 ± 61.7) and IFC + 100 (169.9 ± 64.2) conditions (p > 0.05). Notably, the HC and AUD groups did not differ in control SSRT as assessed by the IFC + 100 condition (p > 0.05). However, the HC group (−21 ± 31.2) as compared to the AUD group (14.5 ± 54.7) showed a significant facilitation in SST performance in the IFC + 4 condition, t(38) = −2.52, p = 0.02, CIs: −64.01 to −7; Figure 2B).
Within- and between-group SST performance of the larger HC group (N = 35) adhered to that of the matched HCs (Fig. 2C; data found in online Supplementary Materials Section 3), while the abstinent AUD (N = 15) performance adhered to that of the full AUD (N = 20) group (online Supplementary Materials Section 4.1).
ANCOVA model of SST performance controlling for psychiatric factors
We used a one-way ANCOVA model with mean SSRT difference score as our dependent variable and group (i.e. AUD [N = 20] v. HC [N = 20]) and testing session order (i.e. IFC + 4 first [N = 20] v. IFC + 4 s [N = 20]) as fixed factors, controlling for nicotine use severity, depression, anxiety, and negative and positive urgency impulsivity as covariates of no interest. No covariate was correlated to another greater than rho = .67. We showed a significant main effect of group (F(1,31) = 9.57, p = 0.004, CIs: −68.64 to −14.11), with an insignificant group*testing order interaction and no covariates related to SSRT difference score (p > 0.05). These results were confirmed in two separate ANCOVAs, using: (1) the abstinent AUD group (N = 15) with the same covariates (F(1,26) = 7.67, p = 0.01, CIs: −94.64 to −13.99), and (2) the larger HC group (N = 35) controlling also for age and years of education (F(1,44) = 11.84, p = 0.001, CIs: 18.64–71.37).
Logistic regression of SSRT improvement in IFC + 4 cPAS
Fourteen of 20 (70%) HC compared to 5 of 20 (25%) AUD improved SST performance in the experimental condition. A stepwise logistic regression model using observed SSRT performance improvement in the IFC + 4 condition (i.e. yes = 1/no = 0) as a dependent variable and group as a fixed factor was statistically significant (X 2 [38, N = 40] = 8.42, p = 0.004); this model explained 25.3% (Nagelkerke R2) of the variance in SSRT improvement between-groups and correctly classified 72.5% of cases. Overall, HC group designation was associated with more than twice higher likelihood of SSRT improvement in the IFC + 4 cPAS condition (HCs: 72% v. AUD: 13.6%; odds ratio: 2.33; p = 0.006; CIs: −3.34 to −0.55, Fig. 3). A separate stepwise logistic regression analysis performed using the abstinent AUD group (N = 15) produced comparable results (model summary found in online Supplemental Materials Section 4.2).
Proportion of those improved in stop signal reaction time (SSRT) in healthy control (HC- left) and alcohol use disorder (AUD- right) groups after cortical paired associative stimulation (cPAS) intervention. Solid black lines traced from solid black dots indicate an improvement (and the slope indicates the extent of the improvement) in SSRT in the IFC + 4 cPAS condition, while dotted black lines traced from open dots indicate an impairment (and the slope indicates the extent of the impairment) in SSRT in the IFC + 4 cPAS condition. HC group designation was associated with significant improvement in SSRT in the IFC + 4 cPAS condition, with a majority of HCs showing this effect.
SST performance relationships to impulsivity measures
SSRT in the IFC + 4 and IFC + 100 conditions, as well as the SSRT difference score between these conditions, were unrelated to any UPPS-P subscale or MCQ scores (all − 0.12 < rho <0.24 and p > 0.22).
