Rockwood_2008_Int.J.Geriatr.Psychiatry_23_207

BACKGROUND: Lacking long-term placebo-controlled trials of cholinesterase inhibitors (ChEIs) in Alzheimer's disease, how decline is modeled is crucial to interpreting the effects of long-term, open-label use. We aimed to: 1) understand changes in cognition and function in people taking galantamine; (2) identify treatment subgroups.
METHODS: This is an open-label long-term extension (up to 48 months) of two placebo-controlled clinical trials from Europe and Canada. At baseline, 240 patients with mild-moderate Alzheimer's disease at baseline had received galantamine continuously for 36 months. The trajectory of scores on the Alzheimer's Disease Assessment Scale-Cognitive Subscale (ADAS-Cog) and the Disability Assessment for Dementia (DAD) was modeled over 48 months. Goodness of fit was compared using normalized residuals and by calculating correlation coefficients between observed and fitted data. Iterative K-means clustering identified subgroups.
RESULTS: Of 240 patients, 51 (21%) withdrew between months 36-48 (nine died and 12 with adverse events). The mean ADAS-Cog worsened from 22.6 (mean)+/-8.6 (SD) at baseline to 31.3+/-3.1, with similar changes in the DAD. The logistic function fit better (ADAS-Cog norm of residuals=2.05, r=0.993; DAD=2.21, r=0.998) than the commonly used Stern equation (ADAS-Cog=4.22, r=0.944; DAD=7.51, r=0.986). K-means clustering identified three stable groups; least decline (n=82), persistent progression (n=75), and an intermediate group (n=82). Initial conditions were important: the mean ADAS-Cog score for those with least decline was 15+/-8, better than both the intermediate group (20+/-9) and persistent progressors (30+/-10; p<0.001).
CONCLUSIONS: Alzheimer's disease patients show variable long-term decline that is best modeled as a logistic function, not as linear decline. Over 48 months, about one-third of patients showed little decline, while one-third showed important decline.