HOW ADHERENCE DYNAMICS CHANGE POLICY CONCLUSIONS: MODELING COLORECTAL CANCER SCREENING WITH THE LONGITUDINAL ADHERENCE TRAJECTORY (LAT) FRAMEWORK
Author(s)
Mohammad Dehghani, PhD1, Vahab Vahdat, MSc, PhD2, Burak Ozbay, PhD2, Chris Estes, MPH2, Christopher Tyson, PhD2, Joyce Luo, BSE3, Oguzhan Alagoz, PhD4;
1Northeastern University, Boston, MA, USA, 2Exact Sciences, Madison, WI, USA, 3Massachusetts Institute of Technology, Cambridge, MA, USA, 4University of Wisconsin-Madison, Middleton, WI, USA
1Northeastern University, Boston, MA, USA, 2Exact Sciences, Madison, WI, USA, 3Massachusetts Institute of Technology, Cambridge, MA, USA, 4University of Wisconsin-Madison, Middleton, WI, USA
OBJECTIVES: Colorectal cancer (CRC) screening reduces incidence and mortality, yet effectiveness is constrained by imperfect adherence. Current models often assume fixed or perfect adherence, ignoring dynamic participation across repeated screening rounds. We developed the Longitudinal Adherence Trajectory (LAT) model, a framework representing adherence as a continuous trajectory shaped by age and prior screening behavior. This study details the validation of the LAT model and its calibration for evaluating CRC screening strategies.
METHODS: The LAT model integrates two behavioral mechanisms. The first (i) is a growth component, modeled as a logistic growth function, reflecting how adherence strengthens with continued participation. The second (ii) is a decay component, modeled as a Hill decay function, capturing how missed screenings reduce future participation. Parameters were calibrated to match three population-level targets: overall participation, uptake at screening initiation, and proportion of never-screeners. The validated CRC-AIM microsimulation model then evaluated outcomes across three screening strategies using published real-world adherence to the screening and follow-up colonoscopy: annual fecal immunochemical test (FIT), triennial multi-target stool-DNA (mt-sDNA), and decennial colonoscopy. Outcomes included life-years gained (LYG) under perfect and LAT adherence. Sensitivity analyses evaluated impacts of varying inputs and calibration targets.
RESULTS: The LAT model accurately reproduced empirical adherence targets (objective function values 0.001-0.025) and generated realistic longitudinal participation dynamics across modalities. Incorporating LAT into CRC-AIM substantially altered comparative outcomes. Under perfect adherence, colonoscopy provided the greatest LYG (363 per 1,000 individuals); however, when LAT were applied, LYG for colonoscopy was 265, compared to 283 for mt-sDNA and 193 for FIT. Sensitivity analyses demonstrated that results were robust to parameter uncertainty.
CONCLUSIONS: The LAT model offers a practical framework for modeling longitudinal adherence that addresses key limitations of existing approaches. These findings underscore that screening effectiveness is determined not only by test performance but also by how adherence behavior unfolds across the screening lifespan.
METHODS: The LAT model integrates two behavioral mechanisms. The first (i) is a growth component, modeled as a logistic growth function, reflecting how adherence strengthens with continued participation. The second (ii) is a decay component, modeled as a Hill decay function, capturing how missed screenings reduce future participation. Parameters were calibrated to match three population-level targets: overall participation, uptake at screening initiation, and proportion of never-screeners. The validated CRC-AIM microsimulation model then evaluated outcomes across three screening strategies using published real-world adherence to the screening and follow-up colonoscopy: annual fecal immunochemical test (FIT), triennial multi-target stool-DNA (mt-sDNA), and decennial colonoscopy. Outcomes included life-years gained (LYG) under perfect and LAT adherence. Sensitivity analyses evaluated impacts of varying inputs and calibration targets.
RESULTS: The LAT model accurately reproduced empirical adherence targets (objective function values 0.001-0.025) and generated realistic longitudinal participation dynamics across modalities. Incorporating LAT into CRC-AIM substantially altered comparative outcomes. Under perfect adherence, colonoscopy provided the greatest LYG (363 per 1,000 individuals); however, when LAT were applied, LYG for colonoscopy was 265, compared to 283 for mt-sDNA and 193 for FIT. Sensitivity analyses demonstrated that results were robust to parameter uncertainty.
CONCLUSIONS: The LAT model offers a practical framework for modeling longitudinal adherence that addresses key limitations of existing approaches. These findings underscore that screening effectiveness is determined not only by test performance but also by how adherence behavior unfolds across the screening lifespan.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
MSR121
Topic
Methodological & Statistical Research
Disease
SDC: Gastrointestinal Disorders, SDC: Oncology