The Effects of Anti-Vascular Endothelial Growth Factor Loading Injections on Retinal Microvasculature in Diabetic Macular Edema

Abstract

Purpose: To evaluate changes in the retinal microvasculature using widefield swept-source optical coherence tomography angiography (SS-OCTA) following three anti-vascular endothelial growth factor (anti-VEGF) loading injections for diabetic macular edema (DME).

Methods: Thirty-four treatment-naïve patients with DME received an initial three loading injections, followed by injections on an as-needed basis. Macular ischemia was evaluated based on the foveal avascular zone (FAZ) area, perfusion density, and vessel density on a 3 × 3-mm SS-OCTA image. Midperipheral ischemia was analyzed by dividing a 12 × 12-mm image into 16 boxes to compare changes in the nonperfusion area (NPA). Participants were categorized as aggravated, stable, or improved based on changes in the NPA after three injections.

Results: Of the 34 included patients, eight (23.5%) demonstrated aggravation of the NPA, 23 (67.6%) remained stable, and three (8.8%) exhibited improvement. Although FAZ area, perfusion, and vessel density increased, the differences were not significant compared to baseline. The number of injections and glycated hemoglobin (HbA1c) levels in the NPA aggravation group were significantly higher than in the stable and improvement groups. Logistic regression analysis revealed that NPA aggravation was independently associated with the number of anti-VEGF injections.

Conclusions: Changes in NPA following anti-VEGF loading injections varied among patients with DME and were significantly associated with HbA1c levels and injection frequency. Worsening mid-peripheral NPA after the anti-VEGF injections independently risked DME recurrence.

Translational Relevance: We revealed that worsening mid-peripheral retinal ischemia after anti-VEGF loading injections contributes to the recurrence of DME.

Introduction

Retinal ischemia is a central pathology in diabetic retinopathy (DR), initiating a chain of events that ultimately leads to vasculopathy.¹ Although capillary nonperfusion is one of the main pathological characteristics of DR, the present DR grading system relies primarily on the Early Treatment Diabetic Retinopathy Study (ETDRS) classification of color fundus photographs (CFPs). However, this does not adequately reflect signs of ischemia. Given the importance of capillary ischemia as a diagnostic indicator of DR, a new grading system is needed that incorporates microvascular changes in the retina, addressing the limitations of CFPs or fluorescein angiography (FA) findings. Furthermore, Silva et al.² demonstrated that peripheral lesions and the severity of DR are correlated with nonperfusion, identified using ultra-widefield FA. The aggravation of retinal ischemia observed on optical coherence tomography angiography (OCTA), including retinal capillary loss and the emergence of nonperfusion areas (NPAs), can serve as an indicator of DR severity and disease progression.^3,4 

Intravitreal anti-vascular endothelial growth factor (VEGF) injections are widely utilized for the management of diabetic macular edema (DME) and have proven effective in inducing the regression of new vessels and improving the Diabetic Retinopathy Severity Scale level.⁵ However, the effects of anti-VEGF injections on retinal ischemia in DR remain controversial, given the various competing factors that can either improve or worsen macular perfusion. Previous studies have reported different results when using FA to assess the effects of anti-VEGF injections on peripheral retinal perfusion.⁶ One possible reason for this discrepancy is the challenge of accurately delineating the NPA on FA images. The development of swept-source OCTA (SS-OCTA) offers a novel, noninvasive method for assessing retinal capillary perfusion in the posterior pole and mid-peripheral retina, providing a viable alternative to FA. 

Numerous studies have recently established a correlation between DR severity and mid-peripheral NPA on OCTA as a novel biomarker of retinal ischemia.^7,8 However, the clinical association between retinal ischemia and the development of DME remains unclear. Therefore, the primary aim of this study was to evaluate changes in NPA on SS-OCTA following three anti-VEGF loading injections and explore potential associations with DME recurrence. 

Methods

This retrospective study included treatment-naïve patients diagnosed with DME who received three monthly anti-VEGF loading injections between January 2019 and July 2021 at Kyung Hee University Hospital. The protocol of this study adhered to the tenets of the Declaration of Helsinki and was approved by the Institutional Review Board of Kyung Hee University Hospital. 

