Figure 3.

Adequately precise identification (API) and type I/II errors in a simulation study of stop-screen trials. Model performance over a range of values for the indolent fraction $(\mathrm{\psi })$ and the progression rate of progressive cancers $(\mathrm{\lambda })$ is visualized as (top row) percentages of 100 simulations achieving joint API for all 4 model parameters and (bottom row) percentages of 100 simulations that reject the null hypothesis, $H_0:\mathrm{\psi }=0$. Performance is visualized for 50,000 women screened annually at ages 50–54 years (5 screenings each) with follow-up to age 55 years (left column) or age 60 years (right column), assuming a constant risk of onset of preclinical cancer of $w=0.0025$ per year and a sensitivity of screening to detect preclinical cancer of $\mathrm{\beta }=80\%$.