Polygenic risk score and its role in cancer susceptibility.

Abstract

BACKGROUND: Polygenic risk score (PRS) has the ability to stratify inherited susceptibility to cancer and, as a complement to monogenic testing, can identify individuals at increased genetic risk even when no pathogenic variant is detected in high- or moderate-penetrance genes. It reflects the combined additive effects of a large number of low-penetrance variants across the genome, and represents a continuum of genetic susceptibility with an approximately normal distribution. Clinically relevant differences are typically observed in individuals in the highest and lowest percentiles of the PRS distribution, while relative risk gradients depend on the cancer type, the specific PRS model, and the reference population used. PRS is not a single test but rather a family of statistical models that differ in their design, predictive performance, and transferability across populations, underscoring the need for external validation and population-specific calibration of absolute risk. Broader implementation is thus still held back by differences between individual PRS models, limited transferability, and the lack of harmonized guidance on indication, reporting, and clinical decision-making. Consequently, clinical use in the European Union remains largely confined to pilot studies and local projects. Within these initiatives, PRS is most commonly applied in two main ways - either as a triage tool for intensified diagnostics or screening in higher-risk groups, or as a component of multifactorial absolute-risk models (e. g. BOADICEA/CanRisk) that integrate PRS with other risk factors such as pathogenic variants in moderate-penetrance genes (e. g. ATM or CHEK2), family history, or lifestyle factors. By refining absolute-risk estimates, PRS may shift individuals across clinical decision thresholds for more intensive surveillance and preventive strategies. AIM: This review summarizes the principles of PRS, the main sources of variability between models, and its potential applications in risk stratification and personalized cancer screening. It also addresses limitations in transferability, the need for calibration, and the currently limited evidence for improvements in hard clinical outcomes.