P3-032: Subjective item variability in the ADAS-cog

Abstract

The Alzheimer's Disease Assessment Scale – Cognitive part is the most widely used cognitive outcome measure in current clinical research of Alzheimer's Disease. The most commonly used version of this scale contains 11 items – 8 of them assessed objectively, while the scoring of the other 3 is subjective in nature. All 3 subjective items – Spoken Language Ability, Word-Finding Difficulty in Spontaneous Speech and Comprehension – are scored solely on the basis of the rater's clinical observation. Rater variability between objective and subjective items is compared. ADAS-Cog rater training was conducted in 18 separate clinical trials that used the 11-item ADAS-Cog. Each of the training programs included the scoring of demonstration ADAS-Cog videos. 3039 raters completed the video scoring a total of 3891 times. (Some raters were tested in more than one protocol.) In assessing each trials' demonstration video scores, the percent agreement to the scoring mode was calculated for both the objective and subjective items. An aggregate count of modal agreement for objective and subjective items was also calculated. Overall, rater agreement to the mode was 69% for subjective items, and was 85% for objective items. The lowest project rate of modal agreement on subjective items was 35% and the highest was 92%. The lowest project rate of modal agreement on objective items was 72% and the highest was 96%. There is widespread variability regarding the scoring of subjective items on the ADAS-Cog. An effect size of 3.5 points on the 11-item ADAS-Cog has been shown to be statistically significant in previous Alzheimer's Disease research. Rater standardization, especially in regards to subjective assessments, should be considered for trials that will utilize the ADAS-Cog.