First IBS Channel Network Conference Rolduc , The Netherlands , 8 - 11 May 2007

In many developed countries, a cervical cytology screening programme has been implemented in order to reduce the incidence of cervical cancer. Although cytology screening has proven to be effective, its performance is not optimal. Several studies have shown that the cytology test reveals a high rate of false negative rates. Since infection with human papillomavirus (HPV) is the primary cause of virtually all cervical cancer cases, HPV-DNA testing has been prompted as an adjunct or primary screening tool (Kuhn et al, 2000). In an ongoing large screening trial of 44,000 women in the Netherlands, the performances of HPVDNA testing and cytology are compared. Determining the accuracy of HPV-DNA testing and cytology from the Dutch screening trial is hampered because women with less than moderate cytological abnormality are not immediately verified but retested during the first and second year to allow them to spontaneously clear the cervical lesion. Current methods (Cecil et al,1996; Begg and Greenes,1983; Kosinsky and Barnart, 2003; Zhou, 1993; Baker,1995; Alonzo,2005) are designed for cross-sectional data and cannot be applied straightforwardly because loss-to follow-up depends on intermediate test results and disease status may change over time. In this presentation, the results of a longitudinal latent class log-linear model with partially missing data will be discussed which is used for the estimation of the sensitivity and specificity of HPV DNA testing and cytology. Both ignorable and non-ignorable models will be fitted to the data.