Multimodal Imaging of the Initial Stages of Diabetic Retinopathy: Different Disease Pathways in Different Patients

Diabetic retinopathy (DR) is the leading cause of blindness among working adults in the United States (1). Furthermore, Narayan et al. (2) demonstrated that the prevalence of diabetes in the United States is expected to increase dramatically.

Our group has proposed three phenotypes of mild nonproliferative DR (NPDR) with different risks for development of vision-threatening complications (3). These phenotypes may be characterized by the predominance of one of three main disease pathways occurring in diabetic retinal disease: neurodegeneration, edema, and ischemia (4).

A forward-looking approach consistent with the concept of personalized medicine would be to develop quantitative assessments of these different disease pathways to enable early and individualized treatment (5).

We therefore used a noninvasive, multimodal imaging approach to examine the prevalence of different disease pathways in the initial stages of the disease. Neurodegeneration can be identified by thinning of the retinal tissue (retinal nerve fiber layer [RNFL] and ganglion cell layer plus inner plexiform layer [GCL–IPL]). Retinal edema may be characterized by increases in retinal thickness in the full retina and in individual layers by using spectral domain optical coherence tomography (SD-OCT). Finally, ischemia may be identified by vessel density metrics using OCT angiography (OCTA).

For this study, we examined eyes of patients with type 2 diabetes showing the initial stages of NPDR. All eyes underwent 7-field color fundus photography for classification by Early Treatment Diabetic Retinopathy Study (ETDRS) criteria, as well as SD-OCT and OCTA to quantify the prevalence of changes in retinal layer thickness and vessel density.

In this retrospective cross-sectional study of data from an observational study (ClinicalTrials.gov identifier: NCT03010397), participants with diabetes in the initial stages of NPDR were included from February 2016 to April 2018. The tenets of the Declaration of Helsinki were followed, and approval was obtained from the Institutional Ethical Review Board. Written informed consent to participate in the study was obtained from all individuals after all procedures were explained.

All participants underwent a full ophthalmological examination, SD-OCT and OCTA imaging using Cirrus AngioPlex (Carl Zeiss Meditec, Dublin, CA), and color fundus photography. A total of 142 eyes from 142 patients with diabetes, 1 eye per patient, were classified into three groups according to the ETDRS severity level: no or minimal NPDR, levels 10–20 (54 eyes); mild NPDR, level 35 (54 eyes); and moderate NPDR, levels 43–47 (34 eyes). The DR severity level was determined by two independent graders within the context of an experienced reading center and was based on the 7-field protocol using the ETDRS classification. The eye showing the more advanced ETDRS grading in any given patient was chosen to be the study eye.

Age, duration of diabetes, and hemoglobin A_1c (HbA_1c) levels were collected for each participant from their patient records. Visual acuity was measured for each eye using the ETDRS protocol and Precision Vision charts at 4 m.

Exclusion criteria included any previous laser treatment or intravitreal injections, or the presence of age-related macular degeneration, glaucoma, or vitreomacular disease and high ametropia (spherical equivalent greater than −6 and +2 diopters), or any other systemic disease that could affect the eye, with special attention for uncontrolled systemic hypertension (values outside normal range: systolic 70–210 mmHg and diastolic 50–120 mmHg) and history of ischemic heart disease.

A control group of 106 eyes from healthy control subjects (36 women and 25 men; mean [SD] age 47.6 [12.8] years) was used to compare with the study group.

The Macular Cube 512 × 128 acquisition protocol, consisting of 128 B-scans with 512 A-scans each, was used to assess the subjects’ central retinal thickness (CRT) and the average thickness value of the GCL-IPL, collected from the standard Cirrus examination reports. Segmentation of retinal layers was performed on the SD-OCT data using a segmentation software implemented by the Association for Innovation and Biomedical Research on Light and Image (AIBILI) (6). Segmentation results were reviewed by a masked grader. None of the eyes in this series of patients were observed to have cystoid macular edema or disorganization of the retinal inner layers that could have influenced the segmentation analysis.

