FormalPara Key Summary Points

Little is known about how the posterior vitreous interface may contribute to the development and progression of neovascular glaucoma in eyes with prior retinal vascular occlusions.

This study investigated the relationship of posterior vitreous detachment and other potential factors on the risk of developing neovascular glaucoma in eyes with retinal vascular occlusions.

Black race, Hispanic/Latino ethnicity, CRVO, diabetic retinopathy, and absence of posterior vitreous detachment were associated with a significantly higher likelihood of developing NVG after retinal vascular occlusion.

Patients with these characteristics may benefit from more frequent disease screening with gonioscopy and a lower threshold for fluorescein angiography following an ischemic event.

Introduction

Neovascular glaucoma (NVG) is a secondary glaucoma characterized by an increase in intraocular pressure (IOP) due to abnormal growth of new blood vessels on the iris and iridocorneal angle. NVG cases are most commonly precipitated by retinal ischemia [1] with the most common etiologies including diabetic retinopathy (DR) (33%), ischemic retinal vein occlusions, specifically central retinal vein occlusions (CRVO) (33%), and ocular ischemic syndrome (OIS) (13%) [2]. Retinal ischemia upregulates angiogenic cytokines such as vascular endothelial growth factor (VEGF), which induces angiogenesis. These new vessels form a fibrovascular membrane that occludes the trabecular meshwork (TM) and obstructs aqueous outflow leading to elevated IOP. Despite aggressive treatment, visual prognosis is poor, leading to a decrease in quality of life for many patients [3].

Previous studies suggest that intraocular surgery, such as cataract extraction or vitrectomy, increases the risk of developing neovascular glaucoma in patients with proliferative diabetic retinopathy (PDR) [1]. The increased neovascularization risk in these patients may be related to the increased likelihood of posterior vitreous detachment induction (PVD) at the time of surgery—intentional in the case of vitrectomy and coincidental in the case of cataract surgery. Other studies suggest that the presence of PVD may be protective against posterior segment/retinal neovascularization formation, specifically neovascularization of the disc or elsewhere (NVD/NVE) in patients with PDR. These patients lack a vitreous scaffold thus preventing neovascularization [4, 5]. Hikichi et al. found that posterior pole neovascularization was less likely to occur in patients with ischemic CRVO and complete PVD compared to patients with partial or no PVD [6]. Other studies assessing the relationship of PVD to non-neovascular glaucoma found mixed results [7, 8]. However, none of these studies directly investigated neovascularization of the iris, angle, or NVG, and to our knowledge, no study has specifically explored the relationship between PVD status and risk for NVG. The most common cause of NVG is a prior ischemic CRVO. NVG may develop any time between 2 weeks and 2 years after the initial CRVO, though it commonly develops within 90 days, thus earning the nickname “90-day glaucoma” [2]. Extent of ischemia, diabetic retinopathy, central retinal artery occlusion (CRAO), and carotid artery occlusive disease are other risk factors associated with NVG [1]. Understanding other risk factors for NVG can help assess which patients with prior ischemic retinopathy are at greatest risk and guide disease monitoring/screening and subsequent treatment. Thus, this study sought to investigate the relationship of PVD status and the risk of developing NVG in eyes with prior retinal arterial or venous occlusion (RAO and RVO, respectively).

Methods

This retrospective case-control study used electronic medical record data from the Atrium/Wake Forest Baptist Health translational data warehouse. Institutional Review Board (IRB)/Ethics Committee approval was obtained (IRB Number: IRB00093436), and this study adhered to the tenets of the Declaration of Helsinki. The initial Informatics for Integrating Biology & the Bedside (i2b2) query identified 1688 patients with a prior ICD-10 code consistent with either RVO or RAO between January 1, 2016, and December 31, 2022. The i2b2 software is a searchable electronic database designed for researchers to identify study cohorts from the Atrium/Wake Forest Baptist Hospital translational data warehouse. This software allows researchers to extract data using the Data Puller Tool with IRB approval. Inclusion criteria were defined as follows: adults ≥ 18 years old, two or more follow-up visits, history of RVO/RAO, and either a diagnosis of NVG (cases n = 101) or no diagnosis of NVG (controls n = 202) (Fig. 1). For the control group, patients with a history of any form of glaucoma (i.e., primary open-angle glaucoma) and patients < 18 years old were excluded. Cases were age and sex matched 1:2 to controls. PVD status was determined based on review of OCT macula when available and/or chart notes from prior dilated fundus examinations (DFEs).Additional information obtained included demographics such as age, sex, race, and ethnicity. Clinical variables, including prior history of diabetic retinopathy, anti-VEGF injections, panretinal photocoagulation (PRP), retinal surgery, and systemic hypertension, were also recorded. Clinical variables were binary (yes/no) since follow-up time varied among patients. For patients with NVG, it was noted whether they had a history of any clinical variables prior to NVG diagnosis. For example, in these patients, it was assessed whether they had received anti-VEGF injections to treat the ischemic retinopathy prior to a diagnosis of NVG. In patients without NVG, whether they had a prior history of any of the clinical variables was assessed.

