Childhood Mental Health and Body Mass Index as Mediators of Genetic Risk for Eating Disorders

This longitudinal study of over 5,000 children reveals that genetic liabilities for anorexia nervosa and binge eating contribute to early disordered eating behaviors through distinct metabolic and psychosocial pathways, with BMI mediating risks across sexes and ADHD or anxiety/depression symptoms serving as additional mediators specifically in females.

The Gist

Imagine your genetic code as a blueprint for a house. For decades, scientists have known that eating disorders (like anorexia or binge eating) are partly built into these blueprints. But there was a big mystery: How does this blueprint turn into actual behavior in a child? Does the genetic risk just sit there waiting until adulthood, or does it start showing up early? And does it build the house the same way for boys and girls?

This study acts like a detective story, using a massive database of 5,600 children (the ABCD Study) to track how these genetic blueprints play out between ages 9 and 12.

Here is the story of what they found, broken down into simple concepts:

1. The Genetic "Weather Forecast"

Think of Polygenic Scores (PGS) as a "genetic weather forecast." It doesn't say "It will rain tomorrow," but rather, "There is a 70% chance of rain based on the clouds."

• The Finding: The study found that children with a "rainy forecast" for Anorexia (AN) started showing early signs of restrictive eating (worrying about weight, trying to control it).
• The Twist: Children with a "rainy forecast" for Binge Eating (BE) were like a storm that could go either way. Their genetic risk showed up as both binge eating and anorexia symptoms. It's as if the BE genetic blueprint is a "wildcard" that can manifest as different types of eating trouble.

2. The "Middleman" (Mediators)

Genetics doesn't usually pull the strings directly. Instead, it pulls the strings of other things first, which then lead to the eating disorder. The study looked at who these "middlemen" were.

• The Body (BMI):
  • For Girls: The genetic risk for Anorexia actually made them thinner (lower BMI). Paradoxically, in this young age group, being thinner seemed to act as a "shield" that slightly reduced the severity of the symptoms. It's like the genetic risk tried to build a thin house, but the thinness itself didn't immediately cause the full-blown disorder yet.
  • For Boys: The genetic risk for Binge Eating was linked to higher BMI. Here, the "body" was the main middleman.
• The Mind (Mental Health):
  • For Girls: This is where it gets complex. For girls, the genetic risk didn't just go through the body; it also went through Anxiety, Depression, and ADHD. It's like a relay race: The genes pass the baton to anxiety, which passes it to social struggles, which finally passes it to the eating disorder.
  • For Boys: The race was much shorter. For boys, the genetic risk mostly went straight through BMI (body weight) without needing the "Anxiety/Depression" middleman.

3. The "Two Different Playbooks"

The biggest takeaway is that boys and girls play by different rules, even with the same genetic risks.

• Girls' Playbook: It's a complex web. Genetics + Body + Mental Health (Anxiety/ADHD) + Social Problems = Eating Disorder risk.
• Boys' Playbook: It's more direct. Genetics + Body Weight = Eating Disorder risk.

4. Why This Matters (The "So What?")

Imagine you are a parent or a doctor trying to stop a fire before it starts.

• Old Way: We waited until kids were teenagers or adults to look for genetic risks.
• New Way: This study says, "Look at the kids now."
  • If you see a girl with high genetic risk, you shouldn't just watch her weight. You need to watch her anxiety levels, her ability to focus (ADHD), and her social struggles. Fixing those might stop the eating disorder before it starts.
  • If you see a boy with high genetic risk, keeping an eye on his weight trajectory is the most critical early warning sign.

The Bottom Line

Genetic risk for eating disorders isn't a "switch" that flips on at age 18. It's a seed that starts growing in the soil of childhood.

• For girls, that seed needs water from mental health struggles and social pressure to grow big.
• For boys, that seed grows mostly based on physical body changes.

By understanding these different "gardening" needs for boys and girls, we can build better, earlier safety nets to catch kids before they fall into the trap of an eating disorder.

Technical Summary

1. Problem Statement

Eating disorders (EDs) are highly heritable, yet the developmental pathways through which genetic liability manifests in early childhood remain poorly understood. Previous research has yielded inconsistent results regarding the association between Anorexia Nervosa (AN) polygenic scores (PGS) and childhood disordered eating behaviors (DEB), and there is a scarcity of data regarding Binge Eating (BE) genetics in youth. Furthermore, it is unclear whether these genetic risks operate through direct pathways or are mediated by metabolic (e.g., BMI) and psychosocial factors (e.g., ADHD, anxiety/depression), and whether these mechanisms differ by sex. This study aims to bridge the gap between adult genetic findings and childhood phenotypes by investigating these longitudinal associations and mediation pathways.

2. Methodology

Study Design and Population:

• Data Source: The study utilized data from the Adolescent Brain Cognitive Development (ABCD®) Study, a multisite, population-based cohort.
• Sample: The primary analysis focused on 5,618 participants of genetically inferred European (EUR) ancestry (mean age ~9.9 years at baseline). Exploratory analyses were conducted on a diverse full sample of 9,132 participants (including African, Admixed American, East Asian, and South Asian ancestries).
• Longitudinal Design: A three-wave temporal design was employed using data from Baseline (T0), Year 1 (T1), and Year 2 (T2) follow-ups.

