Effect of Intravitreal Injections on Retinal Imaging Metrics in Glaucomatous and Non-Glaucomatous Eyes

Abstract

Purpose of Review:

To summarize the available literature on retinal imaging metrics in the context of intravitreal injections in glaucomatous and non-glaucomatous eyes.

Recent Findings:

The retinal nerve fiber layer (RNFL) in injected non-glaucomatous eyes appears to thin at a similar rate to uninjected fellow eyes. A total of four studies evaluating RNFL thinning in injected glaucomatous eyes yielded mixed results, with more recent longitudinal investigations suggesting a potential association. The ganglion cell-inner plexiform layer is also being studied as a potential endpoint in both glaucomatous and non-glaucomatous eyes following intravitreal injections.

Summary:

There appears to be minimal association between intravitreal injections and RNFL thinning among non-glaucomatous eyes. However, in glaucomatous eyes, fewer studies have been published, and results have varied due to challenges in study design. Additional prospective studies with longer follow-up periods are warranted.

Introduction

Glaucoma is the most common cause of irreversible blindness in the world [1]. Elevated intraocular pressure (IOP) is a key risk factor in glaucomatous optic neuropathy. Wet age-related macular degeneration (AMD), diabetic macular edema (DME), and retinal vein occlusion (RVO) are retinal diseases that impact central visual acuity, and are typically treated with intravitreal anti-vascular endothelial growth factor (VEGF) injections [2, 3]. Anti-VEGF injections have become one of the most common procedures performed in medicine, with an estimated 5.9 million injections being performed in the USA alone in 2016 [4].

The safety profile of intravitreal injections in glaucoma patients has been evaluated closely [5]. Studies demonstrated a transient but sizable effect of intravitreal injections on IOP, which led to concerns regarding potential irreversible effects of post-injection IOP elevations [6–10]. Many studies have assessed whether these recurrent IOP spikes may damage retinal structures and whether intravitreal injections may accelerate the progression of glaucoma [11]. In this review, we aim to summarize the available literature regarding the impact of intravitreal injections on the retinal nerve fiber layer (RNFL) and the inner retina. We have attempted to focus on studies that have utilized spectral-domain ophthalmic coherence tomography (SD-OCT) imaging to describe the effect of intravitreal injections on retinal and optic nerve head structures.

Retinal Thinning Following Intravitreal Injections in Non-Glaucomatous Eyes

Intravitreal injections have generally been regarded as safe procedures, with transient rises of IOP found to be the most common adverse effect [7–10]. Some studies have shown that a transient IOP spike can lead to a sustained IOP elevation in a small percentage of injected eyes [12–19]. Since many patients receive monthly injections, repeated IOP spikes become clinically relevant. Studies have evaluated whether such repeated IOP elevations following intravitreal injections lead to damage of the RNFL and the ganglion cell layer (GCL)/ganglion cell-inner plexiform layer (GCIPL) [20–25].

In 2016, Shin et al. performed a meta-analysis reviewing the available literature on the effects of intravitreal injections in non-glaucomatous eyes [20–26]. The studies included in this meta-analysis were conducted in four countries (the USA, Spain, Korea, Turkey), with three of the six studies originating in Turkey [20, 23, 24]. The study sample ranged from 41 to 148 injected eyes with a follow-up period of 1 to 3 years. The number of injections ranged from 4.8 to 16, with an average of 8.5 injections. The authors found no statistically significant difference between the RNFL thicknesses of injected eyes and control eyes [26]. However, four of the six studies were retrospective. The authors then performed a subgroup analysis in which they assessed the two prospective studies and found a statistically significant decrease in RNFL thickness among treated groups [24–26]. Ultimately, they recommended additional prospective studies to provide further guidance. We would encourage interested readers to review the meta-analysis for further details [26].

