The Externalizing Spectrum and Suicide Risk: Insights from Genomics and Electronic Health Records in over 500,000 Veterans

This study of over 500,000 veterans demonstrates that the externalizing spectrum, characterized by behavioral disinhibition and shared genetic factors, is a significant and often overlooked predictor of suicide-related outcomes, including ideation, attempts, and death.

The Gist

The Big Picture: A Hidden Connection

Imagine the human mind as a complex city. Usually, when we think about the risk of someone hurting themselves (suicide), we focus on the "sad neighborhoods"—places of depression, anxiety, and low mood. This study suggests we have been ignoring a different part of the city: the "Wild West" district.

This "Wild West" represents Externalizing Disorders. These are conditions where people struggle to control their impulses, like substance abuse (drinking or drugs), ADHD, or reckless behavior. The researchers wanted to know: Is this "Wild West" district actually a major highway leading to tragedy, even if the people there don't feel "sad"?

The Detective Work: Using a Massive Database

The researchers acted like super-detectives using a massive library called the Million Veteran Program (MVP). This library contains the DNA and medical records of over 500,000 U.S. Veterans.

Think of the Veterans' records as a giant puzzle. The researchers didn't just look at one piece (like "alcoholism"); they built a master key (called a "latent factor") that unlocked the connection between all the different "Wild West" behaviors (drugs, alcohol, smoking, ADHD) to see if they shared a common genetic blueprint.

The Three Clues They Found

1. The Genetic Blueprint (The "Family Resemblance")

The team looked at the DNA of the veterans to see if the genes that make someone prone to "Wild West" behavior (impulsivity) were the same genes that put them at risk for suicide attempts.

• The Analogy: Imagine two different sports teams: a rowdy soccer team (Externalizing) and a team that tries to quit the game entirely (Suicide). The researchers found that these teams share a huge amount of the same "player roster" in their DNA.
• The Finding: There is a very strong genetic link. If you have the genetic "ingredients" for impulsivity and substance issues, you are significantly more likely to have the genetic risk for attempting suicide. This link was found in both veterans of European and African ancestry, though the data was stronger for the European group due to sample size.

2. The Brain Map (The "Traffic Jam")

Next, they looked at what happens inside the brain after someone has died by suicide. They used a special microscope to look at the "traffic" of genes in the brain cells.

• The Analogy: Think of the brain as a city with different types of traffic lights. Some lights tell cars to speed up (excitatory neurons), and some tell cars to slow down or stop (inhibitory neurons).
• The Finding: In the brains of people who died by suicide, the genes related to "Wild West" behavior were most active in the "Brake Pedal" cells (inhibitory neurons). It's as if the brain's ability to hit the brakes on impulsive actions was genetically compromised. When the brakes fail, the car (the person) is more likely to crash.

3. The Real-World Prediction (The "Crystal Ball")

Finally, they looked at the medical records to see if a current diagnosis of these "Wild West" problems could predict a future tragedy.

• The Analogy: Imagine a weather forecast. If you see dark storm clouds (a diagnosis of substance abuse or ADHD) in the past year, how likely is a tornado (suicide death) in the next five years?
• The Finding: The forecast was very clear. Veterans who had four or more recent diagnoses related to impulsivity or substance use were five times more likely to die by suicide in the next five years compared to those with no such diagnoses.
• The "Deaths of Despair" Factor: When they looked at a broader category of "deaths of despair" (suicide, alcohol poisoning, drug overdoses), the risk jumped to nearly 20 times higher for those with many recent diagnoses.

Why This Matters: A New Warning Sign

For a long time, doctors have been looking for "sadness" as the main warning sign for suicide. This study says: "Stop looking only at the sadness. Look at the chaos."

If a patient walks into a clinic acting impulsively, using substances, or unable to sit still, they aren't just "acting out." They might be flashing a giant red flare that says, "I am at high risk for a suicide attempt."

The Takeaway

This research is like discovering a new, hidden road on a map. We knew about the road of depression leading to suicide, but this study shows us a parallel road of impulsivity and externalizing behavior that leads to the same dangerous destination.

By recognizing this "Wild West" road, doctors and families can build better guardrails. Screening for substance abuse and ADHD isn't just about treating addiction; it's a critical step in preventing suicide.

---

Important Note: This study was a preprint (a draft shared before final peer review), meaning it is cutting-edge science that hasn't been fully double-checked by other scientists yet, but it offers a very promising new direction for understanding suicide risk.

Technical Summary

1. Problem Statement

Suicide remains a leading preventable cause of death, with rates rising significantly in the U.S. While mood disorders (internalizing) are well-established risk factors, the role of externalizing disorders—characterized by behavioral disinhibition, impulsivity, and conditions like substance use disorders (SUD), ADHD, and conduct problems—is less understood mechanistically. Previous studies suggest a genetic overlap between externalizing traits and suicide attempts, but the nature of this overlap, its biological underpinnings in brain tissue, and its prospective clinical relevance for mortality remain unclear. This study aims to disentangle the relationship between the externalizing spectrum and suicide-related outcomes (ideation, attempts, and death) using a multi-level approach: genomic, transcriptomic, and epidemiological.

2. Methodology

The study leveraged data from the Million Veteran Program (MVP) and two independent validation cohorts (COGA and Add Health), employing a multi-modal analytical framework:

• Cohorts:

  • Discovery: MVP (N ≈ 591,818), stratified into European-like (EUR-like, N=467,101) and African-like (AFR-like, N=124,717) ancestry groups.
  • Validation: Collaborative Study on the Genetics of Alcoholism (COGA, N=10,986) and National Longitudinal Study of Adolescent to Adult Health (Add Health, N=6,883).
  • Post-Mortem: 40 donors from the VA National PTSD Brain Bank (24 suicide deaths, 16 controls) with single-cell RNA sequencing of the dorsolateral prefrontal cortex (DLPFC).

