Introduction
Addiction is characterized by the obsession with addictive substances or behaviors despite harmful consequences, and the exclusion of other activities (Campbell, Reference Campbell2003). It can be broadly divided into two categories: substance dependence (SD) and behavioral addiction (American Psychiatric Association, 2013). SD is characterized by problematic substance use (e.g. alcohol, cannabis and cocaine), which results from repeated drug administrations and leads to physical disturbances when the substance is withdrawn (NIDA, 2019). Gaming disorder is one behavioral addiction classified as a medical illness in International Classification of Diseases (World Health Organization, 2019). It refers to the persistent involvement with video games and the inability to reduce or quit gaming. Addiction exerts serious negative impacts on people's physical health (Degenhardt et al., Reference Degenhardt, Charlson, Ferrari, Santomauro, Erskine, Mantilla-Herrara and Griswold2018; Evren, Evren, Dalbudak, Topcu, & Kutlu, Reference Evren, Evren, Dalbudak, Topcu and Kutlu2020) and mental well-being (Evren et al., Reference Evren, Evren, Dalbudak, Topcu and Kutlu2020; Kaptsis, King, Delfabbro, & Gradisar, Reference Kaptsis, King, Delfabbro and Gradisar2016). A sound understanding of the corresponding pathophysiology is vital to develop effective interventions and treatments for addiction.
A significant body of research has revealed disturbances in response inhibition in SD (Brewer & Potenza, Reference Brewer and Potenza2008; Zilverstand, Huang, Alia-Klein, & Goldstein, Reference Zilverstand, Huang, Alia-Klein and Goldstein2018) and behavioral addiction (Argyriou, Davison, & Lee, Reference Argyriou, Davison and Lee2017; Moccia et al., Reference Moccia, Pettorruso, De Crescenzo, De Risio, Di Nuzzo, Martinotti and Di Nicola2017). Response inhibition is a core sub-process of cognitive control and is one of the more extensively studied components of cognitive control in healthy populations (Zhang, Geng, & Lee, Reference Zhang, Geng and Lee2017) and individuals with addiction (Smith, Mattick, Jamadar, & Iredale, Reference Smith, Mattick, Jamadar and Iredale2014). Response inhibition refers to the ability to withhold a prepotent motor response (Chambers, Garavan, & Bellgrove, Reference Chambers, Garavan and Bellgrove2009; Nigg, Reference Nigg2000), and is often assessed using paradigms including the Go/No-go task, the stop-signal task and the Stroop task (Meule, Reference Meule2017; Stahl et al., Reference Stahl, Voss, Schmitz, Nuszbaum, Tüscher, Lieb and Klauer2014; Verbruggen & Logan, Reference Verbruggen and Logan2008). A common process underlying these tasks is that participants are required to selectively respond to target stimuli while ignoring distracting stimuli. For example, in a Stroop task (Stroop, Reference Stroop1935), participants need to report the color a word is presented in while avoiding reading the color the word describes. In a Go/No-go task or a stop-signal task (Donders, Reference Donders1969; Logan, Reference Logan, Dagenbach and Carr1994), participants are required to respond to certain stimuli (e.g. ‘K’) and suppress a response when other stimuli are presented (e.g. ‘X’). For SD, it is hypothesized that chronic intake of drugs, stimulants in particular, may damage the dopaminergic prefrontal-subcortical pathways, which are crucial for successful behavioral inhibition (Smith et al., Reference Smith, Mattick, Jamadar and Iredale2014). Meanwhile, deficits in response inhibition may predate or exacerbate substance use by making it difficult for people to abstain from drug administrations (Moeller, Bederson, Alia-Klein, & Goldstein, Reference Moeller, Bederson, Alia-Klein and Goldstein2016). For behavioral addiction, impairments in response inhibition are often associated with poor self-regulation and high impulsivity (Argyriou et al., Reference Argyriou, Davison and Lee2017), which may lead to problematic gaming. Considering the validated deficits in response inhibition in addiction, it is necessary to reveal the neural pathophysiology underlying impaired response inhibition.
