Integrating genome-wide polygenic risk scores and non-genetic risk to predict colorectal cancer diagnosis: a cohort study in UK Biobank

Objectives To evaluate the benefit of combining polygenic risk scores (PRS) with the QCancer-10 (colorectal cancer) non-genetic risk prediction model to identify those at highest risk of colorectal cancer (CRC). Design Population based cohort study. Six different PRS for CRC were developed (using LDpred2 PRS software, clumping and thresholding approaches, and genome-wide significant models). The top-performing genome-wide and GWAS-significant PRS were then combined with QCancer-10 and performance compared to QCancer-10 alone. Case-control (logistic regression) and time-to-event (Cox proportional hazards) analyses were used to evaluate risk model performance in men and women. Setting and participants UK Biobank Study. A total of 434587 individuals with complete genetic and QCancer-10 predictor data were included in the QCancer-10+PRS modelling cohorts. Main outcome measures Prediction of colorectal cancer diagnosis by genetic, non-genetic and combined risk models. Findings PRS derived using the LDpred2 program performed best, with an odds-ratio per standard deviation of 1.58, and top age- and sex-adjusted C-statistic of 0.733 (95% confidence interval 0.710 to 0.753) in logistic regression models in the validation cohort. Integrated QCancer-10+PRS models out-performed QCancer-10 alone. In men, the integrated LDpred2 (QCancer-10+LDP) model produced a C-statistic of 0.730 (0.720 to 0.741) and explained variation of 28.1% (26.3% to 30.0%), compared with 0.693 (0.682 to 0.704) and 21.0% (18.9% to 23.1%) for QCancer-10 alone. Performance improvements in women were similar. In the top 20% of individuals at highest absolute risk, the sensitivity of QCancer-10+LDP models for predicting CRC diagnosis within 5 years was 47.6% in men and 42.5% in women, with respective 3.49-fold and 2.75-fold absolute increases in the top 5% of risk compared to average. Decision curve analysis showed that adding PRS to QCancer-10 improved net-benefit and interventions avoided, across most probability thresholds. Conclusions Integrating PRS with QCancer-10 significantly improves risk prediction over QCancer-10 alone. Evaluation of risk stratified population screening using this approach is warranted. Summary Box What is already known on this topic Risk stratification based on genetic or environmental risk factors could improve cancer screening outcomes No previously published study has examined integrated models combining genome-wide PRS and non-genetic risk factors beyond age QCancer-10 (colorectal cancer) is the top-performing non-genetic risk prediction model for CRC What this study adds Adding PRS to the QCancer-10 (colorectal cancer) risk prediction model improves performance and clinical benefit, with greatest gain from the LDpred2 genome-wide PRS, to a level that suggests utility in stratifying CRC screening and prevention