Numerous options exist for cervical cancer prevention with new screening and vaccination modalities. Microsimulation models are used to investigate these options and predict the cost-effectiveness of integrated strategies but are computationally intensive. Besides, because the outcomes of micro-simulation models are not available in closed form, the execution of probabilistic sensitivity analyses, or parameter estimation, by Bayesian procedures remains challenging. Compartmental models may also be considered. They are fast and provide numerical closed form solutions, but the number of compartments rapidly increases with the number of HPV types and complexity of the screening algorithm.
We developed a compartmental mathematical progression model that integrates natural history of cervical carcinogenesis along multiple HPV types and screening interventions. The full model was compressed to achieve executable and efficient model, even when dealing with complex screening. The model outcomes are available in closed form so that the parameter uncertainty can be assessed by Bayesian procedures. The models were compared to a microsimulation approach in terms of overall and type-specific HPV prevalence of intermediate stages, screening outcomes and cancer incidence.
The outcomes of the compressed compartmental model were stable over different levels of compression and stayed between the simulation error bounds of the microsimulation model for all overall HPV infection states. A small difference was observed in a fraction susceptible versus immune.
Good approximation properties of compressed compartmental models enable us to assess uncertainties surrounding the natural history of cervical carcinogenesis and screening decisions in a computationally undemanding way.