Background/aims Split-spectrum amplitude decorrelation angiography for spectral-domain optical coherence tomography has enabled detailed, non-invasive assessment of vascular flow. This study evaluates choriocapillaris and retinal capillary perfusion density (CPD) in diabetic eyes using optical coherence tomography angiography (OCTA).
Methods Records of 136 eyes that underwent OCTA imaging at a single institution were reviewed. Eyes were grouped as non-diabetic controls (37 eyes), patients with diabetes mellitus (DM) without diabetic retinopathy (DM without DR, 31 eyes), non-proliferative diabetic retinopathy (NPDR, 41 eyes) and proliferative diabetic retinopathy (PDR, 27 eyes). Quantitative CPD analyses were performed on OCTA images for assessing perfusion density of the choriocapillaris and retinal plexus for all patients and compared between groups.
Results Eyes with NPDR and PDR showed significantly decreased choriocapillaris CPD compared with controls, while DM eyes without DR did not show significant change. Choriocapillaris whole-image CPD was decreased by 8.3% in eyes with NPDR (p<0.01) and decreased by 7.1% in eyes with PDR (p<0.01). Choriocapillaris parafoveal CPD was decreased by 8.9% in eyes with NPDR (p<0.01) and decreased by 8.2% in eyes with PDR (p<0.01). Compared with controls, only eyes with PDR showed significantly decreased retinal CPD, as well as significantly increased foveal avascular zone (FAZ) area. In those patients, retinal whole-image CPD was decreased by 9.7% (p<0.01), retinal foveal CPD was decreased by 20.5% (p<0.01) and retinal parafoveal CPD was decreased by 11.4% (p<0.01). FAZ area was increased by 50.9% (p<0.01).
Conclusions Choriocapillaris and retinal CPD are reduced in diabetic retinopathy, while FAZ area is increased in eyes with PDR. Vascular changes captured by new imaging modalities can further characterise diabetic choroidopathy.
- optical coherence tomography
Statistics from Altmetric.com
Vascular dysfunction in diabetes mellitus (DM) can lead to diabetic retinopathy (DR), which is one of the leading causes of visual loss worldwide, affecting approximately 35% of all patients with diabetes.1 Research has historically focused on injuries and changes in the retina, but more recently the focus has expanded to the choroid, the highly vascularised tissue layer that supplies the outer retina and that is abnormal in diabetic choroidopathy (DC).
Histological studies of patients with diabetes revealed pathological vascular changes in the choroid, including neovascularisation, capillary narrowing, atrophy and endothelium destruction.2 Early imaging using fluorescein angiography (FA) and indocyanine green angiography techniques revealed many abnormalities in the retinal and choroidal vasculature of diabetic eyes. For example, a ‘salt-and-pepper’ pattern of hypofluorescent and hyperfluorescent spots due to irregular filling of the choriocapillaris has been described in patients with non-proliferative diabetic retinopathy (NPDR), in contrast to the ground-glass pattern of healthy eyes.3
More recently, improvements in spectral-domain optical coherence tomography (SD-OCT) and enhanced-depth imaging OCT software have enabled better evaluation of the choroid structure. In healthy eyes, choroidal thickness (CT) has been shown to differ with age, axial length and time of day.1 Conclusions about CT in diabetes vary, with the majority of studies reporting decreased CT in eyes without diabetic macular oedema (DME).1 A pattern of increased central CT and progressive nasal and temporal thinning has also been reported in diabetes.4
While many studies of DC focus on CT, few reports address the vessel density changes in the choriocapillaris plexus using optical coherence tomography angiography (OCTA). The advent of OCTA and split-spectrum amplitude decorrelation angiography analysis has allowed for non-invasive, rapid in vivo imaging of detailed microvasculature at distinct depths.5 Using OCTA, Carnevali et al reported early vascular alterations of diabetic patients without DR compared with healthy individuals, but no difference in choriocapillaris vessel density between the two groups.6 In contrast, Dimitrova et al reported a decrease in both retinal and choriocapillaris density in diabetic patients without DR compared with healthy individuals using OCTA.7 Agemy et al reported that capillary perfusion density (CPD) across nearly all microvascular posterior layers was decreased in eyes with DR compared with healthy eyes.8
Capillary density reflects vascular tree integrity and may be useful as a measure to evaluate and monitor the evolution of diabetic ocular disease. In addition, changes in capillary perfusion might theoretically explain why patients with diabetes develop non-perfusion and ischaemia from the outer retina as opposed to inner retinal ischaemia. Herein, this study’s purpose was to assess choriocapillaris and retinal perfusion density changes in diabetic eyes with various levels of retinopathy versus controls using OCTA.
