Article structure
This article is organized around the clinical problem, study design, main quantitative findings, interpretation for precision psychiatry, major limitations, and implications for research and practice. This structure fits the translational nature of the paper, which sits at the intersection of psychiatric genetics, risk prediction, and population-based mental health research.
Highlight
In a Danish population-based case-cohort study of 94 276 unrelated individuals of European ancestry, recurrent copy number variant (rCNV) carriage was associated with higher risk of autism spectrum disorder (ASD), attention-deficit/hyperactivity disorder (ADHD), and schizophrenia spectrum disorder (SSD), but not major depressive disorder (MDD).
Polygenic scores (PGSs) for each disorder were independently associated with risk of the corresponding diagnosis, and at similar levels of absolute risk, PGSs generally identified more individuals than rCNV carriage, except in ASD.
The joint analysis suggested complementarity rather than redundancy: PGSs stratified risk among rCNV carriers, while some data hinted that low PGSs may attenuate risk associated with certain rCNVs.
A statistically significant negative interaction was observed between 16p13.11 duplication and ADHD-PGS on ADHD risk, and across multiple tests there was an overall tendency toward negative rCNV-PGS interaction coefficients.
Background
Psychiatric disorders such as ADHD, ASD, SSD, and MDD are highly heritable but genetically heterogeneous. Two major classes of inherited or de novo genomic risk have dominated contemporary psychiatric genetics. The first is common-variant burden, usually summarized as a polygenic score derived from genome-wide association studies. The second is rare structural variation, including recurrent copy number variants, which can have substantially larger per-variant effects but occur in far fewer individuals.
Clinicians increasingly encounter questions about whether a pathogenic or susceptibility CNV meaningfully changes prognosis, and whether a high or low polygenic score adds actionable information. This question is not merely academic. It affects genetic counseling, interpretation of newborn or pediatric genomic data, and the future design of precision psychiatry frameworks. Yet evidence on how rare high-impact variants and distributed polygenic burden jointly shape psychiatric risk remains limited.
rCNVs are particularly relevant because several loci, including 22q11.2, 16p11.2, 1q21.1, and 16p13.11, have been repeatedly linked to neurodevelopmental and psychiatric outcomes. Their penetrance is incomplete, however, and phenotype varies widely even among carriers of the same CNV. That variability raises a clinically important possibility: common genetic background may help explain why some carriers develop severe psychiatric illness while others do not.
The study by Vaez and colleagues directly addresses this issue by estimating absolute risk associated with rCNVs and PGSs independently and jointly in a large, nationally ascertained Danish sample from iPSYCH.
Study design
Population and setting
This was a genetic association study using the Lundbeck Foundation Initiative for Integrative Psychiatric Research, or iPSYCH, case-cohort sample. The source population comprised individuals born in Denmark from 1981 through 2008 and followed through 2015. The analytic sample included all individuals with a hospital diagnosis of ADHD, ASD, SSD, or MDD, plus a randomly drawn subcohort from the underlying population.
The primary analysis included 94 276 unrelated individuals of European ancestry. Mean age at follow-up was 21.9 years, and 53.7% were male.
Exposure assessment
Carrier status was determined for 27 autosomal recurrent CNV loci using neonatal dried blood spot DNA genotyped on microarrays. The investigators also generated PGSs for psychiatric and other outcomes using summary statistics from published genome-wide association studies.
The study therefore compared a relatively rare class of variants with moderate to large effect sizes against a common-variant summary measure that is individually low impact but cumulatively informative.
Outcomes
The primary outcomes were hospital-diagnosed ADHD, ASD, MDD, and SSD during follow-up. Absolute risk was estimated using a weighted survival analysis framework appropriate for the case-cohort design. Joint effects of rCNV carriage and PGSs were assessed using generalized linear models.
Comparators and analytic focus
The central comparison was not simply whether rCNVs or PGSs were associated with disease, since prior literature already supports both. Rather, the clinically important comparison was how much absolute risk each approach captured, how many individuals each approach identified at given risk thresholds, and whether polygenic background modified risk among rCNV carriers.
Key findings
Independent associations of rCNVs with psychiatric outcomes
As an aggregate group, rCNV carriage was associated with increased risk for three of the four major psychiatric outcomes studied. The reported effect estimates were strongest for ASD, followed by ADHD and SSD, while no clear association was observed for MDD. Specifically, aggregate rCNV carriage was associated with ASD risk with beta 0.33 (95% CI, 0.27-0.39), ADHD with beta 0.29 (95% CI, 0.23-0.35), SSD with beta 0.25 (95% CI, 0.17-0.33), and MDD with beta 0.04 (95% CI, -0.03 to 0.11).
