Introduction

Diabetic macular edema (DME) is the leading cause of visual impairment in patients with diabetes mellitus, occurring in approximately 35% of patients with diabetic retinopathy (DR) [1, 2]. In the pathogenesis of DME, hyperglycemia promotes structural changes in retinal blood vessels and disruption of the blood–retinal barrier, resulting in accumulation of fluid [3]. Besides the vasogenic theory, there is some degree of neuronal death implicated in DR [4] and photoreceptors’ involvement, leading to their neurodegeneration [5]. Otani et al. have described three major patterns of DME based on optical coherence tomography (OCT); sponge-like swelling, cystoid macular edema (CME) and serous retinal detachment (SRD) associated with CME [6]. Nowadays, the advent of spectral-domain OCT (SD-OCT), besides the description of the morphological pattern of DME, has also facilitated the detailed study of the status of retinal photoreceptors, allowing identification of the external limiting membrane (ELM), the photoreceptor ellipsoid zone (EZ) and the interdigitation zone (IZ) [7].

Vascular endothelial growth factor (VEGF) is the most potent angiogenic factor and a prominent mediator of vascular permeability in patients with DME [8]. Therefore, the gold standard in the treatment of DME is anti-VEGF agents, whose safety and efficacy have been proven in large pivotal clinical trials, as well as in real-life studies [9, 10]. Only a few studies have examined whether the response to anti-VEGF treatment changes according to the morphological type of DME [11,12,13,14]. Diffuse macular edema has showed the greatest improvement in central retinal thickness and visual acuity after intravitreal bevacizumab [11, 12], while Seo et al. found no difference in treatment response between DME types using intravitreal ranibizumab [13]. In addition, a significant restoration of photoreceptors after intravitreal ranibizumab treatment, correlating with visual acuity, has been reported [15]. The purpose of the current study is to analyze the photoreceptors’ changes after anti-VEGF treatment, stratifying patients according to the DME pattern on SD-OCT.

Methods

In this retrospective study, the medical records of 58 adult patients (58 eyes) with type 2 diabetes mellitus and treatment naïve DME (Table 1), who were diagnosed and treated at 2nd Department of Ophthalmology, University of Athens, Greece, between January 2016 and June 2018, were reviewed and analyzed. All patients received intravitreal ranibizumab injections and were followed-up for at least 12 months. In case where both eyes of the same patient presented DME, we included only one eye per patient. Specifically, the right eyes of patients with bilateral DME were selected, so as to avoid selection bias and intercorrelation of measurements in the same patient. Patients with other retinal diseases, vitreomacular traction, macular hole, proliferative diabetic retinopathy, ocular inflammation, myopia > 6D, uncontrolled glaucoma, previous vitrectomy, ocular trauma or cataract surgery within the last 6 months and media opacities were excluded from the study. The study was in accordance with the Tenets of Helsinki Declaration, and written informed consent was obtained by all patients to use their data.

Table 1 Demographic and clinical characteristics of the study sample at baseline
Full size table

Data related to demographic characteristics, medical history, smoking status, comorbidities (hypertension, hyperlipidemia, heart failure, nephropathy) and HbA1c were recorded for all patients. All participants had underwent a complete ophthalmological examination at the time of DME diagnosis (baseline), including best-corrected visual acuity (BCVA) measurement using Snellen charts, biomicroscopy, dilated fundoscopy, spectral-domain optical coherence tomography (SD-OCT) and fluorescein angiography (FFA) using Spectralis (Spectralis HRA + OCT, Heidelberg Engineering, Germany).

For each patient, BCVA was converted to the logarithm of the minimum angle of resolution (logMAR) for statistical purposes. Regarding SD-OCT examination, six radial scans 3-mm long were performed at equally spaced angular orientations centered on the foveola. An OCT volume scan was performed on a 20° × 20° cube, consisting of 49 horizontal B-scans with 20 averaged frames per B-scan centered over the fovea, each containing 1064 pixels, separated by 125 mm. OCT scans were evaluated for macular thickness, which was automatically measured at central area of ETDRS map, the type of DME (diffuse, CME and SRD combined with CME, as it is shown in Fig. 1), the location of DME (predominantly in inner or outer layers), the extent of DME (involving the fovea or extending outside the fovea) and the presence of epiretinal membrane (ERM-defined as a thin, hyperreflective band on the surface of the retina above the internal limiting membrane). The quantitative measurement of EZ defect size at fovea was taken manually using the OCT-caliper of the SD-OCT machine by two trained investigators (IP and ED). The structural condition of ELM and IZ at the fovea was categorized qualitatively as intact (if it was continuous and completely visible), or disrupted (if it was disrupted, attenuated because of pathological changes or absent) by the same trained investigators (IP, ED). If reliable evaluation of the three zones (ELM, EZ and IZ) could not be performed due to poor image quality or due to DME, the patient was excluded from the study.

