Genome-wide meta-analysis of quantitatively measured generalized anxiety symptoms in individuals of European ancestry

Anxiety is heritable and exists on a continuum, with symptoms ranging from adaptive threat response to clinical disorder. Here we performed a genome-wide association meta-analysis of generalized anxiety symptom severity in 693,869 individuals of European ancestry from 14 cohorts. We identified 80 independent genome-wide significant variants within 74 loci, 39 of which were newly associated with anxiety. SNP-based heritability was 5.9% (posterior s.d. = 0.15%). Polygenic scores were significantly associated with anxiety symptom severity and disorder in European, African and South Asian ancestry samples (R2 = 1.2–2.9%). Significant genetic correlations (rg) were estimated with mental and physical health traits, including case–control anxiety, neuroticism and depression (rg = 0.71–0.85), irritable bowel syndrome (rg = 0.57), coronary artery disease, endometriosis and migraine (rg = 0.20–0.27). Gene-based and pathway analyses implicated synaptic and axonal processes, with enriched expression in the brain. These findings highlight the discovery power gained from analysing a quantitative trait rather than a case–control phenotype in anxiety genetics. Similar content being viewed by others Main Anxiety disorders are the most prevalent mental health conditions worldwide1 and rates are rising2,3,4. Anxiety is associated with reduced quality of life5, elevated mortality6,7 and is frequently comorbid with other mental8,9,10,11,12 and physical13 health conditions, ranging from irritable bowel syndrome to cancer. When co-occurring with other conditions, anxiety symptoms can exert an independent and sometimes greater impact on quality of life and functioning than the primary diagnosis8, as has been reported in autism14 and bipolar disorder15. Twin and family studies estimate the heritability of anxiety disorders at 20–60% (ref. 13), with measures capturing stable anxiety typically showing higher heritability16. Early case–control genome-wide association studies (GWAS)17,18,19, which aggregated across anxiety subtypes, identified a handful of risk loci. More recent large-scale efforts using a range of analytical approaches have reported between 14 and 51 associated loci20,21,22. A recent GWAS23 of anxiety disorders from the Anxiety Disorders Working Group of the Psychiatric Genomics Consortium (PGC-ANX) identified 58 independent loci from over 120,000 cases. Across these studies, single nucleotide polymorphism (SNP)-based heritability estimates ranged from 5% to 10% (refs. 20,22,23). Fear and worry serve an evolutionary function by promoting vigilance and caution in response to potential threats24. Variation in threat sensitivity across individuals may be adaptive at the group level and, consistent with this, anxiety symptoms exist in the population along a continuum of frequency and severity. Clinical anxiety represents a practical threshold at the upper extreme of this distribution based on levels of distress and functional impairment. GWAS of anxiety using quantitative symptom scores capture genetic variation across the full phenotypic range, not only at clinical thresholds. This approach can offer greater statistical power25 and a more comprehensive representation of genetic associations with anxiety traits. The degree of genetic overlap between quantitative anxiety symptom severity and disorder status has received little study. A GWAS of a 2-item anxiety scale in the European ancestry subsample of the Million Veteran Program (MVP) identified 5 significant loci26 and moderate-to-strong genetic correlations with lifetime anxiety disorder (rg = 0.59–0.87)13,26. However, the brevity of the measure may have limited the potential to comprehensively capture phenotypic variance. Additional support for shared genetic signal across the anxiety continuum, including above and below diagnostically relevant thresholds, comes from a UK Biobank study19 reporting high genetic correlations between mild, moderate and severe anxiety symptom groupings (rg = 0.76–0.98). Among individuals with lifetime anxiety or depression, anxiety symptom severity has also shown a significant genetic correlation with functional impairment ratings (rg = 0.79)27, which have clinical relevance. These findings provide preliminary evidence that a well-powered GWAS of quantitative anxiety would identify variants relevant to both symptom severity and clinically defined anxiety. Aside from the MVP analysis of a two-item scale, previous quantitative anxiety GWAS28,29 have incorporated other psychiatric or personality traits to maximize statistical power, resulting in findings relating to a broader construct than anxiety symptoms specifically. To address this, we performed a genome-wide association meta-analysis of generalized anxiety disorder (GAD) symptom severity in 693,869 individuals of inferred European ancestry across 14 cohorts from PGC-ANX, building on our previous study23 of individuals with anxiety disorder. Most cohorts used the GAD 7-item scale (GAD-7) or conceptually similar self-report measures of recent GAD symptoms. In addition to identifying associated loci, we performed variant- and gene-level investigations, estimated genetic correlations with relevant traits and evaluated polygenic score prediction in independent samples of European, African and South Asian ancestry.

Genome-wide association meta-analysis We performed a genome-wide association meta-analysis of GAD symptom severity in 693,869 individuals of European ancestry across 14 cohorts from 8 countries (see Supplementary Tables 1–3 for cohort, phenotype and GWAS details). The analysis identified 80 independent genome-wide significant variants (P < 5 × 10−8) across 74 loci (Fig. 1, Supplementary Fig. 1 and Supplementary Table 4. Effect sizes (β) represent the associated change in standard deviation units of generalized anxiety symptom score per additional copy of the effect allele. The top signal came from an intergenic locus near the long non-coding RNA RP4-598G3.1 on chromosome 1 (lead SNP rs7546305; P = 3.8 × 10−15; β = −0.014; 95% confidence interval (CI), −0.017 to −0.010), followed by a locus within an intron of PCLO on chromosome 7 (lead SNP rs1476548; P = 4.9 × 10−15; β = −0.014; 95% CI, −0.017 to −0.010). These loci have both previously been associated with internalizing traits, including anxiety (Supplementary Table 5). Of the 74 loci, 16 had no previous associations with internalizing trait GWAS, including neuroticism, anxiety or depression, at any variant in linkage disequilibrium (LD; defined as r2 > 0.1) with the lead SNP (Methods and Supplementary Table 5a). Thirty-nine loci were novel for anxiety specifically, measured either as a diagnosis or symptom severity phenotype. Of the 58 loci previously identified in the PGC-ANX anxiety disorder study23, all showed the same direction of association in the present analysis, with 19 (33%) reaching genome-wide significance and a further 33 (57%) showing nominal significance (P < 0.05) (Supplementary Table 5b). Nine of the 14 cohorts included here also contributed to the PGC-ANX anxiety disorders study, although cohort sample composition differed somewhat due to the availability of symptom score versus diagnostic information. Power calculations30 confirmed increased power in the present analysis relative to the anxiety disorder study (Supplementary Table 6). Among the 6,012 genome-wide significant SNPs, between-study heterogeneity accounted for a moderate proportion of variance in effect size estimates (measured with I2), with a median I2 across cohorts of 29.7%. In contrast, the median I2 for the 80 LD-independent genome-wide significant SNPs was 0%, indicating highly consistent effect size estimates across cohorts. Heterogeneity statistics for the lead SNPs are presented in Supplementary Table 4. Genetic correlations between sufficiently powered cohorts, estimated using LD score regression (LDSC), ranged from 0.64 to 0.97 (significant at a Bonferroni-corrected threshold P < 4.6 × 10−4; Supplementary Table 7). As a sensitivity analysis, we categorized cohorts by generalized anxiety symptom severity measure (for example, GAD-7; six subgroups) and by ascertainment method (community and clinical subgroups) (Supplementary Table 2). Subgroup meta-analyses were performed as per the main analysis, and genetic correlations between subgroups were estimated with LDSC. While most subgroup comparisons were underpowered, including by ascertainment method, the genetic correlation between the 2 most frequently used measures was high (GAD-7 and GAD-2, rg = 0.86; 95% CI, 0.79–0.92; P = 9.6 × 10−166, significant at a Bonferroni-corrected threshold P < 5 × 10−3; Supplementary Table 8). SNP-based heritability and genetic correlations with external traits The SNP-based heritability estimate from SBayesRC was 5.90% (posterior s.d. = 0.146%). SBayesRC was selected due to the known underestimation of heritability from LDSC31. LDSC32 indicated that genomic inflation was largely not attributable to confounding (intercept = 1.03; 95% CI, 1.01–1.05). To characterize the broader genetic architecture of GAD symptoms, we used LDSC33 to estimate genetic correlations with 105 traits spanning mental and physical health, personality, cognitive and socio-economic domains (Supplementary Table 9). Following Bonferroni correction (P < 4.8 × 10−4), 64 associations remained significant (Fig. 2, omitting 4 traits represented by multiple