Materials and methods
A comprehensive chart review was performed to assess ophthalmic data. All study-related procedures were performed in accordance with good clinical practice (International Conference on Harmonization of Technical Requirements of Pharmaceuticals for Human Use E6), applicable US Food and Drug Administration regulations and the Health Insurance Portability and Accountability Act.
The primary end point of this study is to assess choriocapillaris and retinal CPD changes in patients with diabetes using OCTA. The secondary end point is to correlate ocular characteristics such as level of retinopathy, visual acuity (VA), axial length and history of ocular treatment with choriocapillaris and retinal CPD.
Participants and image acquisition
Slit-lamp bio microscopic exam was used to identify diabetic patients without any signs of DR (DM without DR), patients with NPDR, patients with proliferative diabetic retinopathy (PDR) and healthy patients without any signs of retinal disease (controls). Even though patients with DME could be included, no specific group was created for these eyes. Patients were excluded if they had the presence of any other vitreoretinal diseases that may affect ocular circulation (including retinal vascular occlusion, central serous retinopathy or macular dystrophies) or hampered the vascular density analysis (ie, epiretinal membranes). Additionally, patients with DME whose anatomical abnormality (ie, hard exudates or media opacity) was preventing OCTA visualisation were also excluded. Variables such as history of anti-vascular endothelial growth factor (VEGF) treatments, focal laser, panretinal photocoagulation (PRP), high myopia (refractive error ≥6 dioptres) and VA were collected for each patient. A single eye of each patient was included. Eyes were imaged with the Optovue Avanti RTVue XR SD-OCT (figure 1)) from January 2015 to October 2017. The specifics of the scanning protocol have been previously published.9 Quantitative CPD analysis was performed with ReVue software V.2017.1.0.129 (Optovue, Fremont, California, USA). The built-in software calculates the CPD by computing the percentage area occupied by detected OCTA vasculature. Vessels pixels area is 'true' while background/noise is 'false'. This binary mask is then used to generate a 2D local density map and automatically calculated density values in grid sectors. Auto-segmentation was used to define anatomical borders for CPD analysis and for foveal avascular zone (FAZ) area demarcation. Artefacts can interfere with the CPD analysis; however, the Avanti system employs motion correction technology and 3D projection artefacts reduction, algorithm based on the recording several OCT volumes obtained with orthogonal scanning axes that reduces motion and projection artefact in all posterior layers and B-scans. Manual corrections were made if any scan errors were identified.
Categorical variables were described using frequencies and percentages, while continuous variables were described using means, SD and ranges. Relationships between categorical variables were assessed using Kruskal-Wallis tests (for ordered variables), while relationships between continuous variables were assessed using t-tests, one-way analysis of variance test (for normally distributed variables) or Kruskal-Wallis tests (for non-normally distributed variables). Analyses were performed using SPSS Statistics software V.25.
In total, 136 total charts were reviewed, of which 27 were patients with PDR, 41 were patients with NPDR, 31 were diabetic patients without retinopathy and 37 were non-diabetic control patients. Demographics and clinical characteristics can be identified in table 1.
Choroidal and retinal thickness values and vascular indices
Choroidal and retinal thicknesses are displayed in table 2. Compared with controls, DM without DR showed a 3% decrease in CT (309.8±16.1 vs 297.8±18.2, p=0.007). There was no significant difference in CT between controls and NPDR or PDR, and there was no significant difference in retinal central subfield thickness (CST) between control eyes and any of the other groups.
