Sensitive periods for the link between childhood maltreatment and brain aging during adulthood

Abstract

Background: Childhood maltreatment (CM) is associated with the early onset of psychopathology and accelerated biological aging. However, outcomes vary widely among individuals with CM history. This variability may, in part, be explained by differences in age of exposure to CM. In this study, we examined whether CM was associated with accelerated brain aging, depending on the timing of exposure (i.e., 'sensitive periods').

Method: A machine learning algorithm of brain age trained prior to the current study in 3377 healthy individuals was employed in a CM dataset of 150 adult postnatal women, 92 of whom provided MRI data. CM history was retrospectively assessed from birth to age 18 years. Brain-predicted age was calculated from T1-weighted MRI scans. Brain age gap (BAG) was quantified as the disparity between brain-predicted age, relative to chronological age. Sensitive periods were identified using random forest regression with conditional inference trees.

Results: CM severity was associated with greater BAG (β = 0.34, p < 0.001). The most robust type/time risk factors for greater BAG were parental verbal abuse between ages 7 and 15 years, parental physical abuse between ages 4 and 6 years, witnessing sibling violence between ages 4 and 15 years, and sexual abuse between ages 4-6. Parental verbal abuse (7-15 years) and parental physical abuse (4-6) were the variables that were the most important predictors above and beyond duration, multiplicity (number of exposures), and cumulative maltreatment severity.

Conclusion: These findings suggest a link between CM and accelerated brain aging, with certain developmental periods appearing more sensitive to these effects. To the best of our knowledge, this study is the first to identify sensitive periods for the link between CM and brain aging in adults, and the first to examine the link between CM and brain aging in postnatal women. Together, these results suggest that CM's association with brain development is complex and warrants nuanced approaches to investigating the possible mechanisms underlying its effects.

Fig. 1

Childhood trauma and brain age acceleration in adulthood. The accuracy of the brain age metric was confirmed by calculating the mean absolute error (MAE) of the brain age r model in our dataset. We obtained good model fit as demonstrated by low MAE values (4.68 ± 3.37 years) and a significant correlation between brain and chronological age (Pearson's R = 0.52, p < 0.0001) (panel A). The distribution of the brain age gap values is shown as a histogram in panel B. The association between composite maltreatment severity and brain age was also assessed and is shown in panel C. The regression model revealed that greater maltreatment severity was significantly associated with larger brain age gap before (β = 0.34, p < 0.001) and after controlling for covariates of income, age, age-squared, total cortical volume, PTSD, postnatal depression, and smoking status (β = 0.27, p < 0.05).

Fig. 2

Sensitive periods by maltreatment subtypes. A random forest regression with conditional inference trees was conducted to gain a more detailed view into how exposure to specific subtypes of maltreatment influences brain aging (brain age gap) during adulthood. Variable importance is shown for the four types of maltreatment that were significantly associated with BAG. The horizontal axis for each plot shows the recalled age window of exposure to each maltreatment in years. Vertical axes show mean variable importance, defined as the average rise in the mean square error of the model's fit after permutation of each individual variable. Importance levels are shown for the 3 global maltreatment characteristics (cumulative severisty, duration, and multiplicity). Parental verbal abuse between ages 7–15 and parental physical abuse between ages 4–6 years were more significant predictors of BAG than any of the global maltreatment metrics. Although significant predictors, witnessing sibling violence and sexual abuse were not more significant predictors of BAG than the three global metrics.

Fig. 3

Dose-response effects. Dose-response plots are shown to depict the directionality of effects for the significant predictors from the random forest regression. Horizontal axes represent severity scores while vertical axes represent scaled, predicted model values. Overall, greater severity was primarily associated with faster brain age acceleration during adulthood. Physical abuse (4–6 yrs), however, showed an inverse effect in which greater severity was associated with decelerated brain age. Abbreviations are as follows: PVA, parental verbal abuse; PPhys, parental physical abuse; WSibA, witnessing violence toward siblings; SexAb-sexual abuse.