New regression models for long-term cognitive change in Donepezil treated Alzheimer patients.

Objectives: To analyse and present regression models that can describe the long-term change in cognitive function (MMSE and ADAS-cog) of patients treated with donepezil for three years (n=435). Methods: The Swedish Alzheimer Treatment Study (SATS) is an open, 3-year, multicentre study in a routine clinical setting. Patients with the diagnosis of Alzheimer’s Diseases received the cholinesterase inhibitor donepezil. The patients were assessed with several rating scales including MMSE, ADAS-cog at baseline and every 6 months for a total period of three years. In the first model we used a 6-month interval method. A regression model was fitted from the variables change in score between visits and the baseline score. For the 269 patients with at least 4 assessments each we also calculated a longitudinal regression model for the dependent variable ADAS-cog score at different intervals. In addition, we compared the scores from our model with those predicted by the Stern equation. Results: There was a significant quadratic component between baseline scores on MMSE and ADAS-cog and 6 months mean cognitive change (p<0,05). The 6 months rate of cognitive change in donepezil treated patients was strongly dependent on the baseline scores. The longitudinal regression model for the 269 patients showed that multiple linear regression analysis could be used to predict ADAS-cog score at different intervals (p<0,001). A different slope was observed between our model and the Stern equation. Conclusions: Regression models can predict the outcome of cognitive changes in donepezil treated patients. We also observed different outcomes compared to previous models concerning untreated patients.