Participants

This study included patients 20 years of age or older with diabetes and fovea-involving DME, defined as a central subfield thickness >300 µm. All patients were followed up clinically for ≥2 years after receiving their first anti-VEGF loading injection and had a best-corrected visual acuity (BCVA) of ≥1.0 logMAR. The exclusion criteria were as follows: (1) previous treatment with focal macular laser, intravitreal or periocular therapy, or pan-retinal photocoagulation; (2) active proliferative DR or vitreous hemorrhage; (3) cataract surgery within 90 days or any history of other intraocular surgery or buckling; (4) ocular inflammation or history of uveitis; (5) structural damage to the center of the macula including atrophy, fibrosis, or scarring, significant macular ischemia, or organized hard exudates; (6) concurrent disease, other than DME, in the affected eye that could compromise visual acuity; (7) high myopia (equivalent ≤ −8.0 D or axial length ≥26.0 mm); or (8) poor image quality (signal strength index < 7/10). Anti-VEGF therapy (0.5 mg ranibizumab or 2 mg aflibercept) was initiated with three monthly loading injections, followed by treatment on an as-needed basis. Retreatment criteria included the presence of the following: (1) new or persistent intra- or subretinal fluid, and (2) persistent diffuse edema with a central subfield thickness ≥ 300 µm. 

Image Acquisition and Analysis

Ultra-widefield fundus photography (CLARUS 500; Carl Zeiss Meditec, Jena, Germany), spectral-domain OCT (CIRRUS 5000; Carl Zeiss Meditec), and SS-OCTA (PLEX Elite 9000; Carl Zeiss Meditec) were performed at all visits, including the initial visit and again 1 month after the third anti-VEGF injection. In patients with severe macular edema at baseline, we utilized the 3 × 3-mm OCTA images from the examination conducted 1 or 2 weeks after the initial injection. Figure 1 demonstrates 3 × 3-mm and 12 × 12-mm SS-OCTA images used for the measurement of OCTA metrics. The baseline NPA on a 12 × 12-mm image of the total retinal slab was quantitatively assessed with customized MATLAB software (MathWorks, Natick, MA) according to a previously published semi-automatic method.⁵ Advanced Research and Innovation (ARI) Network algorithms (Macular Density Algorithm 0.7.3) were employed to perform quantitative measurements of macular ischemia, including perfusion density (PD), vessel density (VD), and foveal avascular zone (FAZ) area. PD was defined as the total area of perfused vasculature per unit area and VD as the total length of perfused vasculature per unit area in the region of measurement. 

Figure 1.

 

Generated SS-OCTA images (3 × 3 mm and 12 × 12 mm) for the measurement of OCTA metrics. (A) En face OCTA 3 × 3-mm image of the SCP. (B) Binarized image of the SCP for the calculation of perfusion density. C) Skeletonized binarized image of the SCP for the calculation of vessel length density. (D) En face OCTA 3 × 3-mm image of the DCP. (E) Binarized image of the DCP. (F) Skeletonized binarized image of the DCP. (G) En face OCTA 12 × 12-mm image of the whole retinal slab. (H) The 12 × 12-mm SS-OCTA images divided into 16 identical boxes to determine the changes in nonperfusion area.

Paired 12 × 12-mm SS-OCTA images, obtained at baseline and after administration of the three loading injections, were aligned and cropped to retain only the common part of the fundus. After this alignment, all fundus images were divided into 16 identical boxes (4 × 4) for semiquantitative analysis. Signal reduction artifacts, identified using structural en face OCT images, were excluded from the NPA. Two blinded retinal specialist separately reviewed the SS-OCTA image sets. In the event of any discrepancies, the images were re-evaluated to reach a consensus. Changes in the NPA between the baseline and post-injection periods were subsequently assessed by each specialist using the 12 × 12-mm SS-OCTA images. 

Outcome Measurements

To compare the NPA between baseline and 1 month after the third loading injection, the participants were categorized into three groups: improved, stable, and aggravated. When at least one of the 12 regions demonstrated a change in NPA where capillaries disappeared or reappeared, the case was classified into the improved or aggravated group, respectively. If no change was observed, the case was classified as stable. If both an increase and a decrease in NPA were noted, the case was classified into the group with the greater number of changes. The PD, VD, and FAZ area in the superficial capillary plexus (SCP) and deep capillary plexus (DCP) at baseline and 1 month after the third loading injection were evaluated in the macular area on 3 × 3-mm OCTA images. The baseline NPA was calculated for the total retinal slab on a 12 × 12-mm image. The BCVA was obtained at baseline and after 2 years. Moreover, the total number of injections over 2 years was obtained for each patient. 

Statistical Analyses

Statistical analyses were performed using SPSS Statistics 18.0 (IBM, Chicago, IL). Comparisons among the three groups were performed using an analysis of variance followed by Tukey's post hoc analysis to correct for multiple comparisons. Logistic regression analysis was employed to identify variables associated with aggravated NPAs. 

