OBJECTIVES: To model disease progression across multiple domains in Alzheimer's disease (AD) patients treated with a cholinesterase inhibitor in clinical practice. METHODS: 435 AD patients starting treatment with donepezil in 11 centers in Sweden were followed up to 3 years. In 6-month intervals data was collected on cognitive function (ADAS-Cog) physical function (IADL and PSMS scales), care setting and resource utilization. Regression modelling was used to identify determinants of disease progression rates and to establish equations predicting progression across multiple domains. A dynamic panel approach was used to model the 6 months change in cognitive function. For physical function a random-effects model was fitted using ADAS-cog and lagged ADAS-cog as explanatory variables. In both models other patient characteristics (e.g. sex, age, disease duration and ApoE-genotype) were included when significant. RESULTS: The progression in ADAS-cog was estimated to increase with higher progression in the previous 6 months period, i.e. a one point higher progression in the previous period would translate into 0.4 points higher progression in the present cycle. Also, patients having at least one ApoE e2 but no ApoE e4 allele were estimated to have about 2 points higher progression. Both IADL and PSMS scores were estimated to decrease with higher present and lagged ADAS-cog scores (between 0.03 and 0.06 points for each ADAS-cog point) and higher disease duration (0.14 to 0.2 points for each year). Also, male patients were estimated to have 0.46 points lower IADL scores. CONCLUSION: The estimated regression functions can be used in a regression model for simulation of long term disease progression. Adding treatment effects and resource utilization linked to disease severity this enables a dynamic framework for economic evaluation of any treatment intervention.