Attention bias modification in depression: A randomized trial using a novel, reward-based, eye-tracking approach

Highlights

• A novel, reward-based eye-tracking paradigm modified attention bias in depression.

• Visual feedback using an eye-tracker reduces time spent viewing negative information.

• Participants who receive training disengage more quickly from sad stimuli.

• Modifying attention from sad information can influence recall of valenced words.

Abstract

Background and objectives

Biased attention to negative information is a mechanism for risk and relapse in depression. Attentional bias modification (ABM) paradigms manipulate attention away from negative information to reduce this bias. ABM results have been mixed due to inconsistent methodologies and stimuli design. This randomized controlled trial used a novel approach to modifying attentional bias.

Methods

An eye tracker manipulated stimuli in response to participants’ fixations to preferentially reward attention to positive stimuli by obscuring or enhancing image quality of negative and positive stimuli, respectively. Participants with major depressive disorder completed three 35-min sessions of active (n = 20) or sham (n = 20) ABM training. Attentional bias, memory for emotional words, and mood were assessed pre- and post-training.

Results

Training reduced negative attentional bias; relative to sham, active training participants focused significantly more on positive compared to negative stimuli in a free-viewing eye-tracker task (p = .038, η_p² = 0.109) and, at trend, disengaged from sad information more quickly in a computerized task (p = .052, η_p² = 0.096). Active training participants remembered more happy than sad words in an emotional word learning task, indicating a distal transfer of training to emotional memory (p = .036, η_p² = 0.11). Training did not significantly affect mood in the one-week trial.

Limitations

Future studies should build on this proof-of-principle study with larger sample sizes and more intensive treatment to explore which mechanisms of training may lead to improvements in mood.

Conclusions

Attention biases in depression are modifiable through reward-based, eye-tracking training. These data suggest generalizability of training to other cognitive faculties - recall for affective information.

Introduction

Biased attention towards negatively-valenced information is thought to be a significant contributing factor in the development, maintenance, and recurrence of depression. Individuals with depression tend to avoid allocating attention toward positive stimuli and preferentially attend to negative stimuli (Gotlib et al., 2004; Koster et al., 2005; Joormann & Gotlib, 2007; Peckham et al., 2010, Duque & Vázquez, 2015; Lazarov et al., 2018). These patterns of attention have been associated with perseverative processing of negatively valanced information in the form of rumination, and impaired affective regulation due to difficulty disengaging from negative stimuli (Sanchez et al., 2013; Arditte & Joormann, 2014; Sanchez-Lopez et al., 2019; Yaroslavsky et al., 2019). Recent models have linked attentional disengagement and rumination, proposing that the inability to attentionally disengage from sad information promotes the maintenance of ruminative thinking (Whitmer & Gotlib, 2013; Yaroslavsky et al., 2019).

Negative biases in depression have been found to extend to other cognitive functions, such as information processing and memory (Duque & Vázquez, 2015). Compared to healthy participants, positive information is maintained for less time in working memory of depressed participants, whereas negative information is integrated faster and maintained for longer (Levens & Gotlib, 2010). Individuals with depression reliably show a bias towards the preferential processing, recall, and memory of sad information (Bradley et al., 1996; Joormann et al., 2007; Lloyd & Lishman, 1975). Thus, initial biased attention towards sad information may be compounded by difficulty disengaging from that information, ruminative thinking, and remembering more sad than happy information. Modifying initial biased attention has therefore been identified as a potential treatment focus, as early modification may be able to help prevent subsequent ruminative thought patterns and associated memory biases.

Attempts at intervention have typically focused on the use of dot-probe attention bias modification (ABM) paradigms. These procedures attempt to train attention towards positive stimuli and away from negative stimuli. In ABM, attention is diverted away from negatively valanced stimuli by preferentially placing a probe in the same location as a neutral or positive stimulus, thus influencing the focus of attention to be more frequently on the neutral or positive stimulus and less frequently on the negative stimulus (MacLeod et al., 2002). ABM paradigms show some effectiveness at decreasing attentional bias for negative information, increasing attention allocation for positive stimuli, and decreasing depressive symptoms (Wells & Beevers, 2010; Beevers et al., 2015; Li et al., 2015; Yang et al., 2015; LeMoult et al., 2016). However, the diverse methodology, stimuli, and design of ABM research has led to conflicting and inconsistent results, with other studies not significantly manipulating attention or improving mood (Kruijt et al., 2013; De Voogd et al., 2014; Everaert et al., 2015; Platt et al., 2015). Meta-analytic research has not supported the effects of dot-probe ABM for depression, finding limited evidence for symptom or mood improvements (Mogoaşe et al., 2014). The limitations of the dot-probe include its transient nature (capturing only momentary attention rather than naturalistic attention) and exclusive use of motor-based, indirect reaction time data to assess bias, which may be confounded (e.g., measuring motor impairment, processing speed, sustained attention) (Armstrong & Olatunji, 2012; Mogoaşe et al., 2014; Burris, Barry-Anwar, & Rivera, 2017). Poor internal consistency and test-retest reliability are also commonly reported (Schmuckle, 2005; Staugaard, 2009; Kappenman et al., 2014). Moreover, single-session dot-probe paradigms often fail to achieve significant effects, yet they remain common approaches (Kruijt et al., 2013; Arditte & Joormann, 2014; Everaert et al., 2015; Platt et al., 2015). The lack of positive results in such brief paradigms may be a combination of insufficient training duration and the use of unreliable paradigms such as the dot-probe (Everaert et al., 2015).

