Effects of Intravitreal Anti-VEGF Therapy on Glaucoma-like Progression in Susceptible Eyes

Supplemental Digital Content is available in the text. Précis: Intravitreal anti-vascular endothelial growth factor (VEGF) injections may accelerate glaucomatous change in patients with preexisting glaucoma or ocular hypertension (OHT). The safety of long-term injections in this specific population may be reflected in the need for additional glaucoma interventions. Purpose: The purpose of this study was to investigate whether repeated anti-VEGF injections accelerate structural and functional glaucomatous change in eyes with preexisting glaucoma or OHT. Materials and Methods: This is a retrospective, observational study of injected and noninjected fellow eyes. A total of 28 patients with preexisting glaucoma or OHT, who received ≥6 unilateral anti-VEGF injections for concurrent neovascular retinal disease, were selected for chart review. Primary outcome measures were rate of visual field loss in dB/year, rate of change in retinal nerve fiber layer (RNFL) thickness in microns/year, and need for additional glaucoma medications, surgery, or laser. Results: The number of eyes requiring additional glaucoma surgery or laser was 8 of 28 (28.6%) for the injected group and 2 of 28 (7.1%) for the noninjected group. A significantly greater proportion of injected eyes required invasive glaucoma intervention (P=0.034). Average rate of decline in mean deviation and change in pattern standard deviation were both significantly greater in injected eyes (P=0.029; P=0.019). Estimated mean rate of global retinal nerve fiber layer change was −4.27 µm/y for the injected group and −1.17 µm/y for the noninjected group and was significant only for injected eyes (P=0.014). Only the superior quadrant exhibited thinning that was significantly different between groups (P=0.030). Conclusions: Intravitreal injections were associated with accelerated functional and structural glaucoma-like change in susceptible eyes. Clinicians should assess the need for glaucoma medications or other interventions over the course of anti-VEGF therapy.

I ntraocular pressure (IOP) elevation is a well-known short-term effect of intravitreal anti-vascular endothelial growth factor (VEGF) injections. 1 The safety profiles of ranibizumab, bevacizumab, and aflibercept are well supported for treating diabetic retinopathy, neovascular age-related macular degeneration (nAMD), and retinal vein occlusion (RVO). 2,3 However, patients with preexisting glaucoma or ocular hypertension (OHT) who receive repetitive injections may be at greater risk of experiencing prolonged pressure elevation. 4 Although pressure elevations immediately after treatment are short-lived, 5 their longterm effects are unknown, and prolonged therapy could furthermore increase the risk of sustained OHT. 6,7 A current topic of debate is whether repeated pressure spikes or sustained OHT secondary to anti-VEGF treatment may lead to glaucomatous progression. Overall, evidence on the effect of injections on retinal nerve fiber layer (RNFL) thinning is controversial, but a few studies investigating patients with preexisting glaucoma have found that anti-VEGF therapy can be associated with structural change. 8,9 However, to our knowledge no prior studies have assessed both structural and functional markers of glaucomatous change specifically in glaucoma patients while controlling for natural disease progression.
The goal of our study was to investigate whether repeated anti-VEGF injections accelerate glaucomatous progression in injected eyes by using noninjected fellow eyes as a control. Through our assessment of this particular population, we hope to shed light on the safety of long-term anti-VEGF therapy for treating neovascular retinal disease in susceptible eyes.

Study Design
This is a retrospective, observational study of patients who received ≥ 6 unilateral anti-VEGF injections between January 2010 and May 2018. Institutional review board permission was granted to obtain data from patients who were treated for age-related macular degeneration, RVO, or diabetic retinopathy, with a concurrent ICD diagnosis of glaucoma or OHT.

Inclusion Criteria
Anti-VEGF-injected eyes and noninjected fellow eyes with at least 2 consecutive Humphrey visual field tests or 2 consecutive RNFL thickness measurements obtained by Heidelberg spectral domain optical coherent tomography were identified for chart review. Baseline studies were designated as first test obtained before or within 3 weeks of first injection. Subjects were excluded if the time interval from baseline to last documented study was <12 months. Subjects who received intravitreal steroid therapy or had neovascular glaucoma were also excluded. Follow-up period was defined as the date of first injection to most recent visit before June 2018.
For statistical analysis, injected eyes and noninjected fellow eyes were separated into 2 groups. Primary outcome measures were rate of visual field loss in dB/year, rate of change in RNFL thickness in microns/year, and need for additional glaucoma medications, surgery, or laser. Combination drug preparations were counted as 2 medications. Secondary outcome measures were changed in best-corrected visual acuity (BCVA) over the follow-up period, and maximum IOP documented by applanation or rebound tonometry at the start of clinic visits. BCVA was measured by Snellen chart and converted to logarithm of the minimum angle of resolution (logMAR) scale. Heidelberg software was used to calculate mean global and the quadrant RNFL thicknesses. Standard Automated Perimetry using the Humphrey Visual Field Analyzer (Carl Zeiss Meditec) generated mean deviation (MD) and pattern standard deviation (PSD) values. Exploratory subgroup analyses of primary outcome measures were conducted for the subgroup of all subjects treated for nAMD (n = 14) and for nAMD subjects with OHT only (n = 9).

