With the ongoing development of a range of treatments for dementia, there is a strong imperative to develop non-invasive methods of assessing surrogate endpoints of disease. With the eye as a visible part of the CNS, ocular OCT is a natural candidate. This paper reports a subgroup of the population-based Rotterdam Study investigating outcomes across a diverse range of conditions. The aim was to investigate the association between OCT parameters and prevalent and incident dementia in a population-based study. A large number of the total cohort were excluded from the analysis; there were relatively large exclusions for those who did not have an OCT or had ungradable imaging; exclusions for ocular comorbidities including retinal changes, glaucoma, high myopia, and optic nerve pathology; and for systemic reasons including history of stroke or other neurological conditions. Of the 7918 eligible patients, the analysis was performed on 3289 patients (42%) for the optic nerve head parameters, and 2998 patients (38%) for the macula inner retinal layer (38%).
Large exclusions may increase data quality and precision of patient phenotype, but they introduce a potential for bias that may alter the outcomes of analysis. The excluded were in fact significantly different from the included based on age, diastolic BP, use of blood pressure lowering medications, presence of diabetes mellitus, cholesterol parameters, current smoking status and education status.
An association between Ganglion Cell-Inner Plexiform layer thickness and prevalent dementia was reported. However, the result was borderline at a p = 0.05 level in a simplified model and non-significant in the full model at this level. The longitudinal analysis was based on a very small number of individuals who developed new dementia throughout the study period. A thinner retinal nerve fiber layer at baseline was associated with increased risk of developing dementia over the study period (adjusted HR, 1.43 [95% CI, 1.15-1.78]) with a p < 0.001. This was only found with the retinal nerve fiber layer, not with other inner retinal layers.
The prevalence data has to be interpreted with some caution given the large number of exclusions and the p value. While a p value cut-off of 0.05 is fine for a small clinical study, it is rather too generous for a study of 3000 subjects. Borderline significance at this level suggest at least a great deal of noise in the data signal and at most a chance finding. The longitudinal data has a more appropriate p value and a good rate of follow-up, making these exclusions are less relevant. There does appear to be an association between a thinner retinal nerve fiber layer at baseline and developing dementia, but this would need to be confirmed in larger, appropriately designed studies.
Borderline significance at this level suggest at least a great deal of noise in the data signal and at most a chance finding
The eye is a part of the CNS and so it remains very plausible that OCT can inform on CNS disease, however, further studies will be required to clarify more precisely the nature of that information.