Keywords: Genome-wide association studies, cancer, screening, comparative effectiveness
Genome-wide association studies (GWAS) have identified hundreds of common single nucleotide polymorphisms (SNPs) associated with cancer incidence. Although each of these SNPs is only weakly associated with cancer risk, taken together they can describe a broad gradient in risk. This polygenic risk could be used to design precision screening strategies that reduce overdiagnosis and cancer mortality. Here I review statistical challenges that arise when translating the results of GWAS to precision screening: (i) building powerful risk models; (ii) assessing model calibration; and (iii) evaluating the effectiveness of polygenic-risk-stratified screening strategies. I illustrate these issues using the latest genome-wide association results from breast, pancreas and colorectal cancer. Generally, polygenic risk discrimination is a function of GWAS sample size and genetic architecture, while the effectiveness of targeted screening depends on screening modality and risks and effectiveness of early intervention.