Choroidal and retinal vascular indices, including whole-image CPD, foveal CPD and parafoveal CPD, are displayed in table 2. Eyes with NPDR and PDR showed significantly decreased choriocapillaris whole-image and parafoveal CPD compared with controls, while eyes without DR did not show significant change. Choriocapillaris whole-image CPD was decreased by 8.3% in eyes with NPDR (63.8±8.2 vs 69.6±4.6, p<0.001) and decreased by 7.1% in eyes with PDR (64.7±8.2 vs 69.6±4.6, p=0.005) compared with controls. Choriocapillaris parafoveal CPD was decreased by 8.9% in eyes with NPDR (63.1±9 vs 69.3±4.4, p<0.001) and decreased by 8.2% in eyes with PDR (63.6±8.8 vs 69.3±4.4, p=0.003) compared with controls. Only eyes with PDR showed significantly decreased retinal whole-image, foveal and parafoveal CPD. Retinal whole-image CPD was decreased by 9.7% (45.7±5.6 vs 50.6±5.3, p<0.001), retinal foveal CPD was decreased by 20.5% (23.6±7.7 vs 29.7±7.5, p=0.003) and retinal parafoveal CPD was decreased by 11.4% (47.5±5.6 vs 53.6±5.6, p<0.001) in eyes with PDR compared with control.
FAZ area and perimeter values are displayed in table 2. Only eyes with PDR showed significantly increased FAZ values. FAZ area was increased by 50.9% (0.415±0.2 vs 0.275±0.1, p=0.001) and FAZ perimeter was increased by 2.3% (2.865±1 vs 2.084±0.5, p=0.001) compared with controls.
Correlation with ocular characteristics
Among all subject eyes, higher VA was significantly associated with greater retinal whole-image and retinal parafoveal density (p<0.01). A history of at least one anti-VEGF injection was significantly associated with an increased CT (p=0.02), and with decreased retinal whole-image and decreased retinal parafoveal CPD (p<0.01 for both). A history of PRP was significantly associated with decreased retinal whole-image, foveal and parafoveal CPD (p<0.01 for all). High myopia and a history of focal laser were not significantly associated with any differences in choroidal or retinal thickness values or vascular indices.
Results from this study examining choroidal and retinal vascular indices in different stages of diabetes suggest that, compared with non-diabetic controls, eyes with NPDR and PDR demonstrate decreased choroidal vascular density (CVD), and that eyes with PDR additionally demonstrate decreased retinal vascular density and an increased FAZ area. These findings further support the idea that as DR progresses the integrity of the choroidal and retinal vasculature is increasingly altered.
This study adds relevant data to the limited current scientific knowledge regarding the choroidal component of diabetic ocular disease. Wang et al, in a prospective cross-sectional study of 233 eyes, used OCTA to compare CVD in diabetic eyes and controls and similarly found decreased choroidal vascular indices in patients with diabetes.10 Wang et al reported a 6% reduction in the average CVD in eyes with NPDR and 12% in eyes with PDR, while this research found that eyes with DR had a reduction of approximately 7.5% in the choriocapillaris perfusion density compared with control. A fundamental difference between these studies is that the previously mentioned study used a swept-source OCT, while our study used SD-OCT. Nesper et al, in a retrospective study of 137 eyes, quantified retinal and choriocapillaris microvascular changes in the eyes of patients with diabetes compared with healthy subjects using OCTA, and reported significant choriocapillaris flow differences between the control and the PDR groups.11 In their analysis, a third-party software was used to calculate choriocapillaris adjusted flow index, a surrogate for CPD analysis. In contrast, in our study, the same built-in software used to image eyes was also capable of automated CPD assessment and thus was used to evaluate all choriocapillaris and retinal layers.
No significant difference was found in choriocapillaris CPD between NPDR and PDR groups in this study. This result may seem unusual since patients with PDR experience worse ocular ischaemia. However, Cao et al, in a histopathological study which measured the extent of choriocapillaris degeneration in patients with diabetes, showed that eyes with severe chronic ischaemia might develop choroidal neovascularisation (CNV).12 Hua et al also reported a case of CNV in a diabetic patient without signs of retinal microvascular proliferation or drusen, hypothesising that the neovessels found might be an intraocular microvascular complication of DM.13 Therefore, it is reasonable that the similar choriocapillaris CPD in patients with PDR might be a result of CNV formation caused by this natural compensation mechanism manifested in these patients. Another explanation is the lack of subdivision groups of patients with NPDR. A large number of severe NPDR in this study may have hampered the comparison between these two levels of DR.
