Introduction:

Polygenic risk scores (PRS) aggregate variants at multiple genetic loci to calculate the likelihood of developing a specific disease. These scores are developed using databases of individuals who have undergone testing, most from European ancestry. This may reduce accuracy for patients of underrepresented ancestral backgrounds. Economic models assessing the value of PRS without exploring the accuracy when used in a diverse population may inaccurately estimate the value of testing, leading to inappropriate coverage, reimbursement, and clinical implementation decisions. We conducted a scoping review examining population discordance in published PRS models.

Methods:

We searched PubMed and EMBASE using terms related to PRS and economic evaluation through November 30^th, 2021. Snowball sampling identified additional manuscripts from references. Data were extracted using a standardized template. We used race and ethnicity as a surrogate for genetic ancestry, identifying population discordance when authors did not state race or ethnicity of the population modeled, or when the population modeled did not match the race or ethnicity of the population used to derive PRS.

Results:

We identified 11 papers or preprints. Oncology (8 73%), glaucoma (1, 9%), nephropathy (1, 9%), and cardiovascular disease (1, 9%) were represented. All models exhibited population discordance, with most papers using a PRS developed in a subset of the population and extrapolating findings to the entire population. No papers included adjustments for discordance in their primary analysis, while one paper included adjustments in sensitivity analyses.

Conclusion:

Economic analysis of PRS to date have not explicitly address discordance between the populations used to develop the score and used in the model. Future modeling approaches should be transparent about this difference and characterize the impact of PRS performance on clinical and economic utility in diverse populations.