Mediation of Cognitive Bias Modification for Alcohol Addiction via Stimulus-Specific Alcohol Avoidance Association

What is it about?

A new kind of therapy for addiction involves computerised training. However, there are fundamental questions about the involved mechanisms. This study looked at a previously published dataset and analysed whether training effects were related to achieving an automatic alcohol-avoidance association.

Featured Image

Why is it important?

There are various methodological factors of interest in this exploratory study: First we, decided to consider post-test state, rather than pre-post change, as a mediating variable. Note that if it is true that the automatic association is what affects relapse, it doesn't really matter how much it *changed*, it matters what it *is*. This touches on a more general point that anecdotal experience with reviewers and editors suggests isn't always correctly understood. With random assignment (and that's the essential point), tests of post-test differences between conditions are just as valid as tests of pre-post change scores between conditions. Second, we looked at effects on the Implicit Association Test that were specific to alcohol stimuli, instead of averaged over all trials. This follows from very general theoretical considerations - you expect an automatic process to be stimulus driven, so you should consider trials when such a stimulus appears. This makes the results exploratory, certainly, but the above reason for doing it is arguably obvious. Finally, the effect that the data show is that stronger avoidance associations are related to lower relapse risk. This is in accordance with theory and the overall rationale and results of the kind of Alcohol-Avoidance training that we were studying; it contrasts with some recent correlational results. While interesting, as has been pointed out in an earlier commentary (Wiers, Gladwin & Rinck, 2013), correlation is not causation, and mediation analyses such as we did aim to get closer to the causes of effects. There may also be interesting nuances in what kind of avoidance is being measured, which future work may be able to disentangle.