In the past two decades, neuroimaging techniques, especially functional magnetic resonance imaging (fMRI), have been widely used to reveal the altered neural activity during response inhibition in individuals with addiction. The neural basis underlying intact response inhibition involves a wide range of brain regions. For example, the fronto-parietal network (FPN) and the ventral attention network (VAN) are two core neural systems in response inhibition (Zhang et al., Reference Zhang, Geng and Lee2017), which are crucial for attention, working memory and goal-directed response selections (Nee, Wager, & Jonides, Reference Nee, Wager and Jonides2007; Simmonds, Pekar, & Mostofsky, Reference Simmonds, Pekar and Mostofsky2008). Meanwhile, communications among cortical and subcortical areas as well as the cerebellum were shown essential for successful inhibition (Rae, Hughes, Anderson, & Rowe, Reference Rae, Hughes, Anderson and Rowe2015). Luijten and colleagues (Reference Luijten, Machielsen, Veltman, Hester, de Haan and Franken2014) reviewed and summarized results from various studies on the neural pathophysiology of impaired response inhibition in SD and they found an overall pattern of hypoactivities in the FPN and VAN in individuals with SD (see also Moeller et al. Reference Moeller, Bederson, Alia-Klein and Goldstein2016; Morein-Zamir & Robbins, Reference Morein-Zamir and Robbins2015; Zilverstand et al. Reference Zilverstand, Huang, Alia-Klein and Goldstein2018), especially in response to non-addiction-related stimuli (Zilverstand et al., Reference Zilverstand, Huang, Alia-Klein and Goldstein2018). For example, most studies found hypoactivities in the inferior frontal gyrus (IFG), anterior cingulate cortex and dorsolateral prefrontal cortex in individuals with SD during inhibitory control (Luijten et al., Reference Luijten, Machielsen, Veltman, Hester, de Haan and Franken2014). Others found hypoactivities in the occipital lobe (Li et al., Reference Li, Huang, Yan, Bhagwagar, Milivojevic and Sinha2008) and the insula (Fu et al., Reference Fu, Bi, Zou, Wang, Ye, Ma and Yang2008). However, several researchers reported hyperactivations of regions within the FPN (Hester, Nestor, & Garavan, Reference Hester, Nestor and Garavan2009) and VAN (Luijten et al., Reference Luijten, Veltman, Hester, Smits, Nijs, Pepplinkhuizen and Franken2013). For behavioral addiction, there has not yet been a review summarizing the results on the neural abnormalities during response inhibition, but previous studies have reported mixed results. Some researchers observed greater activation in the FPN (e.g. Ding et al. Reference Ding, Sun, Sun, Chen, Zhou, Zhuang and Du2014; Dong, DeVito, Du, & Cui, Reference Dong, DeVito, Du and Cui2012) and the fronto-striatal pathway (Ko et al., Reference Ko, Hsieh, Chen, Yen, Chen, Yen and Liu2014) in individuals with behavioral addiction during response inhibition while others reported lower activity of these regions (e.g. De Ruiter, Oosterlaan, Veltman, Van Den Brink, & Goudriaan, Reference De Ruiter, Oosterlaan, Veltman, Van Den Brink and Goudriaan2012; Wang et al. Reference Wang, Hu, Xu, Zhou, Lin, Du and Dong2017). The inconsistencies between hypo- and hyper-activities in previous literature may be due to several factors. Abstinence or treatment status (e.g. Moeller et al., Reference Moeller, Tomasi, Woicik, Maloney, Alia-Klein, Honorio and Sinha2012), addiction duration (e.g. Claus, Feldstein Ewing, Filbey, and Hutchison, Reference Claus, Feldstein Ewing, Filbey and Hutchison2013) and the substance of abuse of individuals with SD may modulate response inhibition. The experimental tasks and stimuli (e.g. addiction-related v. addiction-unrelated stimuli) may also influence task performance and/or the corresponding neural activity in individuals with addiction (Luijten et al., Reference Luijten, Machielsen, Veltman, Hester, de Haan and Franken2014; Moeller et al., Reference Moeller, Bederson, Alia-Klein and Goldstein2016).
Considering the inconsistencies in the altered neural activity during response inhibition in addiction, a quantitative meta-analysis is needed to unravel the conflicting results. This study aimed to use meta-analysis to reveal consistent neural alterations in response inhibition in adults with addiction. Because individuals with addiction are characterized with excessive use of addictive substances or engagement in behavioral addiction, which are behavioral manifestations of impaired response inhibition (Argyriou et al., Reference Argyriou, Davison and Lee2017; Smith et al., Reference Smith, Mattick, Jamadar and Iredale2014), we hypothesized that individuals with all addictions would show reduced activity in regions in the FPN and VAN, core neural systems for successful response inhibition. Additionally, because SD and behavioral addiction are likely distinct in nature, it is reasonable to expect different patterns of neural abnormalities between these two addiction subtypes.
