Vision loss associated with diabetic retinopathy (DR) is most commonly caused by diabetic macular edema (DME) (1). The Diabetes Control and Complications Trial (DCCT) reported that 27% of patients with type 1 diabetes developed macular edema within 9 years of diabetes onset (2). Other studies indicate that in type 2 diabetes patients, the prevalence increases from 3% within 5 years of diagnosis to 28% after 20 years (3). Although several treatment options are available, no consensus on DME treatment based on patient status has been achieved.

Vascular endothelial growth factor (VEGF) is an important mediator of abnormal vascular permeability in eyes with DME (4). Anti-VEGF injections are generally proposed as first-line therapy for center-involved DME and are effective in improving visual acuity (VA), with 10%–40% of patients achieving significant improvement in VA after 1 year of treatment (5, 6). However, a considerable proportion have unsatisfactory response to anti-VEGF agents; 40% of eyes with DME do not or have suboptimal response to anti-VEGF treatment (7, 8). Nonetheless, there is little information to date about the prognostic factors of poor responders.

Optical coherence tomography (OCT) images are readily available to physicians and provide detailed information. Structural changes presumably reflect part of the complex pathophysiologic processes occurring in DME. Furthermore, anatomical measures on spectral-domain (SD) OCT can predict treatment success or failure of various therapies (9). Distinct structural changes identifiable on SD-OCT could reflect part of the intraocular pathophysiologic process change after anti-VEGF treatments and help predict the treatment response.

Among patients with DME refractory to anti-VEGF therapy after a loading dose of three consecutive monthly injections, those who were switched to other treatment modalities (e.g., corticosteroids) had better visual and anatomical outcomes at 12 months than did those who continued with anti-VEGF therapy (10). Post hoc analysis from the DRCR.net Protocol I study also indicates that early central macular thickness (CMT) response to anti-VEGF is a significant prognostic indicator of medium to long-term anatomical outcomes in DME (11). Accordingly, the early identification of patients who would not benefit from first-line treatment with anti-VEGF therapy is critical. Real-world studies have become increasingly important in providing evidence of treatment effectiveness in clinical practice. They can therefore provide information on the long-term safety, particularly of rare events, and efficacy of drugs in large heterogeneous populations, as well as information on utilization patterns and health and economic outcomes (12). We aimed to investigate whether the characteristics identified on SD-OCT could be predictive markers of treatment response after three monthly anti-VEGF therapies in DME patients.

In this real-world evidence-based study, we identified partially continuous IS-OS layers as biomarkers that predict better response to anti-VEGF therapy in DME. In contrast, ERM is a significant predictor of poor response. DME has a complex pathogenesis, with multiple factors contributing to its pathophysiology, including angiogenic, inflammatory, hypoxic, and hemodynamic processes that lead to the breakdown of the blood-retinal barrier (BRB) and leakage of intraretinal fluid (13). Anti-VEGF injections are generally recommended as the first-line therapy for DME; however, refractory cases are not uncommon. A post-hoc analysis of the DRCR.net Protocol I reported an approximately 40% prevalence of refractory DME after 2 years of monthly intravitreal ranibizumab treatment (7). Combined data from the RIDE/RISE trial found that 23% of eyes receiving intravitreal ranibizumab had persistent macular edema at the end of the study period (14). The real-world prevalence of refractory DME may be higher than estimated in these studies, as rigorous enrolment and follow-up protocols in clinical trials are unlikely to be fully replicated in everyday practice (15). An important issue is the possibility of early identification of patients who would not benefit from first-line anti-VEGF therapy.

VEGF is significantly higher in all types of DME than that in the eyes of non-diabetes patients, indicating that VEGF is equally important for any morphological changes in eyes with DME (16). However, evidence indicates that bioactive factors such as cytokines are also released into the retina. The proposed pathophysiology of each type is quite different; thus, each DME type has its own morphological and topographic characteristics (17). interleukin (IL)-6, a pro-inflammatory cytokine, intraocular levels of IL-6 were significantly higher in eyes with SRD than in eyes with DRT or CME, implying active inflammation. A recent study showed a better response to dexamethasone implants in eyes with SRD (9). The predictive value of SRD at baseline for the treatment response to anti-VEGF agents in DME remains controversial. Although some studies reported a significant improvement in VA in patients with SRD at baseline (18, 19), others found no difference or even worse functional results (20, 21). Univariate analysis to assess whether SRD is responsive to anti-VEGF agents in the current study showed that SRD significantly responded to anti-VEGF agents. However, in the multivariate survival analysis, treatment response to anti-VEGF regimens was not a significant influencing factor in patients with SRD. This can happen when SRD and other covariates are highly correlated. Furthermore, additional variables (ex. ERM) may explain more of the variance in the outcome variable, and thus reduce the impact of the initially significant of SRD.

