Magnetic Resonance Imaging Measures Predict Cognitive Decline in Taiwanese Elderly

Background: The proportion of elders in Taiwan increase rapidly and thus age-related health issues (e.g., dementia) become an important topic in the elderly. Brain structural changes begin at the early stage of dementia, and magnetic resonance imaging (MRI) technique can identify these subtle changes. Previous studies using MRI on brain regions of interest, and thus the remaining regions are less explored. Therefore, this study aims to explore the association between MRI measures and cognitive decline in Taiwanese elders. Methods: This cohort study recruited 316 participants aged 65 or older from the annual Elderly Health Checkup Program at National Taiwan University Hospital from 2011 to 2013 (baseline) and followed up from 2013 to 2015. The outcome of this study is global and domain-specific (memory, verbal fluency, executive function and attention) cognitive decline from baseline to follow-up. MRI scans were performed at baseline with a total of 114 volume variables and 68 thickness variables. Factor analysis was performed to identify cognitive factors. Slice inverse regression was performed for dimension reduction of MRI variables to identify MRI factors. Multivariable linear and logistic regressions were used to assess the association between MRI measures and cognitive decline. Stratified analyses were performed by important confounders, which included age groups (<75 and ≧75 years old), sex, and apolipoprotein E (APOE) ε4 status. Results: Four cognitive factors (memory, verbal fluency, executive function and attention factors) were identified by factor analysis. One unit increase in MRI factors 1 (including limbic, insula and frontal regions) protected against the decline of global cognition (MoCA-T, β= 0.002); one unit increase in MRI factor 3 (including limbic and frontal regions) showed similar finding [MoCA-T, adjusted odds ratio (AOR) = 0.75]. One unit increase in MRI factors 3 (including limbic, frontal and temporal regions) was associated with decline of memory factor (β= -0.34, AOR = 1.79); one unit increase in MRI factor 2 (including parietal, frontal and temporal regions) showed similar finding (AOR = 1.59). One unit increase in MRI factor 1 (including limbic and temporal regions) was associated with the decline of verbal fluency factor (β= -0.20). While one unit increase in MRI factor 3 (including frontal and temporal regions) protected against the decline of verbal fluency factor (β= 0.18, AOR = 0.58). One unit increase in MRI factors 1 (including frontal and temporal regions) was associated with the decline of executive function factor (AOR = 0.65). One unit increase in MRI factors 1 (including accumbens, temporal and frontal regions) protected against the decline of attention factor (β= 0.29, AOR = 0.49), while one unit increase in MRI factors 3 (including 5th ventricle, frontal and temporal regions) was associated with decline of attention factor (β= -0.11). Conclusion: This study used hypothesis-free approach to identify MRI markers. Most of our findings are consist with previous studies (e.g., frontal, temporal, and limbic regions) except limbic, accumbens, ventricles, pericalcarine and pars orbitalis regions have not been related to early cognitive decline in non-demented elderly previously. Finding remain significant especially in older group (ages≧75), women, and APOE ε4 carriers. Our findings are useful for early detection and prevention of dementia in the elderly.