Bridging Decision Analytic Modelling with a Cross-Sectional Study: Application to Parkinson's Disease

The ideal study design for demonstrating the possible health outcomes and costs associated with a new drug would be a naturalistic prospective study. However, it is not often feasible to derive the required information from scientifically sound prospective studies. In these cases, decision analytic models may provide some of the missing information. However, the use of a Delphi panel to gather data for these models is a major concern because of potential bias and data accuracy. Because reimbursement of pharmaceuticals is often based on economic data derived from modelling studies, it is obvious that potential bias due to the use of Delphi panels should be minimised. In this manuscript we present an alternative data source for modelling studies: the cross-sectional study. Data from such studies can be used to yield costs and utilities for Markov health states. The overall combined design may be considered a hybrid between a naturalistic prospective study and a modelling study by maximising the pros and minimising the cons of both types of design, including an increase of external validity. This hybrid design is based on bridging the probabilities derived from the literature and clinical trials with information on costs and utilities from a cross-sectional study. This design also has logistical advantages, namely a shorter required study duration compared with prospective naturalistic studies for chronic diseases. This combined design was illustrated using a Markov model for Parkinson's disease.