These findings align with the broader literature suggesting that recurrent structural variants have their most consistent effects in neurodevelopmental and psychotic phenotypes, while their relationship to depression appears weaker and more variable. From a clinical standpoint, this pattern reinforces the view that rCNV findings are especially informative in developmental psychiatry and schizophrenia-spectrum risk assessment.
Independent associations of disorder-specific polygenic scores
Each disorder-specific PGS was positively associated with the corresponding psychiatric outcome. The effect sizes were beta 0.14 (95% CI, 0.12-0.16) for ASD-PGS and ASD risk, beta 0.28 (95% CI, 0.26-0.30) for ADHD-PGS and ADHD risk, beta 0.28 (95% CI, 0.26-0.30) for SSD-PGS and SSD risk, and beta 0.38 (95% CI, 0.36-0.40) for MDD-PGS and MDD risk.
These estimates indicate that common-variant burden remains a robust predictor across disorders, including MDD, where aggregate rCNV burden was not significantly associated. This difference between variant classes is clinically notable: common-variant liability may provide broader but shallower signal across psychiatric phenotypes, whereas rCNVs may deliver more concentrated signal for selected disorders.
Absolute risk and population reach
One of the most important translational observations in the study is that PGSs identified more individuals than rCNV carriage at comparable levels of absolute risk, except for ASD. This means that while individual rCNVs may confer larger relative effects, their rarity limits the number of people they can capture in a risk-stratified framework. PGSs, by contrast, can spread moderate risk information across a much larger segment of the population.
For clinicians and health systems, this is a crucial distinction. A rare high-impact variant may be highly informative for a given patient or family, but from a population-level prevention perspective, a polygenic tool may identify more people who cross a meaningful risk threshold. The exception seen in ASD suggests that some rCNVs may be especially efficient markers of substantial neurodevelopmental risk.
Joint effects of rCNVs and PGSs
The joint analysis is where the paper makes its most original contribution. Rather than showing simple additive accumulation alone, the data suggested that polygenic background may modulate the expression of some rCNV-associated risks. A statistically significant negative interaction was observed between 16p13.11 duplication and ADHD-PGS on ADHD risk, with beta -0.51 (95% CI, -0.86 to -0.16).
In addition, across aggregated rCNV groups and the 9 most common individual rCNVs, 27 of 39 interaction tests were negative, an imbalance unlikely to be due to chance alone (binomial P = .01). Although most individual tests were not definitive, the directional consistency suggests that the penetrance of certain rCNVs may be lower among individuals with relatively low polygenic burden for the same disorder.
This does not mean rCNVs are benign in low-PGS individuals. Rather, it argues against a simplistic deterministic model. The more accurate view is that psychiatric risk may emerge from layered genetic architecture, where rare large-effect variants and diffuse common-variant burden both shape outcome, sometimes in ways that are not purely multiplicative.
Why the negative interaction signal matters
Many clinicians might intuitively expect rare high-impact variants and high polygenic burden to amplify one another. That may still be true in some settings. However, a tendency toward negative interaction coefficients suggests a threshold or liability model in which either one source of risk can partly substitute for the other. In other words, individuals carrying a high-impact CNV may need less additional common-variant burden to cross a disease threshold, while those with lower polygenic burden may show somewhat attenuated penetrance despite carrier status.
This interpretation is biologically plausible and consistent with multifactorial threshold models of neuropsychiatric disease. It also helps explain variable expressivity among carriers of the same structural variant.
Clinical interpretation
What this means for psychiatric genetics clinics
For psychiatric genetics and neurodevelopmental clinics, the study supports using rCNVs and PGSs as complementary rather than competing tools. If a patient carries a recurrent susceptibility CNV, that result remains clinically meaningful, especially for ASD, ADHD, and SSD-related risk discussions. But it should not be interpreted in isolation. Polygenic background may help refine the expected range of psychiatric risk and perhaps future prognosis.
Conversely, a high PGS in the absence of a known rCNV may still be relevant, particularly at the population level. Because PGSs can identify many more individuals at comparable absolute risk, they may ultimately be more useful in stratified prevention frameworks, assuming ethical, technical, and ancestry-related challenges can be addressed.
Counseling implications
The findings are especially relevant for families confronting incidental or targeted CNV findings. A recurrent CNV increases risk, but incomplete penetrance remains the rule, not the exception. This study provides evidence that common genetic background may partly explain that uncertainty. In practical terms, counseling may need to move away from single-variant narratives toward integrated genomic risk communication.
That said, current clinical use of psychiatric PGSs remains limited. Most professional groups do not yet recommend routine use for diagnosis or treatment selection, largely because predictive performance, transferability across ancestries, and ethical implementation remain unresolved. This study advances the science, but it does not by itself justify widespread clinical deployment.