In DME, the concentration of intraocular VEGF is significantly correlated with IL-6 levels (16). Anti-VEGF therapy reduces intraocular subclinical inflammation, and the aqueous humor concentration of IL-6 is decreased after anti-VEGF treatment (22). This could explain the response to anti-VEGF therapy in eyes with SRD in the current study. Our result was also consistent with that in the study by Sophie et al. (18) who reported that suppression of VEGF effectively eliminated subretinal fluid. Future prospective comparative investigations of the efficacy of anti-VEGF and of dexamethasone implants in eyes with SRD are required to optimize patient management.

In the current study, 45.8% of all patients presented with ONL cysts, and the majority had large ONL cysts (100–200 µm), mainly occurring at a relatively late stage of the disease. We found that large ONL cysts at baseline are less likely to be refractory group after anti-VEGF treatment at 4 months. Previous studies have reported that large ONL cysts negatively affect macular function and are predictive of worse VA outcomes after anti-VEGF therapy (23, 24). Elevated VEGF levels in DR affect the inner BRB, leading to increased vascular permeability, decreased osmotic gradient, extracellular fluid accumulation, and cyst formation (25). Furthermore, liquefaction necrosis of Müller cells and related inflammatory factors result in fluid accumulation in the cystic space (17). However, unlike SRD, IL-6 and IL-8 levels were not significantly increased in eyes with cystic changes (16). This indicated that the eye may not be in an active inflammatory state; rather, it could be a remnant of a previous inflammatory reaction (16). Anti-VEGF agents have been shown to decrease permeability and improve inner BRB by interacting with junctional proteins in the vascular endothelium (23). Our results support this finding and are consistent with a recent report that throughout anti-VEGF treatment, significant regression of ONL cysts accompanied notable improvement of macular function with a substantial decrease in their size (23). Nevertheless, we could not find an association between large ONL cysts and a mean reduction of CMT after anti-VEGF treatment at four months. Furthermore, only 3 eyes presented with giant ONL cysts (>200 µm), and we were unable to find an association between treatment outcomes and anti-VEGF agents in patients with giant ONL cysts.

The pathogenesis of DRT involves the persistent breakdown of the inner BRB and impairment of fluid absorption by Müller cells (17). DRT can be localized or more diffusely encompass the macula. Previous studies have reported that intravitreal anti-VEGF therapy is more effective for the DRT type than for the other types of DME (20, 26). Nevertheless, no study assessed whether the degree of the DRT would interfere with the treatment response or not. To clarify the relationship between DRT and response to anti-VEGF treatment, we first qualitatively evaluated the width of the DRT on a standard horizontal 6-mm B-scan OCT centered through the fovea and further stratified it into three subgroups (≤ 1, 1–3, and 3–6 mm). The results showed that the width of DRT was inversely proportional to the odds of poor response. Specifically, there was a trend indicating that the degree of DRT was proportionally associated with better response, although this was not significant in multivariate analysis.

There is a high incidence of vitreomacular interface abnormality (VMIA) among DME patients (27). The current study found a 20.8% incidence of ERM in our cohort. Although DR and its severity are risk factors for developing secondary ERM (27, 28), cases of VMIA are excluded from major clinical trials, even though DME is associated with this condition in 25% of patients (29). Nevertheless, knowledge regarding the effect of VMIA on the response to anti-VEGF treatment in patients with DME has not been thoroughly investigated. Ercalik et al. retrospectively evaluated 56 eyes with or without ERM and found a negative effect of ERM on intravitreal anti-VEGF treatment (30). Wong et al. conducted a prospective study of 104 eyes with DME treated with anti-VEGF and found that ERM was associated with a worsened visual and anatomic response (31). Notably, neither study considered other OCT biomarkers; thus, the findings might not completely represent the true impact of ERM in eyes with DME.