studies). The strongest correlations were observed with quantitative internalizing traits, including neuroticism, depressive symptoms and a genetic anxiety factor (rg = 0.83–0.86), and with case–control phenotypes for anxiety and depression (rg = 0.71–0.80). Moderate-to-strong estimates were also found for post-traumatic stress disorder (rg = 0.64), insomnia (rg = 0.49), irritable bowel syndrome (rg = 0.57) and chronic pain (rg = 0.56). Negative correlations were observed with socio-economic status indicators including household income (rg = −0.43). Smaller but significant correlations were identified with coronary artery disease, endometriosis, hypothyroidism and migraine (rg = 0.20–0.27). Most other associations with physical traits and illnesses were comparatively weaker and included positive correlations with resting heart rate, rheumatoid arthritis and atopic dermatitis, and a negative correlation with lung function (absolute rg = 0.06–0.13). Conditional analyses To assess whether genome-wide significant loci for generalized anxiety symptoms represented associations unique to this phenotype, we performed multi-trait conditional and joint analysis (mtCOJO)34, conditioning on highly genetically correlated traits. mtCOJO uses Generalized Summary-data-based Mendelian Randomization (GSMR) to estimate the effect of a conditioning trait on an outcome (bxy). Conditioning on neuroticism (67 instruments, GSMR bxy = 0.47 ± 0.02) resulted in 9 remaining genome-wide significant SNPs across 2 loci (P < 5 × 10−8; Supplementary Table 10). Conditioning on case–control anxiety (11 instruments, bxy = 0.22 ± 0.02) yielded 66 significant SNPs across 6 loci, and conditioning on case–control depression (98 instruments, bxy = 0.31 ± 0.01) produced 76 SNPs across 6 loci. Approximately half of SNP effect sizes were attenuated and standard errors had a median inflation of 7–9%, suggesting the loss of significance was not solely attributable to reduced power. Sign changes were observed in 17–20% of variants for each trait, consistent with instability, as expected under high genetic correlations. The LDSC SNP-based heritability (h2) estimates of the conditioned summary statistics were significant (neuroticism h2 = 0.014; 95% CI, 0.012–0.016; P = 1.02 × 10−42; case–control anxiety h2 = 0.018; 95% CI, 0.016–0.020; P = 1.54 × 10−59; case–control depression h2 = 0.017; 95% CI, 0.015–0.019; P = 7.34 × 10−51), but block jack-knifing confirmed that each was significantly lower than the original estimate (Z-statistic = 33.24, P = 3.3 × 10−242; Z = 28.61, P = 5.0 × 10−180; Z = 30.48, P = 4.24 × 10−204, respectively). Polygenic risk scores To evaluate the within- and cross-ancestry generalizability of our GWAS findings, we generated a polygenic risk score (PRS) for GAD symptom severity using SBayesRC35 and tested its association with both quantitative and case–control anxiety outcomes in independent samples across ancestry groups (Supplementary Table 11). A Bonferroni-corrected significance threshold of P < 1.7 × 10−2 was applied, accounting for 3 ancestry groups tested per outcome. The PRS significantly explained 2.9% of the variance (R2) in generalized anxiety symptom scores in an independent European ancestry sample (β = 0.55; 95% CI, 0.44–0.65; P = 8.9 × 10−24), 1.4% in an African ancestry sample (β = 0.48; 95% CI, 0.28–0.67; P = 2.9 × 10−6) and 1.2% in a South Asian ancestry sample (β = 0.45; 95% CI, 0.27–0.64; P = 2.0 × 10−6). For case–control anxiety, assuming 20% prevalence, the PRS explained 1.8% (β = 0.25; 95% CI, 0.14–0.36; P = 1.2 × 10−5) of the variance on the liability scale in a European ancestry sample, 1.7% (β = 0.27; 95% CI, 0.10–0.44; P = 2.0 × 10−3) in an African ancestry sample and 1.4% (β = 0.25; 95% CI, 0.10–0.41; P = 1.3 × 10−3) in a South Asian ancestry sample. These R2 values represent the variance explained by the PRS beyond that accounted for by covariates. Positional and functional annotation Functionally informed fine-mapping using PolyFun36 and SuSiE37 identified 4 putative causal variants with posterior inclusion probabilities (PIP) ≥0.95. Two of these were index variants for genome-wide significant loci: rs2392289 at locus 31 (PIP = 0.983) and rs72676302 at locus 58 (PIP = 0.978). The remaining two variants were in loci that did not reach genome-wide significance. One variant, rs72676302, had a high Combined Annotation Dependent Depletion (CADD) score of 19.78, suggesting a deleterious effect (above the suggested threshold of 12.37 (ref. 38)), although there was little biological evidence that it is within a regulatory element (regulomeDB score = 5). In addition, we identified 24 small credible causal sets (<10 variants each) that cumulatively met the PIP threshold (Supplementary Table 12). SNP-level gene annotation was performed using FUMA39 based on positional, expression quantitative trait loci (eQTL), and chromatin interaction (Hi-C) mapping (Supplementary Table 13). We identified genes annotated by more than one method, with this convergence providing greater support for their involvement. This approach again highlighted PCLO and CRHR1, a key regulator of the stress response, along with TMEM106B, which has been repeatedly associated with anxiety18,19,23 and with depression40. SORCS3 was implicated by both positional and chromatin interaction mapping and has also been reported in a previous anxiety GWAS23. In addition, multiple genes were identified that have previously been implicated in depression (for example, ERBB4, GRM7, VRK2, DCC, LRFN5, PCLO and NEGR1) and schizophrenia (for example, ERBB4, VRK2 and NEGR1). MAD1L1 (refs. 17,18,19,26) was supported only by positional mapping in our analysis but has been linked to anxiety phenotypes in previous GWAS, lending further support to its potential relevance. Although detailed functional information was limited for many mapped genes, several are posited to play roles in key neurotransmitter systems, including glutamatergic (GRM7 and HOMER1), GABAergic (ERBB4) and dopaminergic (DRD1) signalling. In contrast, gene annotations for loci novel for internalizing traits did not map clearly to defined neurobiological pathways relevant to anxiety. For example, rs58179213 is proximal to multiple genes implicated in diverse biological processes, including lung function (SFTPC), intestinal barrier function and immunoprotective inflammation (FHIP2B), and hair growth (HR). While such processes could influence anxiety via indirect pathways (for example, gut–brain interactions or somatic symptoms), independent replication will be required to provide the greater power necessary to determine their robustness and relevance. Gene-based associations and enrichment Gene-based association analysis was conducted using MAGMA41, which aggregates trait–SNP associations across all SNPs within a gene while accounting for LD. In total, 197 genes across 80 independent loci surpassed the Bonferroni-corrected significance threshold (P < 2.5 ×10−6; Supplementary Table 14). The top associated gene was PCLO (P = 2.5 × 10−21), mirroring the SNP-based results. To assess biological pathways significantly enriched for associations with GAD symptom severity, we performed pathway analysis in MAGMA with predefined gene sets (MsigDB42; curated and Gene Ontology terms; Supplementary Table 15,a). Six gene sets passed Bonferroni correction (P < 2.9 × 10−6): postsynaptic membrane (271 genes, P = 2.6 × 10−8), synaptic membrane (383 genes, P = 7.1 × 10−8), axon (627 genes, P = 5.3 × 10−7), neurogenesis (1,627 genes, P = 8.6 × 10−7), postsynapse (648 genes, P = 9.2 × 10−7) and generation of neurons (1,412 genes, P = 1.0 × 10−6). These results were unaffected by the exclusion of the major histocompatibility complex region, although the number of genes per set slightly decreased (Supplementary Table 15b). Gene–tissue expression analysis was conducted using catalogues of gene expression levels across different human tissues (Supplementary Fig. 2–5), with a Bonferroni correction applied for the number of tests within each tissue set. Among brain samples from 11 developmental stages (BrainSpan), only the 2 prenatal samples were significant (P = 6.3 × 10−4 and 2.5 × 10−3). In adult tissues (GTEx v.8 (ref. 43)), significant enrichment was observed in brain (P = 3.0 × 10−9) and pituitary (P = 4.9 × 10−5) tissues. Analysis of more specific tissue types revealed enrichment in 11 brain regions, most strongly in the frontal cortex, cortex, cerebellum, anterior cingulate cortex and nucleus accumbens (all P < 9.3 × 10−4; Supplementary Table 16). Drug targets To explore therapeutic relevance, we ran DrugTargetor44 to identify drug priorities with potential utility for clinical anxiety, based on their relevance for GAD symptom severity. We did not observe any significant enrichment of anxiety associations for the 1,551 drug targets tested (Supplementary Table 17). However, at the drug class level, we identified significant associations for Anatomical Therapeutic Chemical classifications N06 ‘psychoanaleptics’, including N06A antidepressants, N02A ‘opioids’ and N06B ‘psychostimulants’ (Bonferroni-adjusted P < 0.05; Supplementary Table 18). With the exception of psychostimulants, these classes include drugs that have been used clinically to treat anxiety45. The enrichment for psychostimulants probably reflects shared dopaminergic and noradrenergic pathways involved in arousal and vigilance and underlying both anxiety and attention, consistent with pleiotropic genetic associations rather than direct therapeutic action.

