“…Because longitudinal interpretation typically requires a minimum of several years of data to estimate slope, the traditional cross-sectional approach was used for the first 4 years, whereupon time-variant improvement in the sensitivity and specificity was incorporated. Empirical studies 1,20,21 have shown that the pattern of technical improvement of test performance over Uhe ϭ utility healthy exposed; Ued ϭ utility early disease; Uie ϭ utility identified early; Uad ϭ utility advanced disease; Ufp ϭ utility false positive. time is statistically complex; we used a simple algorithm with improvement in sensitivity and specificity logarithmically related to the number of testing episodes to date (eg, at any screen, performance would be better with semiannual testing than with biannual testing).…”