A total of 45 eyes of 40 patients with NVG secondary to DM were included in this study, of which 6 eyes (13.3%) had ME at the time of NVG diagnosis. Five patients had bilateral eye involvement: four had no ME in either eye (control group), and one patient had ME in one eye and no ME in the fellow eye.

The mean CRT was 352.33 ± 74.56 μm in the ME group and 233.31 ± 42.22 μm in the control group (Table 1 and Fig. 1A, 1B). The mean age was 57.24 ± 7.86 years in the ME group and 58.63 ± 9.52 years in the control group (p = 0.711). No statistically significant differences were observed between the two groups in sex, systemic diseases (hypertension, hyperlipidemia, ischemic heart disease, cerebrovascular accident, and chronic kidney disease), smoking history, or BMI (p > 0.05) (Table 2).

When NVG manifested in only one eye of a patient, the CRT of the contralateral eye without NVG was compared with that of the eye with NVG. In the control group, the CRT of the contralateral eye was significantly higher (289.1 μm vs. 242.2 μm, p = 0.045), whereas in the ME group, the CRT of the contralateral eye was lower, though not significantly (324.5 μm vs. 375.8 μm, p = 0.293).

Before being diagnosed with NVG, 16 out of 39 eyes (41.0%) in the control group and 2 out of 6 eyes (33.3%) in the ME group received intravitreal anti-VEGF injections at least once (p > 0.999). PRP was performed in 21 of 39 eyes (53.8%) in the control group and 2 out of 6 eyes (33.3%) in the ME group (p = 0.414) (Table 1). The time of the last intravitreal anti-VEGF injection ranged between at least 2.1 and 20 months before establishing NVG diagnosis (average of 7.3 months in the control group and 6.75 months in the ME group, p = 0.866). The time of the last PRP ranged between at least 2.1 and 72 months before making NVG diagnosis (average of 13.3 months in the control group and 13.0 months in the ME group, p = 0.983). No patients received intravitreal anti-VEGF injections at regular intervals; the frequency, timing, and total number of injections varied in each patient.

A total of 18 patients with chronic kidney disease were included, of which 16 patients were in the control group and 2 patients were in the ME group (Table 2). Among 16 patients with chronic kidney disease in the control group, 4 patients were undergoing dialysis: 3 patients were undergoing hemodialysis, and 1 patient was undergoing peritoneal dialysis. Two patients with chronic kidney disease in the ME group were not undergoing dialysis.

Regarding the comparative analysis of systemic disease incidence between the two groups, no significant differences were noted in any parameter. Logistic regression analysis was initially planned but not performed due to the small sample size of the ME group and absence of significant findings in univariate analysis.

Of the 40 patients (45 eyes) who developed diabetic NVG in this study, those with ME (6 eyes, 13.3%) were fewer than those without it (39 eyes, 86.7%). Moreover, no significant differences were observed in the BCVA, IOP, lens state, systemic disease prevalence, and demographic characteristics, such as age and sex, between the groups with and without ME.

Comparing the CRT of the contralateral eye without NVG with that of the eye with NVG, we observed that the eye with NVG had a significantly lower CRT in the control group. Conversely, although in the ME group, the contralateral eye without NVG had a lower CRT, all eyes presented with ME with a CRT of ≥300 μm. These findings suggest a limited correlation between the development of NVG and ME. However, highlighting the limitations of this comparison is important, as it does not account for differences in the severity of diabetic retinopathy between the two eyes, preexisting ME, or prior treatment of ME.

In this study, the proportion of eyes that received intravitreal anti-VEGF injections and PRP before NVG diagnosis was comparable between the control and ME groups. Treatments were administered sufficiently long before NVG diagnosis; in the case with the shortest interval, the last intravitreal anti-VEGF injection or PRP was performed 2.1 months prior to diagnosis. In all cases, bevacizumab was the last anti-VEGF agent administered prior to NVG diagnosis. A pharmacokinetic and pharmacodynamic modeling study estimated that the proportion of intraocular free VEGF remained below 0.001% for up to 8 weeks with intravitreal injections of ranibizumab or bevacizumab, 12 weeks with aflibercept, and 10 weeks with brolucizumab [7]. Regarding PRP, no significant difference in vitreous VEGF levels was found between eyes that received PRP and those that did not in high-risk proliferative diabetic retinopathy [8]. Another study indicated that PRP did not consistently reduce serum VEGF levels, with reductions observed only when diabetic retinopathy regressed after PRP [9]. Based on these findings, it is unlikely that the history of intravitreal anti-VEGF injection and PRP significantly affected intraocular VEGF concentrations at the time of NVG diagnosis in this study. However, the biological effects of anti-VEGF therapy may not be strictly limited to their pharmacokinetic duration. Residual or indirect mechanisms may persist beyond the period of VEGF suppression, and interindividual variability in drug response may also exist. Therefore, although we assumed a minimal effect on VEGF levels at NVG onset, the influence of prior anti-VEGF treatments cannot be completely excluded.