In this genome-wide association meta-analysis of 693,869 individuals from 14 cohorts, we identified 80 genome-wide significant variants across 74 loci associated with quantitative measures of GAD symptom severity. Approximately half of the identified loci replicated associations reported in previous anxiety GWAS19,20,21,22,23,46, while the remainder were newly identified associations. The strongest intragenic association was estimated for rs1476548 within PCLO, which was also implicated through positional mapping, eQTL mapping and gene-based association testing. PCLO encodes a protein involved in regulating presynaptic structure and neurotransmitter release. This gene has long been of interest in major depressive disorder47, with recent evidence also linking it to anxiety disorders20,23. Another gene of interest from our analysis was SORCS3, which was supported by multiple lines of evidence in the recent PGC-ANX case–control anxiety GWAS23. SORCS3 plays a role in postsynaptic functioning and glutamate receptor regulation, particularly in the hippocampus48. It has been linked to memory and learning processes, specifically synaptic depression and fear extinction49, and mental health and neurodevelopmental conditions, including major depressive disorder, Tourette syndrome, attention deficit hyperactivity disorder and autism50,51. Our SNP-based heritability estimate (5.9%) aligned with an existing GWAS of GAD symptom severity (5.6%)26 while remaining lower than liability scale estimates from case–control anxiety meta-analyses23. In contrast to traditional case–control phenotyping, which aims to maximize clinical specificity through diagnostic thresholds, our approach leveraged the full spectrum of symptom variability, increasing power for discovery52 and capturing genetic risk relevant to both subclinical and clinical presentations. The key differences between clinical diagnoses and symptom severity measures relate to the presence of distress and impairment and to symptom duration. While subclinical symptoms can still cause distress and impairment, this is not always true for lower levels of anxiety severity, potentially contributing to some diagnosis-specific genetic variance. Similarly, diagnostic tools often assess lifetime occurrence and require symptoms to be present for a minimum period of time (6 months for GAD), whereas symptom severity scales typically capture recent experiences (for example, the past 2 weeks), introducing greater susceptibility to transient fluctuations and measurement noise. Consistent with this, GWAS53,54,55 of depression symptom severity typically yield lower SNP-based heritability estimates than case–control analyses. We aimed to partially address and control for temporal fluctuations and better approximate a stable underlying trait16 by incorporating assessments from multiple time points into our analysis, where available. In addition, our meta-analysis combined data from multiple cohorts with differences in phenotype definitions, genotyping arrays, imputation methods, quality control procedures and population structure adjustments, which may have introduced further heterogeneity and reduced the observed SNP-based heritability. By contrast, single-cohort studies with individual-level data are more homogeneous and can implement alternatives to summary-statistics-based methods. Despite our lower heritability estimate, the strong genetic correlation observed between our phenotype and case–control anxiety suggests that GAD symptom severity captures much of the same genetic risk. This finding is consistent with a recent analysis of obsessive–compulsive symptoms56. Quantitative, symptom-based approaches may be particularly well suited to genetic studies of anxiety, given the high burden of anxiety symptoms observed across other mental health conditions8,9,10,11. In this context, efforts to isolate ‘pure’ anxiety cases may be both methodologically challenging and capture an unusual clinical phenotype that is unrepresentative for most individuals with anxiety. There was a broad range of significant genetic correlations across both mental and physical health conditions, consistent with the frequent co-occurrence with anxiety symptoms and widespread pleiotropy. A strong genetic correlation was observed with neuroticism, a well-established risk factor for anxiety23. This probably reflects both genuine shared liability and conceptual or item-level overlap between the measures. Conditional analyses indicated extensive overlap between loci associated with generalized anxiety symptoms and those implicated in neuroticism, case–control anxiety and depression. Many genome-wide significant loci may therefore index a broader neuroticism-related liability, although the presence of anxiety-specific associations cannot be excluded due to the statistical noise introduced by conditioning on highly genetically correlated traits. Many of the genetic correlations aligned with findings from a genomic structural equation modelling analysis of anxiety symptoms and disorder22, including strong associations with irritable bowel syndrome and chronic pain, and a moderate association with migraine. These results do not necessarily indicate horizontal pleiotropy but could instead arise from the experience of these conditions eliciting uncertainty and worry that contribute to anxiety. Polygenic scores derived from our genome-wide association meta-analysis demonstrated within- and cross-ancestry generalizability, significantly explaining 1.2–2.9% of the variance in GAD symptom severity in European, African and South Asian ancestry samples. Across these samples, the PRS also accounted for 1.4–1.8% of the variance in case–control anxiety on the liability scale. While broadly consistent with the 0.5–2.3% range reported in the PGC-ANX case–control analysis23, direct comparisons are limited by methodological differences in PRS construction and target sample composition. A key limitation of the present meta-analysis is its restriction to cohorts of European ancestry, due to lack of data for other ancestries at sufficient scale for GWAS analysis. Our PRS findings support a degree of shared genetic architecture with African and South Asian populations. Ancestry-specific modelling across diverse populations remains necessary to identify population-specific risk loci, such as that identified in a previous analysis26 of African American participants, and ensure equitable benefits from genetic discoveries. Overall, these findings provide additional evidence that quantitative phenotyping can effectively capture genetic signal relevant to clinical anxiety. The GAD symptom severity measures and ascertainment methods varied across contributing cohorts, although most assessed symptoms using the GAD-7. While widely adopted across clinical and research contexts, the GAD-7 does not comprehensively assess all GAD symptoms from the Diagnostic and Statistical Manual of Mental Disorders (DSM-5)57—omitting sleep, fatigue and concentration problems—and is not designed to capture symptoms of fear-based anxiety (that is, phobias and social anxiety disorder) or panic disorder. This limits the generalizability of our findings across anxiety disorders, particularly in light of evidence for partially distinct phenotypic and genetic contributions to GAD compared with fear-based disorders58,59. Expanding future studies to incorporate a broader range of anxiety symptom measures will enable more robust, transdiagnostic translation of these findings. Furthermore, future work could examine the genetic architecture of individual symptom domains, such as cognitive versus physiological symptoms, to better understand the biological specificity of these. Our subgroup analyses based on measure and ascertainment method were largely underpowered to reliably estimate SNP-based heritability or correlations. Although sufficiently powered comparisons indicated high genetic overlap, we cannot be certain that all cohorts captured the same underlying genetic architecture. While population-based cohorts allow assessment of the full range of symptoms, symptom severity measures typically better distinguish variation at the upper end of the distribution. This results in highly skewed symptom severity scores, as most participants report few or no symptoms, whereas individuals in clinical cohorts typically report more symptoms. Combining population-based cohorts with studies selecting on diagnosis introduces some heterogeneity and may limit generalizability to broader populations, but it can help yield a more normally distributed phenotype for GWAS analysis. This combination may have improved statistical power for detecting associations in our study. Our quantitative GWAS of GAD symptom severity identified more genome-wide significant loci than a slightly larger and mostly overlapping case–control anxiety study (N = 852,222; 122,341 cases)23, with many loci replicated across the 2 methods. This aligns with expectations under the liability-threshold model when considering common conditions such as anxiety, whereby quantitative traits generally offer greater statistical power than case–control designs of equal sample size52. Beyond identifying anxiety-associated loci, our results implicate key neurobiological pathways, including synaptic function and neurotransmission, and notable genes such as PCLO and SORCS3. These findings demonstrate that a quantitative anxiety symptom-based phenotype can reveal biologically meaningful signals and complements insights from case–control designs. Clinically ascertained samples remain essential for identifying disorder-specific biology and mapping genetic risk to diagnostic presentations; however, obtaining clinical cases at sufficient scale for binary genome-wide analyses is challenging. Although electronic health records offer an efficient option, these are limited to individuals seeking and receiving medical attention. Quantitative, symptom-based approaches within biobanks and population studies therefore offer a promising scalable alternative for advancing the field of anxiety genetics. Moving forwards, the combination of these with deeply phenotyped clinical cohorts will be crucial for translating genetic insights into diagnostic and therapeutic advances. Together, these approaches can elucidate the biological continuum of anxiety, from healthy stress responses to debilitating disorder. Given the high and rising rates of anxiety, especially in young adults, it is more important than ever to improve our ability to identify and understand sources of risk. Despite its public health impact, progress in anxiety genetics lags behind other major mental health conditions. We hope our findings encourage genome-wide investigations leveraging existing but potentially underutilised anxiety severity data in genotyped cohorts, accelerating our progress in understanding the genetic architecture of anxiety.