Additional studies evaluating the effect of intravitreal injections have been published since this meta-analysis. We reviewed six studies that also investigated the relationship between intravitreal injections in non-glaucomatous eyes and RNFL thinning, as well as ganglion cell complex (GCC)/GCIPL changes (Table 1). Four were prospective studies [27–30], while two were retrospective [31, 32]. The majority of treated eyes had wet AMD, while two studies also included DME and RVO [28, 31]. The sample sizes ranged from 20 to 49 eyes with an average of 29 eyes in the treatment and control groups. The follow-up period ranged from 11.2 to 96 months [28–31]. The number of injections ranged from 5 to 31.5, the majority being 5–8 injections [27–32]. Of note, Beck et al. evaluated eyes with the most injections, an average of 31.5 injections over 3.75 years [32]. In contrast to the prospective studies included in the meta-analysis by Shin et al., 3 of the 4 recent prospective studies found no statistically significant difference between the RNFL thickness of injected eyes and untreated fellow eyes [27, 29, 30]. In total, five of the six studies evaluated found no significant difference in the RNFL thickness of treated versus control eyes [28]. However, the small sample sizes of the studies, the small number of injections, and the short follow-up period limit the strength of the conclusions. Many wet AMD patients require injections for years; this degree of longitudinal evaluation was not captured in most of these studies.

Table 1.

Studies evaluating imaging metrics following intravitreal injections in non-glaucomatous eyes.

Author, Year	Country	Study Design	Number of Injected Eyes	Study Controls	Diagnoses	Primary Outcomes	Anti-VEGF medication	Mean Number of Injections	Mean Follow Up Period	OCT System	Study Findings
Horsley et al., 201021*	USA	Retrospective	41	None	Wet AMD	RNFL thickness	Ranibizumab, Bevacizumab, Pegaptanib	16	2.25 years	Stratus	No significant difference in RNFL thinning
Martinez-de-la-Casa et al., 201225*	Spain	Prospective	49	27 fellow eyes(no AMD or dry AMD)	Wet AMD	RNFL thickness	Ranibizumab	4.8	1 year	Spectralis	Greater RNFL thinning in injected eyes
Sobaci et al., 201323*	Turkey	Retrospective	65	None	Wet AMD	RNFL thickness	Ranibizumab, Bevacizumab	5.65	1.17 years	Cirrus	No significant difference in RNFL thinning
Shin et al., 201422*	Korea	Retrospective	148	183 age matched eyes (AMD, DME, or RVO)	Wet AMD, DME, RVO	RNFL thickness	Ranibizumab, Bevacizumab	5.7	1.75 years	Spectralis	No significant difference in RNFL thinning in all subgroups
Demirel et al., 201520*	Turkey	Retrospective	29	27 age matched eyes (healthy) or fellow eyes (dry AMD)	Wet AMD	RNFL thickness	Ranibizumab	13.9	3.25 years	Cirrus	No significant difference in RNFL thinning
Parlak et al., 201524*	Turkey	Prospective	44	44 fellow eyes (no AMD or dry AMD)	Wet AMD	RNFL thickness	Ranibizumab	4.86	1 year	Spectralis	No significant difference in RNFL thinning
Beck et al., 201632	Switzerland	Retrospective	34	34 fellow eyes (unspecified diagnosis)	Wet AMD	Macular RNFL thickness, GCL thickness	Ranibizumab, Aflibercept, Bevacizumab	31.5	3.75 years	Spectralis	Greater GCL thinning in injected eyes. No significant difference in temporal RNFL thickness (measured at outer segment of ETDRS ring)
Jo et al., 201629	Korea	Prospective	20	20 fellow eyes (healthy)	Wet AMD	RNFL thickness	Ranibizumab	5	1 year	Cirrus	No significant difference in RNFL thinning
Zhang et al., 201631	China	Retrospective	49	49 fellow eyes (unspecified diagnosis)	Wet AMD, DME	RNFL thickness	Conbercept	8	1 year	Spectralis	No significant difference in RNFL thinning
Zucchiatti et al., 201730	Italy	Prospective	24	None	Wet AMD	RNFL thickness, GCC thickness	Ranibizumab	5.3	1 year	Cirrus	No significant difference in RNFL or GCC thickness between baseline and after 12 months
Gomez-Mariscal et al., 201928	Spain	Prospective	13	24 fellow eyes (unspecified diagnosis)	Wet AMD, RVO, DME	RNFL thickness, ONH structures	Aflibercept, Ranibizumab	8	11 months	Spectralis	Enlarged BMO-MRW, deepened cup depth and width, greater RNFL thinning in injected eyes with ≥ 6 injections
Valverde-Megias et al., 201927	Spain	Prospective	20	9 fellow eyes (Dry AMD)	Wet AMD	RNFL thickness	Ranibizumab	21	8 years	Spectralis	No significant difference in RNFL thinning