In addition to the use of the dot-probe and single-session designs, limitations in ABM research stem from the tendency to recruit either dysphoric or undergraduate populations with non-clinical levels of depressive symptoms (Ferrari et al., 2016; Möbius et al., 2018; Wells & Beevers, 2010; Yang et al., 2015). According to a meta-analysis by Armstrong and Olatunji (2012), individuals who meet the cut-off scores for major depressive disorder show more robust bias toward negative stimuli compared to those with lower or subclinical scores.

More recently, ABM studies have used eye tracking as a measure of real-time attentional biases with greater ecological validity than previous paradigms. Eye-tracking improves upon the dot-probe paradigm by providing a continuous measure of attention via eye movements, which are more proximally related to attention than manual key presses and thus may improve the efficiency of attention measurement (Arditte & Joormann, 2014; Armstrong & Olatunji, 2012). Eye-tracking paradigms have been found to have higher reliability than dot-probe measures (Wechsler, 1997a; Christiansen et al., 2015). Eye-tracking has been applied to the measurement of attentional bias through free-viewing tasks (Sanchez et al., 2013), as well as used in combination with the dot probe to determine when participants fixate on positive and negative stimuli; this ensures that attention is redirected towards positive stimuli and away from negative stimuli before a task progresses (Ferrari et al., 2016; Möbius et al., 2018). Beyond providing a more robust measure of bias, a lesser-explored application of eye-tracking is its potential as a treatment tool to directly modify biased attention. There is ample evidence to suggest that human emotional response can be modified by adapting the contingencies around a person based on their response (e.g., traditional biofeedback paradigms). More recently, different technologies (e.g., eye-tracking, fMRI) have been used to modify brain and behaviour response; for example, Young et al. (2017) found that fMRI-measured amygdala response to positive images could be modified using principles of biofeedback. A potential method for overcoming limitations of previous ABM research is to provide real-time feedback from tracking eye movements to modify the environment and shape behaviour. While the dot-probe leads to the shifting of attention as designated by the probe, eye-tracking can allow a person's initial behaviour (e.g., looking at a negative or positive stimulus) to be either reinforced or punished, ultimately shaping their behaviour through the intentional disengagement from negative stimuli and processing of positive information, rather than simply redirecting their attention. Only two studies to date have explored modifying attention in depression using eye-tracking-based feedback; Sanchez et al. (2016; Sanchez-Lopez et al., 2019) instructed participants to allocate attention to text that contained positive words, and provided feedback in the form of green or red frames surrounding positive or negative words, respectively. That study, conducted among non-clinical undergraduates, found that eye-tracking-based feedback led to more sustained attention on positive information and reductions in state rumination (Sanchez-Lopez et al., 2019). Thus, there is evidence that eye-tracking feedback can manipulate negative attention bias, but this has not yet been explored using facial affect or among depressed populations.

The effectiveness of affective ABM in depression is mixed. Few studies have addressed the aforementioned weaknesses in design and implementation to examine whether the issue is a problem of concept or execution. We aimed to examine whether ABM might be more effective in a paradigm that addresses these shortcomings.

The aim of the present proof-of-principle randomized controlled trial was to explore the effects of a novel, reward-based, free-viewing ABM paradigm on attentional bias, emotional memory, and depressive symptoms. The design of this study improves upon past research in three ways: 1) overcoming the limits of the dot-probe through a free-viewing, eye-tracking training paradigm; 2) multiple training sessions; and 3) recruiting only individuals with major depressive disorder (MDD). We hypothesized that ABM, compared to sham control, would reduce negative information bias and affect bias (i.e., memory for valenced items), and improve depressive symptoms and quality of life. Exploratory analyses examined whether trait rumination was related to training outcomes.