A Framework to Assess Uncertainties of Polygenic Risk Scores in Disease Screening and Prevention in Cost-Effectiveness Analyses

OBJECTIVES: Polygenic risk scores (PRS), risk prediction algorithms based on genome-wide association studies, show promising potential to predict disease risk and guide personalized screening and treatment. However, PRS risk prediction is uncertain and by extension has uncertain health economic value. It is unclear how previous cost-effectiveness analyses (CEA) handled the varied sources of uncertainty related to PRS risk prediction. This study aims to fill this knowledge gap and develop a framework to guide future CEA for PRS.

METHODS: We conducted a targeted review for PRS CEA published between January 2018 and June 2023 by searching the PubMed database using the keywords of PRS and “cost-benefit/utility/effectiveness”. We then assessed how published CEA considered uncertainty of PRS in sensitivity analyses and discussed the sources of PRS uncertainty that should be taken into account in CEA.

RESULTS: Our targeted review identified 16 studies: 13 modeled cancer screening scenarios, 3 modeled preventive treatment scenarios. We found inconsistent approaches to quantifying PRS uncertainty: 4 studies used an area under the curve (AUC) approach to represent PRS predictability, and 2 studies directly examined the impact of uncertain PRS-predicted disease risk on PRS’s value. The rest did not quantify or examine the uncertainty of PRS risk prediction. To address this issue, we propose that 3 layers of PRS uncertainty should be explicitly incorporated in future CEAs: (1) the heritability of a given disease, which is the upper limit of PRS’s predictability, (2) the observed predictability of PRS represented by metrics such as AUC, and (3) the downstream predicted disease risk by PRS.

CONCLUSIONS: To date, CEA studies have inconsistently examined the uncertainty of PRS-predicted risk. Our proposed framework illustrates different layers of uncertainty of PRS. Future CEAs should explicitly incorporate these sources of uncertainty to assess the value of PRS in disease screening and prevention.