New research suggests that reliance on the standard test Alzheimer's Disease Assessment Scale—Cognitive Behavior Section (ADAS-Cog) to measure cognitive changes in Alzheimer’s patients is a bad idea. The test is the most widely used measure of cognitive performance in clinical trials.
Using a sophisticated method of analysis ("Rasch analysis"), analysis of ADAS-Cog data from the AD Neuroimaging Initiative (675 measurements from people with mild Alzheimer's disease, across four time points over two years) revealed that although final patient score seemed reasonable, at the component level, a ceiling effect was revealed for eight out of the 11 parts of the ADAS-Cog for many patients (32-83%).
Additionally, for six components (commands, constructional praxis, naming objects and fingers, ideational praxis, remembering test instructions, spoken language), the thresholds (points of transition between response categories) were not ordered sequentially. The upshot of this is that, for these components, a higher score did not in fact confirm more cognitive impairment.
The ADAS-Cog has 11 component parts including memory tests, language skills, naming objects and responding to commands. Patients get a score for each section resulting in a single overall figure; different sections have different score ranges. A low total score signals better cognitive performance; total score range is 0-70, with 70 being the worst.
It seems clear from this that the test seriously underestimates cognitive differences between people and changes over time. Given that this is the most common cognitive test used in clinical trials, we have to consider whether these flaws account for the failure of so many drug trials to find significant benefits.
Among the recommended ways to improve the ADAS-Cognitive (including the need to clearly define what is meant by cognitive performance!), the researchers suggest that a number of the components should be made more difficult, and that the scoring function of those six components needs to be investigated.