McGeoch L, et al. - In this updated review of a previous review, researchers focused on the discovery of risk prediction models for colorectal cancer including genetic variants. So that models including at least one single nucleotide polymorphism (SNP) and applicable to asymptomatic individuals in the general population could be identified, they searched Medline and EMBASE from the end date of the previous review (January 2014) to February 2019. Twenty-three novel models were discovered, giving a total of 29. The estimated AUROC in external or split-sample validation was 0.56-0.57, 0.61-0.63, and 0.56 to 0.70 for models incorporating SNPs alone, for SNPs in combination with other risk factors, and on incorporating age, respectively, among those in which the SNP selection was based on published GWASs. For only four, calibration was reported. Discrimination increased by 0.01-0.05 when SNPs were added to other risk factors. By public health modelling studies, the range of beginning ages for screening would be several years greater if determined by risk models vs using family history alone.