A genetically informed Registered Report on adverse childhood experiences and mental health

Abstract

Children who experience adversities have an elevated risk of mental health problems. However, the extent to which adverse childhood experiences (ACEs) cause mental health problems remains unclear, as previous associations may partly reflect genetic confounding. In this Registered Report, we used DNA from 11,407 children from the United Kingdom and the United States to investigate gene-environment correlations and genetic confounding of the associations between ACEs and mental health. Regarding gene-environment correlations, children with higher polygenic scores for mental health problems had a small increase in odds of ACEs. Regarding genetic confounding, elevated risk of mental health problems in children exposed to ACEs was at least partially due to pre-existing genetic risk. However, some ACEs (such as childhood maltreatment and parental mental illness) remained associated with mental health problems independent of genetic confounding. These findings suggest that interventions addressing heritable psychiatric vulnerabilities in children exposed to ACEs may help reduce their risk of mental health problems.

Figure 1. Associations between polygenic scores and ACEs in ALSPAC.

Note. Data are presented as odds ratios +/- 95% CIs, obtained from logistic regression models. Panel A shows associations between polygenic scores for mental health problems and ACEs, Panel B shows associations between negative control polygenic scores and ACEs. P-values for individual associations between polygenic scores and ACEs are from two-sided tests and are false discovery rate (FDR) corrected. The sample size for ALSPAC analyses was n=6,411.

Figure 2. Associations between polygenic scores and ACEs in ABCD.

Note. Data are presented as odds ratios +/- 95% CIs, obtained from logistic regression models. Panel A shows associations between polygenic scores for mental health problems and ACEs, Panel B shows associations between negative control polygenic scores and ACEs. P-values for individual associations between polygenic scores and ACEs are from two-sided tests and are FDR corrected. The sample size for ABCD analyses was n=4,996.

Figure 3. Pairwise differences between polygenic scores in their association with ACEs.

Note: Data are presented as log odds differences +/- 90% CIs. Positive effect sizes reflect the first labelled polygenic score having a stronger positive average association with ACEs than the second polygenic score. Red dashed lines show the pre-specified equivalence bounds. 90% confidence intervals are presented and p-values are for the difference in log odds ratio between polygenic scores (two-sided tests). n=6,411 in ALSPAC and n=4,996 in ABCD.

Figure 4. Pairwise differences between ACEs in their association with polygenic risk for mental health problems.

Note: Data are presented as log odds differences +/- 90% CIs (two-sided tests). Positive effect sizes reflect the first labelled ACE having a stronger positive association with pooled polygenic risk for mental health problems; negative effect sizes reflect the second labelled ACE having a stronger positive association with pooled polygenic risk for mental health problems. The red dashed lines show the p re-specified equivalence bounds. n=6,411 in ALSPAC and n=4,996 in ABCD.

Figure 5. Diagrams showing structural equation models to estimate the genetic contribution to the associations between ACEs and mental health.

Note. In all diagrams, ACE represents the adverse childhood experience, MH represents the mental health outcome (e.g., internalising problems or externalising problems) and PGS represents the polygenic score, with one polygenic score shown in panels A and B, and all 8 polygenic scores (PGS_1-PGS_8) shown in panel C. Panel A depicts the underlying conceptual model, in which the polygenic score is treated as a confou nder, whereas panel B depicts the statistical model to calculate the genetic confounding effect, in which the polygenic score is treated as a mediator. Note that conceptually, the polygenic score cannot be a mediator in the association between ACEs and mental health because genetic variants are set at conception and do not change throughout the lifespan. However, statistically, we can estimate the genetic confounding effect by treating the polygenic score as a mediator and calculating the indirect effect of ACEs on mental health through the polygenic score. Panel C represents the statistical model in which all 8 polygenic scores are included as mediators. Though not depicted in the figure to aid clarity, we will account for correlations between polygenic scores in the model.

Figure 6. Genetic confounding of the associations between ACEs with internalising and externalising problems.

Note. Data are presented as standardised beta coefficients +/- 95% CIs for associations between ACEs and mental health outcomes, before accounting for polygenic scores (yellow circles), and after accounting for (i) observed polygenic scores for mental health problems (red points), and (ii) a latent polygenic score capturing SNP heritability in the outcome (blue points). Panel A shows the associations between ACEs and internalising problems in ALSPAC; Panel B shows the associations between ACEs and externalising problems in ALSPAC; Panel C shows the associations between ACEs and internalising problems in ABCD; Panel D shows the associations between ACEs and externalising problems in ABCD. Tests were two-sided. Confidence intervals could not be reliably computed for associations attenuated to zero and therefore these estimates should be interpreted with caution. n=6,411 in ALSPAC and n=4,996 in ABCD.