Participants and measures We meta-analysed data from 14 international cohorts (N = 693,869): 13 PGC-ANX studies with generalized anxiety symptom data and summary statistics from a pre-existing GWAS26. Details of each study and sample descriptives are provided in Supplementary Table 1. We performed a meta-analysis with access to individual participant data60, such that each PGC-ANX cohort performed genome-wide association analyses specifically for this study and shared summary statistics with the core analytical team. The majority of the sample (70%) had completed the GAD-7 or closely related brief self-report measures assessing recent anxiety symptoms. The remaining 30% used other brief self-report anxiety scales (Supplementary Table 1), each available in at least 3,000 individuals. We analysed total sum scores, with higher scores indicating greater severity of symptoms. If participants were missing data on <25% of measure items, the missing scores were imputed with the participant’s mean score of the other items. Participants with ≥25% missing data were excluded from analysis. Several cohorts had assessed anxiety symptoms on two or more occasions. Longitudinal twin studies have shown that symptom stability is primarily driven by genetic factors16,61,62 and that stability extracted from repeated assessments yields higher heritability estimates than single time points16. For cohorts with anxiety assessments from 3 or more time points (12% of the sample), a latent factor was created in R with the package lavaan63, the predict function and a maximum likelihood estimator. For cohorts with 2 time points (44%), a mean score was calculated. All scores, whether single time point, mean or factor score, were standardized to have a mean of zero and a standard deviation of one. Given the high comorbidity of anxiety and other mental health conditions, no additional exclusions were applied beyond those defined by each study. For two cohorts—Genetic Links to Anxiety and Depression Study (GLAD+) and the UK Biobank—individual-level data were merged before the GWAS. Participants from clinical cohorts had been recruited based on a lifetime history of depression or anxiety, as assessed by self-reported diagnostic questionnaires. Meta-analysis Supplementary Table 3 provides details of the studies that contributed to this meta-analysis, which were: Australian Genetics of Depression Study (AGDS)64, Avon Longitudinal Study of Parents and Children (ALSPAC)65,66, CoLaus|PsyCoLaus67, Estonian Biobank68, Generation Scotland69, NIHR Bioresource GLAD+70, Lifelines71, MEGA TRR5872, MVP26, Norwegian Mother, Father and Child Cohort Study73, Providing Tools for Effective Care and Treatment of Anxiety Disorders (PROTECT-AD)74, Twins Early Development Study (TEDS)75, Tracking Adolescents’ Individual Lives’ Survey (TRAILS)76 and UK Biobank77. Each cohort performed genotyping using microarray platforms and imputed genotypes using ancestry-matched panels, primarily the Haplotype Reference Consortium78. Standard quality control procedures were applied, including filters on sample and variant call rates, sex concordance and excessive heterozygosity (full details in Supplementary Table 3). The one set of pre-existing summary statistics was from an analysis in the MVP, obtained through the database of Genotypes and Phenotypes (dbGaP; phs001672). Most groups adopted a mixed linear model approach and retained related individuals in their GWAS. Where applicable, covariates such as ancestry principal components and genotyping batch were included. We did not include age or sex as covariates, as they are not confounders of genetic effects and may represent effect moderators of interest warranting dedicated follow-up investigation, rather than variables to be adjusted for. All resulting summary statistics were on the GRCh37 genome assembly (build 37, hg19). Before meta-analysis, variant-level quality control was performed across the summary statistics, excluding those with minor allele frequency <1% or imputation accuracy score <0.6. Variants present in fewer than half of the contributing cohorts were excluded, resulting in a total of 7,499,431 autosomal SNPs. X-chromosome data were analysed from 7 cohorts (166,852 variants), with male genotypes coded as diploid (0/2) and sex included as a covariate. The meta-analysis was conducted in METAL (v.2020-05-05)79 using an inverse-variance-weighted, standard-error-based approach. The β values from the meta-analysis represent the associated change in standard deviation units of generalized anxiety symptom score per additional copy of the effect allele. All statistical tests conducted in this study were two tailed. Heterogeneity across cohorts was assessed by inspecting the P values from Cochran’s Q test as implemented in METAL. In addition, we calculated the median I2 values (HetISq in METAL) for SNPs reaching genome-wide significance (P < 5 × 10−8) and for independent lead SNPs. We also estimated genetic correlations between contributing cohorts using LDSC33. The inclusion of clinical alongside community-based cohorts offered greater representation across the full range of anxiety symptom severity, increasing statistical power as evidenced in a recent depression GWAS80. However, due to the risk of bias or confounding from differences in study design and phenotyping, we performed subgroup meta-analyses stratified by anxiety measure and excluding clinical cohorts (Supplementary Table 2). Meta-analyses of the measure and ascertainment subgroups were performed in METAL, and genetic correlations between the groups estimated using LDSC33. For both cohort and subgroup genetic correlations, most pairwise comparisons were not sufficiently powered (that is, heritability z-scores <4 for 1 or both sets of summary statistics81) to draw conclusions. To identify LD-independent significant SNPs and loci, clumping was performed in FUMA (v.1.6.5)39. The r2 threshold for independent significant SNPs was 0.1, and was 0.05 for lead SNPs, within a 500-kb window. Genome-wide significance was defined using the conventional threshold (P < 5 × 10−8). To identify novel loci in our results, we cross-referenced significant loci with published trait associations from the GWAS Catalog82 using LDTrait83 (date accessed 2 September 2025), applying an r2 threshold of >0.1 and a 500-kb window. Novelty was strictly defined as having no previous associations with internalizing traits including anxiety, depression, neuroticism and worry. To supplement this, we compared our results with recent anxiety and depression studies23,80 not yet available in the GWAS Catalog. Overlapping significant loci were identified with BEDtools (v.2.31.0)84 and LD assessed using a threshold of r2 > 0.1. The investigation of novel loci also revealed the extent to which our results replicated previous findings. We also determined novel loci specifically for anxiety, whether assessed as symptom severity or a case–control phenotype. Many of the 14 cohorts in our meta-analysis overlap with previous case–control anxiety meta-analyses, with the exception of GLAD+, Lifelines, PROTECT-AD, TEDS and MEGA (approximate N = 110,000). In some instances, the cohort sample composition differs due to the availability of quantitative versus diagnostic information. SNP-based heritability and genetic correlations with external traits We estimated SNP-based heritability via SBayesRC (v.0.2)35. This provided an estimate of the proportion of variance in quantitative anxiety attributable to variation in the common SNPs present in this meta-analysis. We inspected the LDSC32 genomic inflation factor (λGC) and intercept to evaluate the contribution of potential confounding relative to polygenicity. Genetic correlations were also computed using LDSC with 105 GWAS summary statistics covering a broad range of phenotypes and applying a Bonferroni-corrected P value threshold of 4.76 × 10−4. High genetic correlations, such as those often observed between anxiety, depression and neuroticism, do not necessarily indicate identical biology; even when most loci are shared, traits may involve different biological pathways, tissue enrichments or show individual patterns of relationships with other traits. Identifying unique genetic influences on anxiety is important to better understand its specific aetiology and inform potential treatment pathways. However, conditional analyses, especially in the presence of strong genetic correlations, are statistically challenging and require substantial power, and therefore should be interpreted cautiously. We used the mtCOJO34 tool from GCTA (v.1.94.1), which performs conditional analyses between summary statistics to provide marginal effect estimates for the trait of interest. We conditioned our anxiety meta-analysis on depression diagnosis80 (98 index SNPs), neuroticism85 (67 index SNPs) and anxiety diagnosis23 (11 index SNPs). Depression symptoms86 was underpowered (two index SNPs) for the analysis. We also estimated the SNP-based heritability of each set of conditioned summary statistics with LDSC and used block jack-knifing to compare these to the unconditioned heritability estimate. PRSs To evaluate the within- and cross-ancestry validity of our GWAS, we calculated GAD symptom severity PRSs in independent samples from the UK Biobank77 and Prospective Imaging Study of Ageing (PISA)87. We then performed regressions between our PRS and quantitative anxiety, using GAD-7 scores, and case–control anxiety, as defined by a self-reported diagnostic questionnaire or self-report of a diagnosis from a health professional. Specifically, we used SBayesRC35 to calculate PRSs in European, African and South Asian ancestry samples, excluding related individuals. SBayesRC is a Bayesian regression method that uses GWAS summary statistics to estimate SNP effect sizes while accounting for LD and polygenic architecture. It extends the SBayesR framework by incorporating functional annotations or prior biological information, improving the detection of probable causal variants and enhancing predictive accuracy for complex traits. We conducted linear regressions to assess the variance explained in GAD symptom severity by the PRS in each sample (European N = 3,452; African N = 1,581; South Asian N = 1,813). For case–control status, we performed logistic regressions to estimate associations in the target samples (European total n = 3,107, case n = 407; African total n = 1,303, case n = 218; South Asian total n = 1,549, case n = 265). We calculated Nagelkerke’s R2 for our PRS on the liability scale using the observed linear regression R2 and