Discussion
Deficits in inhibitory control are commonly observed in addiction disorders (Domínguez-Salas et al., Reference Domínguez-Salas, Díaz-Batanero, Lozano-Rojas and Verdejo-García2016; Verdejo-García et al., Reference Verdejo-García, Lawrence and Clark2008), with impaired performance in Go/No-Go and SST (Chambers et al., Reference Chambers, Garavan and Bellgrove2009). Under these paradigms, convergent multimodal evidence has implicated fronto-striatal circuitry, including two PFC subregions − the rIFC and pre-SMA − which extend to the STN; this ‘hyperdirect’ network is believed to be essential for the successful application of fast, reactive stopping behaviors (Aron, Reference Aron2007, Reference Aron2011). The specific contributions of these areas to inhibitory control have been theorized; it is suggested that the rIFC is crucial for salience detection of the stopping cue (Cai et al., Reference Cai, Chen, Ide, Li and Menon2017), and the pre-SMA for action monitoring (Bonini et al., Reference Bonini, Burle, Liégeois-Chauvel, Régis, Chauvel and Vidal2014), with rIFC expediting the stop signal prior to the pre-SMA (Duann, Ide, Luo, & Li, Reference Duann, Ide, Luo and Li2009). The STN − which integrates this critical input from the PFC (Aron, Reference Aron2007; Obeso et al., Reference Obeso, Wilkinson, Teo, Talelli, Rothwell and Jahanshahi2017) − then performs response selection to cancel the prepotent action (Bastin et al., Reference Bastin, Polosan, Benis, Goetz, Bhattacharjee, Piallat and David2014; Kühn et al., Reference Kühn, Brücke, Hübl, Schneider, Kupsch, Eusebio and Brown2008).
Here, we used cPAS − a repetitive paired-pulse TMS protocol purported to induce cortical synaptic plasticity (Stefan et al., Reference Stefan, Kunesch, Cohen, Benecke and Classen2000) − to strengthen the efficiency of this inhibitory network in AUD and HC adults, and compared group differences in responsivity on the SST. We showed that, post-cPAS, HC adults demonstrated an overall decrease in SSRT − the key measure of inhibitory control (Logan et al., Reference Logan, Zandt, Verbruggen and Wagenmakers2014) − thereby replicating findings of two studies from our research group (Kohl et al., Reference Kohl, Hannah, Rocchi, Nord, Rothwell and Voon2018; Mandali et al., Reference Mandali, Tsurumi, Popa and Voon2021). It was posited that rIFC stimulation delivered 4 ms prior to pre-SMA stimulation had primed the pre-SMA to STN connection critical to reactive stopping (Obeso et al., Reference Obeso, Wilkinson, Teo, Talelli, Rothwell and Jahanshahi2017), thereby improving response inhibition (Kohl et al., Reference Kohl, Hannah, Rocchi, Nord, Rothwell and Voon2018). Importantly, inhibitory control facilitated by this protocol in our prior studies appeared to be specific to older (⩾30 years) adults, in line with findings that pre-SMA to STN anatomical connectivity strength more robustly predicts SST performance as age increases (Coxon, Impe, Wenderoth, & Swinnen, Reference Coxon, Impe, Wenderoth and Swinnen2012).
Conversely, AUD adults matched for age failed to show similar levels of improvement in SST performance after the same cPAS intervention, with a small minority of AUD showing decreased SSRT post-cPAS. These results may reflect the progressively detrimental effects of chronic alcohol use on PFC circuitry (for a review, see Moselhy, Georgiou, & Kahn (Reference Moselhy, Georgiou and Kahn2001)). In AUD, diminished recruitment of the PFC has been observed in a range of executive functioning tasks (Mann, Günther, Stetter, & Ackermann, Reference Mann, Günther, Stetter and Ackermann1999), with weaker functional connectivity between the PFC and striatum shown during SST performance (Courtney, Ghahremani, & Ray, Reference Courtney, Ghahremani and Ray2013). AUD adults also present regional atrophy (Cardenas et al., Reference Cardenas, Durazzo, Gazdzinski, Mon, Studholme and Meyerhoff2011; Chanraud et al., Reference Chanraud, Martelli, Delain, Kostogianni, Douaud, Aubin and Martinot2007; Makris et al., Reference Makris, Oscar-Berman, Kim, Hodge, Kennedy, Caviness and Harris2008), as well as abnormalities in frontal lobular blood flow (Noël et al., Reference Noël, Sferrazza, Van der Linden, Paternot, Verhas, Hanak and Verbanck2002), associated with the development and perseveration of alcohol-seeking behaviors (Goldstein & Volkow, Reference Goldstein and Volkow2002), and probability of relapse (Noël et al., Reference Noël, Sferrazza, Van der Linden, Paternot, Verhas, Hanak and Verbanck2002).
