Integrating polygenic risk scores with clinical covariates significantly improved the prediction of survival beyond 95 years, increasing the model AUC from 0.913 to 0.956.
Case-Control (n=3,319)
Yes
Does the integration of polygenic risk scores with clinical covariates improve the prediction of exceptional longevity in a Taiwanese population?
Genetic risk captured by polygenic risk scores contributes additional predictive information for extreme longevity beyond clinical factors in Taiwanese populations.
Absolute Event Rate: 0.956% vs 0.913%
p-value: p=<10^-8
Human longevity arises from complex genetic and environmental interactions, yet the genetic basis of survival to extreme old age remains underexplored in Asian populations. We performed genome-wide association studies (GWAS) in a Taiwanese cohort, defining survival thresholds at ≥ 85, ≥ 90, and ≥ 95 years, and validated significant loci in an independent cohort. Multiple loci, including ZNF806, NUAK1, TANC1, SLC22A3, PTPRD, and PCSK2, were associated with longevity, of which 14 replicated with consistent effect directions (82% concordance). Allele frequencies aligned with East Asian references but diverged from prior Han Chinese studies, reflecting sub-ethnic variation. Polygenic risk scores (PRSs) alone showed limited predictive ability but provided statistically significant incremental improvement when integrated with clinical covariates. In the external validation cohort, adding PRS modestly improved model discrimination (AUC from 0.900 to 0.904 for ≥ 85 years and from 0.893 to 0.912 for ≥ 90 years) and yielded the largest improvement in the ≥ 95 group (AUC from 0.913 to 0.956; DeLong P < 10⁻⁸), with corresponding gains in reclassification metrics. These findings suggested that while clinical factors remained the primary predictors of survival, genetic risk captured by PRS contributed additional information, particularly at extreme longevity thresholds. Together, the results highlighted an age-dependent genetic architecture enriched for neural, cardio-metabolic, and stress-response pathways and supported the use of genetics-informed models as complementary tools for precision aging research in Taiwanese populations.
Hsieh et al. (Wed,) conducted a case-control in Exceptional longevity (n=3,319). Polygenic risk score (PRS) vs. Clinical covariates only was evaluated on Model discrimination (AUC) for survival to ≥95 years (p=<10^-8). Integrating polygenic risk scores with clinical covariates significantly improved the prediction of survival beyond 95 years, increasing the model AUC from 0.913 to 0.956.