Abstract
BACKGROUND: Global cognitive metrics are useful for tracking cognitive aging, yet there is no consensus on how to best define and measure global cognition. The aim of this study was to compare scoring methods for global cognition against validity indicators. METHOD: Using harmonized factor scores of four cognitive domains (memory, executive function, language, visuospatial) in the United States and Korea (Table 1), we compared scoring methods for global cognition with validity indicators of cognition, functioning, and structural neuroimaging at baseline. Scoring approaches included combining cognitive domains with mean z‐scoring, congeneric confirmatory factor analysis (CFA; maximum likelihood robust estimation), principal component analysis (PCA), and weighting domains based on the Mini‐Mental Status Examination (MMSE(Composite)). Correlations (r) were used to compare composites against continuous validity indicators. The Clinical Dementia Rating Scale Sum of Boxes (CDR‐SB) and MMSE were used to measure cognition and functioning. Neuroimaging measures included brain volume relative to cerebrospinal fluid (BV/CSF) and intracranial volume‐normalized measures of brain volume (BV/ICV), ventricular volume (VV), frontal lobe volume (FL), and medial temporal lobe volume (MTL). Receiver operating characteristic (ROC) curves assessed the ability to detect cognitive impairment (CDR‐Global≥0.5). RESULT: Cognitive composites were highly intercorrelated (Figure; r range:0.94,1.00), correlated with CDR‐SB (r range:‐0.72,‐0.65) and MMSE (r range:0.75,0.77), and detected cognitive impairment in the good range (AUCs=0.84‐0.88). Cognitive composites were similarly correlated with neuroimaging metrics, with differences of at most r = 0.02 across comparisons, except for FL, where MMSE(Composite) had an r 0.04 lower than the others. Composites yielded stronger correlations with most neuroimaging metrics compared to standalone measures except for BV/ICV with CDR‐SB (r = ‐0.27). MTL and VV demonstrated the strongest associations with cognitive composites and other indicators of cognition/functioning. CONCLUSION: Regardless of scoring method, global cognitive composites were similarly related to validity indicators of cognition/functioning and structural neuroimaging in a harmonized sample. MMSE(Composite) showed the strongest relationships with CDR but was less strongly associated with frontal lobe volume. Composites demonstrated stronger relationships with structural neuroimaging compared to standalone measures. Future work aims to evaluate similarities and differences between cohorts and evaluate relationships longitudinally.