Method
Study selection
We searched Scopus, PubMed and Web of Science for articles published in English before 15 November 2020, using the following terms and their derivatives: ‘functional magnetic resonance imaging’ OR ‘fMRI’; AND ‘addiction’ OR ‘drug use’ OR ‘drug addiction’ OR ‘substance addiction’ OR ‘substance dependence’ OR ‘cocaine’ OR ‘marijuana’ OR ‘cannabis’ OR ‘thc’ OR ‘methamphetamine’ OR ‘amphetamine’ OR ‘ecstasy’ OR ‘mdma’ OR ‘heroin’ OR ‘opiate’ OR ‘polysubstance’ OR ‘nicotine dependence’ OR ‘alcohol abuse’ OR ‘alcohol dependence’ OR ‘alcohol addiction’ OR ‘nicotine addiction’ OR ‘gambling’ OR ‘gamblers’ OR ‘gaming addiction’ OR ‘gaming disorder’; AND ‘response inhibition’ OR ‘inhibitory control’ OR ‘interference resolution’ OR ‘action withholding’ OR ‘action cancellation’ OR ‘stop signal’ OR ‘go nogo’ OR ‘countermanding’. The reference lists of relevant review articles were also examined to include additional papers.
We included a study if it: (1) was published in English, in a peer-reviewed journal; (2) used fMRI; (3) compared neural activation between adult human healthy controls (HCs) and adult human participants with SD, gambling disorder or gaming disorder; (4) used tasks that required participants to inhibit prepotent responses and (5) conducted whole-brain analyses in the form of three-dimensional coordinates in standard stereotactic coordinate space (i.e. Talairach or Montreal Neurological Institute).
We excluded a study if it: (1) was conducted in non-human or non-adult participants; (2) included comorbid participants; (3) did not include a HCs group; (4) included occasional users (e.g. occasional smokers) for the addiction group and/or the control group; (5) used the same patient data as other included studies; (6) was a connectivity study or a diffusion tensor imaging study; (7) did not investigate task-based neural activation (e.g. resting-state fMRI study); (8) did not conduct comparisons between the addiction group and HCs and (9) only included ROI findings. Reviews and meta-analytic studies were also excluded.
Quality assessment of each study included was conducted with a 9-point checklist (online Supplementary Table S1). The current study was performed according to the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines (Stroup et al., Reference Stroup, Berlin, Morton, Olkin, Williamson, Rennie and Thacker2000). See Fig. 1 for the PRISMA flow diagram on the study selection for this meta-analysis.
Fig. 1. PRISMA flow diagram of study selection.
Data analysis
Voxel-wise meta-analysis
We used the signed differential mapping (SDM) software package (version 5.15 for Windows; http://www.sdmproject.com/software) to perform meta-analyses on the different neural activation patterns for people with addiction and HCs. The SDM method allows the combination of statistical parametric maps and peak coordinates originally reported in individual studies (for reviews see Radua et al., Reference Radua, Rubia, Canales, Pomarol-Clotet, Fusar-Poli and Mataix-Cols2014b; Radua & Mataix-Cols, Reference Radua and Mataix-Cols2009). Briefly, we first extracted peak coordinates and effect sizes (i.e. t values) of different patterns of brain activity between addiction groups and HCs from each individual study. Note that z scores reported as effect sizes were converted to t values using an online converter (http://www.sdmproject.com/utilities/?show=Statistics). Second, a standard MNI map of the activation differences was re-created by applying an anisotropic Gaussian kernel for each included study. We used the anisotropic kernel in order to improve the plausibility of the maps by allocating different values to distinct voxels of a peak contingent on relevant spatial correlations and used an isotropic full-width at half-maximum = 20 mm for smoothing to control false-positive results (Radua et al., Reference Radua, Mataix-Cols, Phillips, El-Hage, Kronhaus, Cardoner and Surguladze2012, Reference Radua, Rubia, Canales, Pomarol-Clotet, Fusar-Poli and Mataix-Cols2014b). Third, we applied a random-effects general linear model to create the mean map after accommodating the effect size maps. Consequently, the included studies were weighted differentially based on their sample sizes, and between-study heterogeneities and intra-study variances, amplifying the contributions of the studies with larger sample size or smaller variance (Radua & Mataix-Cols, Reference Radua and Mataix-Cols2012).
