Background
Alcohol use disorder (AUD) is a major public health problem which has a substantial impact on patients’ psychological, physiological and social functioning [
1‐
4]. Severe AUD often remains a lifelong condition due to patients’ susceptibility to relapse [
5‐
7]. Despite the development of valuable treatment programs [
8,
9], one third of patients relapse within 1 year after treatment, highlighting the need for effective new interventions to be added to these programs.
Current models of substance use disorders (SUDs) are based on neuroscientific research into subcortical alterations (e.g., [
10‐
12]) and psychological investigations on implicit processes [
13], and postulate on an imbalance between subcortical, implicit processes and cortical control processes. Briefly, these models assume that automated (subcortical) cue-reactivity is enhanced, thus leading to a strong drive to consume, while the opponent (cortical) executive control is weakened, making it difficult to inhibit this impulse (e.g., [
11,
14‐
16]). The neuroscientific findings that underlie these models include enhanced reactivity to substance-related cues in brain regions related to reward prediction and saliency processing (e.g., nucleus accumbens, amygdala), habits and motivation (dorsal striatum, orbitofrontal cortex) and alterations in brain regions related to inhibitory control (dorsolateral prefrontal cortex, inferior frontal cortex and anterior cingulate cortex) [
17,
18]. There is an extensive body of literature indicating that SUD is associated with impaired performance in inhibitory control tasks [
19‐
24]. Until recently, these tasks did not differentiate between inhibition in a substance-related versus neutral context. However, this distinction seems to be clinically relevant, given that the central malfunction leading to relapse is the inability to exert inhibitory control in a substance-related context. Recent studies have suggested that alcohol-related inhibition demands additional neuronal resources in patients with AUD and heavy drinkers [
25‐
28].
To translate the findings of impaired inhibitory control and enhanced cue-reactivity into a therapeutic context, an inhibition training for eating disorders [
29‐
31] has been adapted into a novel alcohol-specific inhibition training and introduced into the field of AUD research [
32,
33]. During this short, computerized, behavioral training, alcohol-related stimuli are consistently paired with a stopping response. Such an intervention might thus improve inhibitory control by training it on an explicit as well as a habitual level and possibly reduce cue-reactivity through stimulus devaluation.
The effect of one or two of these training sessions on alcohol consumption has been tested in regular drinkers [
34,
35] and non-clinical, heavy drinkers [
32,
33,
36,
37]. When alcohol consumption was assessed immediately after the training session, two studies found decreased consumption [
35,
36], one study observed a non-significant trend in that direction [
33], another study reported longer latency until participants took the first sip [
37], and one study found no effects [
34]. Of those studies assessing alcohol consumption 1 or 2 weeks post intervention, three studies reported decreased consumption after alcohol-specific inhibition training [
32,
33,
37]. Note, however, that in two of these studies, the intervention was compared to a control condition, which could potentially have fostered alcohol consumption within the control group [
32,
33]. Two other studies found no significant effects [
34,
35]. In summary, while the effects of this training are inconsistent in non-clinical samples, it might be more effective in clinical populations where (1) the motivation to change drinking behavior is higher and (2) the impairment targeted by the training is assumed to be more pronounced at baseline. Therefore, it is important to test the feasibility and efficacy of the training in clinical populations with AUD. The present randomized controlled, double-blind study aims to close this gap, and includes follow-up measurements of drinking behavior and related concepts up to 12 months after treatment discharge, allowing assessment of potential clinically relevant long-term effects.
There are two putative working mechanisms by which alcohol-specific inhibition training is thought to be effective. The first is through enhancing inhibitory capacities in the context of alcohol-related cues, which can be assessed with inhibitory tasks, such as the Go-NoGo Task (GNG) or the Stop-Signal Task (SST). The second putative working mechanism regards devaluation of the stimuli. It is based on the
Behavior Stimulus Interaction Theory [
38], which states that the repeated stopping response to the rewarding (alcohol, in this context) stimuli causes a devaluation thereof [
31,
33]. This mechanism is thus thought to alter the automatically attributed appeal of alcohol-related stimuli, an effect which can be measured with the Implicit Association Test (IAT).
With respect to the devaluation mechanism, the first two studies were consistent with predictions of the
Behavior Stimulus Interaction Theory in their reports that participants’ implicit attitudes shifted towards a more negative evaluation of alcohol-related stimuli after alcohol-specific inhibition training [
32,
33]. However, this finding could not be replicated in the four subsequent studies [
34‐
37].
