Keywords: dynamic scheduling, sensitivity, sojourn time, transition probability, lead time, over diagnosis.
We develop a probability model for dynamical scheduling of future screening exams based on an asymptomatic individual's screening history, using data from a breast cancer randomized screening trial. We derive the conditional probability of incidence before the next screening exam, given an asymptomatic subject with a screening history. On the assumption that screen-detected cases have better prognoses than clinically-detected ones (which may be more advanced), the model provides a screening time interval to limit the probability of being an interval case (i.e., no clinical symptoms before the next screen) to a small value, such as 0.05 or 0.10. This conditional probability depends on various factors, such as one's current age, screening history, screening sensitivity, sojourn time in preclinical state, and transition probability into the preclinical state. We then derive the lead time distribution and the probability of overdiagnosis if one would be diagnosed with cancer at the next proposed screening time, so that predictive information can be provided to individuals on how early her disease could be detected and the risk of overdiagnosis if she undergoes exam as suggested.