Background: Considering the scarcity of longitudinal assessments of reliability, there is need for a more precise
understanding of cognitive decline in Alzheimer’s Disease (AD). The primary goal was to assess longitudinal changes in
inter-rater reliability, test retest reliability and internal consistency of scores of the ADAS-Cog.
Methods: 2,618 AD subjects were enrolled in seven randomized, double-blind, placebo-controlled, multicenter-trials from
1986 to 2009. Reliability, internal-consistency and cross-sectional analysis of ADAS-Cog and MMSE across seven visits
Results: Intra-class correlation (ICC) for ADAS-Cog was moderate to high supporting their reliability. Absolute Agreement
ICCs 0.392 (Visit-7) to 0.806 (Visit-2) showed a progressive decrease in correlations across time. Item analysis revealed
a decrease in item correlations, with the lowest correlations for Visit 7 for Commands (ICC=0.148), Comprehension
(ICC=0.092), Spoken Language (ICC=0.044).
Discussion: Suitable assessment of AD treatments is maintained through accurate measurement of clinically significant
outcomes. Targeted rater education ADAS-Cog items over-time can improve ability to administer and score the scale.