Disorder-specific nuance
The contrast across disorders deserves emphasis. rCNVs were clearly associated with ASD, ADHD, and SSD, but not MDD. This pattern may reflect true biological differences, with depression having a more diffuse and environmentally sensitive genetic architecture than classic neurodevelopmental CNV syndromes. It also suggests that clinical utility of structural variant testing may differ by disorder and developmental context.
For pediatric psychiatry and developmental services, these results are more immediately relevant than for adult depression clinics. For schizophrenia-spectrum disorders, the findings reinforce a growing case for integrating CNV information into etiologic evaluation, especially in early-onset, syndromic, or neurodevelopmentally complicated cases.
Strengths and limitations
Major strengths
The study has several notable strengths. First, the iPSYCH design provides a very large, population-based sample with linkage to high-quality national registry diagnoses. Second, the use of neonatal blood spots reduces reverse-causation concerns and ensures genomic exposure assessment precedes clinical outcome. Third, the authors focused on absolute risk, which is more clinically interpretable than odds ratios alone. Fourth, the joint modeling of rCNVs and PGSs addresses a major gap in psychiatric genetics.
Key limitations
Several limitations should temper interpretation. The analysis was restricted to unrelated individuals of European ancestry, limiting generalizability and underscoring the longstanding ancestry bias in genomic prediction research. Diagnostic outcomes were based on hospital diagnoses, which may underascertain less severe or untreated illness and may be influenced by service use patterns. Follow-up ended in 2015, with a mean age at follow-up of 21.9 years, so later-onset disorders, particularly some MDD and SSD cases, may not have been fully captured.
In addition, many individual rCNVs are rare, which limits precision for locus-specific interaction testing. The observed negative interaction trend is intriguing but should be viewed as hypothesis-generating beyond the statistically significant 16p13.11 duplication finding. Replication in other cohorts will be essential.
Finally, the study does not establish immediate clinical utility. Demonstrating statistical association and demonstrating benefit in real-world screening, counseling, or preventive intervention are different steps.
Expert commentary
This study fits with a broader transition in psychiatric genomics from variant discovery to risk architecture and translational interpretation. Over the past decade, work on CNVs has shown that rare structural variants can substantially increase risk for neurodevelopmental and psychotic disorders, while polygenic studies have shown that common variants distribute risk widely across the population. Vaez and colleagues move the field forward by asking how those signals coexist within the same individuals.
The most practice-relevant message is not that one metric is superior to the other. It is that they answer different clinical questions. rCNVs can flag a biologically meaningful susceptibility state, often with pleiotropic developmental implications. PGSs can organize gradations of liability across many more people. Used together, they may eventually support a more nuanced model of penetrance, prognosis, and targeted monitoring.
Still, implementation science is lagging behind genomic discovery. Psychiatric care lacks validated pathways for responding to a high PGS, and even CNV-informed counseling is inconsistently available outside specialty settings. Ethical challenges also remain substantial, especially around screening in children, probabilistic communication, and equity across ancestries and health systems.
Conclusion
In this large Danish study, recurrent CNVs and polygenic scores each contributed meaningful but distinct information about psychiatric risk. rCNV carriage was associated with ASD, ADHD, and SSD, whereas disorder-specific PGSs were associated with all four studied disorders, including MDD. At comparable absolute risk levels, PGSs generally identified more people than rCNV carriage, except in ASD. The joint analyses further suggest that polygenic background may modify penetrance among CNV carriers, with signals pointing toward attenuation of some CNV-associated risk in individuals with lower PGSs.
For precision psychiatry, the central implication is clear: rare structural variants and common polygenic burden should not be viewed in isolation. Their combined interpretation may better reflect the true architecture of psychiatric vulnerability than either measure alone. The next steps are replication, extension to ancestrally diverse cohorts, longer follow-up, and development of clinically responsible frameworks for integrated genomic counseling and risk communication.
Funding and ClinicalTrials.gov
The report was conducted within the Lundbeck Foundation Initiative for Integrative Psychiatric Research (iPSYCH) and associated consortium efforts, including the LINC Consortium, as stated in the publication. No ClinicalTrials.gov registration applies because this was a genetic association study rather than an interventional clinical trial.
Citation
Vaez M, Montalbano S, Waples R, Dybdahl Krebs M, Georgii Hellberg KL, Gådin J, Stow D, Holmans P, van den Bree M, Børglum AD, Helenius D, Werge T, Schork AJ, Ingason A; LINC Consortium. Recurrent Copy Number Variants and Psychiatric Outcomes in the Context of Polygenic Scores. JAMA Psychiatry. 2026 May 27:e261064. doi: 10.1001/jamapsychiatry.2026.1064. Epub ahead of print. PMID: 42201728; PMCID: PMC13217261.
