Data requirements in a model of the natural history of Alzheimer's disease.

The "natural history" of Alzheimer's disease (AD) is discussed in terms of the data required in a general, discrete-time, non-homogeneous Markov model. The proposed model differs from similar models reported in the literature because mortality is disaggregated into AD-specific mortality and competing mortality due to other causes. Data are reviewed from the literature for AD incidence, and rates of disease progression and mortality. We conduct a preliminary sensitivity analysis using the reviewed data as base-case. The model shows that survival is sensitive to the modelling assumptions concerning mortality. This observation could have important consequences for studies that assess the cost of care following therapeutic interventions.