A cross‐cohort cognitive composite for tracking changes in pre‐clinical and prodromal Alzheimer's disease: Neuropsychiatry and behavioral neurology/assessment/measurement of neuropsychiatric/behavioral and psychological symptoms. (7th December 2020)

• Record Type: Journal Article
• Authors: Giorgio, Joseph; Tanna, Ankeet; Jagust, William J.; Baker, Suzanne L.; Landau, Susan M.; Tanaka, Tomotaka; Chen, Christopher P.; Reilhac, Anthonin; Malpetti, Maura; Rowe, James B.; O'Brien, John T.; Kourtzi, Zoe
• Is Part Of: Alzheimer's & dementia. Volume 16(2020)Supplement 6
• Page Start: n/a
• Page End: n/a
• Publication Date: 2020-12-07
• Subjects: Alzheimer's disease -- Periodicals; Alzheimer Disease -- Periodicals; Dementia -- Periodicals; Démence; Maladie d'Alzheimer; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 616.83
• DOI: 10.1002/alz.037928
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: The primary goal of large global studies investigating Alzheimer's Disease (AD) is to make meaningful and robust conclusions on cognitive outcomes. However, this is challenging due to the heterogeneous cognitive test batteries used across studies. Here, we use four international cohorts to derive robust and sensitive cross‐cohort cognitive composites. Methods: Using longitudinal neuropsychological data from four international cohorts 1 (NIMROD, n=90, fu=2+‐1.34years; ADNI, n=2168, fu=3.13+‐3.03years; NUS, n=186, fu=3.44+‐1.47years; BACS, n=188, fu=4.22+‐3.27years) we generated two AD‐related cognitive composites. For cognitively normal (CN) individuals at baseline we generated a Cross‐Cohort Preclinical Alzheimer's Cognitive Composite (CC‐PACC). For individuals with MCI at baseline we generated a Cross‐Cohort Prodromal Alzheimer's Cognitive Composite (CC‐ProACC). We harmonised neuropsychological data across cohorts using k‐nearest neighbour's imputation resulting in a complete set of 139 item level cognitive variables (Figure 1). Using PCA on real and imputed values across all cohorts we derived single scores for memory, executive function and general cognition. We combined these values into either the CC‐PACC or the CC‐ProACC following established methodology[1]. To show AD‐related change in cognition we used linear mixed effects models to test the interaction of cortical amyloid status and time from baseline. We investigate cohort effects testing the 3‐way interaction of cohort, cortical amyloid status and time. For the ADNI preclinical cohort, we contrasted AD‐related change in the CC‐PACC against established measures (PACC[1]). Results: Annual change in the CC‐ProACC from baseline was significantly different for individuals who were amyloid positive vs. amyloid negative (F(1, 2598)=97, β= ‐0.22, P<0.0001) ( Figure 2a), no significant differences were observed between prodromal‐AD cohorts (ADNI‐NIMROD‐NUS) (F(2, 2590)=2.33, P=0.1). Annual change in the CC‐PACC from baseline was significantly different for individuals who were amyloid positive vs. amyloid negative (F(1, 1827)=37.4, β=‐0.14, P<0.0001) ( Figure 2b), no significant differences were observed between preclinical‐AD cohorts (ADNI‐BACS) (F(1, 1823)=1.56, P=0.21). For the preclinical ADNI cohort we show the CC‐PACC was 1.9‐times more sensitive than the PACC for measuring AD‐related cognitive decline (Figure 3). Conclusions: Our approach unifies and consolidates neuropsychological data from four global cohorts by deriving robust cross‐cohort preclinical (CC‐PACC) and prodromal (CC‐ProACC) AD cognitive composites. We demonstrate that the CC‐PACC outperforms the gold standard measure of preclinical‐AD cognitive decline (PACC), with twice the sensitivity for cognitive decline. 1 COHORT, (n)number of individuals, (fu)mean duration of follow‐up+‐std. Reference: [1] Insel(2019), Neurology.