Using G-methods to assess and mitigate bias from coarsening time intervals in evaluating colorectal cancer screening efficiency

Population-based cancer screening has been crucial for increased early detection and reduced cancer mortality in recent decades.This includes colorectal cancer (CRC) screening, which also directly improves cancer prevention through the detection and removal of precancerous lesions. 1 Observational studies are often used to assess screening effectiveness due to the ethical challenges and practical limitations of conducting randomized control trials. 2Unlike most studies of repetitive treatment, time-varying confounders (e.g.detection of precancerous lesions at the last screen) are needed to accurately estimate the effectiveness of cancer screening.Generalized methods (G-methods) can account for such timevarying confounders in observational studies. 3However, precise measurements of screening intervals may be compromised due to data coarsening in records or analysis, with screening classified as annual, biennial or triennial, rather than using the actual length of time between individuals' screening events.In their recent study, 'Bias Due to Coarsening of Time Intervals in the Inference for the Effectiveness of Colorectal Cancer Screening', Karmakar et al. highlight the potential for these data coarsening to bias outcomes and present a sensible approach to understanding and limiting this type of modelling bias. 4