ISCTM 2012 Scientific Meeting – Metric-Based Measurement to Assess Cognitive Outcome in AD Trials Summary
22 Feb 2012
The Working Group titled “Metric-Based Measurement to Assess Cognitive Outcome in AD Trials” (Working Group 7) convened recently at the 8th Annual ISCTM meeting in Washington, DC to continue work towards defining a new cognitive endpoint for Alzheimer’s disease clinical trials as an alternative to the ADAS-Cog.
The overall goals of the Working Group are to:
1.Develop a cognitive algorithm for use in MCI trials that is both based on the ADAS-Cog and expands beyond it, by adding more difficult and relevant cognitive questions from pre-existing assessments
2.Expand the difficulty of cognitive measures for prodromal and mild AD clinical trials
3.Develop tests that are relative to the condition of interest and that are able to adequately discriminate between different cognitive abilities in people enrolled in MCI clinical trials
4.Apply the approach and methods for cognitive assessment to failed trials to determine if a signal can be detected or proof of concept confirmed
The ADAS-Cog is the most commonly used scale in Alzheimer’s disease (AD); it was developed in the early 1980’s as a quantitative conceptualization of the cognitive domains affected in AD specifically for use in clinical trials in people with this disease. The scale was validated, according to the standards that existed at the time, in a single, small study that included only 27 patients with AD and 28 age-matched elderly controls, and was designed with questions appropriate for a more severely affected population of patients than that for which it is currently most often used. Despite these limitations on construct validity, the ADAS-Cog has been the primary cognitive measure for all approved drugs for mild to moderate Alzheimer’s disease and is generally included in trials for new Alzheimer’s drugs. The most commonly used version of the ADAS-Cog includes 11 items assessing four cognitive domains (memory – 3; language – 5; praxis – 2; orientation – 1). Scores range from 0 to 70, with higher scores indicating more cognitive impairment. The ADAS-Cog items were thought to be structured from less difficult to more difficult items, but item analysis has shown this to not be true for all of the subscales. Nonetheless, the ADAS-Cog has been, and continues to be, used in clinical trials across the cognitive continuum of people affected by Alzheimer’s disease. And, given experience and the issues outlined above, among others, a major question still looms regarding the sensitivity of the ADAS-Cog to detect improvement or decrement in cognition in people with mild cognitive impairment (MCI) or early Alzheimer’s disease, since the scale does not represent the cognitive profile in these populations. Can this issue be remediated by including additional questions or new subscales that better reflect the cognition of the subjects in current trials?
Determining an answer to these questions comes in part from analyzing the scores from ADAS-Cog data collected in people with mild to moderate AD (M2M AD) and mild cognitive impairment (MCI). Data from the AD neuroimaging initiative (ADNI), an observational cohort, was reviewed. In people with M2M AD, ADAS-Cog scores ranged from 6 to 57 and MCI scores ranged from 0 to 37. Thus, in MCI clinical trials only one-half of the scale is being utilized; easier questions are not informative and the ceiling effects seen on the harder items reflect this observation.
Given this above preliminary data and knowledge of the other cognitive assessment scales concurrently tested within ADNI, the Working Group has concluded that it may be possible to eliminate items with ceiling effects from the ADAS-cog and add in items from other scales to expand the range cognitive range to generate a tool more sensitive to milder cognitive decline. By using a Guttman analysis to rate the hierarchy of each item a rationale for eliminating too easy items is generated; if the item can’t be rated, then it is not useful. Using a Guttman analysis, it is apparent that performance-based items included in the ADAS-Cog are more relevant that impression- or interpretation-based items. Further, a Rasch analysis of the ADAS-Cog, which generates category probabilities for each item as a measure of scoring function, can identify potential gaps in the domains covered by the ADAS-Cog and provide explicit guidance on how to fill the gaps. For example, according to Rasch analyses, adding delayed word recall and number cancellation to the ADAS-Cog in an effort to boost sensitivity for detecting cognitive change in early populations (i.e., ADAS-Cog 13) does not fill the item gaps for utilization of this scale in MCI clinical trials.
Although the ADAS-Cog is the current ‘gold standard’ for measuring cognition in Alzheimer’s disease, it is not required for registration, which could suggest other endpoints for primary analysis of cognition. Regulatory bodies (FDA and EMA) do not mandate which outcome measures must be used in registration trials, but optimization and use of a new cognitive endpoint will require FDA and EMA agreement. In order to smooth the path towards acceptance of a novel cognitive endpoint by health authorities it will be of key significance to have these regulatory agencies intimately involved. Of importance is the fact that cognitive endpoints are not PROs (Patient Reported Outcomes) and as such a cognitive endpoint based on the ADAS-Cog, which is validated in Alzheimer’s disease, should seen differently from a regulatory perspective; there is a significant amount of neuropsychological data to support use of the individual ADAS-Cog items and other items in the other cognitive assessments in ADNI and elsewhere. The neuropsychological data tells us that item sensitivity for the ADAS-Cog is a product of the cognitive ability of the patient. Developing a single measure to span the spectrum of the diagnosis from MCI to moderate AD is unlikely.
The Working Group’s approach to this is generation of an item bank with stage-specific items that can be selected based on the disease severity of the population being tested such that they may provide a reasonable means of avoiding ceiling and floor effects commonly observed with the ADAS-cog. Key to this approach is the understanding that the item bank depends heavily on the baseline disease severity and rate of deterioration over time; thus, each item may require additional sub-tests that can measure and detect change as the patient progresses.
Many questions remain unanswered and the Working Group has agreed to continue its review work. Item analysis is progressing through work funded by the Foundation for the NIH, which funds ADNI. Ultimately, the working group’s main goal is to review and give advice and feedback on a cognitive item bank that allows for both selection of items relative to the baseline severity and adequate sensitivity to detect change over time. One potential outcome of this work is that it may lead to re-analysis of promising new compounds that have been shelved due to lack of efficacy when measured with the standard ADAS-Cog.