Association of Inner Retinal Thickness with Prevalent Dementia and Brain Atrophy in a General Older Population: The Hisayama Study

PURPOSE: To assess the association of inner retinal thickness with prevalent dementia and regional brain atrophy in a general older population of Japanese. DESIGN: Population-based, cross-sectional study. PARTICIPANTS: A total of 1078 residents aged 65 years or older who participated in an eye examination, a comprehensive survey of dementia, and brain magnetic resonance imaging scanning in 2017. METHODS: The thicknesses of the inner retinal layers, namely, the ganglion cell-inner plexiform layer (GC-IPL) and retinal nerve fiber layer (RNFL)—were measured by swept-source OCT (SS-OCT). The association of these retinal thicknesses with the risk of the presence of dementia was estimated using restricted cubic splines and logistic regression models. Regional brain volumes were estimated separately by applying 2 different methods: voxel-based morphometry (VBM) and analysis by FreeSurfer software. The associations of GC-IPL and RNFL thickness with each brain regional volume were analyzed using multiple regression analysis. MAIN OUTCOME MEASURE: Prevalent dementia and regional brain atrophy. RESULTS: Among the study participants, 61 participants (5.7%) were diagnosed with dementia. The likelihood of the presence of dementia significantly increased with lower GC-IPL thickness after adjusting for potential confounders (odds ratio, 1.62 [95% confidence interval, 1.30–2.01] per 1 standard deviation decrement in the GC-IPL thickness), but no significant association was observed with RNFL thickness. In the VBM analyses with the multivariable adjustment, lower GC-IPL thickness was significantly associated with lower volume of known brain regions related to cognitive functions (i.e., the hippocampus, amygdala, entorhinal area, and parahippocampal gyrus) and visual functions (i.e., the cuneus, lingual gyrus, and thalamus). Meanwhile, the volume of the thalamus significantly decreased with lower RNFL thickness, but none of the brain regions related to cognitive function exhibited a volume change in association with RNFL thickness. The sensitivity analysis using FreeSurfer analysis also showed that lower GC-IPL thickness was significantly associated with lower regional brain volume/intracranial volume of the hippocampus, amygdala, cuneus, lingual gyrus, and thalamus. CONCLUSIONS: Our findings suggest that the measurement of GC-IPL thickness by SS-OCT, which is a noninvasive, convenient, and reproducible method, might be useful for identifying high-risk individuals with dementia.