Neurology · Medical Literature
Consensus meta-analysis of genome-wide association studies for Alzheimer's disease and related dementias.
Abstract
To better characterize the genetic architecture underlying Alzheimer's disease (AD) and related dementias (ADRD), we performed a meta-analysis of European-ancestry genome-wide association studies in 128,681 cases or proxy cases of ADRD and 849,833 (proxy) controls. We identified 91 genetic loci associated with ADRD risk, of which 16 are new and 56 are specifically detected in clinically diagnosed AD cases. We also provide a list of 18 loci (15 new) requiring further external validation. A polygenic score combining the effects of ADRD loci other than APOE was primarily associated with AD rather than non-AD pathology. Individuals in the tenth decile of the score exhibited a twofold increased risk of presenting with Braak neurofibrillary tangles stage of >4 and moderate-to-severe neuritic amyloid plaque pathology at death compared to individuals in the median score group. In conclusion, our study validated a large number of loci associated with the risk of clinically diagnosed AD, while further investigations are required to confirm the impact of the other loci on AD clinical diagnosis and of each locus on AD pathology.
Clinical relevance
Does a large-scale consensus meta-analysis of European-ancestry genome-wide association studies identify novel genetic loci associated with Alzheimer's disease and related dementias, and can a polygenic risk score derived from these loci stratify individuals by underlying AD neuropathological burden? __SNAPSHOT__ {"studyDesign":"Systematic meta-analysis of genome-wide association studies (GWAS) with sensitivity analyses and polygenic score (PGS) validation against neuropathological endpoints","setting":"Multi-consortium, multi-national; pooled data from IGAP, EADB, PGC-ALZ, UK Biobank, and FinnGen — predominantly European institutions","sampleSize":"n=978,514 total (128,681 cases or proxy cases of ADRD; 849,833 [proxy] controls); clinical breakdown: 72,721 clinically diagnosed AD cases, 55,960 proxy ADRD cases, 614,267 controls, 235,566 proxy controls from 52 studies","population":"Adults of European ancestry with clinically diagnosed AD, proxy ADRD cases (individuals reporting ≥1 affected first-degree relative with dementia), and population controls; specific age range not reported in abstract","intervention":"Genome-wide association meta-analysis across 20,045,120 variants; construction and validation of a polygenic risk score (PGS) aggregating ADRD-associated loci excluding APOE","comparator":"Population controls and proxy controls; for PGS analyses, individuals at the tenth decile were compared to individuals at the median score group","primaryOutcome":"Identification of genome-wide significant loci associated with ADRD risk; secondary: association of PGS with post-mortem AD neuropathological measures (Braak neurofibrillary tangle stage >4 and moderate-to-severe neuritic amyloid plaque pathology)","followUp":"Not applicable (cross-sectional genetic association study); neuropathological outcomes assessed at death","fundingSource":"Multi-source; consortium-based funding acknowledged (government, non-profit, and industry-affiliated investigators disclosed); full funding details reported in supplementary materials — specific agencies not enumerated in abstract"}
Key findings
The meta-analysis identified 91 genome-wide significant (Tier 1) loci associated with ADRD risk, of which 16 were novel and 56 were specifically detected in analyses restricted to clinically diagnosed AD cases. An additional 18 loci (15 novel; Tier 2) were identified but classified as requiring external replication. Twenty-five independent secondary signals were identified across 16 known loci, suggesting greater genetic complexity than previously appreciated. Sensitivity analyses demonstrated that 75 of the 91 Tier 1 loci remained genome-wide significant after exclusion of proxy cases, and 56 remained significant in a no-biobank analysis focused exclusively on clinically diagnosed AD. Rare loss-of-function and damaging missense variants in SRC were jointly associated with AD risk (OR=4.23, 95% CI 2.04–8.79, P=1.06×10⁻⁴) in a sequencing sub-study. The ADRD PGS excluding APOE was primarily associated with AD pathology rather than non-AD dementia pathology; individuals in the 10th decile of the PGS exhibited approximately a twofold increased risk of Braak neurofibrillary tangle stage >4 and moderate-to-severe neuritic amyloid plaque pathology at death compared to those at the median score group (exact OR and confidence intervals for PGS-pathology association not reported in abstract). Genetic correlation analyses found that the ADRD phenotype (excluding APOE) was significantly correlated with Lewy body dementia (r=0.64, 95% CI 0.37–0.92), amyotrophic lateral sclerosis (r=0.29, 95% CI 0.18–0.40), and Parkinson's disease (r=0.16, 95% CI 0.06–0.27). NNT/NNH are not calculable from the reported data.