Analyses of Best-corrected Visual Acuity and Intraocular Pressure
Linear mixed model (LMM) was used for comparisons between the injected and noninjected groups, and differences from baseline at 12-, 24-, and 36-month time points after the start of treatment. For baseline comparisons between groups, a P ≤ 0.05 decision rule was established a priori as the null hypothesis rejection criterion. For comparisons at the 12-, 24-, and 36-month time points, 1 set of hypotheses tested the null hypothesis that mean difference from baseline was 0, for which a 2-sided P ≤ 0.05 decision rule was established a priori as the rejection criterion. The other set tested per time point the null hypothesis that mean difference was the same between groups, for which a Bonferroni multiple-comparison corrected 2-sided P ≤ 0.05/3 decision rule was established a priori as the rejection criterion.

Analyses of Visual Field and Optical Coherence Tomography Parameters
Visual field and OCT serial data included MD, PSD, global and quadrant RNFL thickness measurements at baseline and~12 months after start of treatment. Comparisons of baseline measurements and rates of change over 12 months were conducted by LMMs, in which the randomeffect was the patient and the fixed-effect was the eye group. A P ≤ 0.05 decision rule was established a priori as the rejection criterion for testing the null hypotheses that mean baseline values and mean rate of change for each parameter were the same between groups. LMM was also used to compare absolute global and quadrant RNFL thickness change between groups. Post hoc exploratory subgroup analyses of patients treated for nAMD were performed to assess whether the trends observed in the overall sample remained within non-RVO patients. Subjects in these limited cohorts were unaffected by potential RVO-specific decline in visual field and OCT parameters that could be independent of treatment-induced progression.

Analyses of Glaucoma Therapy-related Outcomes
LMM was used to compare the number of antiglaucoma drops prescribed at baseline, whereas the McNemar test was used to compare the number of eyes with a prior history of glaucoma surgery or laser. The frequencies of receiving additional glaucoma medications, surgeries or laser were also compared between groups using the McNemar test.

Statistical Software
LMM statistical analyses were conducted using SAS version 9.4 Mixed Procedure (SAS Institute Inc., Cary, NC).

Patient and Eye Baseline Characteristics
A total of 28 patients were included in this study. Baseline patient characteristics are summarized in Table 1. Treated eyes received an average of 18.9 total injections, at a mean rate of 0.36 injections per month. There were no significant differences between injected and noninjected fellow eyes with regards to baseline IOP, number of prescribed antiglaucoma drops, or number of prior invasive glaucoma interventions. Median follow-up time was 44 months.

Intraocular Pressure
Average baseline IOP was 16 maximum IOP over the follow-up period was 25.0 mm Hg (95% CI, 22.3-27.7 mm Hg) for injected eyes and 26.8 mm Hg (95% CI, 23.5-30.1 mm Hg) for noninjected eyes, with no difference detected between groups (P = 0.398). Noninjected eyes showed significantly negative mean change in IOP compared with baseline at 24 and 36 months (P = 0.017 and 0.021, respectively), but there were no differences between groups at any time point (Fig. 1B).

Visual Fields
For visual field parameters, mean baseline MD was −6.40 dB (95% CI, −8.67 to −4.13 for injected eyes and -6.08 dB (95% CI, −8.39 to −3.77 dB) for noninjected fellow eyes, with no difference between groups (P = 0.829). Mean baseline PSD was 5.20 dB (95% CI, 3.87-6.54 dB) for injected eyes and 4.71 dB (95% CI, 3.35-6.07 dB) for noninjected fellow eyes, with no difference between groups (P = 0.604). Analyses of visual field outcomes are summarized in Table S6 (Supplemental Digital Content 3, http://links.lww.com/IJG/A334). Estimated mean rate of change in MD was −1.07 dB/y for injected eyes and −0.01 dB/ y for noninjected groups (P = 0.027 and 0.006, respectively), with a significantly greater average rate of decline in injected eyes (P = 0.029). Estimated mean change in PSD was 0.90 dB/y for injected eyes and 0.01 dB/y for noninjected eyes (P = 0.019 and 0.096, respectively), with a significantly greater rate of change in the injected group (P = 0.019). On nAMD and nAMD/OHT subgroup analyses, there were no significant differences between injected and noninjected eyes for either MD or PSD rates of change (Table S2, Supplemental Digital Content 4, http://links. lww.com/IJG/A330).

