Jialin Wen

Background: Social anxiety disorder (SAD) is associated with dysfunctions in face processing, a critical aspect of social interaction. However, there is a dearth of biomarkers that could facilitate precise diagnosis and targeted treatment. Aims: This meta‐analysis seeks to investigate abnormalities in P1, N170, LPP, and N2pc during face processing in individuals with SAD. It also aims to examine the role of potential moderators. Methods: A systematic review of the literature yielded 32 relevant studies, encompassing the aforementioned ERP components: P1 (21 studies), N170 (16 studies), LPP (11 studies), and N2pc (6 studies). We performed a random‐effects meta‐analysis to calculate Hedges′ g and assessed publication bias using the trim‐and‐fill method. Risk of bias was evaluated across five domains using an adapted Downs and Black checklist. Results: Risk of bias was generally low, with some concerns regarding external validity. Primary analysis showed enhanced P1 ( g = 0.92) and LPP ( g = 1.07) amplitudes in SAD. However, the P1 effect became nonsignificant ( g = 0.20) after adjusting for publication bias, whereas the LPP enhancement remained robust. No significant differences were found for N170 or N2pc. Mixed‐model analysis indicated that P1 and N2pc exhibited the largest absolute effect sizes. No significant moderators (age, clinical status and task type) were identified. Conclusions: The findings highlight the LPP as the most robust neural correlate of face processing deficits in SAD, reflecting impaired late‐stage attentional disengagement. Although P1 amplitudes appear enhanced, this effect is likely inflated by publication bias. Current evidence does not support consistent alterations in N170 or N2pc. LPP may serve as a promising potential biomarker for the clinical diagnosis and treatment monitoring of SAD.