In regard to an inhibition-related working meachanism, those studies using the respective measures found no effects of GNG-based inhibition training on an SST [
32,
37] nor flanker task [
34]. However, these tasks measure slightly different inhibitory aspects (e.g., action cancellation in case of the SST) than the one affected during the GNG-based training (action restraint). The only study which assessed inhibitory functioning in the same form as targeted during training (i.e., measurement of inhibitory control with a GNG after a GNG-based training) reported no effects on reaction times during Go trials. Unfortunately, data on errors of commission (EOCs), which are a commonly used indicator of inhibitory capacities and an essential and proximal experimental outcome of a GNG-based inhibition training, were not reported in this study [
36] and are thus still missing. Another important aspect concerns Go/NoGo ratio incorporated during training sessions. With one exception [
34], most studies emloyed equiprobable GNG tasks with a Go/NoGo ratio of 50/50 during the training session. Such a ratio is unusually balanced for GNG tasks, where NoGo responses usually only make up 10–20% of the trials (e.g., [
23,
24,
39]) in order to establish a high tendency to respond while making inihibition more demanding. Thus, it is possible that these versions failed to create the context of a prepotent Go response, in which inhibition could effectively be trained. The present study aims to extend prior research on potential working mechanisms by comparing two versions of this alcohol-specific inhibition training differing solely in the Go/NoGo ratio (thus in inhibition difficulty) and by comparing these versions’ effects on inhibitory capacities as well as on implicit associations.
With regard to other clinically important psychological variables, research has shown that outcome expectancies [
40], self-efficacy to remain abstinent [
41], motivation to change alcohol use [
42] and craving [
43] predict drinking outcomes at 1-year follow-up. These variables might thus act as potential mediators of effects in a clinical study. Furthermore, if the alcohol-specific inhibition training influences implicit associations towards alcohol, it might also affect explicit expectancies and/or subjective craving. The present study will, therefore, include these variables as potential outcomes and mediators in explorative analyses.
Taken together, the present study investigates the effects of an alcohol-specific inhibition training for the first time in a clinical sample using a randomized controlled, double-blind design with long follow-up periods. In doing so, the study contributes to research on potential working mechanisms by comparing two versions of the training which differ only in the Go/NoGo ratio and by the comparison of their effects on inhibitory functions and implicit associations. Explorative analyses will target psychological variables, such as craving, outcome expectancies, self-efficacy and motivation to change alcohol use. As the training’s rationale is anchored in neuroscientific research, we will also monitor neurophysiological effects with multi-channel electroencephalography (EEG) in order to investigate whether the training changes the neurophysiological correlates of alcohol-specific inhibition and/or implicit associations. Finally, prior research has shown that endogenous cortisol is a consolidation enhancer [
44], which peaks in the morning [
44,
45] and has been shown to improve effects of other learning-based therapies [
46]. Since training procedures rely on learning processes affected by endogenous cortisol, we are, therefore, interested in whether its levels influence training effects.
Study aim
In the present project, the
Inhibition
Training (INTRA) study, 246 recently abstinent patients with AUD attending an inpatient treatment program will be randomly assigned to one of two alcohol-specific inhibition training groups (each with a different Go/NoGo ratio) or to a control group. Our aim is to examine whether variations of inhibition training have a positive effect on drinking behavior, implicit attitudes, and neurophysiological reactivity to alcohol-related stimuli. Thus, a subgroup of patients will additionally undergo EEG recording before and after the intervention so that neurophysiological effects of the training can be assessed and related to clinical outcomes. In addition, 50 healthy controls (with EEG measurement of 20) will be assessed once to compare patients’ pre-training data. Since training effects rely on learning processes, the influence of endogenous cortisol level (a consolidation enhancer, which peaks in the morning and decreases in the course of the day [
44]) on training outcome will be investigated by varying the time of day in which the training is performed. All patients’ inhibition and implicit associations towards alcohol will be measured immediately before and after the training. We will also measure the training’s effects on proximal outcome variables (e.g., implicit associations, inhibitory control, abstinence-related self-efficacy, craving) post training, and distal outcome variables (e.g., percentage of days abstinent – PDA; heavy-drinking days – HDD; and time to first drink – TFD) at 3-, 6- and 12-month follow-ups.
