Assembling the evidence…
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.
We identified 80 independent genome-wide significant variants within 74 loci, 39 of which were newly associated with anxiety.
Well-powered meta-analysis of 693,869 individuals with standardized QC; novelty defined strictly against catalogued internalizing-trait associations
SNP-based heritability was 5.9% (posterior s.d. = 0.15%).
SBayesRC estimate with tight posterior SD, consistent with a prior quantitative anxiety estimate and discussed against liability-scale values
Polygenic scores were significantly associated with anxiety symptom severity and disorder in European, African and South Asian ancestry samples (R2 = 1.2-2.9%).
Target samples of 1,303-3,452 drawn partly from a discovery cohort with overlap handling unstated; variance explained is small
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).
Bonferroni-corrected LD score regression across 105 traits; correlations are associative and need not reflect horizontal pleiotropy
These findings highlight the discovery power gained from analysing a quantitative trait rather than a case-control phenotype in anxiety genetics.
Locus-count comparison against a largely overlapping, differently composed case-control study is not a controlled contrast
Derived from the full evaluation — not a separate score.
Strengths
Meta-analysis of 693,869 individuals across 14 cohorts in 8 countries yields 80 independent genome-wide significant variants within 74 loci. Median I² for the LD-independent lead SNPs is 0%, and cross-cohort genetic correlations range from 0.64 to 0.97, indicating consistent effect estimates despite differing instruments.
↳ Results, Genome-wide association meta-analysis; Supplementary Tables 4 and 7
mtCOJO conditioning on neuroticism, case-control anxiety and depression is reported in full, including that only two loci survive conditioning on neuroticism, alongside standard-error inflation of 7-9% and sign changes in 17-20% of variants as evidence of instability. The Discussion draws the implication that many loci may index broader neuroticism-related liability.
↳ Results, Conditional analyses; Discussion, paragraph 4
The paper reports that all 58 loci from the prior case-control anxiety GWAS show the same direction of effect here, with 19 reaching genome-wide significance, and names the five non-overlapping cohorts and approximate added sample. This makes the increment over prior work auditable rather than asserted.
↳ Results, Genome-wide association meta-analysis; Methods, Meta-analysis
The European-ancestry restriction, the GAD-7's incomplete DSM-5 symptom coverage, measurement noise from recent-symptom scales, and the underpowered subgroup comparisons are each connected to a specific consequence for what can be concluded. The paper states it cannot be certain all cohorts captured the same underlying genetic architecture.
↳ Discussion, paragraphs 3-5
Limitations
The claim that quantitative phenotyping delivers superior discovery rests on comparing 74 loci here with 58 loci from a larger case-control study, but nine of the 14 cohorts overlap that study and sample composition differs by the availability of symptom versus diagnostic data. The comparison is suggestive rather than an identified contrast.
↳ Abstract, final sentence; Discussion, final paragraph; Methods, Meta-analysis
Cohorts used six different symptom measures and mixed community and clinical ascertainment, but the paper reports that most pairwise cohort and subgroup genetic-correlation comparisons had heritability z-scores below 4. Equivalence of genetic architecture across instruments is therefore not established, only not contradicted.
↳ Results, Genome-wide association meta-analysis; Methods, Meta-analysis
Target samples number 1,303 to 3,452 individuals with 218 to 407 cases, drawn partly from UK Biobank, which also contributes to discovery; how overlap was excluded is not described. The liability-scale conversion assumes 20% population prevalence against observed case fractions nearer 13-17%.
↳ Methods, PRSs; Results, Polygenic risk scores
All 14 discovery cohorts are European-ancestry, and cross-ancestry polygenic scores explain only 1.2-1.4% of variance in African and South Asian samples versus 2.9% in European. Population-specific risk loci cannot be identified from this design.