Retinal Nerve Fiber Layer
OCT studies showed that baseline mean global RNFL thickness was 81.3 µm (95% CI, 73.1-89.5 µm) in injected eyes and 74.9 µm (95% CI, 66.7-83.0 µm) in noninjected eyes, with no difference between groups (P = 0.175). There were no differences in baseline thickness detected between groups for the inferior, nasal, and temporal quadrants (P > 0.2). Mean RNFL thickness of the superior quadrant in injected eyes was significantly thicker compared with noninjected eyes at baseline (P = 0.016). Analyses of OCT outcomes are summarized in Table S7 (Supplemental Digital Content 5, http://links.lww.com/IJG/A335).
The estimated mean rate of global RNFL change was −4.27 µm/y for injected eyes and −1.17 µm/y for noninjected eyes. The mean global RNFL change differed from 0 change for the injected eye group (P = 0.014), but not for the noninjected group (P = 0.145). The between-eye-group mean difference of 3.10 um/y did not reach statistical significance (P = 0.094). Quadrant analyses revealed significant rate of change in thickness only in the superior and inferior quadrants of injected eyes (P = 0.006 and 0.010, respectively), and of these 2 quadrants, the between-group difference was only statistically significant for the superior quadrant (P = 0.030). On subgroup analyses, mean rate of change in thickness in the superior quadrant of injected eyes was −4.67 μm/y for nAMD subjects (P = 0.027) and −2.41 μm/y for nAMD/ OHT subjects (P = 0.021; Table S3, Supplemental Digital Content 6, http://links.lww.com/IJG/A331).
Absolute mean global RNFL thickness change in 12 months was −8.46 µm for injected eyes and −1.52 µm for noninjected eyes. Absolute change in thickness was significant for injected eyes (P = 0.001) and not for noninjected eyes (P = 0.509), but the difference between groups did not reach statistical significance (P = 0.067). RNFL descriptive statistics are provided in the Supplemental Table S1 (Supplemental Digital Content 7, http://links.lww.com/IJG/A329).

Glaucoma Therapies
The average number of prescribed antiglaucoma drops before the start of injections was 1.3 for injected eyes and 1.4 for the noninjected eyes, with no difference between groups (P = 0.908). The number of eyes with a history of glaucoma surgery or laser before first injection was 10 of 28 (35.7%) for the injected group and 8 of 28 (28.6%) for the noninjected group, and the proportions were not different between groups (P = 0.05). Nine of 28 (32.1%) injected eyes and 6 of 28 (21.4%) noninjected eyes required the addition  of at least 1 antiglaucoma medication over the follow-up period and the proportions were not different between groups (P = 0.178). The number of eyes requiring glaucoma surgery or laser over the follow-up period was 8 of 28 (28.6%) for the injected group and 2 of 28 (7.1%) for the noninjected group. A significantly greater proportion of injected eyes required invasive intervention over the followup period (P = 0.034). Subgroup analyses showed no significant between-group differences in proportion of eyes requiring additional glaucoma medications or interventions for either nAMD eyes (P = 0.563 and 0.157, respectively) or nAMD/OHT eyes (P = 0.563 and 1.000, respectively).