Alcohol-related PFC changes at the macroscopic level may be ascribed to the effects of alcohol on the efficacy of synaptic transmission and function (Koob & Volkow, Reference Koob and Volkow2016). In rodent models, acute administration promotes inhibitory processes (Abrahao et al., Reference Abrahao, Salinas and Lovinger2017), while chronic use leads to a compensatory increase in global cortical excitability via a reduction of GABAergic and concomitant upregulation of glutamatergic transmission and NMDA-receptor release (Kalivas, Reference Kalivas2009). Studies in humans applying TMS to index these mechanisms of altered regional cortical excitability during acute (Conte et al., Reference Conte, Attilia, Gilio, Iacovelli, Frasca, Bettolo and Inghilleri2008; Kähkönen et al., Reference Kähkönen, Kesäniemi, Nikouline, Karhu, Ollikainen, Holi and Ilmoniemi2001, Reference Kähkönen, Wilenius, Nikulin, Ollikainen and Ilmoniemi2003; Ziemann et al., Reference Ziemann, Lönnecker and Paulus1995) and after recurrent (Conte et al., Reference Conte, Attilia, Gilio, Iacovelli, Frasca, Bettolo and Inghilleri2008; Naim-Feil et al., Reference Naim-Feil, Bradshaw, Rogasch, Daskalakis, Sheppard, Lubman and Fitzgerald2016; Nardone et al., Reference Nardone, Bergmann, Kronbichler, Caleri, Lochner, Tezzon and Golaszewski2010; Quoilin et al., Reference Quoilin, Wilhelm, Maurage, de Timary and Duque2018) use have provided support for this neuromolecular data. Thus, even during prolonged periods of abstinence (i.e. post-withdrawal), it appears these widespread neuroadaptations in cortical excitability persist in an allostatic manner (Koob, Reference Koob2009), which may result in enduring changes to experience-dependent synaptic plasticity processes including LTP and LTD (Koob & Volkow, Reference Koob and Volkow2016).
The well-documented effects of repeated alcohol exposure on glutamatergic transmission (particularly its actions on NMDA-receptors) in the PFC, and its role in synaptic functioning (Kalivas, Reference Kalivas2009), may provide insight into the current findings. For instance, pathologically high levels of extracellular glutamate may result in excitotoxic changes in morphology such as aberrant synaptic pruning (Ferrer, Galofro, Fobregues, & Lopez-Tejero, Reference Ferrer, Galofro, Fobregues and Lopez-Tejero1989), myelin reduction (Pfefferbaum & Sullivan, Reference Pfefferbaum and Sullivan2005), and selective cellular loss (Chandler, Sutton, Norwood, Sumners, & Crews, Reference Chandler, Sutton, Norwood, Sumners and Crews1997) within PFC circuits; all factors which contribute to the efficacy of neuronal communication. Likewise, it is plausible − given that NMDA receptor activation interferes with LTP (Huang, Colino, Selig, & Malenka, Reference Huang, Colino, Selig and Malenka1992), and that acute alcohol administration reduces short-term LTP-like plasticity in the human cortex (Loheswaran et al., Reference Loheswaran, Barr, Rajji, Blumberger, Le Foll and Daskalakis2016; Reference Loheswaran, Barr, Zomorrodi, Rajji, Blumberger, Foll and Daskalakis2017; Lücke et al., Reference Lücke, Heidegger, Röhner, Toennes, Krivanekova, Müller-Dahlhaus and Ziemann2014) − that a continued state of glutamatergic upregulation may have rendered those with AUD less receptive to stimulation interventions aimed at plasticity induction relative to their HC counterparts (Chiamulera, Piva, & Abraham, Reference Chiamulera, Piva and Abraham2021). Our findings, then, suggest that pre-existing alcohol-related neuroadaptations on the neurotransmitter level (especially that of glutamatergic dysregulation) may have weakened the capacity for cPAS to enhance inhibitory control − a cognitive process highly relevant to the trajectory of AUD (Groman et al., Reference Groman, James and Jentsch2009; Wilcox et al., Reference Wilcox, Dekonenko, Mayer, Bogenschutz and Turner2014).
