Abstract Purpose To investigate the association between mid-peripheral capillary free zones (CFZs) and retinal structural and functional metrics in diabetics without diabetic retinopathy (DR). Methods This cross-sectional study included 45 eyes from 28 diabetics without DR and 46 eyes from 31 controls (mean age in both groups, 59 years). Macular optical coherence tomography (OCT) scans were acquired for retinal nerve fibre layer (RNFL) and ganglion cell layer thickness measurements. Thickness measurements were obtained using the Early Treatment of Diabetic Retinopathy Study grid. Retinal mid-peripheral CFZs were computed from OCT angiography images using custom MATLAB software. Retinal function was evaluated as a pilot exploratory objective using full-field flash electroretinography. Correlations between mid-peripheral CFZs and retinal structure and function were assessed using linear mixed-effect models, accounting for the association between eyes, while receiver operating characteristic curves were used to compare the multimodal models. Results Larger periarteriole CFZs were associated with thinner inner inferior RNFL thickness ( β = −0.48, p = 0.03) in diabetics without DR. Functionally, there was no significant association between the mid-peripheral CFZs and ERG parameters ( p > 0.05) in the no DR group; these findings should be interpreted with caution given the pilot nature of the functional data. The multimodal model of vascular and structural parameters had a modestly improved area under the curve (AUC) and specificity compared to the model of vascular parameters alone (AUC = 0.85 versus 0.83, specificity = 0.65 versus 0.54, respectively). Conclusions These findings demonstrate that enlarged mid-peripheral periarteriole CFZs are associated with thinner RNFL in diabetics without clinical retinopathy. The multimodal model of vascular and structural metrics showed modestly improved diagnostic ability. This study shows early novel retinal vascular and neural associations in diabetics without clinical retinopathy and demonstrates the potential utility of a multimodal model for discriminating this group from healthy controls.
Jonnadula et al. (Wed,) studied this question.