Studies have shown a consistent trend between polygenic risk score percentiles and disease risk in populations, and this evidence is used to support its clinical use. However, in clinic, polygenic risk score values, not percentiles, are used for estimating disease lifetime risk. In this study, we assessed the calibration of polygenic risk score values for three common cancers (prostate, breast and colorectal) by comparing them with the respective observed risks in the large UK Biobank population. A systematic bias was found for each cancer. This bias, however, can be corrected using a scaling approach. We conclude that calibration is necessary and feasible prior to clinical implementation. (By W Kyle Resurreccion, https://jmg.bmj.com/content/early/2020/12/20/jmedgenet-2020-107286 )