Fig. 1
figure 1

Flowchart of case-control study population

Full size image

A total of 1688 initial participants with an ICD-10 diagnosis consistent with a retinal vein occlusion (RVO) or retinal artery occlusion (RAO) were identified and stratified based on presence or absence of an ICD-10 code consistent with glaucoma. After manual review of the patient charts, a total of 101 cases of incident neovascular glaucoma (NVG) after RVO/RAO were identified. These cases were age- and sex-matched 1:2 to controls who never developed NVG.

Retinal vein occlusion (RVO); retinal artery occlusion (RAO); neovascular glaucoma (NVG); primary open angle glaucoma (POAG); pseudoexfoliative glaucoma (PXF); bilateral (OU).

The primary outcome was the development of NVG compared to controls without NVG. A t-test was used to compare the difference between groups for continuous variables, and a Fisher’s exact test was used for categorical variables. Time between the first ischemic event and either the date of NVG diagnosis or the final visit was not normally distributed; a comparison of follow-up time was made with a Wilcoxon rank-sum test. Logistic regression analysis was employed to assess the association between PVD and NVG, without and with adjustment for any demographic or clinical risk factors that were potentially related (p < 0.10) in the bivariate analyses. In addition, sensitivity analyses were run (1) excluding cases with prior RAOs or with avitric eyes and (2) excluding eyes whose PVD status was determined using DFE. In the final multivariable models, a p-value < 0.05 indicated statistical significance.

Results

Table 1 compares characteristics between cases and controls. The groups were well matched on gender and age. Risk of NVG based on eye, lens status, history of hypertension, history of PRP, or history of retinal surgery was not statistically significant (all p > 0.10). Diabetic retinopathy (DR) (p = 0.06) and history of anti-VEGF treatment (p = 0.08) were borderline statistically significant. There was a significant difference based upon race/ethnicity, type of vascular event, and PVD status (all p < 0.05). In addition, patients with NVG were followed for significantly less time than those without NVG (p < 0.0001).

Table 1 Demographic and clinical characteristics stratified by cases (neovascular glaucoma) and controls (no neovascular glaucoma)
Full size table

In the eyes with NVG, 81/101 (80.2%) of eyes had both NVI and NVA clearly documented; 15/101 (14.9%) of NVG eyes had documented NVD and 7/101 (6.9%) with NVE. View of the fundus was limited in N = 16 eyes and N = 15 NVG eyes when assessing for NVD and NVE, respectively. In the control group without NVG (N = 202), two eyes had NVE and one eye had NVD on clinical examination.

A multivariable logistic regression model was used to assess the relationship between PVD and NVG diagnosis. Four covariates (race/ethnicity, ischemic event, history of anti-VEGF, and diabetic retinopathy) were included based upon the bivariate analysis, which suggested these may be confounding risk factors (p < 0.10). Table 2 shows the results from the multivariable model, and Fig. 2 shows the odds ratios from the multivariate analysis. To be classified as having a PVD, patients needed PVD documentation on either dilated fundus examination or OCT macula. Patients without PVD were significantly more likely to develop NVG (OR = 3.070, p = 0.0001) independent of other covariates. As expected, patients with CRVO were more likely than those with BRVO/HRVO to develop NVG (OR = 5.772, p < 0.0001). Patients with diabetic retinopathy were also more likely to develop NVG (OR = 1.980, p = 0.0440). Non-White/non-Hispanics (OR = 2.562, p = 0.0051) and Hispanics (OR = 3.651, p = 0.0288) were more likely than White patients to develop NVG. However, prior anti-VEGF treatment, as expected, was slightly protective for NVG after controlling for other covariates (OR = 0.795, p = 0.4443).