Although this study did not assess longitudinal treatment outcomes, it is possible that ME status at the time of NVG diagnosis was influenced by previous ME episodes and their treatment. While this study focused on diabetic NVG, a previous report on central retinal vein occlusion (CRVO) demonstrated that eyes with longer disease duration, greater numbers of anti-VEGF injections, and more extensive retinal photocoagulation showed increased macular and choroidal atrophy, which in turn lowered the likelihood of ME recurrence [10]. To explore this possibility, we reviewed the treatment indications in eyes with a history of anti-VEGF injections. Of the 18 eyes with prior injections, only 3 eyes in the control group had been treated for DME, while the remainder were treated for proliferative diabetic retinopathy. The time intervals between the final anti-VEGF injection and NVG diagnosis in these three DME cases were 150, 202, and 247 days, respectively. These findings suggest that a prior history of DME was uncommon in our study and that the intervals were sufficiently long to minimize the likelihood that residual effects of past DME influenced ME status at NVG onset. Nonetheless, the potential impact of previous ME and its treatment remains a limitation of our study.

Existing studies indicate an immediate decrease in CRT 1 month after initiating hemodialysis for the first time [11,12]. Takamura et al. [11] reported that the effect of reducing CRT persisted even 1 year after starting dialysis. However, other studies presented conflicting results regarding improvement in ME after dialysis in patients already undergoing regular hemodialysis [13-15]. In this study, patients on dialysis had been undergoing hemodialysis or peritoneal dialysis for at least 1 year before NVG diagnosis. The proportion of patients undergoing dialysis among those with chronic kidney disease included in the present study was not significant (22.2%). Therefore, the effect of dialysis on ME was considered to be nonsignificant.

Although both ME and NVG are known to be driven by elevated intraocular VEGF levels, ME was infrequently observed at the time of NVG diagnosis in our study. In the VIBIM study, aflibercept injections were administered as a treat-and-extend regimen for DME [16], and ME was successfully controlled even when the injection intervals were increased. However, diabetic retinopathy worsened in some cases. In the Diabetic Retinopathy Clinical Research Network (DRCR.net) Protocol S, approximately 60% of patients resumed 0.5 mg ranibizumab injections within 16 weeks of delay, suggesting that anti-VEGF therapy cannot inhibit the progression of diabetic retinopathy for more than 4 months [17]. These findings imply that the progression of diabetic retinopathy and the development of ME may not occur in parallel, even though both are VEGF-driven processes.

This dissociation may also extend to the relationship between ME and NVG. Recent studies on CRVO have shown similar patterns. Lee and Kim [18] reported an NVG prevalence of 21.1% among patients with CRVO with ME treated with as-needed intravitreal bevacizumab injections. In 75% of these patients, the mean CRT was <300 μm at the time of NVG diagnosis, indicating that NVG developed despite ME being well controlled. Other studies also reported that anti-VEGF injections for CRVO did not prevent NVG [19-21]. Taken together, these findings suggest that neovascularization does not always accompany ME, and anti-VEGF injections for controlling ME do not consistently inhibit neovascularization.

This result may be attributed to differences in VEGF concentrations between the aqueous and vitreous humors, their distinct production sites in NVG. VEGF concentration is higher in the vitreous humor than in the aqueous humor in patients with diabetic retinopathy and central or branch retinal vein occlusion [22-24]. In a study by Sun et al. [25], VEGF-A concentration in the aqueous humor was significantly higher than that in the vitreous humor in patients with NVG. This suggests that VEGF is produced locally in patients with NVG. The nonpigmented ciliary epithelium has been reported as an important site of VEGF synthesis in NVG. Furthermore, the ciliary epithelium has been considered the focus of treatment of NVG that does not respond to PRP [6,26]. Therefore, VEGF that causes NVG may exist locally in the anterior chamber and does not affect ME development, and intravitreal anti-VEGF injections to control ME may not have a significant effect on NVG.

VEGF plays important and similar roles in the development of NVG and ME. However, different cytokines mediate their developments. The development of DME involves the angiogenic pathway via VEGF, as well as the inflammatory pathway mediated by various inflammatory cytokines [27,28]. Kwon and Jee [29] investigated the concentration of cytokines in the aqueous humor of patients with DME that did not respond to intravitreal bevacizumab injection. They found that only the pro-inflammatory cytokine interleukin 8 level was significantly higher in the nonresponsive group than that in the responsive group. In NVG, the concentrations of VEGF, inflammatory cytokine interleukin 6, basic fibroblast growth factor, transforming growth factor β, nitric oxide, and endothelin-1 are elevated in the aqueous humor [1]. These pathways are linked to the action of VEGF. However, considering that no association was observed between the development of NVG and ME in previous studies and this study, different mediators are considered.