We conducted meta-analytic comparisons between all addictions and HCs, contrasting conditions where response inhibition was entailed or successful, to conditions where response inhibition was not needed or unsuccessful, from tasks including Stroop tasks, Go/No-go tasks and stop-signal tasks. Only contrasts using neutral or non-addiction-related stimuli were used (for more information, see online Supplementary materials). We calculated the differences between the two groups for each voxel and extracted the statistical significance results from a standard randomization test (Radua, van den Heuvel, Surguladze, & Mataix-Cols, Reference Radua, van den Heuvel, Surguladze and Mataix-Cols2010; Tang et al., Reference Tang, Lu, Zhang, Hu, Bu, Li and Gong2018). The SDM kernel size and thresholds used in this meta-analysis were p < 0.005 with peak height Z > 1 and a cluster size of larger than 10 voxels, which have been validated to optimize sensitivity while correctly controlling false-positive rate in the empirical validation of SDM (Radua et al., Reference Radua, Mataix-Cols, Phillips, El-Hage, Kronhaus, Cardoner and Surguladze2012).
Jackknife sensitivity analysis
We assessed the replicability of the results by conducting a systematic whole-brain voxel-based jackknife sensitivity analysis. This was accomplished through the repetition of main statistical analysis while removing one study each time (Radua & Mataix-Cols, Reference Radua and Mataix-Cols2009).
Analyses of heterogeneity and publication bias
We performed a heterogeneity analysis with Q statistic maps to investigate between-study variability left unexplained (Radua & Mataix-Cols, Reference Radua and Mataix-Cols2012). Additionally, we performed the Egger's test to look for potential publication bias in these findings (Radua et al., Reference Radua, Grau, Van Den Heuvel, De Schotten, Stein, Canales-Rodríguez and Mataix-Cols2014a).
Meta-regression analysis
Two meta-regressions were conducted, regressing the abnormal neural activity on addiction duration and abstinence days, respectively. To reduce spurious results, we used a more conservative threshold (p < 0.0005) and only considered clusters showing a significant slope in addition to a significant difference with HCs at one of the extremes (Radua & Mataix-Cols, Reference Radua and Mataix-Cols2009).
Additionally, we repeated all analyses above for the SD subgroup, but not for behavioral addiction subgroup due to limited number of studies.
Results
The literature search yielded 391 publications in the databases. Based on our inclusion and exclusion criteria (see online Supplementary materials for a detailed description), 23 studies reporting 23 datasets were ultimately identified in the current meta-analysis, including 20 SD datasets (comprising 479 substance users and 456 matched HCs) and three gaming disorder datasets (comprising 38 gamers and 38 matched HCs). The demographic and clinical characteristics of the included studies are shown in Table 1. The quality score of each study and other information including experimental paradigms and image acquisition techniques are presented in the online Supplementary Tables S1–S3.
Table 1. Demographic and clinical characteristics of the dataset included in this meta-analysis
SD, substance dependence; HCs, healthy controls; NA, not available.
Consistent with our hypotheses, all addictions showed hypoactivity in regions within the FPN and VAN. Specifically, compared with HCs, all addictions showed hypoactivity in the right insula (BAs 38, 47, 48), right middle temporal gyrus (MTG; BAs 20, 21, 22), right temporal pole (BAs 20, 21, 38, 48), right IFG (orbital part, BAs 38, 47) and right supramarginal gyrus (BAs 2, 40, 48). Additionally, all addictions exhibited significant hyperactivities in the left cerebellum (hemispheric lobule VIIB). Detailed results are presented in Table 2 and Fig. 2a.
Fig. 2. Meta-analyses results regarding regional differences of task-evoked activation between (a) all addictions and HCs during response inhibition, (b) SD subgroup and HCs during response inhibition. Areas with hypo-activity are displayed in blue, and areas with hyper-activity are displayed in red. The color bar indicates the maximum and minimum SDM-Z values. HCs, healthy controls; SD, substance dependence; SDM, signed differential mapping.
Table 2. Meta-analysis results regarding regional differences of task-evoked activation between participants with all addictions and HCs during response inhibition
HCs, healthy controls; BA, Brodmann area; R, right; L, left.