Fig. 1
figure 1

Patterns of diabetic macular edema in spectral-domain optical coherence tomography: a diffuse macular edema; b cystoid macular edema; c serous retinal detachment combined with cystoid macular edema. Please note external limiting membrane (orange arrow), ellipsoid zone (green arrow) and interdigitation zone (blue arrow)

Full size image

All patients were treated with at least three monthly intravitreal ranibizumab injections as a loading dose. Thereafter, all patients were followed-up at a pro re nata (PRN) basis, with monthly monitoring for at least 12 months. At each monthly visit, all patients underwent BCVA measurement and SD-OCT. Re-injection was performed if the height of macular edema was ≥ 320 μm and BCVA decreased ≥ 1 Snellen line.

Statistical analysis was performed using SPSS 22.0 statistical software (SPSS Inc, Chicago, Illinois, USA). Descriptive statistics, including the mean values, median, standard deviations and percentages, were used to describe the baseline characteristics. Differences between the three groups were evaluated with one-way analysis of variance (ANOVA) for numerical data and with Chi-square test for categorical data. For longitudinal comparisons, the Wilcoxon signed-rank test was performed for numerical data and McNemar’s test for categorical data. Since multiple comparisons took place (baseline vs. month 6 or 12), a Bonferroni correction was adopted, as appropriate. A p value less than 0.05 was defined as statistically significant.

Results

Table 1 shows the demographic and clinical characteristics of the study sample at baseline.

Sub-analysis based on the patterns of DME on SD-OCT is shown in Table 2. Patients having CME had a larger EZ defect size (467 ± 103 μm) compared to those with diffuse macular edema (346 ± 102 μm) and SRD/CME (286 ± 100 μm), (p < 0.001). In cases with diffuse macular edema, the EZ defect extended mainly outside the fovea. Patients with CME presented disruption in IZ in a higher percentage than patients with diffuse macular edema or SRD/CME (p < 0.001), while ELM was significantly more intact in patients with SRD/CME (p < 0.001).

Table 2 Sub-analysis based on the three different patterns of DME at baseline
Full size table

In the whole cohort, there was a statistically significant improvement in BCVA (0.27 ± 0.09 logMAR, p < 0.001) and in CRT (368.0 ± 99.3 μm, p < 0.001) at month 12 post-treatment. The evolution of BCVA and CRT over time in the whole cohort and in the three groups of DME separately showed greater improvement in CME group (Figs. 2 and 3). The number of injections at the end of the follow-up period of 12 months was 7.4 ± 1.3 in patients with CME, 6.1 ± 1.2 in patients with SRD/CME and 5.9 ± 1.3 in patients with diffuse macular edema (p = 0.036).

Fig. 2
figure 2

Evolution of best-corrected visual acuity over time in patients with diabetic macular edema

Full size image
Fig. 3
figure 3

Evolution of central retinal thickness over time in patients with diabetic macular edema

Full size image

The EZ defect size improved significantly at month 12 post-treatment (459 ± 123 at baseline vs. 293 ± 109 at month 12, p < 0.001). Only patients with CME and SRD/CME showed significant difference in EZ defect size at month 12 compared to baseline. Specifically, patients with CME demonstrated greater improvement in the EZ defect size (467 ± 103 μm at baseline vs. 312 ± 102 μm at month 12, p < 0.001), compared to patients with SRD/CME (286 ± 100 μm at baseline vs. 205 ± 101 μm at month 12, p < 0.001) and patients with diffuse macular edema (346 ± 102 μm at baseline vs. 301 ± 113 at month 12, p = 0.102).

There was a statistically significant restoration of IZ at month 6 (69.0% intact IZ at baseline vs. 77.6% at month 6, p = 0.038) and month 12 post-treatment (69.0% intact IZ at baseline vs. 86.2% at month 12, p < 0.001). The restoration in IZ was more prominent in cases with CME (61.3% intact IZ at baseline vs. 87.1% at month 12, p < 0.001). However, in patients with diffuse macular edema, the IZ showed restoration which did not reach statistical significance (76.2% intact IZ at baseline vs. 85.7% at month 12, p = 0.068). In patients with SRD/CME, the condition of IZ remained stable (83.3% intact IZ at both time-points).

