This but a scratch! Comparing frequentist and Bayesian meta-analysis on CBM on addictive behaviours

Authors

  • R. van Beek
  • O. Zerhouni
  • M. Boffo
  • K. Nikolaou
  • R. Wiers
  • M. Marsman

Abstract

Different automatic cognitive biases have been found to underlie the onset and maintenance of addictive behaviours. Cognitive Bias Modification (CBM) interventions target these maladaptive processes by reducing or reversing them. Recent meta-analyses investigating the effectiveness of CBM interventions point in different directions. Using the Cochrane Handbook for Systematic Reviews of Interventions we conducted a systematic review of CBM interventions targeting alcohol and tobacco use. We aggregated published evidence on the effects of CBM on (i) cognitive bias and (ii) consumption in clinical and subclinical alcohol and tobacco users. We quantified the evidence using both frequentist and Bayesian random effects models. Twenty studies investigating the effect of CBM interventions on cognitive biases and a number of clinical outcome measures were included in the analysis. A random-model analysis with the Sidik-Jonkman method showed a medium-sized reduction of bias toward alcohol and tobacco cognitive biases attributable to CBM and a small-sized reduction on actual consumption at post-test with 95% confidence intervals that did not encompass zero. However, the Bayesian random-effects model showed only a small effect on cognitive bias for alcohol. Moderator analysis provided insights on potential moderators underpinning the effectiveness of CBM, such as type of addiction, number of training trials, type of CBM intervention, mode of delivery, and type of participants. While recent evidence has questioned the efficacy of CBM in addiction, we discuss the limitations of current research designs and provide suggestions for improvement.

Published

2017-12-31

Issue

Section

Symposia