This three-arms randomized controlled trial of 145 adults with depression, conducted in the USA (Hsu et al., in press – Psychological Medicine), examined the efficacy of a dot-probe-based attention bias modification treatment (ABMT) in reducing symptoms of depression. Participants with at least moderate depression severity and an attention bias toward dysphoric stimuli (faces) were randomized to active ABMT (80% of trials at a neutral face location and 20% at the sad face location), sham ABMT (50% at neutral and 50% at sad location), or assessments only without training. The training consisted of two 25 minute in-clinic and three 8 minute at-home ABMT sessions per week for 4 weeks (a total of 2224 training trials). Intent-to-treat analyses indicated that, relative to assessment-only, active ABMT significantly reduced self-reported (d = −0.57) and interviewer-rated (d = −0.49) depression, and similar results were noted for active versus sham ABMT (ds = −0.41 and −0.42 for self-reported and interviewer-rated depression, respectively). Sham ABMT did not significantly differ from the assessment-only condition. Eye gaze data suggest that ABMT led to a reduction in the odds of a trial where gaze was greater for sad than neutral stimuli relative to assessment-only and sham ABMT, but this result did not replicate with a more traditional reaction time measure of bias. Taken together, the results of this study suggest that depressed individuals with at least modest negative attentional bias benefitted from active ABMT.
Hsu, K. J., Shumake, J., Caffey, K., Risom, S., Labrada, J., Smits, J. A., … & Beevers, C. G. (in press). Efficacy of attention bias modification training for depressed adults: A randomized clinical trial. Psychological Medicine, 1-9.
Commentary by Daniel Pine
National Institute of Mental Health – USA
Cognitive bias modification remains one of the most promising, exciting topics in research on mental health. Such excitement follows from many factors beautifully illustrated by a new paper authored by Hsu and colleagues. This paper reports results from a randomized controlled trial in 145 adults with moderately severe symptoms of major depressive disorder. The trial uses a parallel-group design, randomizing subjects to either assessment only, active attention-bias modification treatment, or sham attention-bias modification treatment. Relatively clear findings emerge in the study, with greater reductions in depressive symptoms for the active compared to the other two control groups. Moreover, the effect size for this reduction is in the medium range, largely consistent with findings from meta-analyses on efficacy of such training for anxiety disorders. Two aspects of the new findings highlight some of the most trenchant features of research on cognitive bias modification.
First, research in this area bridges neuroscience and novel therapeutics. Hsu and colleagues build a sturdy bridge between these fields by applying rigor in the approach to assessments for both clinical and cognitive-neuroscience variables. They apply state-of-the-art methodology on the clinical side, pre-registering their study and conducting a broad clinical assessment, with a clearly defined outcome measure. They apply equally rigorous methodology on the neuroscience side. They select patients into the trial based on the presence of a measure of attention bias with reasonable reliability that is targeted by the training regimen. Many other studies strive for such rigor; Hsu and colleagues are more successful in their strivings than many of us.
Second, research on cognitive bias modification progresses through cooperation among research groups: many new studies from one group directly consider the strengths and weaknesses of prior studies from other groups. Thus, prior to the work of Hsu and colleagues, other studies selected subjects based on pre-treatment attention bias, but the new report more thoroughly considers psychometrics of the attention measure. Similarly, other studies, like the new one, also use pictorial stimuli, but Hsu and colleagues employ novel stimuli for cognitive training by drawing pictures from both face-photograph data sets and the International Affective Picture System Collection. Finally, other studies use control conditions included in the new paper, but by including two control conditions, Hsu and colleagues can draw firm conclusion concerning the clinically active ingredient that generate the therapeutic effect.
Research on cognitive bias modification is exciting because scientists are working together to learn from the lessons of the past in ways that bring together neuroscientists and mental-health researchers. The work from Hsu and colleagues follows in this tradition in ways that highlight the strengths of this approach. By doing so, we can all learn much by considering the context of the new findings. Not only do the authors present exciting new results, but they do so by using a clinical-trial design that leverages the insights generated from earlier trials.
Commentary by Dana Shamai-Leshem
Tel Aviv University – Israel
Hsu et al (2021) present a randomized controlled trial (RCT) of attention bias modification treatment (ABMT) for depression. Participants in this study were recruited based on both depression severity and the presence of pre-treatment negative attention bias. The authors found that active ABMT was associated with greater reduction in depression symptoms, in remission rates, and in treatment response rate compared to the other study conditions.
The authors present a few novel methodologies that are both interesting and important, especially in the light of mixed clinical findings in prior studies of ABMT for depression. The study highlights the potential importance of pre-treatment attention bias as both a predictor for ABMT efficacy and as am inclusion criterion for ABMT. The authors also overcome some of the psychometric limitations of the dot-probe task by using a more stable eye-tracking-based index as the criterion of pre-treatment negative attention bias. While this strategy is indeed preferable to the traditional reaction-time-based attention bias score, its one-week stability was still not highly acceptable (r=.67). A more reliable attention bias measurement, such as the free viewing task introduced by Lazarov et al (2016, 2018) may be suitable in future studies.
ABMT in this study was administered by a combination of in-clinic and home-delivered sessions. Studies that examined the efficacy of remotely-delivered ABMT have found good compliance with study protocol on one hand, but poor clinical effects on the other hand. Nevertheless, a couple of recent RCTs (Hilland et al., 2020; Jonassen et al., 2019) examined the effectiveness of daily at-home sessions of ABMT for depression and found significant clinical efficacy. It seems that the strategy of Hsu et al to administer relatively high frequency of at-home ABMT together with in-clinic sessions allowed even better clinical outcomes.
One point I noticed is that attrition rate was quite high in this study (19.3%), and that cognitive target engagement was quite weak (as reflected in no differences between the groups in attention bias scores at post-treatment). The authors concluded that the later might be a result of measurement challenges, which may be true. An alternative possibility is that long training sessions were used in this study (25 minutes each), with long stimuli presentation durations (3000-4500ms per trial), which might have caused the training task to be tedious and boring and thereby decreased target engagement. Given that the aforementioned studies (Hilland et al., 2020; Jonassen et al., 2019) found clinical efficacy for depression with shorter presentation durations, future studies could also consider shortening stimuli presentation.
To conclude, the field of ABMT could benefit from adapting the novel ideas presented in this study by Hsu et al. Pre-treatment attention bias can be useful in identifying individuals who could benefit from targeted ABMT. For this, reliable and stable methods to establish the presence of attention bias pre-treatment are needed.