corresponding formula88, assuming a population prevalence of 20%. All regressions included the first ten ancestry-specific principal components and genotyping batch as covariates. The variance explained by the PRS was calculated as the difference in R2 between a full model, including the PRS and covariates, and a null model with only covariates. Positional and functional annotation We used PolyFun (v.1.0.0)36 to estimate per-SNP heritabilities, leveraging a regularized extension of stratified-LDSC applied to the v.2.2UKB baseline-LF model annotations, which captures heritability enrichment related to allele frequency, LD and variant function. These prior causal estimates were then used for fine-mapping in SuSiE (v.0.11.92)37, limiting to a maximum of one causal SNP per locus. We extracted annotations at a PIP threshold of ≥0.95 and created credible causal sets containing the minimum set of ranked variants that cumulatively met this threshold. Unlike standard definitions of credible causal sets in SuSiE, we did not require a minimum pairwise r2 between variants in a set, as the PolyFun + SuSiE pipeline does not incorporate LD estimates. We performed SNP-level gene annotation using FUMA (v.1.6.5)39, integrating three complementary methods: positional mapping (based on physical proximity to genes), eQTL mapping (linking variants to gene expression) and chromatin interaction mapping (using Hi-C data to identify regulatory interactions). eQTL mapping used significant SNP–gene pairs and eQTLs from the brain tissue datasets GTEx v.8 Brain43 (13 regions) and BRAINEAC89 (10 regions), and averaged expressions across these, applying a false discovery rate (FDR) threshold of P < 0.05. Chromatin interaction mapping used Hi-C brain tissue data (dorsolateral prefrontal cortex, hippocampus and left and right ventricles)90 and adult and fetal cortex91, with an FDR threshold of P < 1 × 10−6. These methods differ in their underlying biological rationale and may implicate different genes. Genes identified by two or more mapping approaches were therefore highlighted, as convergence across the methods increased our confidence in the potential functional relevance of a gene. Gene-based associations and enrichment Gene-based association, gene-set and gene–tissue expression enrichment analyses were performed in MAGMA (v.1.08)41 via FUMA (v.1.6.5)39. These analyses aimed to identify genes associated with GAD symptom severity, biological pathways enriched for associated genes and relevant tissues where genes are preferentially expressed, offering insight into the potential biological mechanisms underlying our findings. For gene-based associations, we tested 19,954 genes, applying a Bonferroni-corrected significance threshold of P < 2.5 × 10−6. SNPs were mapped to genes using a 35-kb upstream and 10-kb downstream window. Gene-set analyses were performed using 6,494 curated gene sets (c2.all) and 10,529 Gene Ontology terms (c5.bp, c5.cc and c5.mf) from the Molecular Signatures Database (MSigDB; v.2023.1.Hs)42. Significance was determined by a Bonferroni-corrected threshold of P < 2.9 × 10−6. For tissue enrichment, we tested relationships between trait-associated genes and gene expression in human tissues, using data from BrainSpan (brain samples from 11 general developmental stages and 29 specified ages) and GTEx v.8 (covering 30 general and 54 specific tissue types). Drug targets We examined whether genes associated with GAD symptom severity were associated with individual drugs and drug classes using the DrugTargetor (v.1.3)44 method. DrugTargetor integrates MAGMA gene-level association results with curated drug–gene interaction databases (ChEMBL92,93 and DGIdb94). We used MAGMA (v.1.10) to prioritize associated genes within windows 35 kb upstream and 10 kb downstream. We restricted our analysis to hypothesized drug action within the nervous system, a maximum of 1,551 unique drugs and 163 drug classes. To assess the enrichment of drug classes, we calculated the area under the enrichment curve (AUC), where 50% indicates random enrichment and 100% optimal enrichment, and AUC significance was assessed using one-sided Wilcoxon–Mann–Whitney tests. Ethics This study analysed pre-existing data from cohort studies. Each contributing study obtained ethical approval from the relevant institutional ethics committee, and participants provided informed consent permitting genetic and health-related research. Details of all ethical approvals are provided in Supplementary Table 1, and data application numbers where applicable. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Data availability Summary statistics are available on the PGC download page (https://pgc.unc.edu/for-researchers/download-result). Individual study data can be accessed following review and approval by the individual study cohorts; see https://pgc.unc.edu/for-researchers/individual-level-data-access/ for more information. Summary statistics for the genetic correlations are available following the procedure detailed within the relevant publications (searchable using the PMID in Supplementary Table 9) or via GWAS Catalog (https://www.ebi.ac.uk/gwas/home). Code availability Analytical code is available via GitHub at: https://github.com/megskelton/gad-sympt-metagwas.

GBD 2021 Diseases and Injuries Collaborators. Global incidence, prevalence, years lived with disability (YLDs), disability-adjusted life-years (DALYs), and healthy life expectancy (HALE) for 371 diseases and injuries in 204 countries and territories and 811 subnational locations, 1990-2021: a systematic analysis for the Global Burden of Disease Study 2021. Lancet 403, 2133–2161 (2024). Archer, C., Turner, K., Kessler, D., Mars, B. & Wiles, N. Trends in the recording of anxiety in UK primary care: a multi-method approach. Soc. Psychiatry Psychiatr. Epidemiol. 57, 375–386 (2022). Goodwin, R. D., Weinberger, A. H., Kim, J. H., Wu, M. & Galea, S. Trends in anxiety among adults in the United States, 2008–2018: rapid increases among young adults. J. Psychiatr. Res. 130, 441–446 (2020). GBD 2019 Mental Disorders Collaborators. Global, regional, and national burden of 12 mental disorders in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Psychiatry 9, 137–150 (2022). Rapaport, M. H., Clary, C., Fayyad, R. & Endicott, J. Quality-of-life impairment in depressive and anxiety disorders. Am. J. Psychiatry 162, 1171–1178 (2005). Meier, S. M. & Deckert, J. Genetics of anxiety disorders. Curr. Psychiatry Rep. 21, 16 (2019). Pratt, L. A., Druss, B. G., Manderscheid, R. W. & Walker, E. R. Excess mortality due to depression and anxiety in the United States: results from a nationally representative survey. Gen. Hosp. Psychiatry 39, 39–45 (2016). Wittchen, H. U., Zhao, S., Kessler, R. C. & Eaton, W. W. DSM-III-R generalized anxiety disorder in the National Comorbidity Survey. Arch. Gen. Psychiatry 51, 355–364 (1994). Pavlova, B., Perlis, R. H., Alda, M. & Uher, R. Lifetime prevalence of anxiety disorders in people with bipolar disorder: a systematic review and meta-analysis. Lancet Psychiatry 2, 710–717 (2015). Kessler, R. C., Chiu, W. T., Demler, O., Merikangas, K. R. & Walters, E. E. Prevalence, severity, and comorbidity of 12-month DSM-IV disorders in the National Comorbidity Survey Replication. Arch. Gen. Psychiatry 62, 617–627 (2005). Hollocks, M. J., Lerh, J. W., Magiati, I., Meiser-Stedman, R. & Brugha, T. S. Anxiety and depression in adults with autism spectrum disorder: a systematic review and meta-analysis. Psychol. Med. 49, 559–572 (2019). Achim, A. M. et al. How prevalent are anxiety disorders in schizophrenia? A meta-analysis and critical review on a significant association. Schizophr. Bull. 37, 811–821 (2011). Ask, H. et al. Genetic contributions to anxiety disorders: where we are and where we are heading. Psychol. Med. 51, 2231–2246 (2021). Ozsivadjian, A., Knott, F. & Magiati, I. Parent and child perspectives on the nature of anxiety in children and young people with autism spectrum disorders: a focus group study. Autism 16, 107–121 (2012). Chadwick, P., Kaur, H., Swelam, M., Ross, S. & Ellett, L. Experience of mindfulness in people with bipolar disorder: a qualitative study. Psychother. Res. 21, 277–285 (2011). Cheesman, R. et al. Extracting stability increases the SNP heritability of emotional problems in young people. Transl. Psychiatry 8, 223 (2018). Otowa, T. et al. Meta-analysis of genome-wide association studies of anxiety disorders. Mol. Psychiatry 21, 1391–1399 (2016). Meier, S. M. et al. Genetic variants associated with anxiety and stress-related disorders: a genome-wide association study and mouse-model study. JAMA Psychiatry 76, 924–932 (2019). Purves, K. L. et al. A major role for common genetic variation in anxiety disorders. Mol. Psychiatry 25, 3292–3303 (2020). Li, W. et al. Genome-wide meta-analysis, functional genomics and integrative analyses implicate new risk genes and therapeutic targets for anxiety disorders. Nat. Hum. Behav. 8, 361–379 (2024). Verma, A. et al. Diversity and scale: genetic architecture of 2068 traits in the VA Million Veteran Program. Science 385, eadj1182 (2024). Friligkou, E. et al. Gene discovery and biological insights into anxiety disorders from a large-scale multi-ancestry genome-wide association study. Nat. Genet. 56, 2036–2045 (2024). Strom, N. I. et al. Genome-wide association study of major anxiety disorders in 122,341 European-ancestry cases identifies 58 loci and highlights GABAergic signaling. Nat. Genet. 58, 275–288 (2026). Lee, W. E., Wadsworth, M. E. J. & Hotopf, M. The protective role of trait anxiety: a longitudinal cohort study. Psychol. Med. 36, 345–351 (2006). Altman, D. G. & Royston, P. The cost of dichotomising continuous variables. BMJ 332, 1080 (2006). Levey, D. F. et al. Reproducible genetic risk loci for anxiety: results from ∼200,000 participants in the Million Veteran Program. Am. J. Psychiatry 177, 223–232 (2020). Skelton, M. et al. Genetic overlap between functional impairment and depression and anxiety symptom severity: evidence from the GLAD Study. Psych. Med. 55, e224 (2025). Tesfaye, M. et al. Identification of novel genomic loci for anxiety symptoms and extensive genetic overlap with psychiatric disorders. Psychiatry Clin. Neurosci. 78, 783–791 (2024). Thorp, J. G. et al. Symptom-level modelling unravels the shared genetic architecture of anxiety and depression. Nat. Hum. Behav. 5, 1432–1442 (2021). Visscher, P. M. et al. Statistical power to detect genetic (co)variance of complex traits using SNP data in unrelated samples. PLoS Genet. 