Highlights
– Large, register-linked Danish study demonstrates both shared and distinct clinical and genetic risk architectures for suicide attempt (SA) and death by suicide.
– Chronic, functional health conditions (for example, hearing problems or dyslipidemia) showed stronger associations with SA; severe somatic illnesses (for example, cancer) were relatively more associated with suicide.
– Genetic liability to psychiatric disorders influenced suicide mortality, whereas SA was associated with broader health-related polygenic risk profiles in addition to psychiatric genetic risk.
– Cumulative burden (recurrent attempts) showed dose-response relationships with most measured health conditions and polygenic scores, supporting a graded risk model for persistent suicidality.
Background and clinical context
Suicide remains a major and complex public health problem worldwide. Fatal and nonfatal suicidal behaviour differ in important epidemiological ways: suicide attempts are far more common and concentrated among younger people and women, whereas death by suicide disproportionately affects older individuals and men. Clinicians frequently assume nonfatal suicide attempts and suicide deaths lie on a single severity spectrum, but emerging data suggest partially distinct etiological pathways. Understanding which clinical conditions and genetic liabilities differentially predict attempts versus deaths could refine prevention strategies across medical and psychiatric settings.
Study design and methods
The work by Ge et al. (JAMA Psychiatry, 2025) used two complementary, register- and genetics-based designs nested within the Danish population to dissect shared and distinct risk factors for SA and suicide. Key design features:
Register-based analysis (health conditions)
– Nested case-control design using national Danish registers to identify cases of clinically diagnosed nonfatal suicide attempt (n=81,713) and death by suicide (n=9,362) with matched population controls.
– Inclusion restricted to individuals older than 10 years to reduce misclassification.
– Exposure assessment comprised 28 clinically coded health conditions spanning psychiatric, neurological, pain-related, sensory, and chronic somatic illnesses.
– Conditional logistic regression was used to estimate associations between health conditions and outcomes, and analyses examined cumulative SA burden (recurrent attempts).
Genetic analysis (polygenic scores)
– Genotypes came from the iPSYCH2015 dataset, an epidemiologically ascertained Danish genetic sample nested within the national cohort. The PGS analysis included 8,221 SA cases and 225 suicide cases with matched controls.
– Thirty-five polygenic scores (PGSs) for psychiatric, cognitive, and broad health-related traits were calculated using LDpred2-auto, a contemporary method for deriving genome-wide polygenic predictors while accounting for linkage disequilibrium.
– Associations between PGSs and outcomes were estimated using logistic regression. To formally test whether effect sizes differed between SA and suicide, the authors applied a Bayesian model-based classification and Cochran’s Q heterogeneity tests.
Key findings
The study delivers several principal observations with clinical and research implications.
1) Population sizes, demographics, and contrasting epidemiology
– Register sample: 81,713 SA cases (61.8% female; mean age 32.3 years) with 408,490 matched controls; 9,362 suicide cases (25.2% female; mean age 45.1 years) with 46,749 matched controls.
– Genetic sample: 8,221 SA cases (72.3% female; mean age 19.7 years) and 225 suicide cases (35.6% female; mean age 24.6 years). The markedly smaller number of genomic suicide cases limits power to detect modest genetic associations for mortality.
2) Differential patterns across health conditions
– Many health conditions increased risk for both SA and suicide, confirming a broad multisystem link between poor health and suicidal outcomes.
– Relative predominance: chronic, functional, and disabling conditions (examples cited include dyslipidemia and hearing problems) tended to show stronger associations with nonfatal SA. The authors interpret this as likely reflecting pathways through functional impairment, social isolation, chronic pain, and reduced quality of life that elevate the risk of attempting suicide but not necessarily the transition to fatality.
– Conversely, severe, life-threatening somatic diseases (for example, advanced cancer) showed relatively stronger associations with death by suicide. This pattern is consistent with the influence of existential distress, pain, and prognostic knowledge on the decision to end life.
3) Genetic architecture: psychopathology versus broad health liability
– Polygenic liability to major mental disorders (mood disorders, schizophrenia, etc.) was associated with risk of death by suicide. In contrast, SA was associated with PGSs for psychiatric traits plus additional PGSs indexing broader health-related traits (for example, cardiometabolic or sensory-related genetic predispositions).
– The observed divergence implies that genetic risk for suicide death may be more tightly coupled to psychiatric liability, while genetic susceptibility to attempts reflects a wider constellation of vulnerabilities that includes somatic and functional domains.
4) Cumulative burden and dose-response relations
– For most health conditions and PGSs evaluated, there was a clear dose-response relationship with cumulative SA burden: more recurrent attempts were associated with higher exposure prevalence or greater polygenic loading. This graded relationship supports a model where additive clinical and genetic risks compound vulnerability to repeated suicidal behaviour.
