Contextualizing the Utility of Polygenic Risk Scores using Absolute Risk Models in Diverse Ancestry Populations.

Area of Science:

• Genetics and Genomics
• Epidemiology
• Biostatistics

Background:

• Polygenic risk scores (PRSs) are increasingly used to assess inherited disease risk.
• A significant challenge is the lower accuracy of PRSs in non-European populations, particularly those of African ancestry.
• Previous evaluations often overlooked contextual factors and focused on relative risk or AUC, not absolute risk.

Purpose of the Study:

• To introduce a novel measure, the conditional average derivative estimator (CADE), for evaluating PRS utility in absolute risk models.
• To assess PRS utility across diverse populations and contexts by integrating them with other risk factors.
• To analyze PRS utility for breast and prostate cancer risk using data from the All of Us Research Program.

Main Methods:

• Developed and applied the conditional average derivative estimator (CADE) to quantify variable importance.
• Integrated PRSs for breast and prostate cancer into age-specific absolute risk models.
• Utilized individual-level data from the All of Us Research Program and SEER cancer registry data.

Main Results:

• Despite lower discriminatory accuracy in African Americans (AA), PRSs demonstrated significant utility in specific contexts.
• CADE values indicated greater PRS utility for stratifying prostate cancer risk and mortality in AA compared to European Americans (2-fold and 7-fold higher, respectively).
• The findings highlight context-dependent utility of PRSs beyond traditional accuracy metrics.

Conclusions:

• Conclusions regarding the limited clinical utility of PRSs in non-European populations may be premature.
• Quantifying PRS risk-stratification utility at the absolute-risk level is crucial.
• Future assessments should incorporate disease onset, survival, and broader health/economic factors.