Table 2 Logistic regression analyses of the association between posterior vitreous detachment and other clinical and demographic factors with neovascular glaucoma
Full size table
Fig. 2
figure 2

Odds ratios of the association between posterior vitreous detachment and other clinical and demographic factors with neovascular glaucoma. Posterior vitreous detachment (PVD); anti-Vascular endothelial growth factor (anti-VEGF); central retinal artery occlusion (CRAO); branch retinal artery occlusion (BRAO); branch retinal vein occlusion (BRVO); central retinal vein occlusion (CRVO); hemiretinal vein occlusion (HRVO)

Full size image

A separate sub-analysis was conducted excluding eyes whose PVD status was determined via dilated fundus examination alone (Supplemental Tables 1 and 2) and only retaining eyes that had a reliable OCT, which resulted in 53 cases and 116 control eyes. The key findings were similar to those of the overall cohort, but the analysis was underpowered because of excluding roughly half of the eyes in each group. Risk of NVG based on eye, lens status, prior PRP, history of hypertension, history of anti-VEGF, prior retina surgery, and history of diabetic retinopathy were not statistically significant (p > 0.05). Race/ethnicity and type of vascular event were still significantly associated with NVG (all p < 0.05). PVD status was borderline statistically significant (p = 0.0699), which reflected the reduced sample size. Patients with NVG were still followed for significantly less time than those without NVG (p < 0.0001). The univariable and multivariable logistic regression models also demonstrated consistent associations with PVD, race/ethnicity, and type of vascular event, as seen in the initial analysis (Supplemental Table 2). Patients without PVD were more likely to develop NVG than those with a PVD (OR = 1.752, p = 0.1518) but this did not reach statistical significance in the multivariable model.

The initial multivariable analysis included patients with prior retinal artery occlusions (RAO). In a sensitivity analysis, patients without PVD were still significantly more likely to develop NVG after excluding patients with prior RAO (OR 3.048, p = 0.0002) (Supplemental Table 3). Patients who were avitric were categorized as having “no PVD” in the initial multivariable analysis. In a separate sensitivity analysis excluding avitric cases, patients without PVD were still significantly more likely to develop NVG (OR 3.118, p = 0. 0001) (Supplemental Table 4). Therefore, including cases with prior RAO and avitirc eyes in the multivariable logistic regression analysis did not alter the statistical significance in the relationship between PVD and NVG.

Discussion

This is the first study to demonstrate that the presence of a PVD is protective against development of NVG in eyes with prior retinal vascular occlusions. Additionally, the results demonstrate that being Non-White or Hispanic independently increases the risk for developing NVG after retinal vascular occlusion. Our results also confirmed what has been previously demonstrated: that the presence of CRVO and concomitant DR increase the risk for developing NVG after retinal vascular occlusion [1]. These associations were consistent in analyses where PVD status was confirmed by OCT alone, but we were underpowered to only include analyses using OCT. Regarding the data on neovascularization, 15 eyes had NVD and 7 eyes had NVE among those with NVG. In the control eyes, only two eyes had clearly documented NVE and one had NVD. Overall, this suggests that NVD and NVE were more common in eyes with NVG but were still very low prevalence, so we are not powered to analyze this further. Our findings highlight that the absence of a PVD is another potential risk factor for developing NVG, and it should be added to the list of eyes to be closely monitored for NVG after an RVO or RAO. Such eyes may benefit from more frequent disease screening with gonioscopy and lower threshold for fluorescein angiography in the months following an ischemic event.

An intriguing result of this study was that PVD was protective against NVG development. Those without a PVD were three times more likely to develop NVG. One possible explanation is that there may be a relationship between oxygen diffusion and VEGF signaling which changes following a PVD. A recent study found that patients with a complete PVD on initial presentation of CRVO had lower rates of cystic macular edema (CME) and a lower injection burden at 1 year [9]. If eyes with PVD have greater oxygen diffusion into the vitreous as suggested by Quiram et al. and Stefansson, this could plausibly explain why patients with PVD had lower CME rates and, similarly, in our study had lower NVG rates [10, 11]. In other words, PVD could be protective against edema and neovascularization, possibly because of improved oxygen diffusion and lower VEGF signaling.

