The ADAS-cog in Alzheimer's disease clinical trials: psychometric evaluation of the sum and its parts.
Academic Article
Overview
abstract
BACKGROUND: The Alzheimer's Disease Assessment Scale Cognitive Behavior Section (ADAS-cog), a measure of cognitive performance, has been used widely in Alzheimer's disease trials. Its key role in clinical trials should be supported by evidence that it is both clinically meaningful and scientifically sound. Its conceptual and neuropsychological underpinnings are well-considered, but its performance as an instrument of measurement has received less attention. Objective To examine the traditional psychometric properties of the ADAS-cog in a large sample of people with Alzheimer's disease. METHODS: Data from three clinical trials of donepezil (Aricept) in mild-to-moderate Alzheimer's disease (n=1421; MMSE 10-26) were analysed at both the scale and component level. Five psychometric properties were examined using traditional psychometric methods. These methods of examination underpin upcoming Food and Drug Administration recommendations for patient rating scale evaluation. RESULTS: At the scale-level, criteria tested for data completeness, scaling assumptions (eg, component total correlations: 0.39-0.67), targeting (no floor or ceiling effects), reliability (eg, Cronbach's α: = 0.84; test-retest intraclass correlations: 0.93) and validity (correlation with MMSE: -0.63) were satisfied. At the component level, 7 of 11 ADAS-cog components had substantial ceiling effects (range 40-64%). CONCLUSIONS: Performance was satisfactory at the scale level, but most ADAS-cog components were too easy for many patients in this sample and did not reflect the expected depth and range of cognitive performance. The clinical implication of this finding is that the ADAS-cog's estimate of cognitive ability, and its potential ability to detect differences in cognitive performance under treatment, could be improved. However, because of the limitations of traditional psychometric methods, further evaluations would be desirable using additional rating scale analysis techniques to pinpoint specific improvements.
