Quantifying the relative importance of genetics and environment on the comorbidity between mental and cardiometabolic disorders using 17 million Scandinavians

Individuals with mental disorders (MD) have substantially higher rates of somatic comorbidities, notably cardiometabolic disorders (CMD), contributing to 7–10 years shorter life expectancy and 2–3-fold increased risk of premature death. Epidemiological register studies show bidirectional associations between MD and CMD. However, comorbidity patterns vary widely by specific diagnoses, reflecting a mixture of genetic and environmental causes (e.g., medications, substance use, socioeconomic factors, adverse life events). Prior genome-wide approaches report shared genetic architectures between MD and CMD but are limited by sampling biases, heterogeneity across studies, and model assumptions, and they typically quantify genetic correlation without partitioning comorbidity into genetic versus environmental components. This study aims to quantify the relative genetic and environmental contributions to comorbidity between six MDs (ADHD, AN, ASD, AFF including MDD/BD, BD, SCZ) and 15 CMDs using nationwide health registers and multi-generational genealogies from Denmark and Sweden (17 million individuals). The goals are to estimate heritability and genetic correlations using family-based register data and to decompose observed MD–CMD comorbidity into genetic and environmental components for each disorder pair.

The paper synthesizes evidence that MDs are linked to elevated CMD risks and reduced life expectancy, with bidirectional associations documented in large Scandinavian register studies. Prior GWAS-based analyses indicate substantial genetic overlap between some MDs and CMDs, suggesting shared biology. However, these GWAS-derived estimates can be biased due to case-control sampling, heterogeneity in populations and methods, and reliance on assumptions (e.g., LD panels), and they typically report genetic correlations without estimating the relative weight of genetic versus environmental factors in comorbidity. The authors build on methodological work using liability threshold models and registry-based familial risk to estimate heritability and genetic correlation, addressing gaps left by GWAS-only studies and enabling a direct decomposition of comorbidity.

Design and data sources: The study leverages near-complete, multi-generation genealogies and nationwide health registers from Denmark and Sweden. For Denmark, registers include the Civil Registration System, National Patient Register, and Psychiatric Central Research Register with uniform ICD-coded diagnoses across decades. The Swedish Total Population Register, Multi-Generation Register, and Inpatient/Outpatient Registers provide analogous data. Individuals were followed for MD and CMD diagnoses from the 1970s onward. Analyses combined both countries and also meta-analyzed country-specific estimates. Cohorts: Heritability analyses used full register data (Denmark n≈7.80 million; Sweden n≈13.22 million). Genetic correlations were estimated among those born 1981–2005 with follow-up through 2012 (Denmark n=1,560,901; Sweden n=2,566,100). Molecular genotype data came from the Danish iPSYCH2012 cohort (77,082 unrelated European-ancestry individuals after QC), with neonatal bloodspot genotyping, phasing, and imputation. Case definitions: Six MDs (ADHD, AN, ASD, AFF, BD, SCZ) and 15 CMDs were defined from hospital register diagnoses (ICD-8/9/10; country-specific transitions). Individuals with SCZ, BD, or AFF diagnosed before age 10 were excluded. To minimize left censoring, only individuals born in Denmark or Sweden were retained. Estimating cumulative incidences: Cumulative incidence functions (CIFs) were estimated using the Nelson-Aalen estimator, stratified by birth year, for the general population, individuals with affected full siblings, and individuals with parents affected by cross-disorders (e.g., ADHD risk given parental T2D). Liability-threshold framework: Using year-specific lifetime risk derived from CIFs, the authors estimated additive heritability (h²) from general population versus full-sibling familial risk and genetic correlations (r_g) from cross-disorder parent–offspring risks, following Wray and Gottesman’s liability-threshold approach. Estimates were obtained per birth year and combined using inverse-variance weighted random-effects meta-analysis across years. Decomposition of comorbidity: Observed phenotypic comorbidity was summarized using hazard ratios from prior Danish work, converted into phenotypic correlations (r_p). For each MD–CMD pair with available data (35/90 pairs), r_p was decomposed into genetic (G) and environmental (E) components using the register-based h² and r_g for both traits. SNP-based comparison: For complementary analyses, GWAS summary statistics for MDs (iPSYCH and PGC) and CMDs (public repositories including MVP) were uniformly cleaned and harmonized. SNP-based heritabilities (h²_SNP) and genetic correlations (r_g,SNP) were estimated via LD Score Regression (LDSC) on the liability scale and compared with register-based estimates to assess concordance and implications for comorbidity decomposition.