10, e1004269 (2014). Evans, L. M. & Keller, M. C. Using partitioned heritability methods to explore genetic architecture. Nat. Rev. Genet. 19, 185 (2018). Bulik-Sullivan, B. et al. LD score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat. Genet. 47, 291–295 (2015). Bulik-Sullivan, B. et al. An atlas of genetic correlations across human diseases and traits. Nat. Genet. 47, 1236–1241 (2015). Zhu, Z. et al. Causal associations between risk factors and common diseases inferred from GWAS summary data. Nat. Commun. 9, 224 (2018). Zheng, Z. et al. Leveraging functional genomic annotations and genome coverage to improve polygenic prediction of complex traits within and between ancestries. Nat. Genet. 56, 767–777 (2024). Weissbrod, O. et al. Functionally informed fine-mapping and polygenic localization of complex trait heritability. Nat. Genet. 52, 1355–1363 (2020). Wang, G., Sarkar, A., Carbonetto, P. & Stephens, M. A simple new approach to variable selection in regression, with application to genetic fine mapping. J. R. Stat. Soc. Ser. B 82, 1273–1300 (2020). Amendola, L. M. et al. Actionable exomic incidental findings in 6503 participants: challenges of variant classification. Genome Res. 25, 305–315 (2015). Watanabe, K., Taskesen, E., van Bochoven, A. & Posthuma, D. Functional mapping and annotation of genetic associations with FUMA. Nat. Commun. 8, 1826 (2017). Li, Y. et al. Cross-ancestry genome-wide association study and systems-level integrative analyses implicate new risk genes and therapeutic targets for depression. Nat. Hum. Behav. 9, 806–823 (2025). de Leeuw, C. A., Mooij, J. M., Heskes, T. & Posthuma, D. MAGMA: generalized gene-set analysis of GWAS data. PLoS Comput. Biol. 11, e1004219 (2015). Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl Acad. Sci. USA 102, 15545–15550 (2005). GTEx Consortium. The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science 369, 1318–1330 (2020). Gaspar, H. A., Hübel, C. & Breen, G. Drug Targetor: a web interface to investigate the human druggome for over 500 phenotypes. Bioinformatics 35, 2515–2517 (2019). Ren, L. et al. Anxiety disorders: treatments, models, and circuitry mechanisms. Eur. J. Pharmacol. 983, 176994 (2024). Carey, C. E. et al. Principled distillation of UK Biobank phenotype data reveals underlying structure in human variation. Nat. Hum. Behav. 8, 1599–1615 (2024). Sullivan, P. F. et al. Genome-wide association for major depressive disorder: a possible role for the presynaptic protein piccolo. Mol. Psychiatry 14, 359–375 (2009). Christiansen, G. B. et al. The sorting receptor SorCS3 is a stronger regulator of glutamate receptor functions compared to GABAergic mechanisms in the hippocampus. Hippocampus 27, 235–248 (2017). Breiderhoff, T. et al. Sortilin-related receptor SORCS3 is a postsynaptic modulator of synaptic depression and fear extinction. PLoS ONE 8, e75006 (2013). Yang, Z. et al. Investigating shared genetic basis across Tourette syndrome and comorbid neurodevelopmental disorders along the impulsivity–compulsivity spectrum. Biol. Psychiatry 90, 317–327 (2021). Sun, F. et al. Hippocampal gray matter volume alterations in patients with first-episode and recurrent major depressive disorder and their associations with gene profiles. BMC Psychiatry 25, 134 (2025). Yang, J., Wray, N. R. & Visscher, P. M. Comparing apples and oranges: equating the power of case–control and quantitative trait association studies. Genet. Epidemiol. 34, 254–257 (2010). Direk, N. et al. An analysis of two genome-wide association meta-analyses identifies a new locus for broad depression phenotype. Biol. Psychiatry 82, 322–329 (2017). Cai, N. et al. Minimal phenotyping yields genome-wide association signals of low specificity for major depression. Nat. Genet. 52, 437–447 (2020). Huang, L. et al. Polygenic analyses show important differences between major depressive disorder symptoms measured using various instruments. Biol. Psychiatry 95, 1110–1121 (2024). Strom, N. I. et al. Genome-wide association study of obsessive–compulsive symptoms including 33,943 individuals from the general population. Mol. Psychiatry 29, 2714–2723 (2024). American Psychiatric Association (ed.) Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision (DSM-5-TR) (American Psychiatric Association Publishing, 2022); https://doi.org/10.1176/appi.books.9780890425787 Morneau-Vaillancourt, G. et al. The genetic and environmental hierarchical structure of anxiety and depression in the UK Biobank. Depress. Anxiety 37, 512–520 (2020). ter Kuile, A. R. et al. Genome-wide genetic overlap between fear-based disorders and generalised anxiety disorder. Preprint at medRxiv https://doi.org/10.64898/2026.02.06.26345742 (2026). Zugman, A. et al. Mega-analysis methods in ENIGMA: the experience of the generalized anxiety disorder working group. Hum. Brain Mapp. 43, 255–277 (2022). Funk, J., Morneau-Vaillancourt, G., Palaiologou, E. & Eley, T. Heritability of generalised anxiety stability—a longitudinal twin study among young adults. Psychol. Med. 5, e180 (2025). Gillespie, N. A. et al. Do the genetic or environmental determinants of anxiety and depression change with age? A longitudinal study of Australian twins. Twin Res. 7, 39–53 (2004). Rosseel, Y. lavaan: an R package for structural equation modeling. J. Stat. Softw. 48, 1–36 (2012). Mitchell, B. L. et al. The Australian Genetics of Depression Study: new risk loci and dissecting heterogeneity between subtypes. Biol. Psychiatry 92, 227–235 (2022). Boyd, A. et al. Cohort profile: the ‘children of the 90s’—the index offspring of the Avon Longitudinal Study of Parents and Children. Int. J. Epidemiol. 42, 111–127 (2013). Fraser, A. et al. Cohort profile: the Avon Longitudinal Study of Parents and Children: ALSPAC mothers cohort. Int. J. Epidemiol. 42, 97–110 (2013). Preisig, M. et al. The PsyCoLaus study: methodology and characteristics of the sample of a population-based survey on psychiatric disorders and their association with genetic and cardiovascular risk factors. BMC Psychiatry 9, 9 (2009). Milani, L. et al. The Estonian Biobank’s journey from biobanking to personalized medicine. Nat. Commun. 16, 3270 (2025). Smith, B. H. et al. Cohort profile: Generation Scotland: Scottish Family Health Study (GS:SFHS). The study, its participants and their potential for genetic research on health and illness. Int. J. Epidemiol. 42, 689–700 (2013). Davies, M. R. et al. The Genetic Links to Anxiety and Depression (GLAD) Study: online recruitment into the largest recontactable study of depression and anxiety. Behav. Res. Ther. 123, 103503 (2019). Sijtsma, A. et al. Cohort profile update: lifelines, a three-generation cohort study and biobank. Int. J. Epidemiol. 51, e295 (2022). Deckert, J. et al. GLRB allelic variation associated with agoraphobic cognitions, increased startle response and fear network activation: a potential neurogenetic pathway to panic disorder. Mol. Psychiatry 22, 1431–1439 (2017). Magnus, P. et al. Cohort profile update: the Norwegian Mother and Child Cohort Study (MoBa). Int. J. Epidemiol. 45, 382–388 (2016). Heinig, I. et al. Optimizing exposure-based CBT for anxiety disorders via enhanced extinction: design and methods of a multicentre randomized clinical trial. Int. J. Methods Psychiatr. Res. 26, e1560 (2017). Lockhart, C. et al. Twins Early Development Study (TEDS): a genetically sensitive investigation of mental health outcomes in the mid-twenties. JCPP Adv. 3, e12154 (2023). Oldehinkel, A. J. et al. Cohort profile update: the TRacking Adolescents’ Individual Lives Survey (TRAILS). Int. J. Epidemiol. 44, 76–76n (2015). Bycroft, C. et al. The UK Biobank resource with deep phenotyping and genomic data. Nature 562, 203–209 (2018). McCarthy, S. et al. A reference panel of 64,976 haplotypes for genotype imputation. Nat. Genet. 48, 1279–1283 (2016). Willer, C. J., Li, Y. & Abecasis, G. R. METAL: fast and efficient meta-analysis of genomewide association scans. Bioinformatics 26, 2190–2191 (2010). Adams, M. J. et al. Trans-ancestry genome-wide study of depression identifies 697 associations implicating cell types and pharmacotherapies. Cell 188, 640–652 (2025). Zheng, J. et al. LD Hub: a centralized database and web interface to perform LD score regression that maximizes the potential of summary level GWAS data for SNP heritability and genetic correlation analysis. Bioinformatics 33, 272–279 (2017). Cerezo, M. et al. The NHGRI-EBI GWAS Catalog: standards for reusability, sustainability and diversity. Nucleic Acids Res. 53, D998–D1005 (2025). Lin, S.-H., Brown, D. W. & Machiela, M. J. LDtrait: an online tool for identifying published phenotype associations in linkage disequilibrium. Cancer Res. 80, 3443–3446 (2020). Quinlan, A. R. & Hall, I. M. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics 26, 841–842 (2010). Gupta, P. et al. A genome-wide investigation into the underlying genetic architecture of personality traits and overlap with psychopathology. Nat. Hum. Behav. 8, 2235–2249 (2024). Gilchrist, L. et al. Depression symptom-specific genetic associations in clinically diagnosed and proxy case Alzheimer’s disease. Nat. Ment. Health 3, 212–228 (2025). Lupton, M. K. et al. A prospective cohort study of prodromal Alzheimer’s disease: Prospective Imaging Study of Ageing: Genes, Brain and Behaviour (PISA). NeuroImage Clin. 29, 102527 (2021). Lee, S. H., Goddard, M. E., Wray, N. R. & Visscher, P. M. A better coefficient of determination for genetic profile analysis. Genet. Epidemiol. 36, 214–224 (2012). Ramasamy, A. et al. Genetic variability in the regulation of gene expression in ten regions of the human brain. Nat. Neurosci. 17, 1418–1428 (2014). Schmitt, A. D. et al. A compendium of chromatin contact maps reveals spatially active regions in the human genome. Cell Rep. 17, 2042–2059 (2016). Giusti-Rodriguez, P. M. D. & Sullivan, P. F. Using three-dimensional regulatory chromatin interactions from adult and fetal cortex to interpret genetic results for psychiatric disorders and cognitive traits. Preprint at bioRxiv https://doi.org/10.1101/406330 (2018). Gaulton, A. et al. ChEMBL: a large-scale bioactivity database for drug discovery. Nucleic Acids Res. 40, D1100–D1107 (2012). Zdrazil, B. et al. The ChEMBL Database in 2023: a drug discovery platform spanning multiple bioactivity data types and time periods. Nucleic Acids Res. 52, D1180–D1192 (2024). Freshour, S. L. et al. Integration of the Drug-Gene Interaction Database (DGIdb 4.0) with open crowdsource efforts. Nucleic Acids Res. 49, D1144–D1151 (2021). Gaziano, J. M. et al. Million Veteran Program: a mega-biobank to study genetic influences on health and disease. J. Clin. Epidemiol. 70, 214–223 (2016).