Practice implications
This study establishes the largest and most granular map to date of common genetic variation associated with ADRD and clinically diagnosed AD in European-ancestry populations. The tiered locus classification (Tier 1 vs. Tier 2) provides a principled framework for distinguishing robustly validated signals from those requiring replication — a methodological advance over prior GWASs. The finding that the non-APOE PGS tracks AD-specific neuropathology (tau tangles, amyloid plaques) rather than non-AD pathology supports biological specificity of the identified loci and has potential relevance for risk stratification in preclinical and prodromal research settings. However, several important boundaries apply: (1) The study is restricted to individuals of European ancestry, substantially limiting generalizability to non-European populations who are underrepresented in AD genetic research. (2) Proxy case inclusion, while boosting power, introduces phenotypic heterogeneity — the affected relative may have had non-AD dementia, potentially diluting AD-specific signals. (3) Individual-level predictive accuracy of the PGS remains limited, and the authors explicitly note it is not ready for clinical risk counseling at the individual level. (4) For most novel loci, the causal gene(s) and mechanism(s) remain undefined. (5) Only an abstract was available for initial assessment; full statistical tables, secondary analysis results, and supplementary data are needed for comprehensive critical appraisal. These findings do not yet directly change clinical practice but meaningfully advance the genomic architecture of AD and may influence research trial enrichment strategies and future biomarker development. __LIMITATIONS__ Restricted to European-ancestry participants, markedly limiting generalizability to African, Asian, Hispanic, and other ancestral groups where AD prevalence and genetic architecture may differ,Inclusion of proxy ADRD cases introduces phenotypic noise — dementia diagnosis in the affected relative is unverified and may include non-AD etiologies, potentially blurring AD-specific signals,Biobank-derived ICD-coded AD cases may lack diagnostic precision relative to research-grade clinical or biomarker-confirmed AD, adding heterogeneity,Eighteen Tier 2 loci (15 novel) remain unvalidated externally; their association with AD may not replicate,PGS predictive accuracy at the individual level is acknowledged as limited by the authors; twofold risk is a relative measure in the extreme decile and may not translate to clinically actionable risk thresholds,Causal genes and biological mechanisms for most novel loci are not established; GWAS identifies association, not causation,Neuropathological validation of PGS was performed in autopsy cohorts, which may not represent the broader living clinical population,Full p-values, confidence intervals, and effect sizes for most locus-level findings beyond SRC are not reported in the abstract, limiting independent critical appraisal,Multi-consortium design with overlapping samples across prior GWASs increases risk of sample overlap bias; harmonization methods used to address this are described in full-text only,Industry-affiliated investigators with advisory roles and research support from pharmaceutical companies are disclosed; potential for conflicts of interest requires consideration __BOTTOM_LINE_CATEGORY__ potentially_practice_informing __BOTTOM_LINE__ This landmark consensus GWAS meta-analysis identifies 91 robust and 18 provisional genetic loci for ADRD in European-ancestry individuals, including 16 novel confirmed loci, and demonstrates that a non-APOE polygenic score specifically associates with AD neuropathological burden rather than non-AD pathology. The findings substantially advance the genomic map of AD and provide a biologically grounded framework for future risk stratification and drug target discovery. Clinical application of these findings — particularly individual-level PGS-based risk counseling — is not yet warranted and awaits replication, multi-ancestry extension, and improved predictive calibration.