University of Groningen When attention takes over Heitmann, Janika

(1)

When attention takes over

Heitmann, Janika

DOI:

10.33612/diss.126810192

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from

it. Please check the document version below.

Document Version

Publisher's PDF, also known as Version of record

Publication date:

2020

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):

Heitmann, J. (2020). When attention takes over: attentional bias and its modification in substance use and

addiction. https://doi.org/10.33612/diss.126810192

Copyright

Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

(2)

ATTENTIONAL BIAS AND ITS MODIFICATION

IN SUBSTANCE USE AND ADDICTION

(3)

and addiction

ISBN printed version 978-94-034-2682-2 ISBN digital version 978-94-034-2683-9 Cover design Freshidea / Remco Gaikema

Layout Remco Gaikema

Printing Ridderprint BV, www.ridderprint.nl

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means without express written permission from the author, and, when appropriate, the publisher holding the copyrights of the published articles.

The studies in this dissertation were funded by ZonMw (The Netherlands Organisation for Health Research and Development; 80–84,300–98-61,035), and co-financend by Verslavingszorg Noord Nederland. The funding bodies had no role in the study design, nor in the collection, analysis, and interpretation of the data.

This dissertation is printed on recyled paper in a small batch. Contact the author for a digital version.

(4)

ATTENTIONAL BIAS AND ITS MODIFICATION

IN SUBSTANCE USE AND ADDICTION

Proefschrift

ter verkrijging van de graad van doctor aan de Rijksuniversiteit Groningen

op gezag van de

rector magnificus prof. dr. C. Wijmenga en volgens besluit van het College voor Promoties.

De openbare verdediging zal plaatsvinden op maandag 15 juni 2020 om 12.45 uur

door

Janika Heitmann

geboren op 3 mei 1989 te Hamburg, Duitsland

(5)

Prof. dr. P. J. de Jong

Copromotor

Dr. B. D. Ostafin

Beoordelingscommissie

Prof. dr. M. Rinck Prof. dr. I. Franken Prof. dr. M. H. Nauta

(6)

Chapter 1 7

General introduction

Chapter 2 21

Attentional bias for alcohol cues in visual search: Increased engagement, difficulty to disengage, or both?

Chapter 3 47

Attentional bias for substance cues in outpatients with alcohol or cannabis use disorder measured with an Odd-One-Out visual search task: Engagement bias, disengagement bias, or both?

Chapter 4 73

The effectiveness of attentional bias modification for substance use disorder symptoms in adults: A systematic review

Chapter 5 113

Internet-based attentional bias modification training as add-on to regular treatment in alcohol and cannabis dependent outpatients: A study protocol of a randomized controlled trial

Chapter 6 137

Attentional bias modification training as add-on to regular treatment in alcohol and cannabis use disorder: Clinical effects from a multi-center randomized controlled trial

Chapter 7 173 General discussion References 191 Samenvatting 215 Publication list 223 Dankwoord 224

(7)

(8)

(9)

“Once I stopped smoking cannabis, I saw everyone else doing it.”

Anonymous patient, 2017

ADDICTION AND ATTENTION

Addictive substances such as alcohol and cannabis are frequently abused worldwide, and regular and excessive use of these substances can result in the development of substance use disorders. Substance use disorders are a common health problem with global prevalence rates of alcohol use disorder ranging from 0% to 16%, and drug use disorders ranging from 0% to 3% (WHO, 2004). In the Netherlands, the estimated prevalence rate of adults who meet the criteria of alcohol use disorder ranges from 0.3 to 1.2%. The prevalence rate of drug use disorders in the Netherlands is estimated to range from 0.4 to 1.0%, which consists to 0.1 to 0.5% of adults meeting the criteria of cannabis use disorder (De Graaf, Ten Have, & Van Dorsselaer, 2010). Substance use disorders are known to have negative physical and mental health consequences (Bernardin, Maheut-Bosser, & Paille, 2014; Gates, Gorbin, & Fromme, 2016; Lindgren et al., 2016), and serious societal costs (Goossens, Van Hasselt & Sannen, 2012). Therefore, effective and accessible treatment is highly relevant. In 2015, in the Netherlands 65.000 unique persons were treated for substance use disorders, with alcohol use disorder (approximately 30.000) and cannabis use disorder at the top (approximately 11.000; Wisselink, Kuijpers, & Mol, 2016). However, substance use disorders are known for their persistence and therefore effective treatment can be difficult. The persistence is reflected in high relapse rates after treatment, varying from 40-70% within the first year after treatment (Hunt, Barnett, & Branch, 1971; McLellan, Lewis, O’Brien, & Kleber, 2000), with the result that people often search for treatment multiple times. Given the need for effective and accessible treatment for substance use disorders on the one hand, and high rates of relapse on the other hand, it seems relevant to improve treatment outcome. One explanation for high rates of relapse after treatment could be that current treatments do not target all relevant aspects or underlying mechanisms of addiction, and therefore (often) miss to result in successful and long-lasting treatment effects. One perspective that has promise regarding the contribution of relevant underlying mechanisms is described in the dual process models of addiction. Using these models, the development and persistence of substance use disorders can be explained by the contributing role of controlled as well as automatic cognitive processes (Gladwin & Figner, 2015; Wiers et al., 2007). In these models the paradox of persistent substance use behaviour against a person’s own will has been described as an inability to adequately modulate the effects of automatic processes by means of sufficiently strong controlled processes. The idea that automatic processes are related to the dysfunctional character of addiction is supported by studies showing

(10)

1

that stronger automatic processes are related with the severity of addiction (e.g., Fadardi & Cox, 2006).

When looking more closely into the current evidence-based interventions in addiction care, it is noteworthy that most interventions are based on cognitive behavioural treatment (CBT) techniques which have a restricted focus on methods that are meant to strengthen explicit processes. For example, CBT interventions often involve challenging and changing unhelpful thoughts or beliefs and developing helpful coping strategies (Schippers, Smeerdijk, & Merx, 2014). In line, one could conclude that although automatic processes have a theoretical-based role in the persistence of substance use disorder supported by some evidence, it appears that current interventions do not directly target these processes in treatment. Importantly, current interventions, such as CBT, have been found to have little influence on the strength of automatic processes (Van Hemel-Ruiter, Wiers, Brook, & de Jong, 2016). That is, dysfunctional automatic processes seem to remain stable and unchanged throughout treatment. However, on the basis of the dual process models, failing to target relevant automatic processes may contribute to the high rates of relapse (e.g., Schoenmakers et al., 2010). During the past 20 years, research in the field of addiction has therefore focused on the identification of possible automatic processes that might play a relevant role in the development and persistence of substance use disorders. One of the processes that has been extensively studied is attention.