Results

A total of 34 eyes from 34 patients with diabetes were included in this study. The clinical characteristics of the included patients are summarized in Table 1. Based on the ETDRS severity scale, 27 eyes were classified as having severe non-proliferative DR, with the remaining seven having proliferative DR. The mean BCVA at baseline was 0.319 ± 0.12 logMAR, which improved to 0.215 ± 0.11 logMAR at 1 month after the third loading injection. The mean central macular thickness was 454.8 ± 81.5 µm at baseline, which decreased significantly to 279.57 ± 32.2 mm at 1 month after the third loading injection. All patients were treated with intravitreal ranibizumab, with a mean number of injections of 5.02 ± 2.24 over 2 years, including three loading injections. The mean FAZ area was 0.48 ± 0.35 mm² at baseline, which increased to 0.55 ± 0.40 mm² (P = 0.253) at 1 month following the third loading injection. Both capillary density parameters in the macular area displayed increasing trends; however, the differences were not significant (Fig. 2). The mean PD values at baseline were 0.29 ± 0.02 and 0.19 ± 0.04 in the SCP and DCP, respectively, which increased to 0.30 ± 0.04 and 0.22 ± 0.06 at 1 month after the third loading injection (P = 0.522 and P = 0.633, respectively). The mean VD values at baseline were 12.2 ± 0.9 and 8.58 ± 1.6 in the SCP and DCP, respectively, which increased to 12.7 ± 1.6 and 9.68 ± 2.4 at 1 month after the third loading injection (P = 0.383 and P = 0.412, respectively). 

Table 1.

 

Comparison of Baseline Characteristics Based on NPA Changes After Anti-VEGF Injections

 

Figure 2.

 

Changes in 3 × 3-mm OCTA parameters after three monthly anti-VEGF leading injections. (A) Changes in FAZ area between baseline and after three leading injections. (B) changes in PD between baseline and after three leading injections. (C) Changes in VD between baseline and after three leading injections.

Based on the change in NPA after the three loading injections, 23 eyes (67.6%) were classified into the stable group, three eyes (8.8%) into the improved group, and eight eyes (23.5%) into the aggravated group. Figure 3 displays representative 12 × 12-mm SS-OCTA images displaying eyes with improved and aggravated NPAs. Baseline clinical factors were compared by categorizing the participants into two groups (Table 1), aggravated and stable or improved. Glycated hemoglobin (HbA1c) levels were significantly higher in the aggravated group than in the stable or improved group (8.7 ± 1.5 vs. 7.2 ± 1.5, respectively; P = 0.025), as were the numbers of injections over 2 years (7.3 ± 3.3 vs. 4.1 ± 1.2, respectively; P = 0.029). Furthermore, BCVA, central retinal thickness, and OCTA parameters, including NPA, FAZ area, PD, and VD, did not differ significantly between the two groups. 

Figure 3.

 

Representative 12 × 12-mm SS-OCTA images depicting improvement and aggravation of capillary nonperfusion areas after three monthly anti-VEGF leading injections. Images were aligned to retain the common part of the fundus and were then divided into 16 identical boxes. (A, C) Baseline 12 × 12-mm SS-OCTA before injection. (B) SS-OCTA images displaying capillary reperfusion after three anti-VEGF leading injections (blue boxes). (D) New capillary dropout appears following three anti-VEGF leading injections (red boxes).

Logistic regression was performed to identify the variables associated with the aggravation of NPA (Table 2), the results of which indicated that an increased number of injections over a 2-year period was significantly associated with the aggravation of NPA (odds ratio = 2.55; P = 0.023) after controlling for age, HbA1c, DR severity, baseline NPA, and central retinal thickness. 

Table 2.

 

Logistic Regression Analysis to Identify Variables Associated With Aggravation of NPA

 

Discussion

DR is often characterized by a progressive increase in capillary nonperfusion, which is indicative of macular ischemia. Furthermore, as retinal ischemia remains a key prognostic indicator, an increasing number of studies have focused on its quantification as a potential biomarker. Recent OCTA studies have demonstrated an association between increased DR severity and increased central and peripheral NPAs.^7,8 Although this observation broadly describes the progression of DR, noting that the reperfusion of capillary NPAs has been recognized as a potentially dynamic process that can lead to the resolution of DR is important.⁹ 