Eyes with subclinical and clinical macular edema (center-involving macular edema) were identified following the reference values established by the DRCR.net for Cirrus SD-OCT (7,8).

OCTA data were collected by the AngioPlex device using the Angiography 3 × 3 mm acquisition protocol. This acquisition protocol consists of a set of 245 clusters of B-scans repeated four times, where each B-scan consisted of 245 A-scans over a 3 × 3 × 2 mm³ volume in the central macula. The AngioPlex eye tracking algorithm was used to reduce the effect of eye motion artifacts. Images were evaluated at the time of acquisition for quality, because signal strength greater than 7 was required, with minimal motion artifacts and no evidence of defocus or blur.

Vessel density metrics for the entire 3 × 3 mm² central macular area were computed for the superficial retinal plexus (SRP) and deep retinal plexus (DRP) by a proprietary automated software (Carl Zeiss Meditec, Inc., version 10.0.0.12787). Area and circularity index of the foveal avascular zone (FAZ) for the SRP were also computed using the same software.

Statistical analysis was performed with Stata 12.1 software (StataCorp, College Station, TX), and a P value of <0.05 was considered to be statistically significant.

Reference values were taken considering the mean and SD values of the healthy control population, taking the patient’s sex into consideration.

Changes were considered “absent” if the values were within the normal range of the healthy control groups, “possible” if different by more than 1 SD, and “definite” if different from normal values by more than 2 SDs.

Differences from the control population were categorized by ETDRS severity groups, and the corresponding 95% CI was computed using the Clopper-Pearson method.

Comparison of the ETDRS groups for categorical outcomes was conducted using the χ² test. Continuous outcomes were compared using an ANOVA test. To explore correlations, the Pearson coefficient and the respective significance were computed. ANCOVA of the study outcomes was performed, adjusted for baseline characteristics.

The distribution of characteristics of eyes/patients for each of the stages of the disease is presented in Table 1. No statistically significant differences were found between eyes/patients within different stages of the disease, except for HbA_1c levels that reflected an association between poorer metabolic control and more severe stages of the disease (Table 1).

The measurements of neurodegeneration, edema, and capillary dropout for each of the stages of the disease are presented in Table 2.

In univariate analysis, vessel density in the SRP and DRP was significantly different between ETDRS groups, with eyes in a more severe stage of the disease more likely to have reduced vessel density (P = 0.001 in the SRP and P = 0.020 in the DRP) (Fig. 1). The FAZ circularity index was also significantly reduced in ETDRS group 43–47 (P = 0.010).

No statistically significant differences between these three initial ETDRS groups were found in retina thickness (thinning or increased thickness) (Table 2).

In multivariate regression analysis, considering a model adjusted for age, sex, HbA_1c, visual acuity, and diabetes duration, vessel density remained significantly different between the ETDRS groups and showed significant associations with age (P < 0.001), sex (P = 0.028), and metabolic control (P = 0.034). No significant differences between the ETDRS groups were found in RNFL and GCL-IPL thinning (neurodegeneration) or in the inner nuclear layer (INL) and outer plexiform layer (OPL) thickness (edema) (Fig. 1).

Analysis of the variables representative of capillary dropout, retinal neurodegeneration, and retinal edema showed a wide range of values in each ETDRS grade, demonstrating that there are very different degrees of capillary dropout, neurodegenerative changes, and edema in eyes from different patients with the same retinopathy grade. The results are summarized in the Supplementary Tables.

Definite RNFL or GCL-IPL thinning (2 SD) was found in 24%, 28%, and 21% of the eyes in groups 10–20, 35, and 43–47, respectively (Table 3 and Supplementary Tables).

Increases in retina thickness were predominantly located in the INL or in the OPL. Definite increases were noted in 35% (19 of 54), 35% (19 of 54), and 38% (13 of 34) for the ETDRS groups 10–20, 35, and 43–47, respectively. These increases were significantly correlated with increases in retinal thickness in the full retina (INL: r = 0.77; P < 0.001; OPL: r = 0.48; P < 0.001) (Table 3 and Supplementary Tables).