Exposures (Genetic Liability):

• Polygenic Scores (PGS): AN-PGS and BE-PGS were calculated using summary statistics from the largest available Genome-Wide Association Studies (GWAS) for AN and BE (conducted by the Psychiatric Genomics Consortium).
• Calculation: PGS were computed using PRS-CS, a Bayesian polygenic model that accounts for linkage disequilibrium (LD) variation across ancestries. Scores were standardized within ancestry groups.
• Sensitivity Analysis: A BE-PGS was also generated excluding AN-ascertained cohorts to rule out shared genetic effects driving the results.

Outcomes and Mediators:

• Outcomes (DEB): Disordered eating behaviors were assessed via parent reports using the Kiddie Schedule for Affective Disorders and Schizophrenia (KSADS) ED module.
  • AN Symptoms: Sum of 3 items (preoccupation with weight, vomiting, other control methods).
  • BE Symptoms: Product of frequency and discomfort scores.
  • Longitudinal Analysis: Symptoms pooled across T0, T1, and T2.
  • Mediation Analysis: Symptoms at T2 used as the outcome to ensure temporal sequence.
• Mediators (Assessed at T1):
  • Metabolic: BMI (converted to z-scores based on CDC growth charts) and change in BMI ($\Delta$BMI).
  • Psychosocial: Standardized summary scores from the Child Behavior Checklist (CBCL) for Anxious/Depressed, ADHD, and Social Problems.

Statistical Analysis:

• Association Models: Generalized linear mixed models (GLMM) with random intercepts were used to assess associations between ED-PGS and DEB, controlling for age, sex, ancestry PCs, and genotyping plates.
• Mediation Models: Causal mediation analysis (using the R package mediation) was performed to estimate the Average Causal Mediation Effect (ACME). Models controlled for age and race/ethnicity.
• Robustness: Standard errors and confidence intervals were derived via non-parametric bootstrapping (1,000 resamples). Multiple testing was corrected using False Discovery Rate (FDR).

3. Key Contributions

• Developmental Continuity: Provides robust evidence that genetic risks for EDs identified in adults manifest as observable behavioral symptoms as early as age 9–12.
• Transdiagnostic Nature of BE: Demonstrates that BE-PGS has broad, transdiagnostic effects, predicting both BE and AN symptoms, whereas AN-PGS is more specific to AN symptoms.
• Sex-Specific Pathways: Identifies distinct mediation mechanisms for males and females, highlighting that while BMI mediates risk for both sexes, psychosocial factors (ADHD, anxiety/depression) play a significant additional role specifically in females.
• Paradoxical Metabolic Finding: Uncovers a counter-intuitive finding where higher AN-PGS predicts lower BMI, which in turn acts as a protective factor against early AN symptoms in females, suggesting a complex, stage-dependent interaction between metabolic and psychological risk factors.

4. Key Results

Genetic Associations:

• AN-PGS: Significantly associated with increased AN symptoms in both males ($\beta = 0.026$) and females ($\beta = 0.043$).
• BE-PGS: Showed transdiagnostic associations, significantly predicting both BE symptoms and AN symptoms in both sexes. The effect sizes were generally larger in females.
• Cross-Ancestry: Results were largely consistent in the full, diverse sample, suggesting generalizability beyond EUR ancestry.

Mediation Analysis (Sex-Stratified):

• AN Symptoms:
  • Females: AN-PGS was associated with lower BMI and smaller $\Delta$BMI. Interestingly, lower BMI mediated a reduction in AN symptoms (negative mediation), suggesting that in this early developmental window, the metabolic component of AN risk (low body mass) may be protective against the behavioral expression of AN symptoms.
  • Males: No significant mediation effects were observed.
  • Full Sample (Females): Anxiety/Depression emerged as a positive mediator.
• BE Symptoms:
  • Females: BE-PGS effects were mediated by BMI, ADHD, Anxiety/Depression, and Social Problems. These psychosocial factors accounted for 14–29% of the association with AN symptoms and 15–19% with BE symptoms.
  • Males: BMI was the sole significant mediator in the EUR subgroup, though anxiety/depression and $\Delta$BMI emerged as mediators in the full sample.

5. Significance and Implications

• Early Identification: The study validates the utility of adult-derived PGS for identifying at-risk children, suggesting that genetic screening could be part of early prevention strategies.
• Targeted Interventions: The findings emphasize the need for sex-specific interventions.
  • For males, interventions may primarily focus on metabolic health and BMI trajectories.
  • For females, a multifaceted approach is required, addressing not only weight but also comorbid mental health conditions (ADHD, anxiety, depression) and social functioning.
• Developmental Nuance: The "protective" mediation of low BMI in early childhood challenges the assumption that low BMI is always a risk factor for AN in youth. It suggests that the pathogenicity of metabolic traits may depend on the interaction with emerging psychological stressors and pubertal development.
• Future Directions: Highlights the necessity of including diverse ancestries in GWAS to improve PGS accuracy and the need for longitudinal studies to track when metabolic and psychological risk factors converge to trigger full-blown eating disorders.

Limitations Noted:

• Use of parent-reported KSADS may miss subthreshold or subjective symptoms (e.g., body dissatisfaction).
• Potential sample overlap between AN and BE GWAS (though sensitivity analyses mitigated this).
• The study is observational; causal inference relies on the mediation model assumptions.