Diabetic retinal neurodegeneration: should we redefine ‘retinopathy’ from diabetes?

The pathophysiology of diabetic retinopathy has been an area of intense research for decades. Although early investigations focused on clinically-manifesting retinal vascular abnormalities, emerging evidence using retinal image analysis, visual function tests, and translational models supports degeneration of the inner retina early in diabetes, termed diabetic retinal neurodegeneration (DRN), even before it can be seen as ‘diabetic retinopathy’ on fundus examination.¹ The coordinated structural, glial and vascular degeneration of the retinal neurovascular unit² accompanied by functional deficits (e.g. electroretinogram, dark adaptation, contrast sensitivity, color vision, and microperimetric and perimetric psychophysical abnormalities) as well as peripheral changes on ultrawide field imaging have been recently termed ‘diabetic retinal disease (DRD)’³ to better reflect detectable abnormalities induced by diabetes.

In this article, Lim and colleagues (CITATION OF ARTICLE*) utilize longitudinal spectral-domain optical coherence tomography (OCT) imaging of the optic nerve to demonstrate progressive thinning of the peripapillary retinal nerve fiber layer (pRNFL) in patients with diabetes without (no DR) or with mild to moderate nonproliferative diabetic retinopathy (NPDR) in comparison to age-matched controls. The authors demonstrate that patients with diabetes have about a 3-fold faster rate of pRNFL thinning relative to controls and that this rate is similar to those reported for patients with glaucoma (well done Forest plot in eFigure 3). Provocative to those who maintain that diabetes is primarily a retinal vascular disorder, progressive pRNFL loss was seen even in the group with diabetes but no clinically-apparent diabetic retinopathy.

The prospective study was well-executed and included a balanced control group for comparison with longitudinal follow up of three years. Notably, the study population included a diabetic cohort with early disease that might show slower progression of DRN and avoids any treatment as a confounder on DRN progression. Because of the relatively short study duration, the authors had limited ability to evaluate the influence of systemic factors such as longitudinal hemoglobin A1c on the rate of DRN progression, though they confirmed diabetes duration as a major association.¹

The rate of pRNFL thinning in this study was determined using statistical modeling that assumes a linear rate of change over time, but it is not known whether the rate of DRN is truly linear or instead follows a different curve of decline (e.g., logarithmic) or involves episodes of rapid decline followed by periods of stability (i.e., in patients with “quiescent” retinopathy). Consistent with this notion, those with NPDR had a thinner baseline pRNFL despite no statistically-significant difference in the rate of pRNFL loss between the no DR and NPDR cohort, suggesting that pRNFL loss for patients in the NPDR cohort may have preceded the study period. Although the reported rate of pRNFL loss was about 1 μm/year in the no DR and NPDR cohort, there was a wide 95% confidence interval (up to 14.8 μm/year and 18.0 μm/year for the no DR and NPDR cohort, respectively), suggesting a degree of variability in rate of pRNFL loss. The authors did not explore clinical features of the statistical outliers who may have influenced the average rate of pRNFL loss. It is possible, as has been shown in glaucoma “progressors” versus “non-progressors,” that there are distinctly different sub-populations of patients who have a rapid DRN versus those without.

This longitudinal, prospective study by Lim and colleagues contributes to mounting evidence in support of DRN in patients with diabetes regardless of clinically-apparent diabetic retinopathy. This work builds upon a meta-analysis concluding that preclinical diabetic retinopathy is associated with thinning of pNFL⁴ and further suggests that pNFL thinning can measure neural retinal damage as can macular OCT analysis, similar to work done with glaucoma patients.⁵ pNFL loss of about 1 μm/year in this study and a loss of macular NFL+GCL+IPL layers of 0.54 μm/year due to diabetes with little to no retinopathy¹ is remarkably similar to significant inner retinal loss seen in glaucoma. However, there is still a critical gap in knowledge between glaucomatous visual field deficits and relative lack of visual field defects in patients with longstanding diabetes.

This work strengthens the concept (now extending to those with Asian backgrounds in addition to those with European ancestry) that DRN may manifest before clinically apparent DR. Future studies could evaluate RNFL thinning in relation to loss of the corresponding radial peripapillary network using longitudinal OCT angiography, or similarly, ganglion cell layer thinning in the context of superficial capillary plexus changes.⁶ The clinical relevance of DRN remains to be explored, and future studies are needed to correlate these structural changes with functional tests such as static visual fields or microperimetry. Ultimately, the clinical identification of patients with rapidly-progressive DRN may drive the development and trial of neuroprotective agents for the prevention of DRN and presumed eventual visual loss. As we unravel the relationship between DRN and functional deficits in the context of what we traditionally know as diabetic retinopathy, a more clear picture of ‘diabetic retinal disease’ should come into view.