Modest Improvement in Prediction From Combining Established Colorectal Cancer Risk Model With PRS

A University of Oxford-led team has combined a polygenic risk score with the colorectal cancer prediction model QCancer-10 to gauge whether the two together can better identify people at high risk of disease. As they report in the BMJ, the researchers used data from the UK Biobank to generate six PRSs and chose the best-performing one combine with QCancer-10. QCancer-10 itself takes age, ethnic group, family history, and alcohol and smoking status as well as a few medical conditions into consideration to determine disease risk. The researchers compared the performance of the combined PRS-QCancer-10 tool to QCancer-10 alone, finding that the two together modestly improve risk prediction. For the 20 percent of individuals with the highest risk, the sensitivity and specificity of the integrated tool for predicting colorectal cancer diagnosis was 47.8 percent and 80.3 percent respectively, for men and 42.7 percent and 80.1 percent, respectively, for women. "Although we have shown that risk stratification in some form is likely in principle to improve resource use and performance of colorectal cancer screening, the added benefit of adding PRS to QCancer-10 is modest, and we find no clear justification for implementing PRS based risk stratification at present," the researchers write.