The severity of DR, as graded using CFPs, may potentially improve after anti-VEGF injections.^5,6 However, whether this improvement simultaneously extends to NPAs remains unclear. Additionally, a robust correlation exists between DR lesions on CFPs and capillary nonperfusion, which is established before treatment. A cross-sectional study conducted by Silva et al.² demonstrated a strong association between hemorrhages and microaneurysms and the extent of nonperfusion and ischemia on ultra-widefield FA images. Interestingly, the improvement observed in CFPs did not translate to improvement of NPAs detected using FA, as the NPSs did not exhibit any reperfusion of affected arterioles or venules.⁶ 

Various OCT biomarkers such as submacular fluid, hyperreflective foci, inner segment/outer segment layer integrity, and disorganization of retinal inner layers have been reported in previous studies as being predictive of the clinical outcomes of treatments for DME.^10,11 Several studies have recently assessed the effects of repeated anti-VEGF injections on macular perfusion in the management of DME. This assessment has historically relied on FA images being reviewed by expert human graders. Wykoff et al.¹² reported that the proportion of eyes with improved retinal perfusion over a 100-week period involving regular intravitreal aflibercept injections ranged from 40.0% to 44.7%, whereas that of eyes with worsened retinal perfusion ranged from 8.6% to 9.0%. Furthermore, a dose–response relationship was identified, indicating a reduction in NPA progression with monthly versus quarterly aflibercept injections.¹³ The disparities in the proportions observed in the present study may be attributed to the number of anti-VEGF injections, highlighting the positive effect of repeated injections in managing capillary nonperfusion in DME. Subsequent studies incorporated ultra-widefield FA imaging to evaluate the effects of VEGF inhibitors on peripheral retinal perfusion.^13,14 Despite these additional studies, however, conclusive results highlighting the effects of VEGF inhibition on the macular perfusion status of patients with diabetes were not consistently obtained.^6,14,15 

Although FA serves as the primary method for evaluating retinal ischemia, potential obscuration from leakage makes the procedure a suboptimal diagnostic modality, particularly for the early detection of vascular diseases. Therefore, the interpretation of retinal ischemia on FA remains predominantly semi-quantitative and -qualitative. By contrast, OCTA is a dye-free OCT-based imaging technique that enables the volumetric visualization of the retinal microvasculature. Vascular occlusion leading to retinal ischemia manifests as a flow void signal on OCTA, which is indicative of retinal nonperfusion. The two primary advantages of OCTA over FA are the ability of the modality to capture fine capillary details at a high resolution and the absence of obscuration due to dye leakage. These advantages allow for improved detection and quantification of retinal ischemia. Additionally, widefield SS-OCTA offers a larger field of view with fewer total scans while still visualizing peripheral pathology better than 50° FA. SS-OCTA detected NPA and neovascularization with sensitivities of 96% and 79% and specificities of 100% and 96%, respectively, in patients with DR at rates better than those noted with FA.^16,17 Additionally, recent advances in SS-OCTA technology have made possible the acquisition of single-shot scans with dimensions of 12 × 12 mm, and even going beyond 15 × 15 mm. Widefield SS-OCTA has demonstrated the potential to detect NPA at higher rates than those visible on ultra-widefield FA. Moreover, SS-OCTA is also effective in detecting nonperfusion in the mid-peripheral retina.¹⁵ The quantification of peripheral retinal nonperfusion may prove useful for the early detection and monitoring of proliferative DR.¹⁷ 

Several OCTA studies have recently been conducted to evaluate macular perfusion changes in the eyes of patients with diabetes following anti-VEGF injections. These studies, however, have presented conflicting results, with some demonstrating stable or improved macular perfusion following treatment^18,19 and others indicating aggravation.^20,21 These disparate findings may be attributed to variations in study design, inclusion criteria, image analysis methods, and vascular density quantification. Additionally, the clinical correlations between the resolution of macular edema after an anti-VEGF injection and capillary reperfusion observed on OCTA remain unclear.^22,23 Most earlier OCTA studies utilized a 3 × 3-mm or 6 × 6-mm field of view, potentially missing ischemic changes in the perifoveal area. Similar results have been reported in patients with branch retinal vein occlusion (BRVO), where several individuals with macular ischemia displayed reperfusion in macular NPA but others experienced worsened macular ischemia after ranibizumab injections.²⁴ Of note, the frequent recurrence of macular edema was significantly correlated with decreased retinal perfusion in BRVO,²⁵ and central and parafoveal NPAs in OCTA were correlated with the recurrence of macular edema in patients with BRVO after intravitreal anti-VEGF injections.²⁶ These findings align with the results of our current study, indicating that the worsening of mid-peripheral NPA after anti-VEGF injections is associated with frequent recurrence of DME. Kadomoto et al.²⁷ reported a strong association between parafoveal NPA and visual function in BRVO. However, the present study identified no significant correlation between visual acuity at baseline or after 2 years and changes in NPA. Improvement in visual acuity in patients with DME can be limited despite the resolution of macular edema, as various OCT parameters may influence functional outcomes beyond retinal thickness. 