AMD-Age related Macular Degeneration; BMO-MRW-Bruch’s Membrane Opening Minimum Rim Width; DME-Diabetic Macular Edema; ETDRS-Early Treatment Diabetic Retinopathy Study; GCC-Ganglion Cell Complex; GCL-Ganglion Cell Layer; ONH-Optic Nerve Head; RNFL-Retinal Nerve Fiber Layer; RVO-Retinal Vein Occlusion

^*-Evaluated in Shin et al. meta-analysis²⁶

Other limitations include the possibility of anti-VEGF medication diffusing into the systemic circulation and entering the contralateral eye, which serves as the control eye in most studies [28]. Such an event would naturally become a confounder in these analyses. Effects of intravitreal injections on the uninjected contralateral eye are controversial, with some studies reporting findings in the untreated eye and others finding no effect [33–38]. Additionally, it is important to note that half of the studies used ranibizumab injections only, while the other half used another drug alone or a combination of medications [27–30, 32]. Zhang et al. differed in their use of conbercept, an anti-VEGF medication not commonly used by retinal specialists in the USA [31]. Whether repeated injections of different anti-VEGF medications vary in their effects on retinal structures has not been investigated.

Two studies, Beck et al. and Zucchiatti et al., evaluated changes in the GCIPL as well, a region where initial glaucomatous damage occurs [39–44]. Results of these studies were mixed, with Beck et al. finding decreased GCL thickness in injected eyes compared with fellow control eyes, while Zucchiatti et al. found no significant difference in GCC thickness [30, 32]. Both studies had different limitations; Beck et al. was retrospective, while Zucchiatti et al. was a prospective study that lacked control eyes.

While other studies primarily investigated RNFL and GCIPL thickness, Gomez-Mariscal et al. focused on optic nerve head structures [28]. The authors measured the diameters of Bruch’s membrane opening, lamina cribrosa depth, prelaminar tissue thickness, cup width, and cup depth and found significant changes in all parameters, except the lamina cribrosa depth, following injections [28]. The biggest changes occurred early in the treated eyes; however, even after 1 year, observed changes were noted in those eyes that received six or more injections. The authors concluded that these anatomic changes were likely associated with repeated intravitreal injections [28]. However, it is important to note that this study included patients with DME and RVO, two disease processes that could involve RNFL edema. The investigators also manually completed optic nerve head measurements using metrics which do not have clear reproducibility-variability data available to validate their usage.

Most of the studies above used either the Spectralis (Heidelberg Engineering, GmbH, Dossenheim, Germany) or Cirrus (Carl Zeiss Meditec, Inc., Dublin, CA) OCT systems. Horsley et al. used the older Stratus system (Carl Zeiss Meditec, Inc., Dublin, CA). Both Spectralis and Cirrus are mainstream devices in the ophthalmology community [45–47]. Of note, Kopic et al. (discussed below), used the Copernicus system (Optopol, Zawiercie, Poland), which is an uncommon platform [21, 48]. Differences between these imaging systems should be considered when reviewing the results of these studies. For instance, none of the four studies using Cirrus found a significant association between RNFL thickness and intravitreal injections [20, 23, 29, 30]. In the seven studies that used Spectralis, two found statistically significant thinning of the RNFL [25, 28]. One must consider the interpolation of the circle scan from raster scans to estimate RNFL thickness in the Cirrus system. The Spectralis system completes a circle scan 3.5 mm away from the optic nerve head. Segmentation can be reviewed and edited in the Spectralis system, but not in the commercially available Cirrus system, which is important when underlying retinal pathology may affect automated segmentation. Of note, Beck et al. and Saleh et al. attempted to estimate RNFL thickness using the Early Treatment Diabetic Retinopathy Study (ETDRS) sectors from a Spectralis macula scan, which we refer to as a “macular” RNFL measurement. Such an analysis is uncommon and in contrast to other studies, which calculated peripapillary RNFL.