DISCUSSION
Glaucoma patients receiving serial intravitreal injections exhibited statistically significant structural progression only in treated eyes, with an average rate of global RNFL thinning of 4.27 µm/y. Although the mean difference in global RNFL thinning between groups of 3.10 µm/y did not reach statistical significance (P = 0.094), quadrant analyses revealed significant thinning in the superior and inferior quadrants of injected eyes (P ≤ 0.010), and a statistically significant difference between groups for the superior quadrant (P = 0.030). This pattern of structural change that is more pronounced in the superior and inferior quadrants of treated eyes could be consistent with early glaucomatous optic neuropathy. 10 Treated eyes also showed accelerated decline in visual field parameters compared with fellow eyes (P ≤ 0.029). Our results contribute to the ongoing discussion on the safety of long-term therapy and support that the risk of accelerated glaucomatous change may be associated with anti-VEGF treatment.
Although IOP elevations after anti-VEGF injections are generally transient, 1,11 higher injection frequency and greater total number of injections have been associated with OHT. 12,13 In their recent literature assessment, Hoguet et al 5 found that the data on long-term OHT are mixed, with 7 studies supporting the existence of sustained IOP elevation, and 6 studies concluding no long-term change. Large-scale clinical trials support that therapy is safe, 2,14 but concurrent glaucoma could portend a greater risk for sustained OHT. 4 In the present study of glaucomatous eyes, mean baseline pressures were 16.7 mm Hg for the injected group and 17.7 mm Hg for the noninjected group. Mean maximum pressures were 25.0 mm Hg for the injected group and 26.8 mmHg for the noninjected group, with no difference detected between groups. Compared with baseline, noninjected eyes had significantly lower mean pressures measured at the 24 and 36-month time points (P ≤ 0.021). Even though sustained OHT is defined in various ways within the literature and was not evaluated as a primary endpoint in this study, our findings at 12-month time intervals suggest that prolonged therapy may impact IOP trends and the ability to control pressures in glaucomatous eyes.
The safety of long-term injections in this subset of eyes may be reflected in the need for additional glaucoma medications, surgery, or laser over the duration of treatment. In this study, 32.1% of injected eyes and 21.4% of noninjected eyes required the addition of at least 1 antiglaucoma medication over the follow-up period. 28.6% of injected eyes and 7.1% of noninjected eyes further received glaucoma laser or surgery for pressures uncontrolled on drops alone. Eadie et al 15 reported that a greater number of injections increased the risk of trabeculectomy, glaucoma drainage device, or cycloablative procedure. In line with this, we found that the proportion of injected eyes requiring invasive glaucoma intervention was significantly higher than for noninjected eyes (P = 0.034). These findings highlight the importance of monitoring for elevated pressures and complications in eyes receiving long-term treatment.
Although there is no general consensus on whether intravitreal therapy accelerates structural glaucomatous change, our study specifically examines glaucoma patients and offers a unique perspective by controlling for natural disease progression. A meta-analysis combining 6 studies of nonglaucomatous eyes found that anti-VEGF injections did not have a significant effect on RNFL thinning. 16 Notably, subgroup analyses revealed that the low-biased, controlled experiments did show significant decrease in RNFL thickness. Park et al 9 assessed RNFL thinning in injected eyes both with and without glaucoma, randomized to prophylactic brimonidine before injections or not. They found significant thinning only in nonpretreated glaucomatous eyes, thus concluding that these patients were at risk for injection-associated RNFL thinning. In contrast, Saleh et al 8 found no change in RNFL thickness in glaucoma patients treated for neovascular AMD. The above studies were all limited by a lack of control for natural glaucomatous progression in noninjected eyes, which is addressed in the present study. We found that injected eyes exhibited a significant mean rate of global RNFL thinning (P = 0.014), whereas nontreated fellow eyes did not. Although the difference in thinning between groups did not reach statistical significance (P = 0.148), we believe there might be a relationship between repeated anti-VEGF injections and structural disease progression in susceptible eyes.
Our results further support that injections may be associated with functional glaucomatous change. Saleh et al 8 analyzed visual field decline as a secondary outcome in glaucoma patients receiving long-term therapy and found no significant change in MD or number of absolute scotomata. In the present study, there was significant MD decline in both injected and noninjected eyes, with a greater rate of change detected in injected eyes (P = 0.029). Injected eyes also showed significant worsening of PSD, with a greater rate of change compared with noninjected eyes (P = 0.019). In contrast to conclusions by Saleh and colleagues, we suggest that accelerated functional disease progression may be related to anti-VEGF injections.