The physiological effects of anti-VEGF injections for DME on macular perfusion involve complex factors that lead to both improvement and deterioration. Animal experiments have demonstrated that VEGF suppression induces reperfusion of the closed vessels.²⁸ Increased intraocular VEGF contributes to the expression of intercellular adhesion molecule-1 (ICAM-1) in retinal endothelial cells, promoting the binding of ICAM-1–mediated leukocytes to the vasculature and leading to increased capillary occlusion.²⁹ Additionally, the restoration of normal retinal architecture, decreased apoptosis rates, activation of ischemia-damaged microglia, and normalization of peripheral cells contribute to halting the progression of vessel closure and improving perfusion.¹⁹ Conversely, factors that may result in decreased retinal perfusion following anti-VEGF injections include reductions in retinal vessel diameters and flow velocities, which may be attributed to the inhibition of nitric oxide associated with VEGF suppression.^30,31 Moreover, the loss of pericytes, which provide a protective sheath around mature retinal capillaries, renders them less reliant on VEGF for survival. Additionally, this loss may contribute to a reduction in capillary density following VEGF inhibition in individuals with diabetes.³² Pericyte loss occurs in the early stages of DR, rendering capillary endothelial cells susceptible to VEGF inhibition, leading to the demise of endothelial cells and subsequently, capillary loss.³³ In the context of individual patients, some retinal areas may demonstrate improved perfusion while others worsen, as observed on OCTA following anti-VEGF injections. This finding suggests that different outcomes depend on the competing factors that are predominant after the administration of anti-VEGF injections. In this study, 8.8% of the patients with macular ischemia displayed reperfusion in the NPA, 23.5% exhibited aggravated macular ischemia, and 67.6% demonstrated no significant changes after 3 monthly anti-VEGF loading injections. The results of this study also suggest that retinal capillary reperfusion occurs before irreversible ischemic retinal damage. In cases where reperfusion does not occur, the ischemic areas are either irreversibly damaged or require frequent anti-VEGF injections. This is supported by the finding that patients who presented with aggravation of NPAs after anti-VEGF injections required additional treatments over a 2-year period after the loading series. 

This study has some important limitations. First, only a limited number of participants were enrolled. Second, we utilized a single 12 × 12-mm SS-OCTA image, providing enhanced visualization of microvascular features beyond the posterior pole, in contrast to the conventional angle. Nevertheless, visualization of the mid- to far-peripheral retina remained limited within this single 12 × 12-mm field. Diabetic microangiopathy stems from the peripheral retina, which is more susceptible to diabetes-induced capillary closure than the posterior pole. Further investigations utilizing a large field view are needed to determine whether anti-VEGF treatments can mitigate the worsening of ischemia. Third, this study did not involve the pixel-based quantification of NPA in SS-OCTA images. Instead, a qualitative zone-by-zone comparison of the NPA was performed between the baseline and those after subsequent anti-VEGF injections. Widefield OCTA imaging is susceptible to artifacts due to extended image acquisition times, potentially resulting in challenges related to patient cooperation and motion.³⁴ In particular, shadowed areas or peripheral regions that are out of focus due to retinal curvature may frequently produce low signal artifacts in widefield imaging. A reliable strategy for identifying low OCT signal artifacts involves the comparison of OCT and OCTA en face images, as true NPAs should exclusively manifest in the OCTA signal, whereas signal loss due to artifacts affects both structural and angiographic signals. Despite the subjective nature of our NPA assessments, the direct comparison method was deemed more accurate for determining changes in NPAs and preventing clinical misinterpretation, as opposed to pixel quantification methods. 

In conclusion, DME is a multifactorial disease, making the use of imaging markers alone to predict recurrence or visual prognosis difficult. The results of this study are clinically significant, as changes in retinal ischemia after anti-VEGF injections can be measured by NPA changes on OCTA, which may serve as a predictive marker for DME recurrence. Future investigations are needed to assess whether proactive anti-VEGF treatments can mitigate the progression of ischemia, subsequently preserving long-term vision in patients with DME. 

Acknowledgments

Disclosure: K. Kim, None; J. Lee, None; S.-Y. Yu, None