Assessing the effectiveness of cervical screening

The mathematic models used to assess the benefits and cost-effectiveness of cervical screening reveal little consistency in the definition of disease status, the basic assumptions made, or the data used in the model. Conclusions derived from the models often are model and data dependent. Several authors have used simplified models and unrealistic assumptions, such as the failure to differentiate between different grades of dysplasia, or 100% sensitivity for the screening test. The Markov process assumes that the rate of transfer between states is independent of the duration of time spent in one state, and this assumption may be unsound. The difficulty with all models is in interpreting the appropriateness of the parameter values. Some are well documented, for example, stage-specific survival rates for treated patients or attendance for screening. Many, however, cannot be given a fixed value. The large number of factors that appear implicated in the incidence of and mortality from cervical cancer can lead to many feasible sets of parameter values that generate output that approaches the observed data.