Similarly, the SD subgroup showed hypoactivity in the right insula (BAs 38, 47, 48), right middle temporal gyrus (BAs 20, 21, 22), right temporal pole (BAs 20, 21, 38), right IFG (orbital and opercular parts, BAs 38, 44, 47), right supramarginal gyrus (BAs 2, 40, 48) and right precentral gyrus (BAs 6, 44), compared with HCs. They also exhibited significant hyperactivities in the left cerebellum (hemispheric lobule VIIB). Detailed results are presented in Table 3 and Fig. 2b.
Table 3. Meta-analysis results regarding regional differences of task-evoked activation between participants with SD subgroup and HCs during response inhibition
SD, substance dependence; HCs, healthy controls; BA, Brodmann area; R, right; L, left.
The findings on all addictions and SD group described above remained largely unchanged under the jackknife sensitivity analysis, indicating high robustness (Tables 2 and 3). For all addictions and the SD subgroup, the heterogeneity analysis showed non-significant results for all reported regions, indicating a non-significant unexplained between-study variability. The Egger's test showed no evidence of publication bias for most of the reported regions except for the left cerebellum (p = 0.010 for all addictions; p = 0.012 for the SD subgroup). Detailed results are summarized in Tables 2 and 3.
Finally, the meta-regression analysis revealed that individuals with longer SD duration (available in 18 SD datasets) showed enhanced abnormal activity in the left cerebellum (x = −16, y = −78, z = −38, Z = 3.744, p < 0.0005, 154 voxels). However, abstinence days were not associated with any change in the neural activity patterns.
Discussion
The current meta-analysis revealed several patterns of altered brain activations for adults with addiction during response inhibition. We found that all addictions showed reduced brain activity in the IFG, MTG, temporal pole, insula and supramarginal gyrus, and enhanced brain activity in the cerebellum, compared with HCs. The SD subgroup showed almost the same patterns, with additional hypoactivity observed in the precentral gyrus. A meta-regression analysis showed that longer addiction duration was related to stronger activity in the cerebellum for SD subgroup.
Consistent with our hypothesis, we observed lower activation of regions within the VAN (IFG, insula, MTG and temporal pole) during response inhibition for adults with addiction, compared with HCs (Fig. 2a and Table 2). VAN is specialized for the detection of unexpected yet behaviorally relevant information and for response reorientation based on goals and conflicts (Nee et al., Reference Nee, Wager and Jonides2007; Vossel, Geng, & Fink, Reference Vossel, Geng and Fink2014). Specifically, in healthy populations, IFG and insula are activated in response to unexpected and infrequent stimuli (Shulman et al., Reference Shulman, Astafiev, Franke, Pope, Snyder, McAvoy and Corbetta2009) and during suppression of pre-potent actions (Aron, Fletcher, Bullmore, Sahakian, & Robbins, Reference Aron, Fletcher, Bullmore, Sahakian and Robbins2003). Additionally, the temporal pole and the insula were found to engage in action planning and selection (Kircher, Brammer, Levelt, Bartels, & McGuire, Reference Kircher, Brammer, Levelt, Bartels and McGuire2004; Paulus, Feinstein, Leland, & Simmons, Reference Paulus, Feinstein, Leland and Simmons2005). Our results were consistent with previous literature that reported attenuated VAN activity in individuals with addiction (e.g. Fu et al. Reference Fu, Bi, Zou, Wang, Ye, Ma and Yang2008; Hendrick, Luo, Zhang, & Li, Reference Hendrick, Luo, Zhang and Li2012; Nestor, McCabe, Jones, Clancy, & Garavan, Reference Nestor, McCabe, Jones, Clancy and Garavan2011). The attention of adults with addiction may be automatically oriented to some salient stimuli (e.g. drugs and internet games) even though they do not voluntarily intend so. The observed hypoactivities in this network may also indicate that adults with addiction were less efficient at putting a brake to their actions, even in the presence of a stop signal (i.e. action cancelation; Schachar et al., Reference Schachar, Logan, Robaey, Chen, Ickowicz and Barr2007). This might be one of the reasons why it is rather difficult for people with addiction to abstain from addictive behaviors. Additionally, VAN is often considered to be involved in stimulus-driven or involuntary attention (Asplund, Todd, Snyder, & Marois, Reference Asplund, Todd, Snyder and Marois2010). For individuals with SD in particular, the bottom-up attentional processes were likely associated with continuing substance exposure (Lawrence, Luty, Bogdan, Sahakian, & Clark, Reference Lawrence, Luty, Bogdan, Sahakian and Clark2009; Smith et al., Reference Smith, Mattick, Jamadar and Iredale2014). Importantly, according to the impaired response inhibition and salience attribution model (iRISA; Goldstein & Volkow, Reference Goldstein and Volkow2011; Zilverstand et al. Reference Zilverstand, Huang, Alia-Klein and Goldstein2018), individuals with addiction tend to show a blunted response to non-addiction-related stimuli during diverse cognitive processing including reward processing and inhibitory control. The recruitment of related networks is, however, strengthened during the processing of addiction-related stimuli. Because the stimuli in the included studies were all neutral or non-addiction stimuli (e.g. alphabetical letters), the reduced VAN activity may indicate an impaired bottom-up processing of non-addiction-related stimuli as VAN is perhaps frequently and excessively activated in response to addiction-related cues, for individuals with addiction.
