Embic Shares New Study Outcome at the 14th Annual CTAD Conference

Embic’s newest finding was presented at the 14th annual Clinical Trials on Alzheimer’s Disease (CTAD) conference. The study further validates our mathematical model’s ability to detect pre-clinical AD with item response data from a standard wordlist memory test, and the additional signal detection theory parameters to further characterize cognitive processes. Read full presentation here.

Detection of Pre-clinical Alzheimer’s Disease with Simultaneous Modeling of Underlying Cognitive Processes in Recall and Recognition Tests
Bock JR [1]; Lee MD [2]; Shankle WRS [1-3]; Hara J [1,3]; Fortier D [1]; Mangrola T [1]
[1] Embic Corporation, Newport Beach, CA, USA
[2] Dept. of Cognitive Sciences, University of California at Irvine, Irvine, CA, USA
[3] Pickup Family Neuroscience Institute, Hoag Memorial Hospital, Newport Beach, CA, USA

Abstract
Background: Wordlist memory (WLM) tests are commonly used to detect and monitor cognitive impairment due to Alzheimer’s disease (AD). While traditional scoring methods and analyses of WLM tests (e.g., summary scores) are effective at identifying dementia, they are insufficient for detecting earlier stages of progressive decline, such as pre-clinical AD. This is partly due to the fact that summary scores of WLM test tasks (e.g., immediate and delayed free recall and delayed recognition tasks) do not contain sufficient information to make these more subtle distinctions in cognition. Our previous work demonstrated that using item response data from immediate and delayed free recall tasks, along with a hierarchical Bayesian cognitive processing (HBCP) model, enables greater information extraction from WLM tests, sufficient to characterize differences among subjects in varying stages of severity across the progression of AD. This was achieved by quantifying unobservable (latent) cognitive processes that underlie learning and recall, including encoding, storage, and retrieval of WLM test items. In further work, the HBCP model was applied to immediate and delayed free recall task data from Alzheimer’s Disease Neuroimaging Initiative (ADNI) subjects who were cognitively normal at baseline, a subgroup of which would remain normal, and another which would progress to impairment during follow-up. Using only baseline assessment data, the model successfully characterized and demonstrated statistically significant differences between these two subgroups.

Objective: To expand our HBCP model to incorporate delayed recognition task data in addition to immediate and delayed free recall data, and to identify meaningful differences or similarity in individual cognitive processes between healthy, normal subjects and pre-clinical AD subjects.

Methods: Nine hundred fifteen (915) cognitive assessments, performed on 254 subjects from a community memory clinic between 2002 and 2019, were included in this study. All subjects were cognitively normal by clinical diagnosis, and were given the MCI Screen (MCIS), a battery of cognitive tasks, including multi-trial free recall of a wordlist (with three immediate and one delayed free recall tasks) and a delayed recognition task of the same wordlist (with the addition of foil words). Subjects were classified into two groups: a decliner group (subject n = 92; assessment n = 234), if the subject declined to amnestic mild cognitive impairment or dementia within 2 years; and a non-decliner group (subject n = 162, assessment n = 681), if the subject remained cognitively normal or only subjectively cognitively impaired for at least 2 years. The HBCP model was expanded to include signal detection theory (SDT) parameters, discriminability and criterion, for measurement of recognition task data. This was done in such a way that the existing cognitive processing parameters benefit from both the free recall and recognition task information. We examined cognitive processing parameter posterior samples to characterize patterns in cognitive performance, and we performed Bayes factor analyses of parameter mean differences between decliner and non-decliner groups.

Results: decliner and non-decliner groups. Subjects in the decliner group demonstrated significantly lower encoding parameters for WLM task items in both early- and late-list positions, with unique patterns across specific encoding parameters. However, moderate evidence for statistical similarity between decliner and non-decliner groups was found for the SDT parameters of criterion and discriminability.

Discussion: ognitive processes. Lower encoding processes in the decliner group corroborates our previous findings, and similar levels in subject criterion and discriminability between groups aligns with existing literature pertaining to recognition task performance resilience to cognitive decline. Identifying specific cognitive processes which change before observable cognitive decline occurs, and differentiating them from processes that do not change until later in the progression timeline, is of great value in clinical and research trial settings, both for early detection of AD and ongoing monitoring of cognition during treatment.

Leave a Reply