PART I – ATTENTIONAL BIAS

Attention, defined as the behavioural and cognitive process of selectively focusing on a discrete aspect of information while ignoring other available information, has been suggested to play a role in substance use disorders by means of a so-called attentional bias (AB). AB can be defined as an attentional tendency towards substance-relevant cues in the environment (Fadardi & Cox, 2008). This tendency can result in heightened attentional capture and/or a difficulty to disengage attention from these cues (Posner, 1980; Posner, Snyder, & Davidson, 1980). The development of AB can be explained by classical learning theories. As described in the incentive-sensitization theory by Robinson and Berridge (1993), the use of addictive substances produces a rewarding dopaminergic response in the brain which becomes sensitized with each time the substance is used. When repeatedly used, the substance becomes more salient and cues that relate to the substance (e.g., a bottle of beer) acquire strong incentive motivational properties, grabbing the users’ attention, thereby promoting the development of craving for the substance. Once substance-relevant cues have acquired these high attention-grabbing properties, this attentional tendency might in turn lower the threshold for repeated use, resulting in a self-reinforcing bias-craving-bias cycle.

(11)

This theoretical framework suggests that AB and craving are both expressions of an underlying appetitive motivational process. If this is the case, then AB and subjective craving should be positively related. In line with this, meta-analyses that reviewed addiction studies found a robust, albeit weak positive relationship between AB and craving (Field, Munafò, & Franken, 2009). Further, several experimental studies showed that an experimental induction of subjective craving (e.g., through cue exposure) was accompanied by an increase in AB (Bradley, Garner, Hudson, & Mogg, 2007; Field, Mogg, & Bradley, 2004, 2005; Field & Powell, 2007, Field & Quigley, 2009; Field, Rush, Cole, & Goudie, 2007; Grant, Stewart, & Birch, 2007; Ramirez, Monti, & Colwill, 2015a, 2015b; Schoenmakers, Wiers, & Field, 2008). There are also indications that an experimental reduction in craving (e.g., through substance devaluation) can lead to a reduction in AB (Oh & Taylor, 2013, 2014; Rose, Brown, Field, & Hogarth, 2013; Van Rensberg, Taylor, & Hodgson, 2009).

Regarding the converse direction, for AB to have a causal role in the development and persistence of substance use disorders, one would expect that AB is not only changing as a reaction to a change in craving, but that modifying AB should also influence craving. And indeed, one study found that experimentally manipulating AB towards substance-relevant cues by means of a computerized attention task resulted not only in an increased AB, but also in increased subjective craving (Field & Eastwood, 2005). Further, there are indications that experimentally reducing AB might in turn lead to reduced craving (Kerst & Waters, 2014). In addition to the literature on AB and craving, other studies have supported the idea of AB contributing to the development and persistence of substance use disorders. That is, AB has been found to be related with the severity of addiction (Fadardi & Cox, 2006), treatment outcome (Carpenter, Schreiber, Church & McDowell, 2006), and rates of relapse (Schoenmakers et al., 2010). Taken together, the findings noted above represent compelling evidence for an association between AB and substance use disorder symptoms (craving) underpinning the proposed causal role of AB in substance use disorders (Field & Cox, 2008). Although there is compelling evidence, it is important to keep in mind that there are also studies with findings not in line with this proposed role of AB in addiction (see for example Duka & Townshend, 2004; Waters, Shiffman, Bradley, & Mogg, 2003). However, these inconsistencies in findings might for example be related to motivational processes, as described in a recent review (Field et al., 2016), or methodological issues of the assessment tasks (as described below).

(12)

1

MEASURES OF ATTENTIONAL BIAS

To examine whether attentional processes contribute to the development and persistence of substance use disorders, a variety of paradigms have been developed. The two most often utilized AB assessment tasks are the addiction Stroop task (Cox, Fadardi, & Pothos, 2006), and the visual probe task (Ehrman et al., 2002; MacLeod, Mathews, & Tata, 1986). In the addiction Stroop task participants are instructed to name the font colour of a single word which is presented on each trial, while attempting to ignore the semantic content of that word. The task consists of two word-categories, substance-relevant words and substance-irsubstance-relevant words, which are matched by semantic and perceptual properties. AB is indexed by the mean difference of colour naming between trials showing substance-relevant words and trials showing neutral words. It is assumed that slower colour naming of trials containing substance-relevant words in comparison with trials containing neutral words, indicates automatic processing of the words’ content, and is therefore an indication for AB towards substance-relevant cues. The addiction Stroop task is known to be an indirect measure of AB, as AB is expressed by an interference effect of participants’ performance on a primary task (i.e., colour naming).

In the second task, the visual probe task, on each trial participants are instructed to identify the location of a probe (e.g., dot or cross) which appears in the location of one of two simultaneously presented stimuli (either words or images) immediately after stimuli offset. One of the stimuli contains substance-relevant content whereas the second contains substance-irrelevant content. The words or images are matched by semantic and perceptual properties. AB is indexed by the mean difference in reaction time of trials in which the probe appears in the location of the substance-relevant stimulus and trials in which the probe appears behind the neutral stimulus. Generally, participants react faster to probes that appear in the visual field where they were paying attention to than regions they are not attending (Posner, Snyder, & Davidson, 1980). AB for substance cues is thus expressed as faster reaction times to probes that replace substance-relevant stimuli than those that replace neutral stimuli. Like the addiction Stroop task, the visual probe task can provide an indirect measure of AB. However, in contrast to the addiction Stroop that indexes AB as an interference effect, the visual probe task indexes AB on the basis of selective visuospatial attention. Visuospatial attention involves paying attention to a particular location (Carrasco, 2018). In particular, the visual probe task assesses covert spatial attention, which means that changes in attention are not necessarily accompanied by eye movements. In addition to the addiction Stroop and visual probe task, other indirect assessment tasks have been developed to index AB towards substance-relevant cues, for example the attentional cueing task (see for example Garland, Franken, Sheetz, & Howard, 2012)

(13)

or the flicker-induced change blindness task (see for example Jones, Jones, Blundell, & Bruce, 2002). Furthermore, there have been more recent attempts to assess AB with more direct procedures such as eye-tracking (see for example Pennington, Qureshi, Monk, Greenwood, & Heim, 2019; Soleymani, Ivanov, Mathot, & de Jong, 2020; Wilcockson, & Pothos, 2015), thereby indexing AB in terms of spontaneous selective overt attention towards substance-relevant cues.

METHODOLOGICAL ISSUES OF CURRENT ASSESSMENT TASKS

Using the assessment tasks described above, a large number of studies have investigated the role of AB in addictive behavior. This literature has been synthesized in several reviews, all of which show evidence for the presence of an AB in several substance use disorders, including tobacco use disorder, alcohol use disorder, cannabis use disorder, stimulant use disorder, and opioid use disorder (Cox, Fadardi, & Pothos, 2006; Franken, 2003; Leeman, Robinson, Waters, & Sofuoglu, 2014; Robbins & Ehrman, 2004; Zhang et al., 2018). Despite this body of evidence, there are methodological issues with the majority of the (indirect) assessment tasks which might explain inconsistent findings (e.g., Bauer & Cox, 1998; Charles, Wellington, Mokrysz, Freeman, O’Ryan, & Curran, 2015; Munafò, Mogg, Roberts, Bradley, & Murphy, 2003).