• Genomic Analyses:

  • Latent Factor Construction: Used GenomicSEM to create a confirmatory factor analysis model for a latent externalizing factor (MVP-EXT). Indicators included EHR-derived phecodes for Substance Use Disorder (316), Alcohol Use Disorder (317), Tobacco Use Disorder (318), and ADHD (313.1), plus survey-based smoking and binge drinking.
  • GWAS: Performed univariate GWAS using SAIGE, followed by multivariate GWAS for MVP-EXT.
  • Genetic Overlap: Calculated genetic correlations ($r_G$) between MVP-EXT and suicide attempt/ideation. Used MiXeR to estimate the proportion of shared causal variants.
  • Polygenic Scores (PGS): Generated PGS using PRS-CSx (integrating multi-ancestry data) to predict suicide outcomes in validation cohorts.
  • Genomic Structural Equation Modeling: Tested if the externalizing-suicide link persisted after covarying for depression and schizophrenia.

• Transcriptomic/Cellular Analysis:

  • Applied single-cell Disease-Relevance Scoring (scDRS) to assess the enrichment of MVP-EXT-associated genes in specific cell types within post-mortem brain tissue of suicide decedents vs. controls.

• Epidemiological/Prospective Analysis:

  • Analyzed EHR data to count externalizing diagnoses in the 12 months prior to enrollment.
  • Used competing risk models (Fine-Gray) to estimate Hazard Ratios (HR) for suicide death and "deaths of despair" (suicide, alcohol, drug-related) while accounting for other causes of death.

3. Key Contributions

• Validation of MVP-EXT: Successfully replicated a latent externalizing factor in a massive veteran cohort that strongly correlates ($r_G = 0.87$) with previous externalizing GWAS results, despite MVP data not being part of the original discovery.
• Multi-Ancestry Genomics: Provided one of the largest analyses of externalizing genetics in African-ancestry populations, though noting power limitations compared to European-ancestry groups.
• Biological Mechanism: Identified specific cell-type enrichments (inhibitory neurons) linking externalizing genetics to suicide pathology in post-mortem tissue.
• Clinical Translation: Moved beyond genetic correlation to demonstrate a prospective, time-dependent association between recent externalizing diagnoses and future suicide death, offering a potential target for clinical screening.

4. Key Results

A. Genetic Architecture and Overlap

• Genetic Correlation: MVP-EXT showed strong genetic correlations with suicide attempts in both EUR-like ($r_G = 0.67$) and AFR-like ($r_G = 0.62$) populations. Correlations with suicidal ideation were weak in EUR-like ($r_G = 0.12$) and null in AFR-like.
• Shared Variants: MiXeR analysis revealed that MVP-EXT and suicide attempts share approximately 77% of their influential genetic variants, with high sign concordance (79%), suggesting a shared polygenic architecture.
• Independence: The association between externalizing and suicide attempts remained significant even after statistically controlling for depression and schizophrenia, indicating externalizing is an independent risk factor.

B. Polygenic Score Predictions

• EUR-like: MVP-EXT PGS significantly predicted lifetime suicidal ideation (OR = 1.09) and suicide attempts (OR = 1.20) in validation cohorts.
• AFR-like: PGS associations were null for AFR-like participants, likely due to lower statistical power, differences in linkage disequilibrium, or the "social push" model where environmental stressors (racism, discrimination) may override genetic risk in this population.

C. Cellular Enrichment

• Genes associated with MVP-EXT were significantly enriched in inhibitory neurons, astroglia, and oligodendrocyte progenitor cells (OPCs) in suicide decedents.
• Notably, inhibitory neurons were the only cell type significantly enriched in the AFR-like subset, highlighting a potential convergence on inhibitory circuitry in suicide pathology.

D. Prospective Clinical Risk (EHR)

• Hazard Ratios: Recent externalizing codes were associated with a 1.39x increased hazard for suicide death and a 1.96x increased hazard for "deaths of despair."
• Cumulative Incidence: Veterans with 4+ externalizing codes in the past year had a 5-year cumulative incidence of suicide death nearly 5 times higher than those with no codes.
• Robustness: These associations persisted even after adjusting for comorbid Major Depression and PTSD.

5. Significance and Implications

• Clinical Screening: The study provides strong evidence that externalizing disorders are potent, independent predictors of suicide death, distinct from internalizing disorders like depression. This suggests that clinical intake screenings should prioritize behavioral disinhibition and substance use as critical suicide risk markers.
• Biological Insight: The enrichment of inhibitory neurons suggests that dysregulation in inhibitory neurotransmission may be a core biological mechanism linking impulsivity/externalizing behaviors to suicide risk.
• Health Equity: The findings highlight the complexity of applying genetic risk scores across diverse populations. The lack of PGS prediction in AFR-like populations underscores the need for larger, diverse GWAS and the consideration of social determinants of health (e.g., racism) that may interact with or supersede genetic risk.
• Public Health: Given the high prevalence of externalizing disorders in the veteran population and their strong link to "deaths of despair," interventions targeting behavioral disinhibition could significantly reduce suicide mortality.

Limitations: The study is limited by the predominantly male sample (~85%), the lower statistical power in the AFR-like subgroup for locus discovery, and the reliance on EHR codes which may underrepresent marginalized populations. Future work requires larger diverse cohorts and integration of social determinants.