For the first time, this trial will investigate the therapeutic potential of an alcohol-specific inhibition training as a therapeutic supplement in a sample with patients suffering from severe AUD. In doing so, the impact on training efficacy of time of training and Go/NoGo ratio will additionally be investigated and the underlying neurophysiological mechanisms of the training elucidated. Moreover, potential AUD-related psychological constructs will be explored. The project thus aims to contribute to the improvement of the evidence-based treatment of AUD.
Research questions
Based on prior research, the following five research questions are examined:
1.
Does the alcohol-specific inhibition training (Alc-IT, compared to the control intervention) reduce PDA and HDD and/or extend the time to first drink (TFD) at 3-months’ follow-up (primary outcomes)?
2.
What is the effect of the Alc-IT on behavioral experimental parameters?
(a)
Does Alc-IT decrease positive implicit associations compared to the control training?
(b)
Does Alc-IT enhance response inhibition compared to the control training?
3.
Are the neurophysiological correlates of alcohol-specific inhibition and implicit associations towards alcohol changed by Alc-IT?
4.
Does endogenous cortisol moderate the effect of the Alc-IT?
5.
In addition, explorative questions investigate whether the effects of the training on outcomes are mediated by AUD-related psychological constructs, such as alcohol-related self-efficacy, craving or motivation
Data collection and statistical analysis
Questionnaires
Will be assessed by using paper and pencil and will then be entered into SPSS (Version 24.0, IBM SPSS Statistics for Windows, IBM Corp, Armonk, NY, USA). Behavioral data will be collected with E-Prime 2.0 (EP2Pro2.0.10.356, Psychology Software Tools, Inc., Sharpsburg, PA, USA).
EEG data
Electrophysiological data will be recorded with BrainVision Recorder (Version 2.0, Brain Products GmbH, Gilching, Germany) using 64 active electrodes distributed across the scalp according to the extended 10/10 system. The sampling rate is 500 Hz, online filters are set to 0.016 Hz (high-pass) and 250 Hz (low pass), impedances are kept below 20 kΩ, FCz will serve as on-line reference. Each participant will first undergo a 5-min resting state EEG with alternating epochs of eyes open and eyes closed. Then, the experimental test battery of the T2 and T3 assessment, consisting of a Go-NoGo Task (GNG), an Implicit Association Test (IAT) and a Stop-Signal Task (SST), will be administered while the EEG is still recorded. For details about the tasks, please see section “Experimental tasks.”
Cortisol data
Saliva samples for cortisol analyses will be stored at − 80 °C.
Statistics
All statistical analyses will be conducted by the members of the study team. The main analyses of the training effect will be carried out after data collection is completed to maintain the study’s double-blind design. All behavioral and questionnaire variables will be tested for normal distribution (Kolmogorov-Smirnov test: p > 0.1, for all variables).
Main study
Training effects on our primary outcome measures (PDA and HDD at 3-month follow-up) will be assessed with a 2 × 3 analysis of variance (ANOVA) with the factors time point (T1, T5) and training group (Alc-IT (50/50), Alc-IT (75/25) or control). Further, a Cox regression will be computed to predict the effect of the intervention on TFD (at 3-month follow-up).
Training effects on experimental test parameters will be assessed with a 2 × 2 × 3 repeated measures ANOVA with the factors measurement point (T2, T3), time of day (morning, afternoon) and training group (Alc-IT (50/50), Alc-IT (75/25) or control). Where Mauchly’s test of sphericity indicates heterogeneity of covariance, we will verify repeated measures results with Greenhouse-Geisser corrections.
EEG substudy
All raw EEG data will be pre-processed with BrainVision Analyzer (Version 2.0, Brain Products GmbH, Gilching, Germany) according to current standards including ICA-based correction of eye-movement artefacts, artefact rejection and application of band-pass filtering (see e.g., [
94‐
96]). Event-related potentials (ERPs) will be computed for each stimulus type in the three experiments (IAT, GNG, SST). Epochs from 500 ms (pre-stimulus) to 1500 ms (post-stimulus) will be averaged separately for each stimulus type and measurement point (T2 and T3). ERPs will subsequently be statistically compared for overall amplitude (i.e., global field power, GFP) and topography. For each ERP, GFP [
97] will be calculated as the standard deviation across electrodes, thus measuring momentary global signal strength regardless of topographic modulations. GFP values for each point in time will be extracted and compared (T2 vs. T3) with nonparametric randomization tests, which simultaneously control for multiple comparisons. The analyses will be conducted using the Ragu software [
98,
99].