In our study, we also observed hypoactivation of the FPN (i.e. orbital part of IFG and supramarginal gyrus) in adults with addiction, compared with HCs (Fig. 2a and Table 2). FPN is another important network involved in response inhibition in healthy populations (Zhang et al., Reference Zhang, Geng and Lee2017). FPN is associated with goal-directed behaviors as well as the integration of bottom-up inputs and top-down information (Dodds, Morein-Zamir, & Robbins, Reference Dodds, Morein-Zamir and Robbins2011; Marek & Dosenbach, Reference Marek and Dosenbach2018). IFG, specifically, seems to support context monitoring by modulating the activation of parietal cortices based on task demand (Hampshire & Sharp, Reference Hampshire and Sharp2015). The attenuated FPN activity observed in our study were consistent with previous research on response inhibition in individuals with addiction (e.g. Fu et al. Reference Fu, Bi, Zou, Wang, Ye, Ma and Yang2008; Kaufman, Ross, Stein, & Garavan, Reference Kaufman, Ross, Stein and Garavan2003; Li, Luo, Yan, Bergquist, & Sinha, Reference Li, Luo, Yan, Bergquist and Sinha2009). Lower activation of the FPN may reflect diminished top-down activity and weakened integration of information from different neural resources, which may lead to less effective control initiation (e.g. action withholding; Marek & Dosenbach, Reference Marek and Dosenbach2018) and task adaptation (e.g. action cancelation; Zhang et al., Reference Zhang, Geng and Lee2017). Specifically, compared with HCs, adults with addiction may be less well at withholding responses to an inhibited stimulus (i.e. ‘no-go’) while they were asked to quickly respond to another stimulus (i.e. ‘go’). Perhaps, individuals with addiction were less efficient at voluntarily regulating their actions. This may explain why individuals with addiction tend to increase engagement in addictive behaviors overtime (Jazaeri & Habil, Reference Jazaeri and Habil2012; Miller, Dackis, & Gold, Reference Miller, Dackis and Gold1987) and relapse after treatment (Azevedo & Mammis, Reference Azevedo and Mammis2018; Brecht & Herbeck, Reference Brecht and Herbeck2014). Supporting this, previous research reported more pronounced disruptions in response inhibition in abstinent individuals who experienced strong urge to take drugs, which may later lead to relapse (Verdejo-García et al., Reference Verdejo-García, Lubman, Schwerk, Roffel, Vilar-López, MacKenzie and Yücel2012).
It should be noted that we observed hypoactivation of VAN and FPN, two large-scale networks, rather than any specific module in the brain. Our results may be seen as new evidence for the network perspective on response inhibition (Hampshire & Sharp, Reference Hampshire and Sharp2015). According to this perspective, response inhibition is one component of the broader cognitive control processes and is therefore supported by common networks underlying a wide range of cognitive processes. For example, FPN was found to show altered activity in individuals with cocaine addiction during reward processing (Costumero et al., Reference Costumero, Bustamante, Rosell-Negre, Fuentes, Llopis, Ávila and Barrós-Loscertales2017) and implicit moral processing (Caldwell et al., Reference Caldwell, Harenski, Harenski, Fede, Steele, Koenigs and Kiehl2015). The overall hypoactivation of the FPN and VAN can also be explained by the iRISA (Goldstein & Volkow, Reference Goldstein and Volkow2011; Zilverstand et al., Reference Zilverstand, Huang, Alia-Klein and Goldstein2018). As explained earlier, the attenuation of large-scale neural networks (i.e. FPN and VAN) during non-addiction-related processing in individuals with addiction may function as a compensation for the increased recruitment of these networks during addiction-related processing, maintaining the functional stability of these networks. Supporting this, Czapla et al. (Reference Czapla, Baeuchl, Simon, Richter, Kluge, Friederich and Loeber2017) found that alcohol-dependent individuals showed enhanced activity of FPN during response inhibition to alcohol stimuli (i.e. pictures of alcoholic beverages), compared with HCs. However, they showed reduced FPN activity when the stimuli were non-addiction-related (i.e. geometrical shapes). The iRISA explanation nevertheless remains speculative because we did not investigate brain activity in individuals with addiction during addiction-related response inhibition. Future research can aim to examine brain activity during response inhibition using stimuli with different nature (i.e. non-addiction-related and addiction-related) in individuals with addiction.
