Estimation of the duration of a pre-clinical disease state using screening data

This paper developed a robust model based on the length of the pre-clinical state and the sensitivity of the screening test. The model used data on the detection rate of cancer at each screen, and the rate of interval cancers. The key model parameters were the screening sensitivity and the incidence rate of (symptomic) disease in the absence of screening. A distinction was made between the prevalent (first) screening test and subsequent (incident) tests. This leads to estimates of lead-time bias and length bias. The data from the first breast cancer screening trial (HIP, New York) were used to illustrate the method. When an exponential model was used, the sensitivity was estimated to be 82% and the mean lead time was 1.71 years in this trial.