Retinal capillary and FAZ area changes in DR have been previously reported.6 8 11 14–19 Studies have described decreased retinal CPD and increased FAZ area in patients with DR compared with healthy individuals. In our study, retinal CPD was decreased and FAZ measurements were increased in all groups compared with control. However, only the PDR group showed a significant difference. These findings are consistent with some previous OCTA studies that compared healthy and diabetic patients without DR and were unable to demonstrate a significant difference when examining retinal vascular plexuses and choriocapillaris density measurements.6 18 Some other studies that grouped patients by their levels of NPDR (mild NPDR, moderate NPDR and severe NPDR) reported significant differences in retinal CPD and FAZ area only between moderate/severe NPDR and control eyes.8 14–16 19 Again, the high heterogeneity of ischaemic stages in the NPDR group could hinder the determination of significant differences in retinal CPD between groups compared.
Axial length, gender, age and PRP treatment are confounding factors that are known to influence choroidal vascularisation.20 Correlations between variables, such as high myopia, history of anti-VEGF treatment, laser treatment and VA, were also assessed in our current study. Even though the average CT is reduced in myopic eyes,21 no significant association between axial length and choriocapillaris and retinal thickness measurements was found. History of anti-VEGF treatment was found to be significantly associated with increased choriocapillaris thickness. This correlation can be explained since patients with DME have thicker choroids compared with patients with DR and without macular oedema,1 and anti-VEGF injections are the current main treatment of DME.22 The association between PRP history and decreased retinal CPD was found because this laser treatment is only used for patients with critical stages of DR in which severe damage of retinal plexus is expected. Higher VA was found to be associated with greater choriocapillaris and retinal CPD, which is consistent with the existing finding that disease severity in DR is correlated with retinal and choriocapillaris vascular non-perfusion areas.11
A number of studies have quantified findings of DR objectively.23 24 However, these assessments required manual segmentation remaining too arduous to be practical, findings showed vast variability among diverse patients, and were not evaluated in clinical environment. OCTA is a widely used diagnostic examination, and the findings of this study support previous reports demonstrating that CPD analysis represents an automated, reproducible and reliable tool for detecting and quantifying ocular vascular abnormalities.9 25
Current theories about diabetes ocular involvement fail to explain how patients without DR may report visual defects. Choroidal alterations observed in angiography, histology, Doppler flowmetry and OCT studies may play a significant role as part of an alternative hypothesis on the pathogenesis of diabetic eye disease.2–4 However, it is still unclear if these variations are prognostic, causative, modulatory or unrelated to DR.
The limitations of this study include its retrospective design, the absence of NPDR subgroups (mild, moderate and severe NPDR) and the absence of a DME group. Strengths of this study include the correlation analysis of some confounding factors for choriocapillaris variations, and the use of the most sensitive method for assessing capillary density changes (the built-in AngioVue system) which enables automated image acquisition and quantification of results.26
In conclusion, this study showed that OCTA is capable of detecting capillary perfusion changes in the retina and choriocapillaris of patients with diabetes compared with healthy individuals as well as correlating these capillary density changes with the ocular disease severity. Moreover, this research solidifies the assessment that choriocapillaris density decreases with worsening retinopathy levels in DM and creates an alternative hypothesis on the pathogenesis of diabetic eye disease.
Contributors All authors made substantial contributions to the conception or design of the work or the acquisition, analysis or interpretation of data. They also drafted the work or revised it critically for important intellectual content. Finally, they all gave final approval for the version submitted. All the authors have contributed to the planning, conduct and reporting of the work described in the article.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests RPS received grants and personal fees from Regeneron, Genentech/Roche, Apellis, Optos, Zeiss, Biogen and Alcon/Novartis outside the submitted work. JPE received grants and personal fees from Regeneron, Thrombogenics and Alcon/Novartis; royalties from Leica/Bioptigen outside the submitted work. AVR received personal fees from Allergan, Zeiss and Alcon/Novartis outside the submitted work. SS received grants and personal fees from Allergan and Mallingcrofdt outside the submitted work.
Patient consent Not required.
Ethics approval This retrospective study was performed at Cole Eye Institute, Cleveland, Ohio, after receiving approval from the Cleveland Clinic Investigational Review Board (IRB study number 15-080).
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.