Previous research suggests that complete PVD may protect against posterior segment and/or optic disc neovascularization in eyes with severe CRVO because of the absence of a vitreous scaffold [6, 12]. Our results expand upon these prior studies to suggest an essential role for PVD in preventing neovascularization of the angle. Some studies have suggested a relationship of PVD to other kinds of glaucoma such as open-angle glaucoma. For example, a study by Schwab et al. found that PVD was more prevalent in patients with glaucoma (including open angle, pseudoexfoliation, and pigment dispersion) compared to the control group and that patients with glaucoma were more likely to have an advanced PVD compared to the control group [7]. This study hypothesized that apoptosis of the retinal ganglion cells in glaucoma may lead to oxidative stress in the retina promoting vitreous liquefaction and leading to PVD. While this hypothetical pathway appears to be contradictory to our results, this study did not include patients with NVG and therefore suggests a different role for the vitreous interface depending on the type of glaucoma. Overall, the explanation for our study findings is likely multifaceted, and further research is needed to elucidate the mechanistic link between PVD and NVG [7].

Minority groups were disproportionately affected by NVG in our study independent of other clinical and demographic risk factors. Non-White/non-Hispanic patients were 2.5 times more likely to develop NVG, and Hispanic patients were 3.7 times more likely to develop NVG. These findings are consistent with prior studies that have demonstrated pervasive racial and ethnic disparities and adverse health outcomes for minority groups in a number of medical disciplines. These studies have suggested that minorities had less favorable health outcomes due to the high cost of healthcare, lower socioeconomic status, and language or cultural barriers which delay patient diagnosis and/or decrease treatment adherence [13, 14]. Similar discrepancies are seen in a variety of ocular conditions which create adverse health outcomes for racial or ethnic minorities and in some cases are attributed to late presentation. Reasons for delayed presentations have been shown to be multifactorial in other studies and could be related to financial and transportation constraints. Language barriers, despite the incorporation of interpreters, and cultural barriers may also limit patient education and understanding of the disease process or treatment in many cases. For instance, Black patients were more likely to undergo glaucoma surgery compared to White patients, which may be related to underdiagnosis and late presentation necessitating more aggressive treatment in these patients [15]. Similarly, we have found that Black patients are more likely to present with more advanced signs of keratoconus at our institution, and thus require more invasive surgical management [16]. We have also observed that Black patients are more likely to have missed diagnoses of narrow angles as well as present with more advanced chronic angle closure at the time of treatment with laser peripheral iridotomy [17, 18]. More detailed studies assessing the social determinants of health through mixed methods are necessary to fully understand what factors may be driving the racial/ethnic health disparities observed in this study.

Our results suggest that eyes with more ischemic conditions of the retina were also at greater risk of developing NVG. For example, we found that NVG was 5.8 times more likely in patients with prior CRVO compared to patients with prior BRVO. Moreover, those with concomitant DR were ~ 2 times more likely to progress to NVG. Together, these data support the understanding that increased retinal ischemia is associated with higher likelihood of NVG development. Of note, in this study, 19 eyes with NVG had a prior BRVO. Interestingly, previous research suggests that there is a relatively low risk for developing NVG after a BRVO [19]. Specifically, a study conducted by Hayreh et al. found that of 264 eyes with prior BRVO, none developed NVG [19]. A review article by Hayreh et al. states that neovascularization is triggered when at least half the retina is ischemic, and since BRVO typically affects less than a quarter of the retina, the risk of NVG is minimal [20]. Upon further analysis, 10 of the eyes in our cohort had co-existing bilateral DR; 6/10 eyes had proliferative diabetic retinopathy and diabetic macular edema upon diagnosis of NVG. One eye had mild nonproliferative diabetic retinopathy (NPDR), one with moderate NPDR and one eye with severe NPDR. One eye had NPDR but further staging was not documented. None of the eyes with NPDR had associated macular edema. As previously mentioned, DR is a prominent risk factor for developing neovascularization and likely played a significant role in the development of NVG in these eyes [1]. The remaining nine BRVO eyes without DR still developed NVG in the absence of other obvious risk factors. One possibility is that some of these cases had evolved into hemi-RVO over time, though the retrospective nature of this chart review precluded confirming whether this occurred. Though this finding is uncommon, BRVO can still lead to retinal ischemia and therefore NVG.