Based on these various studies, it appears as though intravitreal injections may not be associated with loss of RNFL thickness. Future studies could monitor changes in RNFL and GCL/GCIPL throughout the study instead of only assessing baseline and final measurements. In addition, systemic effects on the contralateral eye could be adjusted for, possibly with the use of age-matched controls in addition to fellow eyes [20, 28]. These studies assessed varying numbers of injections, IOP measurements at various times, and varied indications for injections [24, 25, 29, 30, 32]. A more uniform assessment may help clarify the variation observed in the reported findings. Lastly, the studies we reviewed were conducted in various countries around the world. This observation is noteworthy, as the pathogenesis of wet AMD (the most common indication for intravitreal injections in these studies) is known to have a genetic component. Such variation may result in different responses to intravitreal injections among genetically distinct populations [49–51]. While the study groups were similar in disease diagnoses, it is unknown whether they are truly comparable across studies due to different races/ethnicities evaluated in each study. Additional prospective investigations with larger sample sizes, longer follow-up periods, and diverse study cohorts would provide valuable data to strengthen these conclusions.

Retinal Thinning Following Intravitreal Injections in Glaucomatous Eyes

While many studies have investigated retinal thinning following intravitreal injections, the presence of glaucoma is often an exclusion criterion. Given the initial insult to the optic nerve in glaucoma, it is hypothesized that these patients may be more susceptible to further retinal injury after repeated post-injection IOP spikes. Identifying whether any observed differences are due to glaucoma, intravitreal injections, or the underlying retinal disease creates challenges in study design [52]. In this review, we are focused on imaging metrics such as RNFL and GCL thickness, although a few studies included perimetry parameters such as mean deviation (MD) and pattern standard deviation (PSD).

A search of the literature yielded three retrospective and one prospective study that aimed to identify a correlation between unilateral progression of glaucoma and intravitreal injections (Table 2). Of note, there were two additional prospective studies that were not included in this review due to the lack of English translations of the publications [53, 54]. All four studies involved injections of either aflibercept, ranibizumab, and/or bevacizumab in glaucoma or ocular hypertensive patients. Wet AMD was the most common indication for intravitreal injections, though Du et al. included RVO and diabetic retinopathy and Kopic et al. included DME [55, 48]. Du et al. also attempted a subgroup analysis of AMD eyes, although the sample size was small. Two of the studies, Lee et al. and Kopic et al., only included eyes with primary open angle glaucoma (POAG) [48, 56], while Saleh et al. and Du et al. included pseudoexfoliative glaucoma, normal tension glaucoma, and ocular hypertension in addition to POAG [55, 57]. The follow-up period ranged from 1 to 4.9 years, with three of the four studies following patients for at least 3 years. The average number of injections ranged from 6.27 to 39 [48, 55–57]. All but one study used the fellow eye as the control; two studies used glaucoma only control eyes (no retinal disease diagnosis discussed) [48, 55], while the third had fellow eyes with glaucoma and dry AMD as the control [56]. Saleh et al. did not include a control group, instead comparing the 16 treated eyes before and after injections [57].

Table 2.

Studies evaluating imaging metrics following intravitreal injections in glaucomatous eyes.

Author, Year	Country	Study Design	Number of Injected Eyes	Study Controls	Diagnoses	Primary Outcomes	Anti-VEGF medication	Mean Number of Injections	Mean Follow Up Period	OCT System	Study Findings
Kopic et al., 201748	Croatia	Prospective	60	60 fellow eyes (glaucoma)	Bilateral POAG + Wet AMD/ DME	RNFL thickness	Bevacizumab	6.27(AMD), 6.5 (DME)	1 year	Copernicus	Greater RNFL thinning among injected DME eyes. No significant difference in RNFL thickness among injected AMD eyes
Lee et al., 201756	Korea	Retrospective	16	16 fellow eyes (dry AMD + glaucoma)	Bilateral POAG + Wet AMD	RNFL thickness, GCIPL thickness, rate of change in GCIPL thickness, VF MD/PSD	Aflibercept, Ranibizumab, Bevacizumab	10.6	4.8 years	Cirrus	Decreased GCIPL thickness & faster rate of GCIPL thinning in injected eyes; no significant difference in RNFL thickness, VF MD/PSD
Saleh et al., 201757	Switzerland	Retrospective	16	None	POAG/ PXG/ NTG + Wet AMD	Macular RNFL thickness, GCL thickness, VF MD	Aflibercept, Ranibizumab	39	3.7 years	Spectralis	Decreased GCL thickness in injected eyes. No significant difference in VF MD or macular RNFL thickness (measured using ETDRS segments)
Du et al., 201955	USA	Retrospective	28	28 fellow eyes (glaucoma)	POAG/ PXG/ NTG/ OHT + wet AMD/ RVO/ DR	Rate of change in RNFL thickness, rate of change in VF MD/PSD	Unspecified	18.9	3.7 years	Spectralis	Faster rate of RNFL thinning in superior quadrant only of injected eyes; faster rate of decline in VF MD/PSD