One important methodological issue of the majority of current assessment tasks is related to their configuration. These tasks are based on reaction times, meaning that AB is indexed by the time taken from the onset of stimulus presentation until the response. Reaction times are considered to represent the outcome of a cognitive process in reaction to a particular stimulus (Johnson & Proctor, 2004). The reaction times delivered by these tasks therefore simply provide information about which stimulus was attended at the time of stimulus offset (see for example Field, Mogg, Zetteler, & Bradley, 2004; Field & Cox, 2008). However, attention has been conceptually distinguished in two underlying processes of attentional selectivity (Posner, 1980; Posner & Peterson, 1990) - engagement and disengagement - which might independently contribute to an AB as exhibited in substance use disorders. This means that AB might be characterized by attentional engagement, reflecting a lower threshold for attention to be shifted towards substance-relevant cues, and/or a difficulty to disengage attention, reflecting an inordinate tendency for attention to remain focused on substance-relevant cues. Therefore, to further understand the role of attention in addiction it seems essential to disentangle the processes that underlie the action of paying selective attention to substance-relevant cues. A better understanding of the underlying processes of AB seems especially relevant as it comes to clinical implications for treatment (see second part of this dissertation). However, the assessment tasks previously described are not configured to differentiate between these two components of attention (e.g.,

(14)

1

Grafton & MacLeod, 2014). Nevertheless, there have been attempts to differentiate between the two processes using the visual probe task by varying the duration of the stimulus presentation (e.g., Field et al., 2004; Noel et al., 2006). For example, stimuli presentation times of 250 ms or 2000 ms are thought to represent relatively greater engagement and disengagement processes, respectively. However, the interpretation of findings that result from these various stimulus presentation durations has been inconsistent across studies, meaning that there was no consensus about which stimulus presentation duration would index attentional engagement, and which duration would index difficulty to disengage (Field & Cox, 2008). As a next step, it seems therefore important to measure AB with a task that can deliver distinct indices for engagement and disengagement bias.

In addition to the fact that the most common AB assessment tasks are not configured to differentiate between processes of attentional engagement and disengagement, other methodological issues of these tasks have been discussed in the literature which might contribute to the inconsistency of the findings. First, several studies found evidence indicating that the reliability of the tasks in terms of internal consistency and test-retest stability is insufficient (Ataya et al., 2012; Brown et al., 2014; Cisler, Bacon, & Williams, 2009; Schmukle, 2005). This might be problematic as only the use of a stable and reliable task would provide confidence in the tasks’ ability to capture individual differences in AB (e.g., as a function of substance use), and its changes over time. Second, it has been questioned whether current assessment tasks adequately model the key features of real-life substance use behaviors. This argument relates to the relatively confined stimulus representation of these tasks in which typically only one or two stimuli are being presented during a particular trial. This restricted number of stimuli may limit the transference of results into real-life substance use-related situations; For example, situations as entering a supermarket where a person is confronted with multiple stimuli rather than only one or two stimuli at the same time (Hertel & Mathews, 2011). Third, and in line with the previous point, this confined representation of stimuli might be problematic as the (simultaneously) presentation of a maximum of two stimuli is suggested to not sufficiently challenge the attentional system (Hertel & Mathews, 2011). Evidence for this idea comes from a study showing that the strength of cognitive biases may depend on the magnitude to which the attentional system is challenged (Evans, Craig, Oliver, & Drobes, 2011). Taken together, when further investigating the role of AB in addiction, it seems critical to use a task that is able to differentiate between attentional processes of engagement and disengagement, is a more appropriate model of real-life substance use behaviours, and adequately challenges the attentional system.

(15)

OUTLINE OF PART I OF THIS DISSERTATION

The first part of this dissertation focusses on the assessment of AB. To follow up on previous studies and to account for the methodological issues of previously used assessment tasks, as described above, we aimed to investigate the utility of two assessment tasks with a more complex task configuration, one of which also delivers two distinct indices for engagement and disengagement. Although there is little substance use disorder research with these tasks, they have been successfully applied in the context of other psychological disorders (Hollitt, Kemps, Tiggemann, Smeets, & Mills, 2010; Melles, Dewitte, Ter Kuile, Peters, & de Jong, 2016; de Voogd, Wiers, & Salemink, 2017).

The first task is the Visual Search task (VST; de Voogd, Wiers, & Salemink, 2014), which is comprised of a stimuli matrix in which participants actively search for a target stimulus. AB is indexed by the difference in reaction times between trials in which one has to identify a substance-relevant target among neutral distractors versus trials in which a neutral target is presented among substance-relevant distractors. It is assumed that stronger AB is reflected in faster reaction times to trials in which the target is a substance-relevant cue compared to trials in which the target is a neutral cue. The second task is the Odd-One-Out task (OOOT; Hansen & Hansen, 1988; Rinck, Reinecke, Ellwart, Heuer, & Becker, 2005), in which one has to indicate whether or not all stimuli in a matrix belong to one stimulus category or whether there is an odd-one-out. As this task includes a trial type in which a neutral cue is presented among neutral distractors from another stimulus category, distinct indices can be calculated for engagement bias and disengagement bias. That is, engagement bias is indexed by the difference in reaction times between trials in which a substance-relevant cue is presented among neutral distractors, and trials in which a neutral cue is presented among neutral distractors from another category. In contrast, disengagement bias is indexed by the difference in reaction times between trials in which one neutral cue is presented among substance-relevant distractors, and trials in which a neutral cue is presented among neutral distractors from another category. To investigate the potential of both the VST and the OOOT as an AB index, we examined the internal consistency and the test-retest reliability, as well as the association between the AB indices and self-reported alcohol use, craving, and alcohol use problems in an analogue student sample. Given that the differentiation between engagement bias and disengagement bias has been suggested to be relevant, as a next step, we further investigated the potential of the OOOT as an assessment task for AB by comparing the AB indices of two clinical samples with two matched healthy samples.

(16)

1

PART II – ATTENTIONAL BIAS MODIFICATION

If AB contributes to the development and persistence of substance use disorders, the direct modification of AB might positively contribute to treatment outcome. However, as has been mentioned above, given that current treatments such as cognitive behavioural therapy do not target AB, AB is likely to remain unchanged throughout treatment (Van Hemel-Ruiter, Wiers, Brook, & de Jong, 2016). As a consequence, new interventions have been developed with the specific aim of directly modifying AB, the so-called Attentional Bias Modification (ABM) interventions. These computerized interventions are aiming at the reduction of the attention-grabbing properties of substance-relevant cues by directly training people’s attention away from these cues. Ideally, an effective ABM intervention therefore results in (i) a reduced AB towards substance-relevant cues, and (ii) clinically relevant changes of substance use disorder symptoms.

