The Effects of Childhood Adversity on Amygdaloid and Hippocampus Volumes: A Systematic Review and Meta-Analysis

Introduction: Research suggests that child maltreatment (CM) is significantly associated with brain morphology. In particular, growing evidence has focused on the effect of CM on the hippocampus and amygdala, two brain structures that mediate socio-emotional and cognitive function across the lifespan (Twardosz, 2010; Teicher, 2016). Research has found inconsistent directions of CM effect on amygdala and hippocampus volumes, which might be due to unexamined moderators across studies (Pechtel, 2014). Thus, the present study conducted a meta-analysis to evaluate the reported effect size of the associations between CM and hippocampus and amygdala volume and test the influence of key moderators.

Methods: The current study followed the PRISMA guideline. Inclusion criteria includes: peer-reviewed studies reporting associations between CM and hippocampus or amygdala volumes, written in English and published between 1997 and 2017. The search was conducted in PsycINFO, Medline, and SSCI, and 20 articles were selected. Cohen’s d was extracted for effect size. MetaSEM package in R (Cheung, 2015) was used for analysis in which a random-effect model in a 3-level meta-analysis framework was constructed. Univariate regression modeling was used to test effects of moderators including laterality, gender, race, CM types, and age.

Results: Results yielded a significantly negative effect size for hippocampus (d = -.76, p < .01, 95%CI [-1.30, -.23]). A significant Q statistic (Q (62) =774.51, p < .001) indicated heterogeneity among the effect sizes. A large I² (.97) at the third level (the study level) and a small I² (<.001) at the second level (within each study) indicated that the heterogeneity can be accounted mainly by the differences of studies. Moderation analysis suggested that those physically abused individuals exhibited smaller volume of hippocampus (B = -.16, p <.05). Fail-safe N indicators (N1 = 5332.01; N+ = 5543.01) suggested that the publication bias in the current study can be “safely ignored.” For amygdala, results suggested a non-significant mean effect size (d = -.35, p = .12, 95%CI [-.79, .09]). A significant Q statistic (Q (19) =106.641, p <.001) indicated heterogeneity among the effect sizes. A large I² (.88) at the third level and a small I² (<.001) at the second level indicated that the heterogeneity can be accounted for mainly by the differences of studies. Moderator analyses revealed no lateral, age, or gender differences.

Conclusions: Child maltreatment is associated with structural changes in the brain that are in turn associated with risks such as alcoholism (De Bellis et al. 2000; Hill et al. 2001). Preventive intervention programs can be designed to address neurobiological impacts of child maltreatment in order to mitigate the sequela of CM.