Discussion

The principal message of this study is that the various DME patterns on SD-OCT present different response after treatment with intravitreal ranibizumab injections, with significant difference in the outer retinal layers. At baseline patients with diffuse macular edema had better photoreceptors’ condition, BCVA and lower CRT than those with CME. EZ size defect was greater in patients with CME, since macular edema was mainly located at the fovea and the outer retinal layers, compared to patients with diffuse macular edema, in whom macular edema extended outside the fovea and was located mainly in the inner retinal layers. Additionally, IZ was more affected in patients with CME than with diffuse macular edema. BCVA improvement and the CRT reduction were observed in the whole cohort, as well as in each group. Controversy exists about how anti-VEGF treatment affects the neuroglial dysfunction in DME and contributes to the photoreceptors’ restoration [5, 13, 15].

In the current study, we measured manually the EZ defect size and found that there was significant improvement in EZ defect size post-intravitreal ranibizumab injections. Significant improvement of EZ defect size was observed in patients with CME and SRD/CME, while patients with diffuse macular edema showed slight improvement that did not reach significance. Regarding the IZ restoration, significant restoration was observed in the whole cohort at months 6 and 12 after intravitreal ranibizumab injections. Patients with CME presented significant IZ restoration at the end of the follow-up of 12 months, while IZ condition did not differ significantly in patients with diffuse macular edema and SRD/CME.

The EZ represents the photoreceptor integrity and is mainly comprised of mitochondria, enabling higher levels of energy consumption in the photoreceptors, while IZ is believed to represent RPE microvilli that surround the cone outer segment terminals [16]. In patients with DME, it is hypothesized that the mitochondrial dysfunction in the foveal photoreceptors results in reduced VA. Cystoid spaces located in the foveal area may compress and deform the photoreceptors [17], as it has been shown in our study in patients with CME. On the other hand, in our study, patients with diffuse macular edema did not present significant restoration in EZ, but only a slight improvement or stabilization of the EZ condition. It seems that the resolution of cysts after anti-VEGF therapy may result in photoreceptors’ layer improvement. As far as IZ is concerned, we found that IZ improvement occurred after EZ recovery, as it was observed previously by Serizawa et al. [18]. The IZ condition was significantly improved in patients with CME, but not in those with diffuse macular edema, in line with the EZ defect size improvement, which was statistically significant in patients with CME and not in patients with diffuse macular edema.

The ELM represents the junctional complex between the Mueller cells and the photoreceptor cells and has barrier properties against macromolecules [19]. The disruption of the ELM might allow blood constituents to pour into the subretinal space and damage the photoreceptors, or vice versa, photoreceptor damage might lead to disruption of the ELM in DME [20]. In our study, restoration of EZ and IZ at the end of the follow-up was correlated with ELM condition and in eyes where ELM was absent no restoration of EZ or IZ was observed. This has been previously described in patients with successful macular hole closure and optic disk pit maculopathy post-treatment, showing that the restoration of EZ is directly related to the integrity of the ELM and to the extent of EZ defect before treatment [21,22,23]. Ito et al. have also reported that ELM, EZ and IZ condition are all positively correlated with VA in patients with DME [24], as it was found in our study as well.

Even though Ramon Y Cajal proposed that when neurons degenerate, they do not regenerate [25], other publications have reported neuronal regeneration in several situations [26]. In patients with DME, it could not be determined whether the restoration of EZ/IZ on the SD-OCT corresponds to regeneration of the photoreceptor cells [13]. Spaide et al. have described that in some eyes with DME, although the ELM and the EZ were disrupted, outer nuclear layer exists just above these lesions [16]. This might explain why restoration of EZ may be observed after anti-VEGF treatment.

Potential limitation of the study pertains to the fact that marked destruction of the retinal layers occurs in the majority of DME cases. Moreover, shadowing artifacts could theoretically be responsible for variable response in different DME patterns. In addition, the EZ defect measurement has been performed manually using the OCT-caliper. It has also to be mentioned that macular edema can fluctuate significantly, depending on intraobserver, interobserver and diurnal variation. However, to our knowledge this is the first study, which examined the impact of various patterns of DME in SD-OCT to the photoreceptors’ response after treatment with intravitreal ranibizumab injections.

In conclusion, this study described the morphological differences on SD-OCT between the three types of DME and was concentrated on photoreceptors’ condition before and after intravitreal ranibizumab treatment, based on different DME patterns on SD-OCT. Intravitreal anti-VEGF injections may lead to regeneration of photoreceptors, which differs between the three patterns of DME. The corresponding anatomical changes in the number and morphology of photoreceptors should be investigated with further histologic or imaging studies.