In this paper, we develop a model and a deconvolution technique to estimate
tile distribution of the age of onset of detectable preclinical cancer, us
ing incidence data and information about cancer growth rate. Our approach c
onsiders tile effect of age on both the growth rate and competing causes of
death. We illustrate our methodology using data on breast cancer, and disc
uss implications for breast cancer screening. (C) 2001 Elsevier Science Ltd
. All rights reserved.