ATTENTIONAL BIAS MODIFICATION INTERVENTIONS

In order to investigate whether the modification of AB actually reduces AB, and whether this reduction is translated into clinically relevant changes of symptoms, thus far the majority of ABM studies used modified versions of the visual probe task (MacLeod, Rutherford, Campbell, Ebsworthy, & Holker, 2002). The modified visual probe task differs from the assessment task with regard to the ratio of probes that appear at the location of the substance-relevant cue. That is, whereas the probe in the assessment version of the task appears equally often at the location of the substance-relevant as well as the neutral cue, in the ABM version the probe mostly or always (different ratios have been used) appears in the location of the neutral cue. Thereby, AB is assumed to be retrained by teaching the participant to shift attention away from the substance-relevant cue and toward a neutral cue. Using this paradigm, some studies have found that the successful modification of AB resulted in changes of substance use-relevant symptoms (e.g., Kerst & Waters, 2014; Schoenmakers et al., 2010), whereas other studies failed to find such effects (e.g., Lee & Lee, 2015; Schoenmakers, Wiers, Jones, Bruce, & Jansen. 2007). Another ABM intervention is the so-called Alcohol Attention Control Training Program (AACTP; Fadardi & Cox, 2009), which consists of three phases of which the third phase is the actual training phase. In this last phase of the training, participants are instructed to identify the coloured outline of a neutral cue (i.e., image) which is simultaneously presented next to a substance-relevant cue. Thereby, participants’ attention is supposed to be re-trained because participants are instructed to attend to the colour around the neutral cue. Thus far, several studies have used this paradigm and found changes of substance use-relevant symptoms with regard to alcohol use (e.g., Fadardi & Cox, 2009), as well as drug use (e.g., Ziaee, Fadardi, Cox, & Yazdi, 2016).

(17)

EFFECTIVENESS OF ATTENTIONAL BIAS MODIFICATION INTERVENTIONS

When looking at the reviews that investigated the effectiveness of the ABM interventions described above, results with regard to clinically relevant changes of symptoms are mixed (Christiansen, Schoenmakers, & Field, 2015; Cristea, Kok, & Cuijpers, 2016). However, there are two factors that may have influenced the results of these reviews. First, it has been argued that it is relevant to distinguish between lab-based studies and studies testing the effects of ABM interventions in clinical samples (see for example Field, 2016; Wiers, 2016). That is, whereas lab-based studies are aiming at the exploration of underlying processes and mechanisms, and usually include non-clinical samples, non-clinical studies are interested in the effectiveness of the intervention in terms of clinical relevance (i.e., clinically relevant symptom changes). Given their distinct research goals, synthesizing studies without distinguishing between these two types of studies may result in a biased representation of results. In line, and maybe even more important, the effectiveness of ABM interventions with regard to symptom changes may be dependent on whether or not the sample is motivated to change (Wiers, 2016). Generally, lab-based studies are conducted in non-clinical samples who are most often uninterested in changing their substance use. Therefore, lab-based studies using ABM procedures may find a successful modification of AB, but are unlikely to change actual substance use behaviour. As a consequence, when aiming to test the effectiveness of ABM intervention in terms of symptom changes, a clinical sample should be included that is motivated to change their substance use behaviour. Second, the emphasis of the reviews is whether the ABM interventions lead to clinically relevant changes. However, when elaborating on the effectiveness of the studies it also seems important to not only focus on whether or not individuals show changes on symptoms, but to also investigate AB at baseline, and its changes from baseline to post-test. If AB does not change (i.e., no successful modification could be achieved), no changes in clinically relevant substance use disorder symptoms might be expected (MacLeod & Grafton, 2016).

Besides the fact that previous synthesis of results may have been influenced by the factors mentioned above, there are other factors to consider which may influence whether or not the modification of AB is effective by means of clinically meaningful changes of substance use disorder symptoms. One potentially important limitation of previously described ABM interventions concerns the simplicity of these interventions that have been employed with the aim of altering real-life AB. That is, the simplistic configuration of these interventions may constrain their therapeutic impact. In line with the described methodological issues of the equivalent assessment tasks, presenting only two stimuli on each trial of the ABM intervention may insufficiently challenge the attentional system, and as a result limit the possibility to achieve long lasting changes. Further, it might also be an inappropriate representation of real-life

(18)

1

substance use situations (Hertel & Mathews, 2011). Thereby, the transfer of changes in AB into symptom changes might be limited. A second factor that may moderate the efficacy of ABM interventions is the number of training sessions. Several lab-based studies have shown that the modification of AB is possible using a single session of ABM intervention. However, these studies have not found co-occurring changes of symptoms (Field et al., 2007; Lee & Lee, 2015; Schoenmakers, Wiers, Jones, Bruce, & Jansen, 2007). This suggests that multiple sessions are likely to be necessary in order to see a transfer of modified AB into changes of substance use disorder symptoms. Consistent with this idea, first studies have shown that multiple sessions of ABM intervention can results in both, modified AB and relevant changes of symptoms (Kerst & Waters, 2014; Schoenmakers et al., 2010; Rinck, Wiers, Becker, & Lindenmeyer, 2018). A third important consideration concerns the context in which the ABM intervention is delivered. Studies in the field of anxiety research have shown that ABM interventions might be effective in a particular context while not being effective in another (Mogg, Waters, & Bradley, 2017). As suggested by the anxiety literature, ABM interventions may be more effective when delivered in an environment in which symptoms are pronounced or relevant (e.g., a social situation in case of social anxiety). In case of substance use disorders, the home-environment seems therefore a more appropriate environment than a clinic in which the use of substances is prohibited.

OUTLINE OF PART II OF THIS DISSERTATION

The second part of this dissertation focusses on the modification of AB and examines whether ABM interventions might be a valuable addition to current addiction treatment. Thus far, reviews investigating the effectiveness of ABM interventions in addiction have reported mixed results. As described above, this might be related to the way these reviews synthesized the literature – not differentiating between lab-based and clinical studies, and not taking baseline AB and its (successful) modification into account. Therefore, the second part of this dissertation starts with a systematic review synthesizing the literature on the effectiveness of ABM in addiction, in terms of clinically relevant changes of symptoms, when taking these two factors into account. Next, taken further into account that (1) the extent of task complexity might not only be relevant in case of the assessment of AB, but also to attain clinically relevant changes after ABM, (2) multiple sessions of ABM might be necessary to see a transfer of modified AB into changes of substance use disorder symptoms, and (3) the environment in which ABM is delivered might be critical, we designed a multicentre randomized controlled trial to test the effectiveness of a novel more complex, multi-session, and home-based ABM intervention. This ABM intervention, called the Bouncing Image Training Task (BITT; based on Notebaert et al., 2018), has a more complex configuration than previously

(19)

used ABM interventions with regard to the number of stimuli that is simultaneously presented, and its dynamic rather than static character. This makes the BITT more likely to adequately challenge the attentional system and to be a better representation for real life substance use situations than the previously described interventions. The BITT was delivered as a multi-session intervention and was provided online so that it could be completed at home. Specifically, the BITT consists of eight images moving across the screen. One of these images contains substance-irrelevant content and participants are asked to track this image with the mouse cursor, whereas the other seven images contain substance-relevant content and should therefore be ignored. The substance-irrelevant cue occasionally changes into a substance-relevant cue, at which point the participant is instructed to find the new position of the substance-irrelevant cue (thereby disengaging attention from the substance-relevant cue). These task elements were designed to improve the effectiveness of ABM, as it has been suggested that ABM interventions consisting of multiple cues and explicitly instructing participants to engage with a disorder-irrelevant cue while ignoring the disorder-relevant distractors, might help to successfully modify AB (Waters & Craske, 2016). In the context of eating behaviour research, the BITT has successfully been used to modify AB for food cues (Jonker et al., 2019). Given that thus far only a limited number of studies have been conducted in clinical samples, and the need for well-powered clinical trials in order to add value to the current state of knowledge, we included treatment-seeking individuals who were diagnosed with either alcohol use disorder or cannabis use disorder.