We also observed hyperactivation of the cerebellum during response inhibition in adults with addiction, compared with HCs (Fig. 2a and Table 2). This is consistent with previous research where cerebellar activity was strengthened in individuals with cocaine addiction during response inhibition (Hester & Garavan, Reference Hester and Garavan2004) and in individuals with alcohol addiction during a working memory task (Desmond et al., Reference Desmond, Chen, DeRosa, Pryor, Pfefferbaum and Sullivan2003). The cerebellum has been validated to regulate voluntary actions over cortical pathways (Brunamonti et al., Reference Brunamonti, Chiricozzi, Clausi, Olivito, Giusti, Molinari and Leggio2014). Our result of the hyperactivity of the cerebellum, as also suggested by previous research (Desmond et al., Reference Desmond, Chen, DeRosa, Pryor, Pfefferbaum and Sullivan2003; Hester & Garavan, Reference Hester and Garavan2004), may constitute a compensatory mechanism to the attenuated activation of other brain regions (i.e. FPN and VAN) in order to successfully perform a task of high demand (i.e. response inhibition) for individuals with addiction. Perhaps, in order to reach a similar level of task performance in healthy participants, individuals with addiction over-relied on the cerebellum when regions specialized for response inhibition could not be properly activated. Furthermore, the compensatory functions of cerebellum may not be restricted to motor aspects. Several researchers have suggested that cerebellum is a modulator in several neural networks that are altered in individuals with addiction (Miquel et al., Reference Miquel, Vazquez-Sanroman, Carbo-Gas, Gil-Miravet, Sanchis-Segura, Carulli and Coria-Avila2016; Moulton, Elman, Becerra, Goldstein, & Borsook, Reference Moulton, Elman, Becerra, Goldstein and Borsook2014).
In addition, a meta-regression analysis revealed that, longer addiction duration was associated with stronger activity in the cerebellum in individuals with SD. This may be a result of overreliance on the cerebellum as a compensatory mechanism overtime in individuals with SD. Those with longer addiction duration may have more experience with activating the cerebellum to inhibit an action optimally. In other words, the brain of an individual with long-time addiction may have a more well-developed and strategic compensatory mechanism during response inhibition, compared with that of a person who has a shorter addiction duration. This is likely because, for an individual who only recently becomes addicted to substance, the activities in FPN and VAN may be less impaired by substance intake than someone with longer addiction history and correspondingly more substance exposure. Consequently, those with shorter addiction duration may be able to, at least partially, activate response inhibition networks, easing the need of cerebellum activation as a compensation.
Furthermore, compared with HCs, the SD subgroup exhibited hypoactivity of the precentral gyrus, in addition to the above-reported activation patterns in all addictions (Fig. 2b and Table 3). This was consistent with previous research that found a negative association of substance use with the activation of the precentral gyrus (Ye et al., Reference Ye, Li, Zhang, Li, Zhu, Chen and Wei2018). This region has been considered as an inhibitory motor region, controlling voluntary movements, and is activated in response inhibition tasks in healthy people (e.g. Criaud & Boulinguez, Reference Criaud and Boulinguez2013). The weaker activation of the precentral gyrus may suggest that individuals with SD had deficits in planning and executing an appropriate action, compared with HCs. Interestingly, most of the studies reported decreased precentral signaling in individuals with SD in the absence of behavioral deficits at task, compared with HCs. Perhaps, impairments in response inhibition are not easily and consistently observed at the behavioral level with the experimental paradigms, especially considering the heterogeneity of tasks used across studies and the variability of behavioral indices reported for even a single paradigm (Meule, Reference Meule2017). It is also worth mentioning that the hypoactivity of the precentral gyrus was not observed in all addictions. Perhaps, the activation of the precentral gyrus and correspondingly action planning and execution were somewhat preserved, if not enhanced, in individuals with behavioral addiction. Indeed, hyperactivity of this region and the cingulate cortex (e.g. Dong et al., Reference Dong, DeVito, Du and Cui2012), another region often showing hypoactivity in SD, was reported in individuals with internet gaming disorder during response inhibition and was interpreted as a result of gaming skill acquisition (Ding et al., Reference Ding, Sun, Sun, Chen, Zhou, Zhuang and Du2014).