Given our findings that injected eyes experienced greater change in glaucoma parameters compared with fellow eyes, there may be some component of anti-VEGF treatment that is damaging and accelerates glaucomatous change outside of natural disease progression. The observed progression could be associated with a combination of underlying glaucoma, short-term pressure fluctuations, and sustained OHT. Immediate postinjection pressure spikes theoretically could play a role, and prophylactic IOP-lowering medications have emerged as a strategy to dampen these fluctuations. 17,18 Although Park et al 9 reported that pretreatment was protective, the overall evidence for this method is lacking. Anterior chamber paracentesis (ACP) is another method that has emerged to address IOP spikes. Enders et al 19 found that patients receiving ACP at each injection demonstrated significantly lower mean RNFL loss than non-ACP patients, hypothesizing that regular paracentesis was able to flatten IOP spikes during anti-VEGF therapy. Sisk et al 20 showed that serial ACP did indeed reduce immediate postinjection IOP, and furthermore that optic nerve changes were stabilized in patients who had sustained elevation of preinjection IOP. Thus, techniques for blunting IOP spikes may hold promise and should be considered for patients receiving repeated injections.
Outside of short-term pressure spikes, long-term therapy may further contribute to sustained OHT. 21 The exact mechanism is unknown; however, theories include silicone or protein particle obstruction of the trabecular meshwork (TM), or direct toxic effects of anti-VEGF molecules on TM cells. 21,22 In addition, anti-VEGF inhibits nitric oxide (NO), which is thought to modulate trabeculocyte size and Schlemm canal vasodilation. 23 Contributors to decreased aqueous outflow facility and sustained OHT in the context of injections are likely multifactorial.
Although the major risk factor for glaucoma is elevated IOP, dysregulation of the vasculature surrounding the optic nerve head (ONH) is another important element to consider. 24 Decreased ocular blood flow and increased vasculature resistance are associated with open-angle glaucoma. 25 Various factors released by the endothelium, including NO, are thought to maintain local perfusion. Thus, inhibition by anti-VEGF may impact myogenic autoregulation and affect hemodynamic parameters. The idea that anti-VEGF therapy could disrupt homeostatic levels of ONH blood flow in susceptible eyes is corroborated clinically. Patients with both AMD and glaucoma exhibited a considerable decrease in ocular blood flow 1 month after anti-VEGF treatment, whereas those without concomitant glaucoma did not. 26 In addition to sustained OHT, the effect of therapy on ONH and retinal vasculature hemodynamics warrants further study.
This study focused on a susceptible subset of patients, but the possibility of glaucoma arising de novo over the course of therapy also requires attention. Changes in IOP dynamics and disruptions to retinal vasculature may alter the risk of developing disease in patients without prior history of glaucoma. The injection procedure itself, by adding volume to a relatively fixed space, could damage the TM in previously healthy eyes. An exploratory subgroup analysis of nAMD patients with only OHT and no diagnosis of glaucoma was conducted and showed that these subjects may have exhibited superior quadrant RNFL thinning (P = 0.027). Although the sample of 9 subjects is too small to make conclusions on significance, we believe that structural changes may be observed in susceptible patients without glaucoma. Concurrent management by a glaucoma specialist may be beneficial for those at risk for either developing or worsening glaucoma. It is also reasonable for clinicians to consider techniques that address immediate postinjection pressure spikes. Although more research is required to develop practice guidelines, such measures could be helpful for certain patients.
This study was limited by its retrospective design, small sample size, and variety of vitreoretinal conditions included. Although our sample is representative of the multiple comorbidities of patients in a typical retina practice, visual deterioration in eyes subsequent to RVO could have impacted the observed changes in glaucoma parameters. As the subjects were all unilateral RVO patients, their inclusion may have introduced some selection bias. On exploratory subgroup analyses of only patients treated for nAMD, changes in MD and PSD were not statistically different between injected and noninjected eyes. Notably, there may have been superior quadrant RNFL thinning only in treated eyes, a pattern consistent across both nAMD subgroups (P ≤ 0.027). We acknowledge that subgroup analyses do not parse out the effects of vitreoretinal comorbidities, and moreover that these results must be interpreted with caution given insufficient sample sizes. However, as the structural findings of patients treated for nAMD appear to trend toward those of the injected group as a whole, we believe that the decline observed in the overall study cannot be explained by RVO alone and could have been at least partially glaucoma-like in nature. It is also possible that neuroretinal toxicity due to repeated drug injections contributed to a clinical picture mimicking disease progression.
In conclusion, intravitreal anti-VEGF injections were associated with accelerated functional and structural glaucomalike progression in eyes with concurrent glaucoma or OHT. Injected eyes were more likely to require invasive glaucoma intervention, thus clinicians must be cautious when administering treatment and monitor for complications. To our knowledge, this is the only study of injection-associated glaucoma-like change in glaucoma patients using the fellow eye as a control. This population appears to be susceptible and should be further studied to elucidate the mechanisms and drivers of disease acceleration. Finally, design of industry trials for novel anti-VEGF agents should consider this population specifically when assessing safety and efficacy.