AMD-Age related Macular Degeneration; DME-Diabetic Macular Edema; DR-Diabetic Retinopathy; ETDRS-Early Treatment Diabetic Retinopathy Study; GCIPL-Ganglion Cell Inner Plexiform Layer; GCL-Ganglion Cell Layer; MD-Mean Deviation; NTG-Normal Tension Glaucoma; OHT-Ocular Hypertension; POAG-Primary Open Angle Glaucoma; PSD-Pattern Standard Deviation; PXG-Pseudoexfoliation Glaucoma; RNFL-Retinal Nerve Fiber Layer; RVO-Retinal Vein Occlusion; VF-Visual Field

Findings in these studies were mixed. Only Du et al. and Kopic et al. found significant differences in the RNFL of injected eyes compared with fellow uninjected eyes [48, 55]. Du et al. found a significantly faster rate of RNFL thinning in the superior quadrant of injected eyes versus uninjected eyes, but not in the rate of overall RNFL thinning [55]. However, the RNFL in the superior quadrant was significantly thicker at baseline in the injected versus uninjected eyes. The authors did not address whether the faster rate of thinning in this quadrant may have been due to the resolution of edema as opposed to intravitreal injections. Of note, Du et al. were the only investigators that used multiple measurements with a linear mixed model to calculate a rate of change in RNFL thickness; this metric was the primary endpoint instead of the difference in RNFL thickness at baseline and final follow-up. Kopic et al. only found a significant difference in RNFL thickness between injected and fellow eyes in those eyes with DME. In contrast, Lee et al. found a similar degree of RNFL thinning in injected and fellow eyes. Saleh et al. found the thinning from baseline to last follow-up to not be statistically significant in injected eyes [57].

Of note, over 40% of cases in Du et al. were RVO, a condition that often involves RNFL edema as part of its presentation with subsequent resolution with time. Thus, RNFL “thinning” in an injected eye after RVO cannot be completely attributed to the intravitreal injection [58]. However, the authors did note a non-significant difference in overall baseline RNFL thickness between the injected and uninjected eyes. Kopic et al. studied both wet AMD and DME as indications for intravitreal injections [48]. As noted, RNFL thickness between the injected and fellow eyes was only significantly different in the DME group, but not in the wet AMD group. Hwang et al. previously reported that DME-only patients have thicker baseline RNFL and later may exhibit RNFL thinning following injections [59]. Intraretinal edema seen in DME may actually extend toward the optic nerve, thereby confounding peripapillary RNFL thickness measurements, especially in the temporal quadrant [22, 59]. Again, this finding suggests that resolution of the DME itself, unrelated to intravitreal injections, may lead to some degree of RNFL “thinning.” These limitations make it challenging to ascertain whether this metric is an accurate measure of glaucoma progression following intravitreal injections in DME and RVO.