The AUD group demonstrated higher levels of trait impulsivity compared to HCs as measured by the UPPS-P; these elevations were specific to the positive and negative urgency subscales, which assess the tendency to act rashly in an intensified euphoric or aversive emotional state, respectively (Cyders & Smith, Reference Cyders and Smith2007). AUD also showed steeper delay discounting on the MCQ. Both decisional impulsivity and mood-based impulsive personality traits are strongly associated with addiction disorders (Amlung, Vedelago, Acker, Balodis, & MacKillop, Reference Amlung, Vedelago, Acker, Balodis and MacKillop2017; Zorrilla & Koob, Reference Zorrilla and Koob2019), yet map onto distinct neural circuitry from that involved in response inhibition (Voon & Dalley, Reference Voon and Dalley2016). Thus, that UPPS-P and MCQ scores were unrelated to SST performance in both control and excitation conditions for either group underscores the dissociability of impulsivity subtypes (MacKillop et al., Reference MacKillop, Weafer, Gray, Oshri, Palmer and de Wit2016), as well as the target specificity of our cPAS intervention (Kohl et al., Reference Kohl, Hannah, Rocchi, Nord, Rothwell and Voon2018).
Trait and decisional impulsivity are relatively stable characteristics, while motor disinhibition can fluctuate naturally and improve with cognitive training (Houben, Nederkoorn, Wiers, & Jansen, Reference Houben, Nederkoorn, Wiers and Jansen2011; Jones, Christiansen, Nederkoorn, Houben, & Field, Reference Jones, Christiansen, Nederkoorn, Houben and Field2013). Furthermore, neuroadaptive changes at glutamatergic synapses which appear to diminish responsivity to plasticity induction can potentially be counteracted with pharmacological agents (Chiamulera et al., Reference Chiamulera, Piva and Abraham2021), such as NMDA channel blocker, ketamine (Fattore, Piva, Zanda, Fumagalli, & Chiamulera, Reference Fattore, Piva, Zanda, Fumagalli and Chiamulera2018). Such agents − which are purported to amplify the malleability of neuronal circuits (Chiamulera et al., Reference Chiamulera, Piva and Abraham2021)− have been combined with TMS protocols to treat symptoms of neuropsychiatric disorders in intractable states (Best, Pavel, & Haustrup, Reference Best, Pavel and Haustrup2019; Pradhan, Parikh, Makani, & Sahoo, Reference Pradhan, Parikh, Makani and Sahoo2015). Preliminary evidence indicates that improved response inhibition attenuates alcohol cue-induced craving (Papachristou et al., Reference Papachristou, Nederkoorn, Havermans, Bongers, Beunen and Jansen2013) and intake (Houben et al., Reference Houben, Nederkoorn, Wiers and Jansen2011); thus, future research is required to delineate the role of noninvasive neuromodulation − possibly paired with pharmacological agents which may increase responsivity to intervention (Fattore et al., Reference Fattore, Piva, Zanda, Fumagalli and Chiamulera2018)− to enhance fronto-striatal integrity underlying inhibitory control and, in turn, mitigate adverse alcohol-related outcomes in AUD.
Limitations and future directions
This study was not without limitations which future research can overcome. First, we obtained a convenience sample of AUD patients from the hepatology clinic with varying stages of sobriety as well as self-reported polysubstance non-use. A subsample analysis confirms that our findings apply to abstinent AUD without the confounding effect of ongoing alcohol consumption. However, as duration of abstinence is related to executive functioning recovery (Kopera et al., Reference Kopera, Wojnar, Brower, Glass, Nowosad, Gmaj and Szelenberger2012; Moselhy et al., Reference Moselhy, Georgiou and Kahn2001), and under-reporting biases associated with self-reported illicit drug use (Macleod, Hickman, & Smith, Reference Macleod, Hickman and Smith2005), larger-scale neuromodulation studies in AUD may include a detailed abstinence timeline screening procedure in addition to a physiological index of alcohol (e.g. phosphatidylethanol) and illicit drug use status for both patient and HC groups as a means of participant exclusion, post hoc statistical covariance, or subgrouping (e.g. short- v. long-term abstinence) to interrogate factors linked to interindividual variation in responsivity to intervention.