5) Statistical evidence for distinctness
– Using heterogeneity tests and Bayesian classification, the investigators found statistically significant differences in effect sizes across a subset of conditions and PGSs between SA and suicide, supporting the notion of partially distinct etiological profiles rather than a single severity continuum.
Expert commentary and interpretation
These results carry several practical and theoretical implications.
Clinical implications
– Broad medical settings matter for suicide prevention. The stronger association between functional/chronic conditions and SA indicates that non-psychiatric clinicians (primary care, otology, cardiology, oncology) should maintain heightened vigilance for suicidal ideation and attempts, particularly among younger patients and women who have higher attempt rates.
– For patients with severe, life-limiting illnesses, proactive psychosocial and palliative care interventions may be especially important given the relative association with suicide mortality.
Genetic and translational implications
– The genetic results reinforce the centrality of psychiatric liability for suicide death and suggest that broader polygenic vulnerability influences who attempts suicide. However, current PGSs are not ready for routine clinical screening: effect sizes remain modest, predictive utility at the individual level is limited, and genetic associations may vary by ancestry.
Mechanistic hypotheses
– Several nonexclusive mechanisms might explain the observed patterns: (1) severe somatic illness can increase suicide risk via pain, functional loss, hopelessness, and access to lethal means; (2) chronic but nonfatal conditions may increase impulsivity, social withdrawal, and repeated help-seeking that manifest as nonfatal attempts; (3) genetic liabilities to psychiatric disorders may influence decision-making, stress responsivity, and suicidal intent, increasing risk of death when combined with adverse environmental events.
Strengths
– Exceptional sample sizes for register-based analyses and population-wide linkage of clinical records strengthen statistical precision and minimize selection bias.
– Integration of genomic data with nationwide registers is a methodological strength that allows simultaneous examination of environmental and genetic determinants.
Limitations and cautions
– The genomic analysis had limited suicides (n=225), constraining power to detect genetic effects specific to death by suicide and increasing risk of type II error.
– Register-based ascertainment relies on clinically coded events and may miss unrecorded attempts or out-of-hospital deaths; misclassification is possible despite rigorous case definitions.
– The cohort is of Danish (predominantly European) ancestry; generalizability to other populations, particularly those with different healthcare systems, social structures, or genetic backgrounds, may be limited.
– Observational design precludes causal inference. Unmeasured confounding (for example, socioeconomic status nuances, access to care, or environmental exposures) could influence observed associations.
Clinical and research takeaways
– Clinicians should consider broad systemic health when assessing suicide risk: both psychiatric and non-psychiatric disorders contribute meaningfully to attempts and, to varying extents, to mortality.
– Suicide prevention strategies need tailoring: interventions to reduce attempts may require targeting functional impairment, chronic pain, and social support, while preventing suicide mortality may require intensive psychiatric treatment and attention to existential distress in severe somatic illness.
– Further work should expand genomic samples of suicide deaths, replicate findings in other ancestries, and explore causality using longitudinal and causal inference approaches (for example, mendelian randomization where appropriate).
Conclusion
Ge and colleagues present robust evidence that suicidal behaviour is underpinned by overlapping but distinguishable clinical and genetic risk architectures. The findings challenge a simplistic severity-only model and encourage a more nuanced approach to risk assessment, prevention planning, and interdisciplinary care that spans psychiatry and general medicine.
Funding and clinicaltrials.gov
The study’s detailed funding statements are reported in Ge et al. (JAMA Psychiatry 2025). This work is an observational register- and genetics-based analysis and was not a clinical trial; therefore, no clinicaltrials.gov registration is expected. Readers should consult the original article for precise funder acknowledgements and declarations of interest.
References
1. Ge F, Wang Y, Agerbo E, Köhler-Forsberg O, Bulik CM, Petersen LV, Vilhjálmsson BJ. Contrasting Risk Profiles for Suicide Attempt and Suicide Using Danish Registers and Genetic Data. JAMA Psychiatry. 2025 Oct 21:e253444. doi:10.1001/jamapsychiatry.2025.3444. PMID: 41118583; PMCID: PMC12541595.
2. Turecki G, Brent DA. Suicide and suicidal behaviour. Lancet. 2016 Apr 23;387(10024):1227-1239. doi:10.1016/S0140-6736(15)00234-2.
3. World Health Organization. Suicide. https://www.who.int/news-room/fact-sheets/detail/suicide. Accessed 2025.
4. Agerbo E, Nordentoft M, Mortensen PB. Familial, social, and socioeconomic risk factors for suicide in young people: nested case-control study. BMJ. 2002 Jul 13;325(7355):74. doi:10.1136/bmj.325.7355.74.
5. Privé F, Arbel J, Vilhjálmsson BJ. LDpred2: better, faster, stronger. Bioinformatics. 2020 Nov 1;36(22-23):5424-5431. doi:10.1093/bioinformatics/btaa1029.
