The aim of the paper is to compare the accuracy and effectiveness of liquid-based versus conventional cytology when used as part of a process of population screening to reduce the incidence of invasive cervical cancer. Data from split-sample studies were analysed by using a hierarchical Bayesian multiparameter evidence synthesis model. Predictions of sensitivity and specificity were derived without having a procedure to detect the lesions. Results suggest that the specificity of both tests is high, such that less than 1% of normal samples are expected to be misdiagnosed. However, the sensitivity may be lower, with around 81% of lesions expected to be correctly diagnosed in smear tests and 88% in liquid-based cytology.