To inspect for topographic differences between ERPs measured before and after the training, a topographic analysis of variance (TANOVA) [
100] will be computed in Ragu. Here, dissimilarities of electric field topographies are identified with a nonparametric randomization test. A significant finding in the TANOVA indicates that activation in underlying brain structures vary in relation to the factor under investigation. Significant TANOVA results will be further explored with the standardized low-resolution electromagnetic tomography (sLORETA) source analysis method [
101]) to determine which brain regions vary in activation in relation to stimulus type and/or measurement time (T2 vs. T3).
Based on prior findings, our ERP analyses will focus on the timeframe from about 150 to about 850 ms. Exact timeframes will be defined based on components as visible in GFP-curves or by microstate analyses [
102,
103]. Following earlier research, care will be taken to include the N2 and P3 components in case of the GNG [
25,
26] and the N2, P3, N4 and LPP components in case of the IAT (e.g., [
104‐
106]).
Cortisol substudy
Salivary cortisol concentrations are determined by a commercially available chemiluminescence immunoassay (CLIA; IBL, Hamburg, Germany). Inter- and intra-assay coefficients of variation are both below 8%. For biochemical analyses of free cortisol concentration, saliva samples will be thawed and spun at 3000 rpm for 10 min to obtain 0.5–1.0 ml of clear saliva with low viscosity.
Power calculation
The intended sample size was calculated with G*Power (Version 3.1.5, Heinrich-Heine-University Duesseldorf, Dusseldorf, Germany). In a-priori analyses (1−β = 0.8, α = 0.05), slightly reduced expected effect sizes were entered compared to earlier studies because of the additional manipulations concerning different training versions and time of training. For the 2 × 3 ANOVA to examine the training effects on the primary outcome measures PDA and HDD (with an expected effect size of f = 0.2 based on earlier studies [
32]), a total sample size of
n = 244 is needed. For the analyses of the training’s effect on TFD (the third primary outcome, estimating a small to medium effect size of w = 0.25 based on earlier studies [
37], and df = 3), G*Power indicated a total sample size of
n = 174 for the Cox regression. For the 2 × 2 × 3 repeated measures ANOVA examining training effects on experimental parameters and also taking into account the effects of time of day of the training (expecting small to medium effects of f = 0.15 and correlations of 0.4 among repeated measures based on earlier studies [
107]), the necessary sample size was
n = 180. Considering the necessary power for all analyses and similar sample sizes in all study arms (82 per intervention group), a total of 246 patients will be recruited.
Discussion
The results of the INTRA study should provide evidence for the efficacy of an add-on treatment to specialized standard care for AUD. In contrast to previous non-clinical studies [
32,
33], this double-blind RCT investigates the effect of a computerized inhibition training in inpatients with AUD. It includes important measures of alcohol consumption and follow-up periods of up to 1 year. With respect to working mechanisms, the study includes a thorough assessment of inhibitory control functions and alcohol-specific implicit associations on both a behavioral and neurophysiological level. Through this investigation, we hope to expand current knowledge about the role of inhibitory functions in AUD and extend previous findings about the effects of this training [
25,
26,
32‐
37]. In order to describe the patients’ baseline measures, including possible inhibitory deficits, and to better interpret the observed modifications, the patients’ behavioral and neurophysiological data will be compared to healthy controls.
In additional to these proximal findings, the effect of the intervention will also be analyzed longitudinally. At 3-, 6- and 12-month follow-ups, important measures of alcohol consumption, our primary outcome, as well as psychological parameters will be collected to inform us about the time after inpatient treatment. Finally, this trial will investigate whether endogenous cortisol might increase possible effects of the inhibition training.
As one potential limitation, one could argue that patients in the control training group undergo a non-context-specific inhibition training, which might enhance their general inhibitory capacities. The choice of the control group is, therefore, very strict and could lead to an underestimation of the effect of the inhibition training. However, control conditions used in previous non-clinical studies [
32,
33] consistently paired alcohol stimuli with Go responses and tended to enhance alcohol consumption in these samples. Therefore, due to ethical considerations, we opted against this control condition in a clinical sample.
Overall, the current double-blind RCT is the first study to investigate the effect of an inhibition training in an inpatient treatment setting in patients with AUD. It allows a detailed proximal and distal evaluation of behavioral, psychological and neurophysiological processes in AUD and of the efficacy of the inhibition training. We therefore hope that it might ultimately contribute to the improvement of evidenced-based AUD treatment.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.