↳ Abstract; Results, Polygenic risk scores; Discussion, paragraph 5
The core discovery result is secure: 693,869 participants, standardized cohort quality control, inverse-variance meta-analysis, and a median I² of 0% across the LD-independent lead SNPs, with cross-cohort genetic correlations of 0.64 to 0.97 supporting consistency despite six different symptom instruments. Assessment turned on how much weight the conditional analyses should carry: conditioning on neuroticism leaves only two genome-wide significant loci, and although the paper reports the standard-error inflation and sign changes that make those results unstable, the finding bounds how far the loci can be read as anxiety-specific. The abstract's closing claim about the discovery power of quantitative phenotyping rests on a 74-versus-58 locus comparison with a study sharing nine of the 14 cohorts, which the Discussion hedges but the abstract does not. Polygenic-score validation is genuine but thin — target samples of 1,303 to 3,452, drawn partly from a discovery cohort, with liability-scale conversion assuming 20% prevalence against observed case fractions nearer 13-17%. The result is a competently executed, substantial discovery paper whose headline framing runs slightly ahead of what its own conditional analyses can carry.
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Strong4.0
Confidence mediumThe meta-analysis of 693,869 individuals yields 80 independent variants across 74 loci with 39 new for anxiety, a clearly delineated extension of the prior case-control discovery base. The score sits mid-band because nine of the 14 cohorts overlap the comparator study and conditioning on neuroticism leaves only two genome-wide significant loci, so the phenotype-specific component of the advance is not established.
Conditioning on neuroticism ... resulted in 9 remaining genome-wide significant SNPs across 2 loci
The pipeline is comprehensive: standardized cohort-level QC, MAF and imputation filters, presence in at least half the cohorts, inverse-variance METAL meta-analysis, Cochran's Q and I², cross-cohort LDSC, an LDSC intercept check, subgroup sensitivity analyses, mtCOJO conditioning with explicit instability caveats, and independent cross-ancestry PRS testing. The score stops mid-to-upper band because the paper states most subgroup and cohort correlation comparisons were underpowered, fine-mapping assumes one causal variant per locus, and PRS target samples are small with discovery-target overlap handling unstated.
most pairwise comparisons were not sufficiently powered
The exposition moves cleanly from discovery through heritability, genetic correlations, conditioning, polygenic scores and annotation, with effect estimates, intervals and corrected thresholds attached to quantitative statements. The abstract does not carry the post-conditioning locus count, though the Discussion states plainly that many loci may index broader neuroticism-related liability.
Many genome-wide significant loci may therefore index a broader neuroticism-related liability
Prior case-control GWAS, the two-item MVP analysis, UK Biobank symptom-grouping work and genomic structural equation modelling results are all engaged and the findings positioned against them. Limitations — European-only discovery, GAD-7's incomplete DSM-5 coverage, instrument and ascertainment heterogeneity — are each carried into an interpretive consequence rather than listed.
This limits the generalizability of our findings across anxiety disorders
Caveats4 of 4 checks
Sample size (693,869), cohort count (14), locus counts, heritability, P values and PRS R² values are reported consistently across Abstract, Results and Methods, and methods map onto reported results. Two minor imprecisions — a gene-based sentence referring to '80 independent loci' where the primary count is 74 loci, and a liability-scale conversion assuming 20% prevalence against observed case fractions near 13-17% — are caveats on wording and on secondary effect-size magnitudes rather than coherence defects.
Ethical approval and informed consent are declared for every contributing cohort, with approval details deferred to supplementary tables. Data access follows cohort-specific policies with MVP summary statistics available via dbGaP; no code-availability or preregistration statement is present, which is standard for consortium GWAS meta-analyses.
Flags: 2 declared / 5 total
95 references in manuscript 89 of 95 checkable references found in an index
No retraction notice found in Retraction Watch.
Sources: Retraction Watch ✓
Where this paper’s evidence sits on the path from initial observation to real-world use.
This is discovery-stage genetic association work: polygenic scores explain 1.2-2.9% of variance and no significant enrichment was found for any of the 1,551 individual drug targets tested. Deployment readiness is distant.
We did not observe any significant enrichment of anxiety associations for the 1,551 drug targets tested
AI-generated, human-governed. Something look off? Contact us to request a review.