IN SUM

The aim of this dissertation is twofold. The first part of the dissertation focuses on the assessment of AB towards substance-relevant cues. Therefore, in Chapter 2, we examined the potentials of two assessment tasks, the VST and the OOOT, as a measure for AB towards alcohol-relevant cues in an analogue student sample. In particular, we were interested whether AB is expressed by heightened engagement, difficulty to disengage or both. Chapter 3 describes a cross-sectional study in which AB was assessed using the OOOT in treatment-seeking individuals diagnosed with either alcohol use disorder or cannabis use disorder. The clinical sample’s indices of engagement bias and disengagement bias were compared with two matched control groups. Given that the second part of this dissertation concentrates on the manipulation of AB and its effectiveness, in Chapter 4 the literature on ABM interventions in substance use disorders was synthesized in a systematic review. In Chapter 5 and in Chapter 6 of this dissertation a multicentre randomized controlled trial is described. In this trial we tested the effectiveness of a multi-session, internet-based and home-delivered

(20)

1

ABM intervention as an add-on to treatment as usual for individuals diagnosed with alcohol use disorder or cannabis use disorder. Finally, in Chapter 7 the findings of the studies as presented in Chapter 2 to 6 are integrated and discussed. Future research directions are discussed and attention is paid to potential clinical implications.

(21)

(22)

Attentional bias for alcohol cues in visual

search: Increased engagement, difficulty

to disengage, or both?

Heitmann, J. Jonker, N. C. Ostafin, B. D. De Jong, P. J.

(23)

ABSTRACT

Background: Cognitive models emphasise the importance of attentional bias in

addiction. However, many attentional bias tasks have been criticised for questionable psychometric properties and inability to differentiate between engagement and disengagement processes. This study therefore examined the suitability of two alternative tasks for assessing attentional bias within the context of alcohol use.

Methods: Participants were undergraduate students (N=169) who completed the Visual

Search Task and Odd-One-Out task, the latter of which is designed to differentiate between engagement and disengagement processes of attention, at baseline and one week later. Participants also completed baseline measures of alcohol consumption, craving, and alcohol use problems.

Results: Internal consistency was adequate for the Visual Search task index, and weak

for the Odd-One-Out task indices. Test-retest reliability was weak for both tasks. The Visual Search task index and the disengagement (but not the engagement) index of the Odd-One-Out task showed a positive association with alcohol consumption.

Conclusions: This study was restricted to a non-clinical student sample. The relatively

high error rate of the Odd-One-Out task might have reduced its sensitivity as an index of attentional bias. Both tasks showed some merit as attentional bias measures, and results suggested that attentional disengagement might be particularly related to alcohol use. However, the reliability of the current measures was inadequate. One potential explanation for the low reliability is that non-clinical samples may have weak and unstable attentional biases to alcohol. Future efforts should be made to improve the psychometric qualities of both tasks and to administer them in a clinical sample.

Keywords: Attentional bias, alcohol addiction, visual search, engagement,

(24)

2

INTRODUCTION

Current cognitive models of addiction emphasise the importance of selective visual attention in the persistence of addiction (Stacy & Wiers, 2010; Wiers et al., 2007). More specifically, increased attention for substance cues has been associated with the intensity of craving (Field, Munafò, & Franken, 2009; Franken, 2003). In turn, craving may increase attentional capture of substances, and/or the difficulty to disengage attention from these cues. As a result, people may enter a self-reinforcing bias-craving-bias cycle that may play an important role in the persistence of addiction. These attentional tendencies, known as attentional bias (AB), have therefore received a lot of interest in the field of substance use disorders (Field et al., 2016). Related research has resulted in an improved understanding of the disorders and also in the development of new treatment tools (e.g., attentional bias modification trainings). However, these developments also draw attention to the limitations of existing measures as indices of individual differences in AB.

One of the most frequently used paradigms to measure substance-related AB is the visual probe task (MacLeod, Mathews, & Tata, 1986). Although the visual probe task is widely used, it is not without its critics. First, the reliability of the task in terms of internal consistency and test-retest reliability has been found to be poor (Ataya et al., 2012; Cisler, Bacon, & Williams, 2009; Schmukle, 2005). A low test-retest reliability is especially problematic when the task is used to measure changes in AB. To account for the low reliability of the visual probe task and other indirect measures of AB, several studies have tested whether more direct measures of attention by the use of eye-tracking can serve as an alternative (Christiansen, Mansfield, Duckworth, Field, & Jones, 2015; Roy-Charland et al., 2017). However, although eye-tracking can measure overt attention shifts (responding by directly moving and focussing the gaze on a target), it does not measure covert attention shifts (responding by seeing something peripherally without directly focussing the gaze on a target). Eye-tracking might be a valuable addition to this field of research, but as also encouraged by other researchers (van Duijvenbode, Didden, Korzilius, & Engels, 2017) a task that can reliably measure the influence of covert attention shifts remains also desirable (see for example attempts to improve the reliability of the visual probe task; Gladwin, 2019). Second, it has been questioned whether the presentation of just one pair of stimuli, such as in the visual probe task, can adequately reflect the complexity of real-life substance use relevant situations in which a person is constantly surrounded by multiple stimuli (Hertel & Methews, 2011). To make attentional bias tasks more compatible with the complexity of real-life substance use situations, it seems essential to more closely mimic this complexity by the use of a more complex stimulus configuration. One paradigm with a more complex stimulus configuration is the flicker-induced change

(25)

blindness paradigm (Rensink, O’Regan, & Clark, 1997), which has been adapted to measure AB towards alcohol cues. Using this paradigm, several studies have found differences in AB between heavy and social/light drinkers (Jones, Bruce, Livingstone, & Reed, 2006; Jones, Jones, Smith, & Copley, 2003). However, the reliability of this task remained unexplored.

Another promising example of a paradigm with a more complex stimulus configuration is the visual search paradigm. In this paradigm participants identify a target stimulus among a series of distractors (Hansen & Hansen, 1988). The visual search paradigm has been successfully applied in the context of anxiety disorders (de Voogd, Wiers, & Salemink, 2017), eating disorders (Hollitt, Kemps, Tiggemann, Smeets, & Mills, 2010), and pain disorders (Melles, Dewitte, ter Kuile, Peters, & de Jong, 2016). Yet, this paradigm has been largely ignored in the field of addiction, with the exception of one promising study assessing AB in smokers (Oliver & Drobes, 2012). The aim of the current study is therefore to examine the potential of the visual search paradigm as an index of AB in the context of alcohol use.

There are two sub-types of the visual search paradigm, which differ with respect to the indices of AB they assess. In the first type of task, often referred to as the Visual Search Task (VST), individuals actively search for a target stimulus in a matrix of distractors (de Voogd, Wiers, Prins, & Salemink, 2014). In the case of an AB for alcohol, participants are expected to be faster in detecting an alcohol cue in an array of non-alcohol distractors than detecting a non-non-alcohol cue in an array of non-alcohol distractors. Due to the typical methods of the task and the scoring procedure, interference effects might be due to either automatic orientation towards an alcohol cue (i.e., attentional engagement), maintaining attention on alcohol cues due to a difficulty to disengage (i.e., attentional disengagement), or both. However, similar to the visual probe task and the flicker-induced change blindness task, the VST is not able to differentiate between these different processes of attention.

