Quantitative Models for Modeling Overdetection in Population-Based Screening for Colorectal Cancer

Background: Overdetection resulting from population-based cancer screening has been debatable over the past decade. However, a systematic approach for quantitative estimation of overdetection has been barely addressed. Contextually, mass screening for breast cancer and prostate cancer has been well documented, but colorectal cancer (CRC) has been scarcely highlighted. Aim: The aim of this study is to develop a systematic approach for quantitative estimation of overdetection and elucidating the mechanism of overdetection in population-based colorectal cancer screening. Methods: The graphic method was first proposed by comparing cumulative incidence curves of advanced and nonadvanced cancers. Then, the modeling approaches used for quantitative methods were developed. We proposed the standardized overdetection ratio (SOR), the index including expected and observed which are simulated by a 3-state Markov model (normal, preclinical detectable phase, clinical phase) based on screened group and control group, respectively. Moreover, we developed the 5-state Markov model in conjunction with SOR to quantify the extent of overdetection. Two randomized controlled trials on guaiac–fecal occult blood test (gFOBT) in U.K. and Denmark and population-based service screening using fecal immunochemical test (FIT) in Taiwan were applied to assess the proposed methods to quantify overdetection in population-based cancer screening. Results: Using the graphic method, the average of overdetection was 31.1% and 24.9% for the UK trial and the Denmark trial, respectively. However, the estimated proportion of overdetection may be biased because the graphic method did not take the natural history of adenoma and high awareness in the routine clinical practice into account. According to the proposed 3-state Markov model, the overdetection of CRC was 9% with gFOBT based on the 2 randomized controlled trials in Europe and 7% with FIT based on a nationwide screening program in Taiwan. The estimated overdetection by the 5-state Markov model were 6.1% and 9.2% for the U.K. and Denmark trial, respectively. Moreover, the main pathway for overdetection in population-based CRC screening is from nonadvanced in the preclinical phase to advanced in the preclinical phase which is supported by the evidence that SOR decreases from 1.29 to 1.16 and 1.16 to 0.97 for the U.K. and Denmark trial, respectively. Conclusion: This study developed the graphic method and the disease natural progression model for quantitative assessment of overdetection in colorectal cancer. The index, SOR, was proposed not only to assess the extent of overdetection but also to elucidate the mechanism to know how overdetection affects the progression of colorectal cancer from the preclinical phase to the clinical phase with information of cancer stage.