Lee et al. and Du et al. also examined visual field parameters to gauge glaucomatous progression [55, 56]. Du et al. calculated the rate of change in MD and PSD with linear mixed models, while Lee at al. noted the difference between MD and PSD at the baseline and final follow-up visits. The mean baseline values for Du et al. were MD – 6.40 dB and PSD 5.20 dB in injected eyes and MD – 6.08 dB and PSD 4.71 dB in uninjected fellow eyes, suggestive of mild-moderate glaucomatous disease [55]. Du et al. found a statistically significant difference in rate of change for the injected eyes versus the uninjected eyes, with injected eyes exhibiting a faster rate of decline in both MD (− 1.07dB/year vs. – 0.01 dB/year, respectively) and PSD (0.90 dB/year vs. 0.01 dB/year respectively) [55]. The authors interpreted these findings as evidence that injections may be associated with the progression of glaucoma, although the underlying retinal disease may have affected these values. Lee et al. reported mean baseline values of MD – 4.75 dB and PSD 4.35 dB in injected eyes, and MD – 6.48 dB and PSD 6.23 dB in uninjected fellow eyes [56]. At last follow-up (approximately 5 years later), these values were MD – 7.29 dB and PSD 5.26 dB in injected eyes, and MD – 8.82 dB and PSD 6.61 dB in uninjected fellow eyes. The authors concluded that there was no statistically significant difference between these perimetric parameters in injected and uninjected eyes, in contrast to Du et al. [56]. It is important to note again that Du et al. evaluated rates of change in addition to absolute differences.

GCL was noted to thin in the injected eyes of the two studies that evaluated this metric [56, 57]. Saleh et al. identified a significant decrease in GCL thickness at final follow-up compared with baseline. Lee et al. found both a significant decrease and a faster rate of thinning of the GCIPL. However, interpretation of GCIPL thickness measurements can be challenging in this setting, as automated segmentation of these layers can be erroneous due to the underlying retinal pathology. Many studies often utilize a manual assessment of the segmentation by a masked investigator; both Lee et al. and Saleh et al. manually evaluated layer segmentation of the GCL and GCIPL. Nonetheless, given these challenges, some studies have suggested discarding GCL assessment altogether and focusing on RNFL thickness alone [30, 60]. One study suggested that even uninjected wet AMD might be associated with inner retinal (i.e., GCL) thinning, possibly as part of the disease pathogenesis [60]. If confirmed, this finding would confound retinal changes after injections even further.

There were additional limitations to the referenced studies. Saleh et al. did not include a control group, and 63% of their patients were diagnosed with glaucoma while receiving intravitreal injections. We found this description challenging to understand, as it is unclear whether injections may have contributed to the delayed diagnosis [57]. Du et al. combined ocular hypertensive eyes with glaucomatous eyes, possibly confounding the findings noted and effectively decreasing the sample size of glaucoma patients in total [55]. Finally, Kopic et al. used Copernicus HR, an OCT system that is not widely used in clinical practice. This study was also the only prospective study identified, but was limited by a relatively short follow-up period of 12 months [48].

Mitigating Post-Injection IOP Elevations

Despite no clear correlation between intravitreal injections and retinal thinning, attempts have been made to mitigate the potential effects of post-injection IOP spikes [61]. Anterior chamber paracentesis, decompression with cotton swabs, and aqueous suppressants have all been used [62–68]. Results seem to demonstrate moderate efficacy, with almost every study recording a statistically significant lower IOP in injected eyes receiving these interventions versus injected eyes without an intervention. While it remains unclear whether IOP itself is the cause of the observed retinal findings [69], Enders et al. reported no statistically significant difference in RNFL thickness when injected eyes were pre-treated with anterior chamber paracentesis compared with uninjected fellow eyes. Conversely, injected eyes that were not pre-treated had statistically significant RNFL thinning compared with uninjected fellow eyes [70]. If a clear association between intravitreal injections and retinal thinning were confirmed, pre-treatment would be an appropriate consideration. As Aref et al. have suggested, guidelines recommending the order in which to perform these interventions may be useful [61]. The use of aqueous suppressants prior to an injection would be easy to implement, while the more invasive paracentesis might be less preferable [66, 68, 71].

Conclusions

The relationship between intravitreal injections and concurrent retinal thinning is complex, as noted above. Based on the current literature, it appears as though there is minimal association between intravitreal injections and RNFL thinning in non-glaucomatous eyes. However, additional prospective studies with longer follow-up and careful study design would be of great value. While there is a significant number of studies focused on non-glaucomatous eyes, fewer studies have evaluated the rate of disease progression in glaucoma patients who require intravitreal injections. As noted above, there are multiple challenges in such studies–underlying disease versus treatment effect, the confounding interaction between injections and glaucoma, identification of the most appropriate metrics, and the length of follow-up. More recent papers involving repeated measurements seem to hint at a potential association in glaucomatous eyes. If a relationship is indeed confirmed with additional studies, pre-injection intervention might serve as a potential solution.