Clinical Manifestations.

Abstract

Balancing the need to standardize cognitive measures with the need to evolve alongside advances in scientific knowledge is an inherent challenge for longitudinal studies of Alzheimer disease (AD). The Uniform Data Set (UDS) offers a solution for this in cognitive batteries and is implemented across Alzheimer Disease Research Centers (ADRCs). Using equipercentile equating, we combined data spanning two versions of the UDS and derived domain-specific and general cognitive composites. We compared the relationships of these measures with common biomarkers of AD, as well as their statistical power to detect cognitive change in clinical trials. Using longitudinal data from the Knight ADRC, domain-specific cognitive composites were created using an exploratory factor analysis. The resulting four factors clustered episodic memory, semantic memory, working memory, and attention/processing speed tasks (Figure 1). We also computed a commonly-used preclinical Alzheimer cognitive composite (PACC), an alternative Knight-ADRC-PACC, which did not include MMSE, and a global composite. Relationships between each cognitive composite and neuroimaging metrics of amyloid, tau, atrophy, and clinical symptoms, were compared cross-sectionally and longitudinally (Figure 2). We also performed power analyses of all composites to calculate necessary sample sizes to observe a medium-sized effect in a four-year clinical trial. Overall, general metrics of cognition slightly outperformed domain-specific composites, although the working memory composite consistently reflected a poor relationship with AD biomarkers (Figure 2). Power analyses revealed that the global, Knight-ADRC-PACC, and attention/processing speed composites would require the smallest sample sizes to detect cognitive change in a clinical trial, while the ADCS-PACC, required two to three times as many individuals (Figure 3). We found that data-driven, domain-specific, composites of cognitive performance can complement already-established general cognitive composites, with the possible exception of the present working memory composite. In line with prior work, we highlight that attention and processing speed may be sensitive metrics of cognitive change in AD. The inclusion of the MMSE, such as in the ADCS-PACC, may greatly reduce the sensitivity and statistical power of summary cognitive composites. Our results have important implications for the development of cognitive outcome metrics for future clinical trials.