- Register-based heritability: Across disorders, narrow-sense heritability estimates ranged from approximately 25% to 75%, with strong concordance between Denmark and Sweden and broad agreement with prior literature. - Genetic correlations: Among 90 MD–CMD pairs, meta-analysis across countries identified 32 significant genetic correlations (Bonferroni p<5.95×10^-6). With the exception of anorexia nervosa (AN), genetic correlations between MDs and CMDs were generally positive, and tended to be larger for ASD and bipolar disorder (BD) than for ADHD and affective disorders (AFF). - Decomposition of comorbidity (genetic vs environmental): Using phenotypic correlations derived from hazard ratios, the environmental component generally equaled or exceeded the genetic component and was often markedly larger. Genetic contributions to comorbidity were negligible for AN with CMD and very low for ASD with CMD. Across MDs, comorbidity with hypertension and atrial fibrillation had lower genetic components than other CMDs. In contrast, some MD–CMD constellations such as AFF with several CMDs showed substantial genetic contributions (around 50%). - Examples of heterogeneity: Pairs with similar genetic correlations (e.g., r_g≈0.1) could have very different genetic contributions to comorbidity depending on trait heritabilities, as illustrated by AFF–hypertension (~50% genetic) versus ASD–T2D (near-zero genetic). - SNP-based comparisons: On the liability scale, median h²_SNP was 0.17 for MDs (lowest BD 0.13; highest ADHD 0.23) and 0.09 for CMDs (lowest HF 0.023; highest unruptured intracranial aneurysm 0.22). Relative to register-based h², h²_SNP was lower for MDs (~63%) and CMDs (~72%). Twenty-one significant SNP-based genetic correlations were identified (Bonferroni p<5.95×10^-4), with strong concordance to register-based r_g. However, because h²_SNP underestimates total heritability, SNP-based decompositions substantially under-estimated the genetic contribution to comorbidity, underscoring the value of family-based register estimates.

The study directly addresses the nature-versus-nurture question underlying MD–CMD comorbidity by combining family-based heritability and genetic correlation estimates with observed comorbidity. Findings demonstrate that most observed comorbidity across MD–CMD pairs is predominantly environmental in origin, though the balance varies substantially by pair. Importantly, genetic correlation alone is insufficient to infer the genetic share of comorbidity; heritabilities of both traits are critical to quantify the genetic component. Disease-pair–specific decomposition reveals that some constellations (e.g., AFF with several CMDs) have substantial genetic contributions, while others (e.g., AN and many CMDs; ASD and CMDs) are mainly environmental. The authors discuss possible mechanisms for environmental effects, including behaviors directly related to MD (e.g., inactivity, sleep irregularities, smoking, overeating) and external exposures (e.g., second-generation antipsychotics like clozapine affecting weight, lipids, and cardiac electrophysiology). They highlight the clinical implications: where environment dominates, interventions should focus on modifiable exposures and care pathways; where genetics contributes, precision medicine approaches (e.g., genetic risk prediction) may add value. The strong agreement between register- and SNP-based genetic correlations supports shared common-variant architectures across several MD–CMD pairs, but SNP-based heritability underestimation leads to underestimation of the genetic share of comorbidity, emphasizing the importance of family-based approaches.

This study provides the first comprehensive quantification of genetic versus environmental contributions to comorbidity between six major mental disorders and 15 cardiometabolic diseases using nationwide genealogies and health registers from Denmark and Sweden, complemented by genotype-based analyses. The main contribution is demonstrating that comorbidity patterns are predominantly environmental overall, yet highly heterogeneous and disease-pair specific, and cannot be inferred solely from genetic correlations. The work lays a foundation for precision prevention and treatment: identifying MD–CMD pairs with substantial genetic components could inform genetic studies targeting comorbidity mechanisms and potential drug discovery/repurposing, while pairs dominated by environmental factors should motivate targeted identification and modification of risk exposures. Future research should: (1) delineate specific environmental exposures driving comorbidity for each MD–CMD pair; (2) identify genetic determinants underpinning comorbid presentations distinct from single-disorder risk; (3) assess gene–environment interplay; and (4) evaluate generalizability beyond Scandinavian settings and across healthcare systems.

- Reliance on secondary care hospital diagnoses over decades introduces potential truncation and under-representation of recently introduced diagnoses in older cohorts; by-year meta-analyses mitigate but may not eliminate bias. - Shared environmental effects among close relatives were not explicitly modeled and could slightly inflate heritability and genetic correlation estimates, though prior work suggests small impact. - The comorbidity decomposition assumes no interactions between genetic and non-genetic factors; violations could bias component estimates. - Hospital records may capture more severe phenotypes than primary care records; this may vary by country and disorder. - Generalizability to non-Scandinavian populations and different healthcare systems is uncertain, despite strong replication across Denmark and Sweden. - SNP-based estimates capture only a fraction of total heritability, affecting SNP-based decompositions; this reinforces, but does not fully resolve, estimation challenges.