Interestingly, neither prior history of anti-VEGF injections nor panretinal photocoagulation (PRP) were significantly protective against NVG. We may have been underpowered to detect the small protective effect of anti-VEGF that was observed since a majority of patients were similarly managed with laser and intravitreal medications in both groups. Whether there were differences in the extent of PRP was not possible to determine given the retrospective design. Similarly, since patients were followed for variable lengths of time and not all patients required anti-VEGF treatment or were treated with the same medicine, only the history of anti-VEGF injection was controlled for rather than the number or type of injections. Of note, the median time from an ischemic event to NVG diagnosis was 6 months, indicating that these patients developed NVG relatively quickly after the ischemic event. This may suggest that characteristics of the patient’s anatomy at presentation may impact risk for NVG.

Although we did not see a difference in the proportion receiving PRP between those with NVG and without in this study, prophylactic or more aggressive treatment with argon laser PRP may be advantageous in preventing neovascularization in patients with more ischemia. A prior study conducted by Magargal et al. found that NVG did not develop in eyes with ischemic CRVO treated with prophylactic PRP unless a secondary ischemic event occurred following PRP laser treatment [21]. Additionally, the Central Vein Occlusion (CVO) study group demonstrated that PRP was beneficial for eyes with at least 2 h of iris neovascularization or any angle neovascularization following CVO and evidence of retinal ischemia [22]. Moreover, prior studies demonstrate that patients with DR who develop NVG in one eye have a high risk of developing NVG in the other eye without prophylactic treatment, and bilateral NVG is most commonly due to DR [2]. This suggests a potential role for aggressive prophylactic argon laser PRP to preserve vision in the non-glaucomatous eye in patients with DR who develop NVG in one eye.

This study has several limitations. It is a retrospective case-control review, which limits the amount and type of data available. Retrospective studies are subject to incomplete documentation, difficulty verifying information, and variation in documentation among providers. This is especially of concern for patients who were referred to this tertiary medical center specifically for treatment and management of NVG but were initially diagnosed or treated by outside physicians. Medical records from outside physicians may be difficult to read if using paper charts or incomplete or missing imaging if using electronic records. Furthermore, data were not collected on whether CRVO cases in particular were ischemic or non-ischemic in etiology. While most NVG cases were due to CRVO (N = 71), not all cases had a history of CRVO; thus, we did not analyze extent of ischemia as it would only apply to CRVOs. Moreover, assessing degree of ischemia necessitates previous fluorescein angiography (FA) studies to review. Given that this is a retrospective study, not all patients received the same testing and not all CRVO eyes had an FA available for us to review.

Another limitation of the study was that we did not have high-quality OCT of the macula available on all eyes, and so dilated fundus examination was used to determine PVD status in some of the patients. Given that PVD is relatively common, the presence of PVD may not have been consistently documented on DFE. While documentation of a Weiss ring on clinical examination indicates completion of a vitreous detachment, this specific examination finding was not consistently documented across providers. Thus, we were unable to assess whether the extent of a PVD may affect the risk of developing NVG in this retrospective chart review. However, when we conducted a sub-analysis only using the eyes with high-quality OCT, we observed a similar association of PVD being protective against NVG. We retained all cases and controls who otherwise met inclusion criteria for the primary analysis since findings were similar (but underpowered) if we excluded those with DFE and since there would also be a selection bias if we excluded patients with DFE.

Conclusion

In conclusion, in this case-control study, several factors were independently associated with higher likelihood of developing neovascular glaucoma following a retinal vein or artery occlusion: absence of a PVD, Hispanic/Latino ethnicity, Black race, CRVO, and DR. Thus, presence of a PVD was protective against NVG, which may suggest some relationship between the vitreous interface and NVG risk. Patients with one or more of these characteristics at the time of presentation for a vascular occlusion may benefit from more frequent monitoring for neovascularization of the angle and intraocular pressure trends. Whether such patients may benefit from more aggressive prophylactic interventions should be considered in future studies.