When analyzing the data for the three ETDRS severity groups, definite (2 SD) decreases in retinal vessel density were present in 22% (12 of 54) of the eyes in ETDRS group 10–20, in 44% (24 of 54) in ETDRS group 35, and in 59% (20 of 34) in ETDRS group 43–47 (Table 3 and Supplementary Tables).

The vessel density in the SRP showed a strong correlation with the vessel density in the DRP (r = 0.77; P < 0.001) and a weakly positive correlation with FAZ circularity (r = 0.23; P = 0.006).

When we examined the data for correlations between the parameters that represent capillary dropout, retinal thinning, and edema, no association was found between these different alterations.

Analysis of the initial stages of diabetic retinal disease shows retinal neurodegeneration in one-quarter of patients. It appears to occur independently of the presence of edema and capillary dropout (9). Edema is present in approximately one-third of the eyes, possibly associated with an alteration of the blood-retinal barrier in the DRP (10,11). Capillary dropout, demonstrated by decreased vessel density occurring in both SRP and DRP, correlates with increases in severity of DR, as judged by ETDRS criteria. It may provide major clinical value by indicating the eyes at increased risk for progression to more severe stages of retinopathy. Previous reports by our group have indicated that metrics of vessel density detected changes in DR earlier in SRL than DRL (12). However, this study shows that the DRL is involved as much as the SRL. Correlation with severity of retinopathy was also found with HbA_1c, but not with age and sex, confirming that metabolic control is relevant for retinopathy progression.

Our findings may have major implications for the management of DR. Our observation that different eyes in the same ETDRS severity level show different predominant disease pathways confirms previous studies suggesting that eyes from different patients may have different phenotypes of disease progression (13).

This study shows that there are eyes with retinal neurodegeneration present from the earliest stages that do not show evidence of edema or ischemia. Other eyes show the presence of edema without neurodegeneration and in the absence of ischemia. Finally, there are eyes with evidence of ischemia identified by decreased vessel density, without neurodegeneration and edema. This phenotype, characterized by the presence of ischemia, appears to be the only one associated with increase in retinopathy severity.

One limitation of this study is the use of automated retinal layer segmentation analyses. These, however, were performed in retinas that remained structurally preserved with no evidence of cystoid changes, and they were reviewed by a masked grader.

In conclusion, eyes in the initial stages of diabetic retinal disease may show neurodegeneration, edema, and decreases in vessel density. These changes, however, occur to different degrees in different eyes, indicating that the predominant mechanism of retinal disease may be different in different patients.

Retinal thinning is an early finding that appears to occur independently of the presence of edema or capillary dropout. Edema may also be identified at the earliest stages of diabetic retinal disease, but its prevalence does not correlate with severity of the disease. Finally, decreases in vessel density involving both the DRP and SRP appear to be the only feature that shows clear association with increase in retinopathy severity, as identified by ETDRS level. Our findings support the concept that capillary dropout is a critical process in DR (14).

Metrics of retinal vessel density obtained with OCTA in a noninvasive manner, allowing repeated examinations, appear to identify the eyes/patients at a higher risk for increase in severity, thus promoting a larger role for precision medicine in the management of DR in individual patients.

Duality of Interest. M.D. is employed by Carl Zeiss Meditec. J.C.-V. reports grant from Carl Zeiss Meditec, outside the submitted work, and is a consultant for Alimera Sciences, Allergan, Bayer, Gene Signal, Novartis, Pfizer, Oxular, Roche, Sanofi, Vifor Pharma, and Carl Zeiss Meditec. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. I.P.M. and A.R.S. collected data and reviewed and edited the manuscript. D.A. and T.S. analyzed data and contributed to writing and editing the manuscript. L.M. assisted in the analysis and interpretation of the data. C.L. and M.D. reviewed and edited the manuscript. J.C.-V. analyzed data and wrote the manuscript. J.C.-V. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.