Using radiation risk models in cancer screening simulations: important assumptions and effects on outcome projections

Abstract

Purpose: To evaluate the effect of incorporating radiation risk into microsimulation (first-order Monte Carlo) models for breast and lung cancer screening to illustrate effects of including radiation risk on patient outcome projections.

Materials and methods: All data used in this study were derived from publicly available or deidentified human subject data. Institutional review board approval was not required. The challenges of incorporating radiation risk into simulation models are illustrated with two cancer screening models (Breast Cancer Model and Lung Cancer Policy Model) adapted to include radiation exposure effects from mammography and chest computed tomography (CT), respectively. The primary outcome projected by the breast model was life expectancy (LE) for BRCA1 mutation carriers. Digital mammographic screening beginning at ages 25, 30, 35, and 40 years was evaluated in the context of screenings with false-positive results and radiation exposure effects. The primary outcome of the lung model was lung cancer-specific mortality reduction due to annual screening, comparing two diagnostic CT protocols for lung nodule evaluation. The Metropolis-Hastings algorithm was used to estimate the mean values of the results with 95% uncertainty intervals (UIs).

Results: Without radiation exposure effects, the breast model indicated that annual digital mammography starting at age 25 years maximized LE (72.03 years; 95% UI: 72.01 years, 72.05 years) and had the highest number of screenings with false-positive results (2.0 per woman). When radiation effects were included, annual digital mammography beginning at age 30 years maximized LE (71.90 years; 95% UI: 71.87 years, 71.94 years) with a lower number of screenings with false-positive results (1.4 per woman). For annual chest CT screening of 50-year-old females with no follow-up for nodules smaller than 4 mm in diameter, the lung model predicted lung cancer-specific mortality reduction of 21.50% (95% UI: 20.90%, 22.10%) without radiation risk and 17.75% (95% UI: 16.97%, 18.41%) with radiation risk.

Conclusion: Because including radiation exposure risk can influence long-term projections from simulation models, it is important to include these risks when conducting modeling-based assessments of diagnostic imaging.