ObjectivesAbnormal retina structures, such as thinner retinal nerve fiber layer (RNFL), have been frequently reported in patients with Alzheimer's disease (AD). However, the association between RNFL and brain structures in cognitively normal adults remains unknown. We therefore set out to conduct a cross-sectional investigation to determine whether RNFL thickness is associated with brain structure volumes in nondemented older adults. MethodsWe measured RNFL thickness by optical coherence tomography and brain structure volumes by 3 T magnetic resonance imaging. Cognitive function was assessed using the Chinese version of Repeatable Battery for the Assessment of Neurological Status. Pearson correlation was initially employed to screen for the potential associations among RNFL thickness, brain structure volumes and cognitive function. And then, multivariable linear regression models were conducted to further examine such associations adjusting for possible confounding factors, including age, sex, years of education and the estimated total intracranial volume (eTIV). Results113 participants (≥ 65 years old) were screened and 80 of them (mean age: 68 ± 5.3 years; 48% male) were included in the final analysis. RNFL thickness in temporal quadrant was associated with medial temporal lobes volumes [unadjusted: r = 0.155, P = 0.175; adjusted: β = 0.205 (0.014, 0.383), P = 0.035], and especially associated with the hippocampus volume [unadjusted: r = 0.213, P = 0.062; adjusted: β = 0.251 (0.060, 0.435), P = 0.011] after adjusted for age, sex, years of education and eTIV. Moreover, it showed that RNFL thickness in inferior quadrant [unadjusted: r = 0.221, P = 0.052; adjusted: β = 0.226 (0.010. 0.446), P = 0.041] was significantly associated with occipital lobes volumes after the adjustment of age, sex, years of education and eTIV, and selectively associated with the substructure of lingual gyrus volume [unadjusted: r = 0.223, P = 0.050; adjusted: β = 0.278 (0.058, 0.487), P = 0.014]. In addition, average RNFL thickness was associated with the cognitive domain of visuospatial/constructional [unadjusted: r = 0.114, P = 0.322; adjusted: β = 0.216 (0.006, 0.426), P = 0.044] after the adjustment in these nondemented older adults. ConclusionsQuadrant-specific associations exist between RNFL thickness and brain regions vulnerable to aging or neurodegeneration in older adults with normal cognition. These findings would promote further investigations into using RNFL as a noninvasive and less expensive biomarker of neurocognitive aging and AD-related neurodegeneration.