Extending Recurrence-Time Models of Periodic Screening to Accommodate Time-Varying Incidence and Sensitivity

• Activity Number: 44
• Type: Contributed
• Date/Time: Sunday, July 31, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Sanda Cristina Oancea*+ and Timothy Church
• Companies: University of Minnesota and University of Minnesota
• Keywords: screening ; cancer ; lead-time ; recurrence-time

Abstract

An important aspect of mass-screening for chronic diseases like cancer is the amount of time by which diagnosis is advanced by a program of testing at regular intervals. Prorok and others have developed models of periodic screening that couple a progressive disease model with a preclinical sojourn time distribution and simplifying assumptions of a constant probability of initiating a preclinical disease phase and a constant sensitivity over the course of the preclinical phase. Such recurrence-time models can be used to produce ML estimates of the parameters of the lead-time distribution gained by screening a population. However, for many applications the assumptions that these are constants are unrealistic. We undertook modification of the recurrence-time model of Prorok to incorporate an age-dependent preclinical incidence and a generation and screen dependent sensitivity function. This allows more flexibly modelling the screening process for a wider array of screening situations. We validate the parameters by using MLE and apply the estimators to data sets from the intervention arm of randomized trials.We illustrate the use of the method on data from a cancer screening trial.