%0 Journal Article
%T Modelling the impact of bias in fecal immunochemical testing on long-term outcomes of colorectal cancer screening.
%A de Jonge L
%A Toes-Zoutendijk E
%A Koopmann BDM
%A van Schrojenstein Lantman M
%A Franken-van Vorsselen B
%A Speijers C
%A van Ingen H
%A Humer E
%A van der Groep P
%A Thelen M
%A Lansdorp-Vogelaar I
%A
%J Clin Chim Acta
%V 561
%N 0
%D 2024 Jul 15
%M 38879061
%F 6.314
%R 10.1016/j.cca.2024.119809
%X BACKGROUND: As the impact of unmanaged bias (i.e. systematic source of inaccuracy) in fecal immunochemical test (FIT) analytical performance on long-term colorectal cancer (CRC) outcomes is unknown, we assessed the impact bias in FIT performance in an ongoing FIT-based CRC screening program.
METHODS: This study consisted of two parts: cross-sectional observational data analysis to estimate change in short-term outcomes and microsimulation modelling to estimate change in long-term outcomes assuming different levels of bias by assuming 15 % lower up to 15 % higher Hemoglobin detected in the stool compared to observed. Two scenarios were considered: bias occurring 1) one-time only, due to the occasional bias associated with the FIT kits used in 2020 and 2) consistently due to a constant bias associated with the FIT kits used from 2020 onwards.
RESULTS: With a hypothetical bias of -15 % to +15 %, we observed a positivity rate ranging from 6.7 % to 7.8 %, and a detection rate for CRC between 0.65 % and 0.68 %. Single biases in FIT performance resulted in less than 0.1 % change in long-term CRC screening outcomes, while consistent biases resulted in a much larger change (up to 1.4 % in CRC cases and CRC-related deaths and up to 2.07 % in total costs). Detecting lower Hemoglobin concentrations resulted in a relatively larger change on long-term CRC outcomes in comparison to positive bias.
CONCLUSIONS: Because of the substantial impact of consistent FIT bias, it is important to set evidence-based acceptance criteria of bias on long-term CRC screening outcomes and in particular, the introduction of an asymmetrical or upward shifted tolerance interval for FIT bias.