Population-specific genetic-risk scores enable improved prediction of mortality within 28 days of sepsis onset: a retrospective Taiwanese cohort study.

Summary

In a Taiwanese cohort of 1,403 sepsis patients, a population-specific polygenic risk score significantly improved 28-day mortality prediction over clinical-only models (AUROC 0.78 vs 0.61) and outperformed models based on European-identified SNPs. Five SNPs reached genome-wide significance and higher PRS was associated with worse survival.

Key Findings

• Five SNPs reached genome-wide significance (p < 5e-8) for 28-day mortality after sepsis onset; 86 SNPs were suggestive (p < 1e-5).
• Adding a Taiwanese population-specific PRS to clinical variables improved discrimination (AUROC 0.78 [0.75–0.80], c-index 0.79 [0.62–0.96]) over clinical-only models (AUROC 0.61 [0.58–0.64], c-index 0.63 [0.45–0.81]).
• Models based on significant SNPs from prior European studies underperformed in this cohort (AUROC 0.60 [0.58–0.63]), highlighting ethnic specificity.
• Kaplan–Meier analysis showed higher PRS groups had significantly worse survival.

Clinical Implications

If externally validated and operationalized, PRS could augment triage and resource allocation, inform intensity of monitoring and organ support, and enable precision trials. Implementation requires genetic testing infrastructure, ethical safeguards, and cost-effectiveness evaluation.

Limitations

• Single-population retrospective cohort; generalizability requires external multi-ethnic validation
• Clinical utility not assessed in a prospective implementation study
• Potential population stratification and unmeasured confounding

Future Directions

Prospective, multi-ethnic external validation; integration with dynamic clinical data; assessment of decision-analytic impact, equity, and cost-effectiveness; development of practice guidelines for genomic risk use in sepsis.