Background: Considerable effort and funding are spent on developing and assessing clinical efficacy of dot probe task (DPT) based Attention Bias Modification (ABM). ABM is regarded as a potential new (online) treatment for anxiety disorders especially. Anxiety disorders are commonly asserted to be characterised by ABM’s treatment target: preferential processing of threatening information. Yet the available meta-analytical evidence for this specific threat- bias in clinically anxious individuals is thin: the largest meta-analysis to date included DPT data for only n = 337 clinically anxious individuals. We reasoned that the baseline bias measures obtained in RCTs for ABM constitute a considerable, hitherto not assessed, body of data on the existence of DPT threat bias in clinically anxious samples.

Method: Baseline ‘threat vs neutral’ DPT summary data for n=860 clinically anxious individuals enrolled in k=11 ABM RCTs were meta-analysed using REML. Additional Bayesian analysis was used to assess support for a series of 1 ms wide bias size intervals.

Results: REML analysis indicated no evidence that mean observed Bias Index (BI) differs from point zero (k= 11, n= 860, mean BI = 1.8, SE = 1.53, p = .229, 95% CI [-1.2 - 4.8]). Bayesian analyses indicated moderate support for the traditional ‘point-zero’ over the ‘not point-zero’ hypothesis (BF01 = 6.7). Interval-based Bayesian analysis suggest that BI most likely falls in the 0-1 ms interval (BFinterval/notinterval = 231) and is almost certainly not larger than +2 ms (towards threat), or -1 ms (away from threat).

Conclusion: Clinically anxious individuals enrolled in RCTs for Attention Bias Modification do not display attention bias towards threat at the start of their trials. This meta-analytical finding casts strong doubt on the common assumption that clinical anxiety is characterized by preferential attention allocation towards threatening information.