Considering the diversity of OCT morphology in DME, we included various OCT biomarker characteristics and considered ERM as a variable in eyes with refractory DME, despite anti-VEGF therapy. Furthermore, we used a multivariate statistical model to analyze the association between treatment response and each OCT biomarker. Our results showed that ERM significantly increased the odds of poor response. According to previous study on ERM pathology in diffuse DME, multilayered membranes are mainly composed of hyalocytes and myofibroblasts. Hyalocytes were shown to produce VEGF and can transdifferentiate into myofibroblasts, known for their contractive properties (32). Furthermore, contraction of the ERM may cause perifoveal capillary leakage and aggravate macular edema. It has also been demonstrated that glial cells in ERM produce various cytokines and growth factors. VEGF and its receptors, as well as IL-6, are localized to cells in the ERM of patients with DR, thus further increasing inflammation and possibly promoting DME persistence (30, 33). Furthermore, ERM may act as a physical barrier and decrease drug penetration after intravitreal injections of anti-VEGF in DME treatment (34).

The connective tissue growth factor (CTGF) is one of the most potent profibrotic factors. It can stimulate fibroblast proliferation and collagen deposition, resulting in fibrosis (35). Anti-VEGF has been reported to cause hypoxia in vascular endothelial cells and increase CTGF expression, which plays an important role in ERM formation (36). As a result, anti-VEGF therapy may potentially aggravate ERM contractions and interfere with the resolution of macular edema in diabetes. These results may explain the increased likelihood of poor response in this group. Nevertheless, consistent with the guidelines for DME management by retinal specialists, PPV is currently recommended as a therapeutic option in cases of DME associated with VMT (37). In the absence of traction formation, there is no consensus on the role of PPV in the actual treatment of diabetic eyes. Our results call for further comparative studies and treatment modalities other than anti-VEGF in DME patients presenting with ERM-impaired visual and anatomic outcomes.

HRF represents subclinical lipoproteins that extravasate after inner BRB breakdown. It is initially present in the inner retinal layers and subsequently migrates to the outer retinal layers (38). HRF is an important imaging marker for retinal inflammation (39). However, the predictive value of HRF for visual outcomes after anti-VEGF treatment in DME is unclear (9). In our study, we did not find that the presence of HRF was associated with treatment response after anti-VEGF therapy. The integrity of outer retinal layers is a direct indicator of the health of the retinal photoreceptors and retinal pigment epithelium. IS-OS integrities is an important factor for predicting VA after treatment. Subjects with long-standing DME may demonstrate focal or diffuse loss of integrity of the IS-OS junction. Previous studies have reported that IS-OS integrity can be expected to recover after anti-VEGF therapy (17, 23). Our results support this finding and show a better response to anti-VEGF therapy in eyes with partially continuous IS-OS layers.

Our study has some limitations. First, its retrospective and non-randomized design precluded a well-matched control enrollment. Second, the sample size was small, which may have hindered the significance of the results. Third, we prescribed three anti-VEGF agents to treat DME in the real-world clinical practice setting. Although most eyes were treated with ranizucimumab (83.3%), we did not assess each anti-VEGF regimen separately, and thus, the different efficacy between each agent may not have been accounted. Fourth, in our study, despite the odds of gaining BCVA ≥10 letters trended toward the OCT predictive value of CMT reduction, no OCT biomarkers showed significant predictive value for good BCVA response at 4 months. We found that the OCT predictive value of CMT reduction cannot fully translate into the change of VA, which was consistent with the study of a post hoc analysis of the protocol T randomized clinical trial (40). They found changes in CMT appear to account for only a small proportion of the total variation in changes in BCVA, and concluded that changes in CMT cannot support as a surrogate for changes in VA in evaluating anti-VEGF for DME. Despite these limitations, an important strength of our study is that we included various common OCT markers in patients with DME, which yielded ample information and helped us tailor timely and individualized treatment during daily practice.

In conclusion, partial IS-OS continuity is the marker that predicts better response to anti-VEGF treatment in eyes with DME. In contrast, the presence of ERM is a significant predictor of poor response. Our results raise the pertinent issue that DME patients with ERM are significant poor responders to anti-VEGF therapy and may benefit more from other therapeutic approaches.

The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author.

The studies involving human participants were reviewed and approved by Institutional Review Board of the Research Ethics Committee of Hualien Tzu-Chi Hospital and Buddhist Tzu-Chi Medical Foundation. The patients/participants provided their written informed consent to participate in this study.

T-CH and M-SH had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Concept and design: M-SH. Acquisition, analysis, or interpretation of data: T-CH, G-HD, Y-CC, and M-SH. Drafting of the manuscript: T-CH, G-HD, and M-SH. Critical revision of the manuscript for important intellectual content: T-CH, F-LC, and M-SH. Statistical analysis: T-CH and G-HD. Administrative, technical, or material support: G-HD, M-SH, and Y-CC. Supervision: T-CH, M-SH. All authors contributed to the article and approved the submitted version.