The second sub-type of the visual search paradigm, called the Odd-One-Out task (Hansen & Hansen, 1988; Rinck, Reinecke, Ellwart, Heuer, & Becker, 2005), is able to deliver two indices that have been supposed to reflect attentional engagement and disengagement, respectively. In the OOOT, individuals indicate whether images in a matrix are from the same category of images or whether there is an odd-one-out (i.e., target image). The OOOT includes trials in which a neutral target is presented among neutral distractors drawn from a different category. This trial type can serve as a personal baseline of how long it generally takes to find a target between distractors, making it possible to calculate separate indices for increased engagement and difficulty to disengage. That is, attentional engagement is inferred from the relative speeding to correctly respond that an odd-one-out is present when an alcohol stimulus is presented

(26)

2

among neutral distractors, compared the time taken to a correct response when a neutral stimulus is presented among neutral distractors (from a different neutral category; cf. Hollitt et al., 2010). Difficulty to disengage from alcohol cues is inferred from the relative slowing to correctly respond that an odd-one-out is present when a neutral stimulus is presented among alcohol distractors, compared the time taken to a correct response when a neutral stimulus is presented among neutral distractors (from a different neutral category; cf. Hollitt et al., 2010).

The differentiation between attentional engagement and disengagement might be relevant, as earlier research in different psychopathologies indicated that different disorders might be characterised by one of these attentional processes or both (Hollitt et al., 2010; Rinck et al., 2005). More detailed knowledge about these attentional processes might not only be essential for the general understanding of the disorder, but particularly important when it comes to treatment indications, for example in terms of testing the efficacy of attentional bias modification interventions that can be tailored to directly target the relevant process to improve treatment outcome. Given that earlier studies in substance use have mainly used tasks that were not able to differentiate between increased engagement and difficulty to disengage, such as the visual probe task, the interpretation of attentional indices may have contributed to the inconsistency of findings across studies (Field & Cox, 2008). A task, such as the OOOT, that is able to make this differentiation seems therefore relevant to further disentangle the AB processes that contribute to substance use.

The aim of the current study was to investigate the potential of both the VST and the OOOT as an index of AB for alcohol cues. As a first step, we examined the tasks’ reliability. Having strong internal consistency and test-retest reliability would provide confidence in the tasks’ ability to capture individual differences in AB and its changes. As a next step, we tested the (predictive) validity of the tasks by examining the association between their indices of AB with self-reported alcohol consumption (i.e., quantity and frequency), craving, and alcohol use problems. In particular, we were interested in whether the two underlying processes of AB assessed by the OOOT– attentional engagement and disengagement – were differentially related to alcohol use. We expected that if the VST and the OOOT are adequate methods for assessing individual differences in AB for alcohol cues, the AB measures will be positively associated with the drinking outcome measures.

(27)

METHOD

PARTICIPANTS

Participants were 169 undergraduate Psychology students (18.3% male) with a mean age of 20.55 years (SD = 2.80; age and gender were unknown for one participant) from the University of Groningen. Initially, 182 participants participated in the study, but 13 participants dropped out after the first assessment. Therefore, their data were not included in the current analyses.

MATERIAL

ALCOHOL USE AND CRAVING

The quantity of used alcohol, the frequency of use, and craving for alcohol were assessed with the Measurements in Addiction for Triage and Evaluation Questionnaire (MATE-Q; Schippers, Broekman, Buchholz, 2011). The quantity of the past 30 days was indexed by the amount of standard glasses of alcohol consumed on a typical Monday, Tuesday, Wednesday, etc. The answers per week were multiplied by four to estimate the consumed standard units of alcohol over the past month. To index the frequency of use, participants were asked to indicate on how many of the last 30 days they consumed alcohol. The Obsessive-Compulsive Drinking Scale (OCDS5) of the MATE-Q was used to assess general craving. It consists of five items measuring the desire for alcohol in the past seven days. Participants indicated their answers on a 5-point Likert scale. Reliability of the OCDS5 as estimated with Cronbach’s alpha was poor (

α

= .49). This seemed to be related to item four of the questionnaire (i.e., How much of an effort do you make to resist these thoughts or try to disregard or turn your attention away from these thoughts as they enter your mind when you’re not using?), which among clinicians is known to be difficult to understand. When excluding this item, Cronbach’s alpha was acceptable (

α

= .75). In addition to the OCDS5, state craving was assessed by asking participants to indicate their amount of craving for alcohol directly before performing the VST and the OOOT on a Likert scale from 1 (no craving) to 7 (extreme craving).

ALCOHOL USE PROBLEMS

Problems with alcohol use were assessed with the Rutgers Alcohol Problem Index (White & Labouvie, 1989) consisting of 18 questions. Participants indicated on a 5-points Likert scale, ranging from ‘never’ to ‘very often’, how often they had experienced

(28)

2

the described situation in the past. Cronbach’s alpha of this questionnaire was good (

α

= .85).

VISUAL SEARCH TASK (VST)

Each trial of the VST started with a red fixation cross appearing in the centre of the screen (de Voogd et al., 2014). Participants located the mouse cursor on the cross to start the trial, after which a 4 x 4 matrix of 16 images (500 x 500 pixels) appeared. Fifteen of the 16 images belonged to the same category of images, whereas one image was different (i.e., target stimulus). Participants clicked as quickly and as accurately as possible on the target image of the search category that was named before the task started. The images belonged to one of the following two categories: alcoholic drinks or non-alcoholic drinks. The next trial appeared after a 1000 ms inter-trial interval. The task was divided into four blocks, two blocks of which the target was an alcohol image in an array of non-alcoholic drinks distractors (i.e., alcohol target trials), and two blocks in which the target was a non-alcoholic drink image that had to be found among alcohol distractors (i.e., alcohol distractors trials). Each block consisted of 18 trials, and per block the target was presented in all possible positions in a random order. The blocks were presented alternately, and the first block was chosen randomly for each participant (Figure 1). AB index was calculated by subtracting the mean reaction time of alcohol target trials from the mean reaction time of alcohol distractors trials. Higher positive scores reflected stronger AB for alcohol.

