Causal Analyses of Associations Between Brain Structure and Suicide Attempt in Adulthood and Late Childhood

Abstract

Objective: Brain markers for suicide risk in adulthood may be detected during childhood and used for earlier detection and initiation of preventive interventions. Genetic instrumental variable analyses were used to determine whether there is evidence of lower brain total cortical surface area and thinner average cortical thickness (ACT) causing increased suicide risk in adults and whether lower measures of similar brain measures can cause increased risk of suicidality and related psychopathology in older children.

Method: Two-sample Mendelian randomization (MR) was used with summary statistics from genome-wide association studies for total cortical surface area, ACT, and suicide attempt in adults to test causal hypotheses. In youth ages 9 to 10 years old, a combined MR and twin-based direction-of-causation approach was applied to the European twin sample (199 monozygotic, 257 dizygotic twin pairs), and a hybrid traditional twin direction-of-causation approach was applied to the full twin sample (308 monozygotic, 397 dizygotic twin pairs) from the Adolescent Brain Cognitive Development (ABCD) Study.

Results: Two-sample MR analyses found a significant negative causal effect of total cortical surface area on suicide attempt risk in adults. MR-direction-of-causation analyses did not find a significant causal effect of any brain measure on suicidality in older children, but found significant negative causal effects of ACT on depression and internalizing psychopathology, and vice versa.

Conclusion: Brain markers of suicide risk may be instantiated differently in adults compared with older children, though lower ACT may be causally related to psychopathology associated with suicidality in these youth.

Keywords: Mendelian randomization; brain structure; causal analyses; suicide attempt; twin direction-of-causation.

Figure 1

Theoretical Framework Linking Causal Processes Between Altered Brain Structure and Suicide Risk in Older Children and Adults Note:Altered global brain structures in older children may cause increased risk of suicidality or worsened psychopathology leading to suicide attempt later in life. Conversely, behavioral problems and psychopathology during childhood may cause changes in brain structure leading to pathological brain development and altered global brain structure in adults that, in turn, may increase risk of suicide attempt. Of note, it is possible that suicide attempt may also cause changes in global brain structure in adults. Thus, detecting causal processes between brain structure and suicide attempt in adults that may mirror causal processes in late childhood may help identify markers of suicide risk earlier in life.

Figure 2

Two-Sample Mendelian Randomization (MR) Estimates of Causal Effects of A) Total Cortical Surface Area (TCSA) and B) Average Cortical Thickness (ACT) on Suicide Attempt Risk Note:The reverse estimated causal effects of suicide attempt on TCSA and ACT are also shown in panels A) and B), respectively. Causal estimates from the inverse-variance-weighted (IVW), weighted-median, MR-Egger, and MR constrained maximum likelihood (MR-cML) approaches are plotted. Error bars represent 95% CIs. Summary statistics for genome-wide association studies of suicide attempt in adults from either a European (EUR) sample or a multi-ancestry (MULT) sample were used. Genome-wide association studies for TCSA and ACT were from adult EUR samples only. MA-AIC = model averaging Akaike information criterion; MA-BIC = model averaging Bayesian information criterion; SNPs = single nucleotide polymorphisms.

Figure 3

Mendelian Randomization Direction-of-Causation (DoC) and Hybrid Twin-Based DoC Causal Estimates Note:Causal influences A) of total cortical brain volume, total cortical surface area, and average cortical thickness on the risk of suicidality and associated behavioral risk factor measures and B) of the reverse causal influences of suicidality and associated behavioral risk factor measures on total cortical brain volume, total cortical surface area, and average cortical thickness in the European Adolescent Brain Cognitive Development (ABCD) Study twin sample using Mendelian randomization DoC are shown. Significant Mendelian randomization DoC causal estimates were assessed with twin-based DoC analyses using the full ABCD Study twin sample and are shown in C). The p values were adjusted (p_adj) for multiple testing using Benjamini-Hochberg false discovery rate correction. Error bars represent 95% CIs. BAS =behavioral activation system.