The increasing use of statistical modelling and lower certainty evidence to expand screening and the aggressive marketing of multicancer early detection tests raises questions about evidence thresholds for updating existing screening recommendations. Since 2018, five of the updated cancer screening recommendations of the US Preventive Services Task Force (USPSTF) have included statistical models (breast, colorectal, lung, cervical and prostate). All five have recommended more intensive screening than the earlier recommendation, either by recommending newer screening technologies or expanding the population eligible for screening. For example, the 2021 colorectal cancer screening recommendation lowered the starting age from 50 to 45 years based on a microsimulation model of hypothetical patient panels.1 2 The models for updating recommendations for all of these cancers relied heavily on either intermediate outcomes or performance characteristics of the screening test. Several blood-base