Alzheimer's disease (AD) shows sex differences in incidence, clinical course, and biomarker profiles. The authors asked whether inherited genetic liability to modifiable dementia risk factors contributes differently to AD susceptibility in females compared with males. To address this, they combined polygenic risk scores (PRSs), genome-wide genetic correlation, and two-sample Mendelian randomization (MR) analyses.
Three complementary genetic methods were applied:
Analyses used summary and individual-level data from two primary resources. Clinical AD diagnosis associations were evaluated in the Alzheimer's Disease Genetics Consortium (ADGC; n = 27,751). Plasma biomarker associations used the Health and Aging Brain Study–Health Disparities (HABS‑HD) cohort with sample sizes ranging from 2,307 to 2,737 depending on the biomarker. Information on GWAS summary statistics for traits and supplementary tables are provided by the authors in the manuscript supplement.
Primary analyses compared female- and male-specific estimates and formally tested sex differences. Multiple testing was controlled with false discovery rate (FDR) procedures. Sensitivity analyses were performed for MR findings; details of instruments and sensitivity results are reported in the manuscript and supplementary material.
PRS analyses identified associations between genetic predisposition and clinical AD for several traits. Notably, PRSs for educational attainment and physical activity were associated with AD diagnosis in both females and males. Some PRS sex differences were observed at a nominal level for educational attainment, total cholesterol, and LDL cholesterol, but none of these sex-difference tests remained significant after FDR correction. Thus, while certain PRSs related to AD, evidence that those relationships differed by sex was limited after correction for multiple comparisons.
Genome-wide genetic correlation analyses revealed sex-differing patterns of genetic overlap that reached FDR-significance for multiple traits. Specifically, sex differences in genome-wide genetic correlation with AD were FDR-significant for social isolation, physical activity, LDL cholesterol, and pulse pressure. These results indicate that overall shared genetic architecture between AD and these traits differs by sex, although such genome-wide differences did not consistently map onto PRS or MR results.
Two-sample MR identified within-sex associations that survived FDR correction in stratified analyses. The primary MR findings were:
However, sensitivity analyses provided inconsistent support for these MR associations. Moreover, comparisons of MR effect estimates between sexes did not yield FDR-significant sex differences. The authors therefore highlight that while some sex-stratified MR associations were observed, evidence for robust sex differences in causal effects was not consistent across methods or sensitivity checks.
PRSs were also tested against plasma biomarkers measured in HABS‑HD. After FDR correction, the authors report the following male-specific PRS associations with plasma markers:
These biomarker associations reached FDR-significance among males but were not described as comparable in females in the primary report.
Across analytic approaches, evidence that genetic liability to modifiable dementia risk factors relates differently to AD by sex was limited and depended on the genetic measure used. Genome-wide genetic correlation suggested several FDR-significant sex differences in shared genetic architecture, but these patterns were not consistently reflected in PRS associations or MR causal inference. The authors emphasize that shared genome-wide architecture does not necessarily translate into sex-specific inherited liability as measured by PRS or into sex-differing causal effects by MR.
The report is a preprint and has not undergone peer review; the authors note inconsistent sensitivity support for MR findings and that no MR sex-difference remained FDR-significant. Specific details on MR instruments, effect sizes, and sensitivity metrics are provided in the manuscript and supplementary material. The authors declare no competing interests.
ADGC and HABS‑HD data are available from the cited consortium and study websites. GWAS summary statistics used for each trait are listed in Supplementary Table S1 of the manuscript. The authors report that appropriate institutional review board approvals and participant consents were obtained for ADGC and HABS‑HD. The manuscript is released under a CC‑BY‑NC‑ND 4.0 license and is presented as a preprint, not yet peer reviewed.