A clinical trial of 300 Australian patients with alcohol use disorder (Manning et al., 2021 – JAMA Psychiatry), tested whether a 4-day protocol of cognitive bias modification based on the Alcohol Approach/Avoidance Task (AAT) delivered during inpatient alcohol withdrawal treatment would reduce the likelihood of relapse during the first two weeks after discharge. A period during which relapse rates are high. Results indicate that relative to sham training, cognitive bias modification designed to reduce alcohol approach bias increased the likelihood of maintaining abstinence during the follow-up period by 11.9% in an intention-to-treat analysis (54.4% vs 42.5%) and by 17% in per-protocol analysis (63.8% vs 46.8%). The CBM task successfully engaged the intended alcohol-related approach bias, with CBM patients’ bias declining significantly over training sessions and reversing to an alcohol avoidance bias, whereas patients in the sham training condition continued to show a similar alcohol approach bias over time. Mediation analyses did not provide evidence that the noted changes in alcohol approach bias mediated the observed group differences in abstinence. Taken together these findings suggest that cognitive bias modification during inpatient alcohol withdrawal treatment may help to prevent relapse during the high-risk period immediately following discharge and that its implementation as an adjuvant intervention in such settings may be warranted. The data also indicate that additional research on the mechanisms underlying the observed clinical effects is needed.
Manning, V., Garfield, J. B., Staiger, P. K., Lubman, D. I., Lum, J. A., Reynolds, J., … & Verdejo-Garcia, A. (2021). Effect of cognitive bias modification on early relapse among adults undergoing inpatient alcohol withdrawal treatment: a randomized clinical trial. JAMA Psychiatry, 78, 133-140.
Commentary by Mike Rinck
Ruhr-University Bochum – Germany and Radboud University, Netherlands
I love this paper. It shows nicely how far at least some areas of CBM have come in terms of clinical application. By now, the training described here has been included in the treatment guidelines for Alcohol Use Disorder in Australia and in Germany. The encouraging results of the study nicely replicate those of earlier studies conducted in Australia and Germany. However, the study is much more than a mere replication. For instance, it shows that such a training can be administered even during withdrawal treatment, and that it has effects early on, within the first two weeks. This nicely complements earlier studies which showed long-term effects. Now it will be important to find out whether this early administration of the training is able to foster long-term abstinence, or whether it should be augmented by later booster training sessions.
The paper also illustrates some of the persistent problems that CBM researchers are confronted with. To mention just a few:
First, clinical effects are often small, which is hardly surprising when CBM is offered as an add-on to existing treatments. As a result, large samples are needed. Luckily, in this study, Victoria Manning and her colleagues were able to recruit a large sample, and they were rewarded by significant effects. However, too many other applications of CBM to clinical populations suffer from insufficient sample sizes. This yields the well-known, but also well-ignored, power problem: Small and medium-sized effects will not be statistically significant, and therefore usually not published. Only randomly inflated large effects will be published, fooling other researchers into believing that replications will be feasible.
Second, as in many other studies, the clinical effects observed in this study were not mediated by training effects, such that patients who were trained better did not have a better chance to remain abstinent. In short, this means that we are doing something useful, but we don’t really know what we are doing and how we are doing it… This may not differentiate us from many other health providers, but it is embarrassing because, unlike many others, we have clear hypotheses about the working mechanisms involved in CBM, and we have the means to test these hypotheses. To explain this contradiction, we often refer to the low reliability of our assessment tasks, which makes it hard to measure individual differences in the effects of our trainings on the targeted mechanisms. This is most likely true for most tasks in most studies. Therefore, we should make it a habit to compute and report the reliability of the bias scores we use in our analyses.
Third, we should be aware that there is yet another potential explanation of differential training effects: The active version of a training may simply be more convincing than the sham version of the same training, thereby creating larger expectations and larger placebo effects, in the absence of larger training effects. Therefore, I urge everyone to pay more attention to the credibility of the different training versions. In clinical trials, we need to make sure (e.g., by inventing good cover stories) that the versions intended as inactive control versions are as credible and as convincing as the active versions. At the very least, we need to check this by asking the participants about their expectations and beliefs regarding the training they received. The three subjective ratings reported by Victoria Manning and her colleagues are a valuable first step into this direction.
Commentary by Johannes Illgner
University of Münster – Germany
I really liked the study by Manning and colleagues, especially since it gives hope that Cognitive Bias Modification (CBM) can supplement existing therapies in a useful way. In particular, the brief and simple administration of CBM training could facilitate future implementation in practice. I was impressed by the good methodological approach with a large sample as well as the detailed documentation of the procedure in the paper with meaningful online supplements. Through the study, I became more interested in two issues: the nature of instruction during CBM trainings and the quality of control conditions in modification studies.
Manning and colleagues used an indirect AAT instruction in which stimuli had to be pulled closer or pushed away based on the orientation of the frame. However, during the training phase, 95% of the alcohol stimuli had to be pushed away and only 5% had to be pulled closer. Thus, participants should have quickly realized that although they were supposed to pay attention to a content-irrelevant property of the stimuli, the goal was that they learn to automatically avoid alcohol stimuli. Other modification studies also use indirect instruction, but I would personally find it exciting to use direct instruction and tell patients clearly why they are doing this training. What the training phase is about becomes clear either way, so you can intentionally enhance the placebo effect.
The study made me aware of the importance of choosing an appropriate control condition. In the experimental condition, the authors used a joystick AAT in which alcohol stimuli almost always had to be pushed away and thus avoided, while a task was used as sham training for the control group in which, among other things, alcohol stimuli had to be pushed to the left and right. It remains questionable whether the authors thus also created a placebo effect in the control group, since no cover story is reported that credibly explains to the patients in the control group that the sham training could actually work. A precise manipulation check seems useful, asking, among other things, what purpose the participants thought was behind the sham training or the actual CBM training. In the future, it will be important to develop sham trainings that reliably trigger a placebo effect but do not go beyond it, so that it is actually possible to conclude that the treatment delivers an effect that goes beyond the placebo effect.
Regarding the portrayal of gender identities, I noticed that there was no mention of a non-binary participant in the abstract, although it is mentioned in the method section that a non-binary person was included. The abstract of the study mentions that 300 people participated and 173 of them were men, which implies to the reader that the remaining 127 people were only women, which was not the case. In the future, people with other gender identities could also be given clear visibility.
Despite the criticisms, I draw a positive conclusion from the study. In the future, it is desirable that such methodologically demanding studies be conducted more frequently and also be used in CBM trainings on other mental disorders. The paper gives hope that CBM forms a useful extension to existing therapies.