AGDS: We are indebted to the participants for giving their time to contribute to this study. We thank all the people who helped in the conception, implementation, beta testing, media campaign and data cleaning. The AGDS was funded by grant 108663 (to N.G.M.) from the Australian National Health and Medical Research Council (NHMRC). This work was supported by NHMRC Investigator Grants 2017176 to B.L.M., 1173790 to N.R.W., 1172990 to N.G.M., 1172917 and 2025674 to S.E.M. and 2016346 to I.B.H. E.M.B. received funding from the PRE-EMPT NHMRC Centre for Research Excellence (1198304 to E.M.B.) and the University of Queensland Health Research Accelerator Program. ALSPAC: We are extremely grateful to all the families who took part in this study, the midwives for their help recruiting them and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses. The UK Medical Research Council (MRC) and Wellcome (grant reference 217065/Z/19/Z) and the University of Bristol provide core support for ALSPAC. This publication is the work of the authors and A.S.F.K. will serve as guarantor for the ALSPAC contents of this article. Genome-wide genotyping data was generated by Sample Logistics and Genotyping Facilities at Wellcome Sanger Institute and LabCorp (Laboratory Corporation of America) using support from 23andMe. Ethical approval for the study was obtained from the ALSPAC Ethics and Law Committee and the Local Research Ethics Committees. CoLaus|PsyCoLaus: The CoLaus|PsyCoLaus study was supported by unrestricted research grants from GlaxoSmithKline, the Faculty of Biology and Medicine of Lausanne, the Swiss National Science Foundation (grants 3200B0–105993, 3200B0-118308, 33CSCO-122661, 33CS30-139468, 33CS30-148401 and 33CS30_177535 to M.P., and grants 3247730_204523 and 320030_220190) and the Swiss Personalized Health Network (grant 2018DRI01). Estonian Biobank: We acknowledge the Estonian Biobank participants and the Estonian Biobank Research Team (A. Metspalu, L. Milani, T. Esko, R. Mägi, M. Metspalu, M. Nelis and G. Hudjashov). Data analysis was carried out in part in the High-Performance Computing Center of the University of Tartu. This research was supported by the Estonian Research Council (grant PSG615 to K.L.) and the Estonian Centre of Excellence for Well-Being Sciences, funded by grant TK218 (to K.L.) from the Estonian Ministry of Education and Research. The research was conducted using the Estonian Center of Genomics/Roadmap II funded by the Estonian Research Council (project number TT17). Analyses were performed using data according to release application 6-7/GI/16880. Generation Scotland (GS): We thank the participants and staff of GS. The work presented is supported by Wellcome Trust (226770/Z/22/Z, 220857/Z/20/Z and 216767/Z/19/Z to A.M.M.) and UK Research and Innovation (MR/Z50354X/1 and MR/Z000548/1 to A.M.M.). Genotyping of GS samples was funded by the MRC and Wellcome Trust (104036/Z/14/Z). GS also received support from the Chief Scientist Office of the Scottish Government Health Directorates (CZD/16/6) and the Scottish Funding Council (HR03006). GLAD+ (including NIHR BioResource): We thank NIHR BioResource volunteers for their participation, and gratefully acknowledge NIHR BioResource centres, NHS Trusts and staff for their contribution. We thank the National Institute for Health and Care Research (NIHR), NHS Blood and Transplant and Health Data Research UK as part of the Digital Innovation Hub Programme. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care. We gratefully acknowledge the participation of the NIHR BioResource Centre Maudsley (grant reference NIHR203318 to G.B.), Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King’s College London volunteers, and we thank the staff for their help with volunteer recruitment. We thank the NIHR Biomedical Research Centre at South London and the Maudsley NHS Foundation Trust and King’s College London for funding. This study represents independent research supported by the NIHR Biomedical Research Centre BioResource at South London and Maudsley NHS Foundation Trust and King’s College London. We gratefully acknowledge capital equipment funding from the Maudsley Charity (grant reference 980 to G.B.) and Guy’s and St Thomas’s Charity (grant reference STR130505 to G.B.). Lifelines Cohort Study: The Lifelines Biobank initiative has been made possible by funding from the Dutch Ministry of Health, Welfare and Sport, the Dutch Ministry of Economic Affairs, the University Medical Center Groningen (UMCG the Netherlands), University of Groningen and the Northern Provinces of the Netherlands. The generation and management of GWAS genotype data for the Lifelines Cohort Study is supported by the UMCG Genetics Lifelines Initiative (UGLI). UGLI is partly supported by a Spinoza Grant from NWO, awarded to C. Wijmenga. We acknowledge the services of the Lifelines Cohort Study, the contributing research centres delivering data to Lifelines and all the study participants. MEGA TRR58: This project was funded by the German Research Foundation (DFG)—project number 44541416 TRR 58 ‘Fear, Anxiety, Anxiety Disorders’, project Z02/1-3 to J.D. and K.D. and project number 499262975 to A.E.-L. and Andre Pittig. This work was supported by the DFG (grants FOR2107 DA1151/5-1, DA1151/5-2, DA1151/9-1, DA1151/10-1 and DA1151/11-1 to U.D.; SFB/TRR 393, project grant number 521379614) and the Interdisciplinary Center for Clinical Research (IZKF) of the medical faculty of Münster (grant Dan 3/022/22 to U.D.). We thank M. Kuhn for help with data acquisition. MoBa: We are grateful to all the participating families in Norway who take part in this on-going cohort study. For generating high-quality genomic data, we thank the Norwegian Institute of Public Health, the HARVEST collaboration, the NORMENT Centre at the University of Oslo, the Center for Diabetes Research at the University of Bergen, deCODE Genetics, the Research Council of Norway, the SouthEastern and Western Norway Regional Health Authorities, the ERC AdG, Stiftelsen KG Jebsen, the Trond Mohn Foundation and the Novo Nordisk Foundation. The MoBa analysis was performed on the Tjeneste for Sensitive Data (TSD) facilities, owned by the University of Oslo, operated and developed by the TSD service group at the University of Oslo, IT-Department (USIT), using resources provided by Sigma2—the National Infrastructure for High Performance Computing and Data Storage in Norway (UNINETT). H.A. and J.H.P. were funded by the Research Council of Norway (RCN numbers 324620 and 336085) and NordForsk (156298 and 230738). A.H. and L.J.H. were funded by the South-Eastern Norway Regional Health Authority (2020022, 2022083 and 2026069) and the Research Council of Norway (336085). E.C.C. was supported by funding from the Research Council of Norway (274611) and the South-Eastern Norway Regional Health Authority (2021045). MVP: We thank MVP staff, researchers and volunteers who have contributed to MVP, and especially participants who previously served their country in the military and now generously agreed to enrol in the study (see https://www.research.va.gov/mvp/ for more details). The citation details for MVP can be found in ref. 95. This research is based on data from the MVP, Office of Research and Development, Veterans Health Administration, and was supported by the Veterans Administration (VA) MVP award number 000. PISA: PISA was funded by the NHMRC of Australia (grant identifier APP1095227 to N.G.M. and M.K.L.). PROTECT-AD: One of nine research consortia in the German federal research programme ‘Research Network on Mental Disorders’, funded by the Federal Ministry of Education and Research (BMBF; http://www.fzpe.de), project P5 (grant number 01EE1402F to J.D. and K.D.). A complete list of project publications can be found at www.fzpe.de. Recruitment of PROTECT-AD was funded by the BMBF (project P1, 01EE1402A to J.H.). The presented work was derived from projects P1 and P5. Principal investigators and individuals responsible for recruitment are: U. Wittchen, A. Pittig, I. Heinig, J. Hoyer (Dresden), A. Ströhle and J. Fydrich (Berlin), A. Hamm and J. Richter (Greifswald), V. Arolt, U. Dannlowski and K. Kölkebeck (Münster), S. Schneider and J. Margraf (Bochum), T. Kircher, B. Straube and W. Rief (Marburg), J. Deckert, K. Domschke, U. Lueken and P. Pauli (Würzburg) and P. Neudeck (Cologne). We thank the following individuals for their help: