A Meta-Analysis of Bias at Baseline in RCTs of Attention Bias Modification: No Evidence for Dot-Probe Bias Towards Threat in Clinical Anxiety and PTSD
A Meta-Analysis of Bias at Baseline in RCTs of Attention Bias Modification: No Evidence for Dot-Probe Bias Towards Threat in Clinical Anxiety and PTSD
复制标题
DOI：
10.1037/abn0000406
复制
发表时间：
2019-08-01
影响因子：
4.6
通讯作者：
Fox, Elaine
中科院分区：
文献类型：
--
作者：
Kruijt, Anne-Wil;Parsons, Sam;Fox, Elaine
人们已经花费了大量的精力和资金来开发注意力偏向修正(ABM)作为焦虑症的治疗方法，理论上是通过减少将注意力定向到威胁的倾向来发挥治疗效果。然而，元分析证据表明，临床焦虑症的特征是威胁相关的注意偏差，这一证据并不充分。迄今为止最大的荟萃分析包括n=337名临床焦虑者的点探头数据。在ABM随机对照试验中获得的有偏见注意力的基线测量形成了以前没有进行过荟萃分析的额外数据主体。本文对13项ABM随机对照试验中的1,005名临床焦虑者进行了基线测量的威胁相关点探针偏倚的荟萃分析。随机效应Meta分析显示，平均偏倚指数(BI)与零无显著差异(k=13，n=1005，平均偏倚指数=1.8ms，SE=1.26ms，p=0.144，95%可信区间[-0.6，4.3]。其他贝叶斯因子分析也支持零点假设(BF10=0.23)，而基于区间的分析表明，临床焦虑的平均偏差不太可能超过0至5ms的区间。从优点(相对较大的样本，可能绕过发表偏倚)、局限性(缺乏对照比较、重新调整数据用途、点探针数据的特异性)以及理论和实践背景等方面对研究结果进行了讨论。我们建议，不应再假定临床焦虑个体的特征是对威胁的选择性注意。在随机对照试验中登记进行注意偏差修正的临床焦虑个体在基线时不具有威胁相关的注意偏差的特征。
Considerable effort and funding have been spent on developing Attention Bias Modification (ABM) as a treatment for anxiety disorders, theorized to exert therapeutic effects through reduction of a tendency to orient attention toward threat. However, meta-analytical evidence that clinical anxiety is characterized by threat-related attention bias is thin. The largest meta-analysis to date included dot-probe data for n = 337 clinically anxious individuals. Baseline measures of biased attention obtained in ABM RCTs form an additional body of data that has not previously been meta-analyzed. This article presents a meta-analysis of threat-related dot-probe bias measured at baseline for 1,005 clinically anxious individuals enrolled in 13 ABM RCTs. Random-effects meta-analysis indicated no evidence that the mean bias index (BI) differed from zero (k = 13, n = 1005, mean BI = 1.8 ms, SE = 1.26 ms, p = .144, 95% confidence interval [-0.6, 4.3]. Additional Bayes factor analyses also supported the point-zero hypothesis (BF10 = .23), whereas interval-based analysis indicated that mean bias in clinical anxiety is unlikely to extend beyond the 0 to 5 ms interval. Findings are discussed with respect to strengths (relatively large samples, possible bypassing of publication bias), limitations (lack of control comparison, repurposing data, specificity to dot-probe data), and theoretical and practical context. We suggest that it should no longer be assumed that clinically anxious individuals are characterized by selective attention toward threat. Clinically anxious individuals enrolled in RCTs for Attention Bias Modification are not characterized by threat-related attention bias at baseline.