There were some limitations in this study. First, due to the small number of studies on behavioral addiction (three), we could not investigate the impaired brain activity in response inhibition for this subgroup. Consequently, we were not able to compare and contrast the brain activity patterns between SD and behavioral addiction subgroup meta-analytically. Future research should gain more insights into response inhibition in behavioral addiction. Second, there are likely subtle differences in the neural abnormality associated with different substances of abuse (e.g. stimulants and depressants). For example, four out of six included studies on depressants (e.g. alcohol) showed hypoactivity of the response inhibition network whereas the results from studies on stimulants (14 studies) were mixed. Again, due to the limited number of included studies, we could not investigate the potential differences across various substance at a meta-analytic level. Further investigations are needed to detect the effects of different substance categories. Similarly, we did not conduct separate meta-analyses based on the types of experimental tasks due to the limited number of studies for each task type. The heterogeneity of tasks may modulate the activation patterns as some tasks may involve subtly different cognitive components, compared with others (Fineberg et al., Reference Fineberg, Chamberlain, Goudriaan, Stein, Vanderschuren, Gillan and Morein-Zamir2014). It is also difficult to discern the extent to which task difficulty levels or cognitive demands modulate the results. Future research on response inhibition should aim to detect potential effects of task heterogeneity on neural activity. Fourth, because we could not obtain the unthresholded activation maps from the included studies, and that there was a substantial variability in the thresholding and correction methods across studies (see online Supplementary Table S3), we did not supplement our analyses with a meta-analysis on unthresholded activation maps, or using a common thresholding or correction technique. It would be desirable if future research can develop and/or follow a standard reporting guideline and make accessible most of their research data for potential reuse purposes. Furthermore, the current study sample limited the ability to investigate potential impacts of reward contingencies on neural abnormalities during response inhibition in addiction. It would be interesting for future research to investigate whether and how reward contingencies modulate neural activity during response inhibition tasks for individuals with addiction. Finally, we only included data from adults with addiction in this meta-analysis. This was because the neural networks for response inhibition in adolescents are rather immature, compared with adults (Vara, Pang, Vidal, Anagnostou, & Taylor, Reference Vara, Pang, Vidal, Anagnostou and Taylor2014). And adolescents undergo developmental changes in brain regions related to response inhibition both structurally and functionally (e.g. Van Leijenhorst et al., Reference Van Leijenhorst, Moor, de Macks, Rombouts, Westenberg and Crone2010). Whether and how these changes may affect adolescents' susceptibility to addictive behaviors is yet unclear. On these grounds, we only investigated adults with addiction in our meta-analysis.
Conclusion
This meta-analysis revealed reduced brain activity in the IFG, MTG, temporal pole, insula and supramarginal gyrus, and enhanced brain activity in the cerebellum during response inhibition in all addictions, compared with HCs. The SD subgroup showed additional hypoactivity in the precentral gyrus. The reduced brain activity in the VAN and FPN implicated altered attention to and inhibitory control for non-addiction-related stimuli during response inhibition tasks for adults with addiction, which may account for their repeated addictive behaviors. Additionally, the enhanced brain activity in the cerebellum may act as a compensatory mechanism to maintain the functional stability of an addicted brain. These results may help to understand the pathology of impaired response inhibition in adults with addiction.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291721000362
Acknowledgements
This study was funded by the National Natural Science Foundation of China (81801685) and the Natural Science Foundation of Guangdong Province, China (2018A030310003). The authors appreciated the critical comments from Yuliang Wang, Department of Psychology, The University of Hong Kong, in preparation of the manuscript.
Conflict of interest
The authors declare that they have no conflict of interests.