Next, the proportion of males in our AUD sample was greater than females; reflective of the well-established trend of males more likely engaging in problematic alcohol use behaviors (Slade et al., Reference Slade, Chapman, Swift, Keyes, Tonks and Teesson2016). Given further evidence of gender-specific differences in cortical reactivity to alcohol administration (Hoppenbrouwers, Hofman, & Schutter, Reference Hoppenbrouwers, Hofman and Schutter2010), the results of our cPAS intervention are perhaps more representative of the variation within the AUD population than would be expected from a gender-balanced sample. However, recent longitudinal meta-analyses have demonstrated a marked decrease in the male-female alcohol use gap (Slade et al., Reference Slade, Chapman, Swift, Keyes, Tonks and Teesson2016). Thus, gendered cohort studies comparing differential capacity for plasticity induction in chronic alcohol are indicated.
Further, depressive and anxious symptoms were more pronounced in AUD compared to HC groups − a psychiatric phenomenon commonly observed in AUD populations (Lai, Cleary, Sitharthan, & Hunt, Reference Lai, Cleary, Sitharthan and Hunt2015). To address this discrepancy, we excluded participants who met the criteria for clinical depression (⩾23) or anxiety (>39); this ensured mean BDI-II and STAI scores in our AUD sample were well within the non-pathological range (Beck et al., Reference Beck, Ward, Mendelson, Mock and Erbaugh1961; Julian, Reference Julian2011). Additionally, all ANCOVA models included BDI-II and STAI scores as covariates, showing neither main nor interactive effects of these potential confounds on SST performance between groups.
Finally, we postulated that the efficacy of cPAS in improving response inhibition is derived from a priming effect of rIFC stimulation on plasticity at the pre-SMA to STN connection. However, the precise mechanisms underlying cPAS, which appear to modulate relationships among these substrates within the inhibitory network, remain to be delineated. Our study did not incorporate physiological TMS measures of intracortical excitation and inhibition or contemporaneous functional imaging, which may provide insight into the processes contributing to differential changes in SST performance. Thus, further studies are required to confirm our preliminary behavioral findings and may integrate TMS or functional imaging paradigms to extend findings mechanistically.
Conclusion
We used cPAS to modulate the rIFC and pre-SMA to STN hyperdirect pathway and modify SST performance in AUD. Our results are two-fold. First, conferring further validity to our novel stimulation method, we replicate previous findings that cPAS targeting the inhibitory network decreases SSRT in healthy adults − presumably by priming the pre-SMA to STN function crucial to the successful execution of stopping behaviors. Second, AUD patients failed to show SST improvement with the same intervention; this may reflect altered cortical excitability resulting from widespread neuroadaptations associated with problematic forms of alcohol use. Thus, we identify a potential marker of impairment with direct implications to a disorder-relevant cognitive process underlying AUD. Further research is required to confirm our preliminary findings, as well as expand the role of noninvasive neuromodulation − potentially paired with pharmacological agents for increased intervention responsivity − in strengthening fronto-striatal networks implicated in inhibitory control in addicted populations.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291723002374.
Data availability
Deidentified participant data is available on reasonable request from the corresponding author.
Acknowledgements
We thank our participants for volunteering, and Liver Disease Clinical Nurse Specialist Aileen Inte for her support in the recruitment of AUD patients for the study.
Funding statement
All research at the Department of Psychiatry in the University of Cambridge is supported by the NIHR Cambridge Biomedical Research Centre (BRC-1215-20014) and NIHR Applied Research Centre. The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care. Valerie Voon is supported by a Medical Research Council Senior Clinical Fellowship (grant number: MR/W020408/1).
Competing interest
None.
Open access
