To quantify the contribution of macula total retinal thickness and peripapillary retinal nerve fiber layer (pRNFL) thickness towards estimating hippocampal and regional brain volumes.

With atrophy and Alzheimer disease pathology detected in the retina as well as the brain, measuring retinal thickness using optical coherence tomography imaging may help estimate the volume of diagnostically and prognostically important brain regions to aid early disease detection.

Participants with either healthy aging, mild cognitive impairment or Alzheimer disease from the Ontario Neurodegenerative Disease Research Initiative (ONDRI) cohort were included. Measures included hippocampal and normal appearing grey and white matter volumes of four brain lobes and remaining basal ganglia / thalamus region, and macula total retinal thickness and pRNFL thickness. Canonical correlation analysis (CCA) measured the correlation of brain and retinal measures. Brain measures were estimated using retinal measures adjusted for head size, age, sex, diagnosis and years of education using partial least squares (PLS) and ordinary least squares (OLS), with effect sizes reported as standardized regression coefficients and f ².

134 participants were included. Non-fovea macula explained about 7% of variability of a latent variable of brain measures based on CCA commonality analysis. Retinal measures only weakly estimated brain measures (f ² typically<0.05 or negative) using PLS or OLS, including for diagnostically and prognostically important brain regions like the hippocampus and temporal lobe. Standardized OLS coefficients for hippocampal volume were small and about -0.05 to +0.05. Effect sizes were larger in women, for example when using OLS to estimate total brain volume from inferior pRNFL (f ²=0.08, std. beta=0.14, p<0.01) or temporal lobe grey matter from total pRNFL (f ²=0.03, std. beta=0.15, p<0.05).

Macula total retinal thickness or pRNFL thickness on their own appear to have a limited ability to estimate hippocampal or large regional brain volumes at baseline.