(29)

ODD-ONE-OUT TASK (OOOT)

In the OOOT, participants focused their attention on a red fixation cross in the centre of the screen for 500 ms (Hansen & Hansen, 1988). Afterwards a 4 x 5 matrix of 20 images (500 x 500 pixels) appeared, and participants indicated whether an odd-one-out image (i.e., target stimulus) was present. Responses were given by pressing the ‘0’ (no odd-one-out) or ‘1’ (yes, odd-one-out present) button of the keyboard. Participants had a maximum of 10 seconds to respond, and the task was divided into three blocks of 24 trials each. The order of trials was random, and for each block the target randomly appeared over the possible positions, but never directly above or below the fixation cross. Each image belonged to one of the following three categories: alcoholic drinks, non-alcoholic drinks, or flowerpots. Given these three different categories of images, nine different combinations of trials were possible. For each block, an equal number of trials was included for all trial types without an odd-one-out (6) and all trial types including an odd-one-out (18; Table 1; Figure 2). Engagement bias (i.e., engagement index) was calculated by subtracting the mean reaction time of the alcohol target trials (i.e., alcohol target in either non-alcoholic drinks distractors or flowerpot distractors) from the mean reaction time of the neutral target in neutral distractors trials (i.e., non-alcoholic drinks target in flowerpot distractors or flowerpot target in non-non-alcoholic drinks distractors). Higher positive scores were expected to reflect more attentional engagement with alcohol cues. Disengagement bias (i.e., disengagement index) was calculated by subtracting the mean reaction time of the neutral target in neutral distractors

trials from the mean reaction time of the alcohol distractors trials (i.e., non-alcoholic

drinks target or flowerpot target in alcohol distractors). Higher positive scores reflected more difficulty to disengage attention from alcohol cues.

STIMULI

Both tasks used the same images of alcoholic drinks and non-alcoholic drinks (van Hemel-Ruiter, Wiers, Brook, & de Jong, 2016; Pronk, van Deursen, Beraha, Larsen, Wiers, 2015). The OOOT additionally contained images of flowerpots (Heitmann et al., 2017). In total, 90 images were used - 30 different images per category (alcoholic drinks, non-alcoholic drinks, and flowerpots). The images were randomly drawn in each trial.

(30)

2

Table 1

Type and amount of trials in the Odd-One-Out task (OOOT)

Trial type Trials per block

1. Alcohol (20) 2

2. Non-alcoholic drinks (20) 2

3. Flowerpot (20) 2

Target Distractors

4. Alcohol (1) Non-alcoholic drinks (19) 3

5. Alcohol (1) Flowerpot (19) 3

6. Non-alcoholic drink (1) Alcohol (19) 3

7. Flowerpot (1) Alcohol (19) 3

8. Non-alcoholic drink (1) Flowerpot (19) 3 9. Flowerpot (1) Non-alcoholic drinks (19) 3

Note. Number of presented images per trial is given in parentheses. Trial numbers 4 and 5 (i.e., alcohol

target trials), trial numbers 6 and 7 (i.e., alcohol distractors trials) and trial numbers 8 and 9 (i.e., neutral target in neutral distractors trials) were included in the current analyses.

(31)

PROCEDURE

The study was approved by the ethical committee of the psychology faculty of the University of Groningen, and data was collected from October to November 2016 in a quiet laboratory. On arrival, participants provided informed consent. They started with indicating their state craving, and thereafter continued with either the OOOT or the VST. The order of the two tasks was counterbalanced. After finishing both tasks, participants filled in the questionnaires. In order to assess test-retest reliability participants returned to the laboratory after exactly one week. At the second visit, participants again completed both tasks in the same order as the first time. After the completion of the study, all participants were debriefed. Participants received course credits or financial compensation in return for their participation.

ANALYSES

To investigate the internal consistency of both tasks, the split-half method was used to calculate Spearman-Brown coefficient between the first half and the second half of the baseline task. To account for a possible learning effect throughout the task a second method to calculate Spearman-Brown coefficient was used by distributing the trials alternately to one of two subsets. The first trial of one particular trial type was randomly allocated to either of the subsets. Split-half reliability was tested for the indices as well as for the trial types of both tasks. The test-retest reliability was also calculated for the trial types and for the indices using Pearson correlations. The estimates for the internal consistency and test-retest reliability were characterised as weak (r <.5), adequate (.5 ≤ r < .8), or good (r ≥ .8) based on commonly reported thresholds (Clark & Watson, 1995). As a second step, repeated measures analyses of variance (RM-ANOVA) were conducted to test whether the performance was stable over time, with trial type (i.e., alcohol target trials and alcohol distractors trials, and neutral target in neutral distractors trials for the OOOT) as independent variable and time (baseline and post-test) as dependent variable. To test to the extent to which alcohol consumption (i.e., quantity and frequency) as assessed with the MATE-Q, general craving as indexed with the OCDS5, and alcohol use problems as assessed with the RAPI-18 were associated with AB as measured with both tasks, bivariate correlational analyses were computed. Consistent with previous research testing the relationship between cognitive performance measures and alcohol use/problems (Wiers, van Woerden, Smulders, & de Jong, 2002), this part of the analyses excluded the data of participants (n = 20) who reported that they did not drink alcohol during the last 30 days. Based on power analyses prior to the study, we aimed for a sample size of at least 150 participants to be able to detect a small to moderate correlation (r = .20) with a power of 80% at an alpha level of 0.05.

(32)

2

The current sample (n = 169) provided 84% power to detect a small-moderate correlation at an alpha of 0.05.

RESULTS

DATA PREPARATION

Participants scoring 3SD’s below the mean percentage correct answers (baseline < 97.53%; post-test < 98.09%) were removed (baseline n = 5; post-test n = 2), because high numbers of incorrect responses might indicate non-serious participation. In line with Hollitt and colleagues (2010), as a next step, incorrect responses were excluded from the analyses (baseline 1.8%; post-test 1.8%). No reaction times were below 200 ms, which were considered anticipation errors. Finally, outliers were calculated based on participants’ average response time per type of trial. Trials scoring 3 SD’s below or above participants’ average response time were removed. This resulted in deleting another 1.83% of trials from the baseline and 1.79% of trials from the post-test.

Identical with the procedure for the VST, participants scoring 3SD’s below the mean percentage correct answers (baseline < 82.40%; post-test < 88.00%) were removed (baseline n = 2; post-test n = 2). As a next step, incorrect responses (i.e., participants indicated that there was no odd-one-out although an odd-one-out was present) were excluded from the analyses (baseline 20.00%; post-test 13.8%). Further examination indicated that the highest number of errors were made in alcohol distractors trials when a non-alcoholic drink image was the target (baseline 42.60%; post-test 31.20%), and in

alcohol target trials in which non-alcoholic drinks images were the distractors (baseline

32.00%; post-test 25.00%). At baseline, participants made significantly more errors when non-alcoholic drinks were used as contrast category compared with the number of errors that were made when flowerpots were used as contrast category (t(332) = 13.28,

p <.001; t(332) = 8.71, p <.001, respectively for alcohol distractors trials and alcohol

target trials). There were no significant differences concerning the number of errors for neutral target in neutral distractors trials (t(332) = 1.36, p = .174). Anticipation errors (reaction times < 200 ms) were removed from the analyses (baseline 1 trial; post-test 8 trials). No further outliers based on participants’ average response times per type of trial, following the 3SD’s rule, were detected.