In one study, VEGF levels in the aqueous humor showed a positive correlation with larger nonperfusion areas in patients with proliferative diabetic retinopathy [30]. In another study, recalcitrant DME was observed in eyes with larger areas of retinal nonperfusion and higher degrees of diabetic retinopathy severity [31]. Given this context, the present study could have determined the relationship between retinal ischemia, diabetic NVG, and ME by measuring the ischemic retinal area at the time of NVG diagnosis. However, diagnosing NVG, fundus fluorescein angiography was simultaneously performed in only 16 of 39 eyes (41.0%) in the control group and 4 of 6 eyes (66.7%) in the ME group, resulting in insufficient data to establish the relationship between retinal ischemia, diabetic NVG, and ME. In all other cases, the date of fundus fluorescein angiography was at least several months before or after the date of NVG diagnosis. Additionally, as the imaging device used in this study was not an ultrawide field device, the imaged range was limited to the posterior to the equator. In the study by DeBoer et al. [32], the ischemic index in NVG patients was significantly higher in the periphery than in the posterior pole. Therefore, comparing between two groups with and without ME solely based on the ischemic index confirmed in the limited range of the posterior pole in this study was not feasible. Nevertheless, when the ischemic index was still calculated in the patients who underwent fundus fluorescein angiography at NVG diagnosis (HRA, Heidelberg Engineering), it was found to be 19.99% ± 11.21% in the control group and 21.43% ± 14.35% in the ME group, without statistically significant difference between the two groups (p = 0.830). Future studies with a larger sample size and widefield angiographic data are warranted to clarify the relationship between macular ischemia and ME in NVG.

In patients with type 2 DM, higher hemoglobin A1c (HbA1c) levels have been associated with increased macular thickness, suggesting that poor glycemic control may contribute to the development or persistence of DME [33]. Moreover, previous research has indicated that certain glucose-lowering agents may influence the risk of DME [34]. A case report demonstrated a significant reduction in DME following 16 weeks of treatment with an sodium-glucose cotransporter 2 (SGLT2) inhibitor [35], and a population-based cohort study reported that long-term use of SGLT2 inhibitors was associated with reduced risk of diabetic retinopathy progression [36]. These findings suggest that not only glycemic control, but also the choice of anti-diabetic medications, may influence ME development. In our study, HbA1c levels were not available in many patients, limiting our ability to evaluate the association between glycemic control and the occurrence of ME. Future studies incorporating HbA1c data and glucose-lowering agents treatment profiles would be valuable to further clarify these associations.

Acan et al. [37] compared patients with DM with and without DME and found that the risk of DME was higher in those with hyperlipidemia, diabetic nephropathy, and diabetic neuropathy and in drinkers. Lopes de Faria et al. [38] reported that hypertension and cardiovascular disease are risk factors for diabetic maculopathy in patients with diabetic retinopathy. In our study, the comorbidities of the patients with diabetic NVG were hypertension (62.5%), ischemic heart disease (10%), hyperlipidemia (20%), cerebrovascular accident (10%), and chronic kidney disease (45%). There was no significant difference in the prevalence of comorbid systemic diseases between the groups with and without ME. Carotid artery stenosis, another possible contributor to ocular ischemia, was not assessed in this study due to the absence of vascular imaging data. Future investigations should consider evaluating carotid status to better understand its relationship with ME development in NVG. However, in patients with severe diabetic retinopathy, confirming the direct effects of each systemic disease is inherently difficult, as poor glycemic control often coexists with unmanageable systemic conditions.

There are several limitations to this study. First, there was patient selection bias due to the retrospective study design. Cases of severe glaucoma accompanied by corneal edema, hyphema, or vitreous hemorrhage due to neovascularization were excluded because CRT could not be measured. Second, the sample size was small (45 eyes of 40 patients), and only 6 eyes had ME, which limited the statistical power to detect significant differences. Given the limited number of ME cases, multivariable regression analysis could not be performed to further explore potential associations with systemic factors. This statistical limitation reduces the strength of the conclusions and underscores the need for larger, prospective studies. Although the initial purpose of the study was to investigate the development of ME and its association with systemic diseases in patients with diabetic NVG, this became difficult due to the small number of cases. Accordingly, we focused on reporting the observed presence of ME and conducting preliminary comparisons. Third, CRT was checked and analyzed only at the time of NVG diagnosis; CRT after diagnosis and treatment were not assessed. Although follow-up CRT measurements were attempted, treatment methods varied among patients, and the time to re-measurement ranged from 1 to 7 months. Nevertheless, this study could provide information on the presence of ME in patients with diabetic NVG and would play an important role in suggesting future research directions.

In conclusion, this retrospective study found that ME was relatively uncommon in patients with diabetic NVG, and no significant associations with systemic diseases were identified. Due to the limited sample size, definitive conclusions could not be drawn. However, the findings provide preliminary insights into the clinical characteristics of ME in this population and highlight the need for larger studies to better understand its underlying mechanisms and risk factors.