J. Dehler, D. Westphal, K. Hummel, J. Hoyer (Dresden), V. Pflug, D. Adolph, C. Mohr, J. Cwik (Bochum), M. Hollandt, A. Pietzner, J. Neubert (Greifswald), C. Konrad, Y. Yang, I. Ridderbusch, A. Wroblewski, H. Christiansen, A. Maenz, S. Tennie, J. Thierschmidt (Marburg), M. Romanos, K. Zierhut, K. Dickhöver, M. Winkler, M. Stefanescu, C. Ziegler, H. Weber (Würzburg), N. Weber, S. Schauenberg, S. Wriedt, C. Heitmann (Münster), C. im Brahm, A. Evers (Cologne), I. Alt, S. Bischoff, J. Mumm, J. Plag, A. Schreiner and S. Meska (Berlin). X. Grählert and M. Käppler of the Coordinating Centre for Clinical Trials (KKS) data centre (Dresden) provided support with the electronic data assessment and data banking. E. Stolzenburg, S. Bologov and K. Bley provided administrative support. TEDS: We gratefully acknowledge the ongoing contribution of the TEDS participants and their families. TEDS is funded by a UK MRC programme grant (MR/V012878/1) to T.C.E. (previously MR/M021475/1 to R. Plomin). TRAILS: This research is part of the TRacking Adolescents’ Individual Lives Survey (TRAILS). Participating centres of TRAILS include various departments of the University Medical Center and University of Groningen, the University of Utrecht and the Parnassia Psychiatric Institute, all in the Netherlands. TRAILS has been financially supported by grants from the Netherlands Organization for Scientific Research NWO (Medical Research Council programme grant GB-MW 940-38-011; ZonMW Brainpower grant 100-001-004; ZonMw Risk Behavior and Dependence grant 60-60600-97-118; ZonMw Culture and Health grant 261-98-710; Social Sciences Council medium-sized investment grants GB-MaGW 480-01-006 and GB-MaGW 480-07-001 (A.J.O.); Social Sciences Council project grants GB-MaGW 452-04-314 (A.J.O.) and GB-MaGW 452-06-004; ZonMw Longitudinal Cohort Research on Early Detection and Treatment in Mental Health Care grant 636340002; NWO large-sized investment grant 175.010.2003.005; NWO Longitudinal Survey and Panel Funding 481-08-013 and 481-11-001; NWO Vici 016.130.002 (A.J.O.), 453-16-007/2735 and Vi.C 191.021 and NWO Gravitation 024.001.003 (A.J.O.)), the Dutch Ministry of Justice (WODC), the European Science Foundation (EuroSTRESS project FP-006), the European Research Council (ERC-2017-STG-757364 and ERC-CoG-2015-681466), Biobanking and Biomolecular Resources Research Infrastructure BBMRI-NL (CP 32) (C.A.H.), the Gratema Foundation, the Jan Dekker foundation, the participating universities, and Accare Centre for Child and Adolescent Psychiatry. Statistical analyses were carried out on the Genetic Cluster Computer (http://www.geneticcluster.org), which is financially supported by the Netherlands Scientific Organization (NWO 480-05-003) along with a supplement from the Dutch Brain Foundation. We are grateful to everyone who participated in this research or worked on this project to make it possible. UK Biobank: This research has been conducted using the UK Biobank Resource under application number 82087. This work uses data provided by patients and collected by the NHS as part of their care and support. 23andme: This article uses summary statistics from pre-existing analyses with 23andme participants to perform genetic correlations with our phenotype. We thank the research participants and employees of 23andMe for making this work possible. Additional: We acknowledge the use of the King’s College London research computing facility, CREATE (https://doi.org/10.18742/rnvf-m076). N.S. received funding from a UKRI Future Leaders Fellowship (grant number MR/T04327X/1) and the UK Dementia Research Institute award number UK DRI-5008 through UK DRI Ltd. N.G.M. acknowledges funding from the NHMRC. O.A.A. received EU, RCN, NIH, NordForsk and SouthEast Health Authority funding. A.S.F.K. is supported by a Wellcome Early Career Award (grant reference 227063/Z/23/Z). E.C.C. is a member of the MRC Integrative Epidemiology Unit at the University of Bristol, which is supported by the MRC and the University of Bristol (MC_UU_00032/1). L.J.H. was supported by funding from the Norwegian South-East Regional Health Authority (grant numbers 2021045, 2019097 and 2022083). A.H. acknowledges the South-Eastern Norway Regional Health Authority (number 2020022) and the Research Council of Norway (numbers 274611 and 336085). J.H. acknowledges NIMH 5R01MH124847. S.E.M. acknowledges funding from NHMRC APP1172917 and APP2025674. B.L.M. was supported by an NHMRC Investigator Grant (APP2017176). A.R.t.K. was supported by the ERC under the European Union’s Horizon 2020 research and innovation programme (grant agreement number 863981). G.M.-V. is supported by a Banting Postdoctoral Fellowship from the Social Sciences and Humanities Research Council of Canada (grant reference 202309BPF-510174-293475). This phase of the Psychiatric Genomics Consortium is part-funded by the US National Institutes of Health (MH124873). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. Author information Authors and Affiliations Consortia Contributions T.C.E., J.R.I.C. and G.B. supervised the study. A core team of T.C.E., J.R.I.C., G.B., M.S., B.L.M., E.A., D.L., A.R.t.K., R.W. and C.A.H. designed and directed the study. T.C.E., J.R.I.C., G.B., M.S., B.L.M, E.A., D.L., G.M.-V., A.R.t.K. and R.W. drafted the article and implemented revisions. M.S., B.L.M, E.A.-L., D.L., G.M.-V., A.E.M., A.R.t.K. and R.W. conducted data analysis and data visualization. M.S., B.L.M., D.L., R.W., M.J.A., E.M.B., E.C.C., P.Z.G., L.J.H., J.H., K.K., A.S.F.K., S.P., J.H.P., G.P., E.C., P.J.v.D.M., O.A.A., A.E.-L., A.H., N.S., B.V., H.W., C.A., H.A., U.D., J.D., K.D., I.B.H., K.L., T.B.L., U.L., M.K.L., S.E.M., A.M.M., A.J.O., M.P., A.R., H.S., J.T.R.W., C.A.H., N.G.M., G.B., J.R.I.C. and T.C.E. provided samples (concept and/or design for an individual study, acquired data for individual study and/or analysis and/or interpretation of results from individual study). All authors discussed the results, provided feedback on the draft of the article and approved the final version. Corresponding author Ethics declarations Competing interests G.B. has received honoraria, research or conference grants and consulting fees from Illumina, Otsuka and COMPASS Pathfinder Ltd. O.A.A. served as a consultant for Precision-Health.ai and Cortechs.ai, and has received speaker’s honorarium from BMS, Lundbeck, Otsuka, Sunovion, Janssen and Lilly. K.D. is a member of the Neurotorium Editorial Board of the Lundbeck Foundation and has received speaker’s honoraria by Janssen Cilag GmbH, not related to the subject of this paper. I.B.H is a professor of psychiatry and the codirector of Health and Policy, Brain and Mind Centre, University of Sydney. He has led major public health and health service development in Australia, particularly focusing on early intervention for young people with depression, suicidal thoughts and behaviours and complex mood disorders. He is active in the development through codesign, implementation and continuous evaluation of new health information and personal monitoring technologies to drive highly personalized and measurement-based care. He holds a 3.2% equity share in Innowell Pty Ltd that is focused on digital transformation of mental health services. A.M.M. has received research support from Eli Lilly, Janssen and The Sackler Trust. A.M.M. has also received speaker fees from Illumina and Janssen. J.D. is a member of the board of the German Society of Biological Psychiatry and is on the scientific advisory boards of non-profit organizations and foundations. The other authors declare no competing interests. Peer review Peer review information Nature Human Behaviour thanks Alexander Neumann, Hon-Cheong So and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available. Additional information Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Supplementary information Supplementary Information (download PDF ) Supplementary Figs. 1–5. Supplementary Tables (download XLSX ) Supplementary Tables 1–18. Rights and permissions Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. About this article Cite this article Skelton, M., Mitchell, B.L., Assary, E. et al. Genome-wide meta-analysis of quantitatively measured generalized anxiety symptoms in individuals of European ancestry. Nat Hum Behav (2026). https://doi.org/10.1038/s41562-026-02476-7 Received: Accepted: Published: Version of record: DOI: https://doi.org/10.1038/s41562-026-02476-7