(33)

DESCRIPTIVE STATISTICS

The average quantity of drinks consumed during the last 30 days was 47.03 (SD = 59.73), and the mean number of days in which alcohol was consumed during the last 30 days was 4.64 (SD = 4.58). General craving for alcohol, as measured with the MATE-Q, was 7.05 (SD = 2.23) – a value under the critical cut-off of 12 for pathological craving (Schippers, Broekman, & Buchholz, 2011), indicating that the current sample had on average a non-pathological level of craving. On an individual level, five participants had a score larger than or equal to the cut-off for pathological craving. The ratings on state craving indicated a generally low level of craving for alcohol directly before the task (M = 1.44, SD = 0.95). General craving of the past seven days was significantly related with state craving before the tasks (r = .37, p < .001). Alcohol use problems were in line with normal values of a nonclinical sample (M = 8.80, SD = 7.54; White & Labouvie, 1989).

For the VST, the mean index of AB was 375.09 ms (SD = 580.56) at baseline, and 286.73 ms (SD = 478.35) at post-test. At baseline, the mean of the engagement index of the OOOT was -516.34 ms (SD =420.92), and the mean for the disengagement index was 823.19 ms (SD = 537.84). At post-test, the mean of the engagement index was -571.49 ms (SD = 444.06), and of the disengagement index 838.84 ms (SD = 518.86). Table 2 shows the mean reaction times of all trial types. See the supporting information for the correlations between the VST index and both OOOT indices (Appendix A).

(34)

2

Table 2 Alcohol target trials, alcohol distractors trials and neutral target in neutral d

istractors trials as measured with the Odd-One-Out task and

the Visual Search Task at baseline and post-test

Baseline

Post-test

Alcohol target trials

Alcohol distractors trials Neutral target in neutral distractors trials Alcohol target trials Alcohol distractors trials Neutral target in neutral distractors trials

OOOT 2524 (590) 2831 (760) 2008 (486) 2178 (518) 2446 (620) 1607 (335) VST 2997 (642) 3372 (794) -2748 (543) 3035 (641) -Note.

Means and standard deviations are given in ms

(35)

INTERNAL CONSISTENCY OF ATTENTIONAL BIAS MEASURES

Internal consistency of each of the trial types was good (Table 3). Internal consistency between the first half and the second half of trials for the AB index at baseline was .57. When trials were alternately distributed to either of the two subsets, Spearman-Brown coefficient was .59.

Internal consistency of each of the trial types was adequate (Table 4). Spearman-Brown coefficient for the attentional indices when comparing the first half with the second half was -.10 for the engagement index, and .22 for the disengagement index. When trials were alternately distributed to either of the two subsets, internal consistency of the engagement index was .34, and .33 for the disengagement index.

Table 3

Internal consistency reported for the split-half and random distribution method per trial type for the Visual Search Task

Method Trial type

Alcohol target trials Alcohol distractors trials

Split-half .83 .84

Random distribution .90 .83

Table 4

Internal consistency reported for the split-half and random distribution method per trial type for the Odd-One-Out task

Method Trial type

Alcohol target trials Alcohol distractors trials Neutral target in neutral distractors

trials

Split-half .65 .66 .60

(36)

2

TEST-RETEST RELIABILITY OF ATTENTIONAL BIAS MEASURES

There was a weak positive correlation between AB at baseline and at post-test that was statistically significant (r = .16, p = .040). Higher scores of AB at baseline were related to higher scores at post-test. Alcohol target trials showed adequate test-retest reliability (r = .58, p < .01); also alcohol distractors trials showed adequate test-retest reliability (r = .68, p < .01). The RM-ANOVA showed that there was a significant effect of time (F(1, 161) = 114.22, p < .01,

η

2_{= .42), and trial type (F(1, 161) = 67.34, p < .01,}

η

2₌

.30). No significant interaction was found (F(1,161) = 2.57, p = .111,

η

2_{= .01), indicating}

similar temporal changes for both trial types. Paired sample t-tests were conducted to indicate whether both trial types differed significantly when comparing baseline with post-test outcomes. There was a significant difference from baseline to post-test for alcohol target trials (t (161) = 5.81, p = <.001), and alcohol distractors trials (t (161) = 7.32, p <.001). Participants became faster from baseline to post-test for both trial types (see Table 2 for the related means). The index of AB did not differ between baseline and post-test (t (161) = 1.60, p = .111).

There was no statistically significant relation between the baseline and post-test engagement indices (r = .09, p = .257). The baseline and post-test disengagement indices showed a weak positive correlation (r = .23, p = .003). Participants showing stronger disengagement bias for alcohol cues at baseline, also showed stronger disengagement bias at post-test. Alcohol target trials showed adequate test-retest reliability (r = .45, p < .001); alcohol distractors trials showed adequate test-retest reliability (r = .52, p < .001). Also neutral target in neutral distractors trials showed adequate test-retest reliability (r = .54, p < .001). RM-ANOVA showed that there was a significant main effect of time (F(1, 165) = 125.71, p < .001,

η

2_{= .43), and a significant main effect of trial type (F(2,}

330) = 429.90, p < .001,

η

2_{= .72). The non-significant interaction between time and}

trial type (F(2, 330) = 0.58, p = .563,

η

2_{< .01), indicated that the temporal changes were}

similar for all trial types. T-tests were conducted to indicate whether all trial types differed significantly when comparing baseline with post-test outcomes. There was a significant difference from baseline to post-test for alcohol target trials (t (165) = 7.86,

p = <.001), alcohol distractors trials (t (165) = 7.39, p <.001), and neutral target in neutral distractors trials (t (165) = 12.56, p <.001). Participants became faster from baseline to

post-test for all three trial types (see Table 2 for the related means). There was no difference between baseline and post-test for the indices of AB (engagement index: t (165) = 1.11, p = .257; disengagement index: t (165) = -0.26, p = .795).

(37)

RELATION BETWEEN ATTENTIONAL BIAS MEASURES AND OUTCOME

MEASURES

To examine whether the AB index of the VST was related with alcohol consumption, craving and/or alcohol use problems, correlational analyses were conducted. At baseline, AB showed a weak but significant relationship with the frequency of alcohol use (r = .20, p = .018), indicating that drinking more regularly was associated with higher scores of AB. All other correlations with the VST were small and non-significant, see Table 5. Also when the outcome variable craving was constructed without the item leading to poor internal consistency, the correlations with AB remained small and non-significant.

Table 5

Correlations between attentional bias indices as measured with the VST and the OOOT and the outcome measures

Quantity Frequency Craving Alcohol use problems

VST AB -.04 .20* .01 -.02

OOOT E -.02 -.01 -.02 -.03

OOOT D .18* .18* .02 .09

Note. VST AB = index of attentional bias measured with the Visual Search Task; OOOT E = index of

engagement as measured with the Odd-One-Out task; OOOT D = index of disengagement as measured with the Odd-One-Out task; *p< .05.

To investigate whether the indices of the OOOT were related with alcohol consumption, craving, and/or alcohol use problems, correlational analyses were conducted (Table 5). We found that specifically the disengagement index showed an association with the quantity of used alcohol (r = .18, p = .028) and with the frequency of consumption (r = .18, p = .026), although both associations were weak and only evident for alcohol consumption but not for craving or alcohol use problems (see supporting information for the results when also the non-drinkers were retained in this analysis, Appendix B). As for the VST, when the outcome variable craving was constructed with the item leading to poor internal consistency, the correlations with the engagement and disengagement index of the OOOT remained small and non-significant.