Quiescent Neovascular Age-Related Macular Degeneration After Endophthalmitis

Abstract

Purpose:

This article describes eyes that achieved extended remission of neovascular age-related macular degeneration (NVAMD) following acute endophthalmitis.

Methods:

Adults who presented to the Duke Eye Center with acute endophthalmitis over a 9-year period and had at least 3 months of follow-up were identified. A retrospective review of medical records was performed to collect clinical data including demographic information, examination findings, etiology, treatment, and outcomes.

Results:

A total of 133 eyes of 130 patients with endophthalmitis were identified. Of these, 15 eyes of 14 patients (11.3%) were receiving intravitreal antivascular endothelial growth factor (anti-VEGF) injections for NVAMD. Six of these 15 eyes (40%) did not require an anti-VEGF injection after endophthalmitis for a mean of 36.2 months. Endophthalmitis was injection-related in 5 of 6 eyes (83%) and Baerveldt glaucoma drainage device–related in 1 of 6 eyes (17%). Two of the 6 (33%) had culture-proven infectious endophthalmitis, whereas 4 of 6 (67%) had culture-negative endophthalmitis. Five of 6 eyes have required no anti-VEGF therapy to date; the remaining eye restarted intravitreal aflibercept therapy 9.3 months after endophthalmitis.

Conclusions:

Acute endophthalmitis may be associated with reduced activity of choroidal neovascularization in a subset of eyes with NVAMD.

Introduction

Age-related macular degeneration is a leading cause of blindness among individuals who are aged 50 years or older in the United States. ^1,2 Neovascular age-related macular degeneration (NVAMD) is characterized by the abnormal growth of new blood vessels, termed choroidal neovascularization (CNV). Currently, the mainstay treatment of NVAMD involves repeated, often monthly, intravitreal antivascular endothelial growth factor (anti-VEGF) injections. Treatment response may be assessed in part with morphologic features such as central subfield thickness or presence of intraretinal fluid (IRF) or subretinal fluid (SRF) on imaging as well as functional outcomes such as visual acuity (VA). ³ Injection interval may vary among treatment strategies, and evidence suggests that a treat-and-extend protocol is noninferior to monthly injections. ⁴ In those undergoing monthly injections as needed for NVAMD, certain factors may predispose to extended periods of disease inactivity. ⁵

Intravitreal injections are one of the most common procedures in ophthalmology, with an estimated 5.9 million injections performed in the US in 2016. ⁶ Owing to a significant increase in the number of intravitreal injections, the incidence of associated adverse events, such as endophthalmitis, has also increased proportionally. Endophthalmitis is a rare, sight-threatening complication that is clinically characterized by a severe inflammatory response within the eye. A recent meta-analysis found that endophthalmitis occurred in 197 of 350 535 (0.056%) intravitreal anti-VEGF injections. ⁷ Intravitreal injection-related endophthalmitis may be secondary to an infectious etiology or may be classified as a sterile inflammatory response, perhaps to the anti-VEGF medication or associated components. ^8,9 Overall, there is no difference among the currently available anti-VEGF agents regarding endophthalmitis risk. ^10,11 Because endophthalmitis from any cause may lead to devastating visual loss, prompt diagnosis and management are necessary.

Eyes that do recover from acute endophthalmitis, either fully or partially, may be subject to biochemical and structural changes that are not yet fully understood. ¹² A few case reports have described extended remission of NVAMD activity following an episode of postinjection endophthalmitis. ^{13 -15} It is plausible that the prominent intraocular inflammatory reaction that occurs during endophthalmitis may influence the activity of CNV in eyes with NVAMD; however, there is limited evidence and the exact mechanisms remain unclear. In our study, we identified cases of endophthalmitis that presented to a tertiary academic medical center over a 9-year period and describe multiple cases in which extended remission of CNV occurred following a single episode of acute endophthalmitis in eyes with NVAMD.

Methods

Patients evaluated at the Duke Eye Center between January 1, 2009, and January 1, 2018, with a diagnosis of endophthalmitis were identified using the Duke Enterprise Data Unified Content Explorer (Duke University Health System). Adults aged 18 years or older with at least 3 months of follow-up after a single episode of endophthalmitis were included. Data collected from review of medical records included demographic information, clinical history, best-corrected visual acuity (BCVA) and other ocular examination findings, medical and surgical interventions, gram stain and culture results, and complications associated with treatment or recalcitrant disease. Cases of endophthalmitis in eyes with coexisting NVAMD were further reviewed to determine prior therapies for NVAMD, treatment responses, and subsequent management patterns following the episode of acute endophthalmitis.

Quiescence of NVAMD was defined as an absence of recurrent disease activity (eg, IRF, SRF, or active CNV) on optical coherence tomography (OCT) imaging at 6 months’ follow-up as well as the clinical decision to discontinue intravitreal anti-VEGF injections. Eyes treated with photodynamic therapy or eyes that ceased injection therapy because of poor visual outcomes after endophthalmitis were not counted as quiescent. In addition, we assessed whether patients had signs of active CNV such as IRF or SRF on the most recent OCT evaluation before developing endophthalmitis. We then tabulated the time since fluid was last seen on OCT for each group among eyes with no fluid before endophthalmitis.

Descriptive statistical analysis using Microsoft Excel (Microsoft, Inc) was performed to characterize the demographic data of our patient population. Visual outcomes were determined by converting Early Treatment Diabetic Retinopathy Study VA to logarithm of the minimum angle of resolution (logMAR) and computing a mean value for each patient group (quiescent vs reactivated CNV due to NVAMD after endophthalmitis) at presentation and at 6 months’ follow-up visit. All other data were collected as medical record review and are reported by the proportion of patients in each group.

Results

A total of 133 eyes of 130 patients with endophthalmitis was identified. Of these, 15 eyes (11.3%) of 14 patients were receiving intravitreal anti-VEGF injections for active NVAMD at the time of diagnosis with endophthalmitis. In these 15 eyes with NVAMD, mean follow-up was 34.1 months, 60% were female, and mean age was 79 years (range, 63-89 years) at diagnosis of endophthalmitis. Thirteen of 15 eyes (87%) were being treated with aflibercept (Eylea, Regeneron), and 2 were being treated with ranibizumab (Lucentis, Genentech). None were receiving compounded bevacizumab (Avastin, Genentech). Fourteen of 15 eyes (93%) were receiving monthly injections and 1 was receiving injections every 2 months. The underlying etiology of endophthalmitis was injection related in 14 of 15 eyes (93%) and attributed to the presence of an exposed Baerveldt glaucoma drainage device in 1 of 15 eyes based on the timing of events. Nine of 15 eyes (60%) had culture-negative endophthalmitis, whereas 6 of 15 eyes (40%) had culture-proven infectious endophthalmitis.

In 6 of 15 eyes (40%) with endophthalmitis that also had comorbid NVAMD, the CNV remained clinically inactive after recovery from acute endophthalmitis over a mean follow-up of 36.2 months, and no further intravitreal injections were needed in 5 of 6 eyes (Table 1). CNV quiescence was defined by clinical examination findings, clinician judgment, and second OCT imaging at 6 months that demonstrated no signs suggestive of active CNV, including interval accumulation of IRF or SRF (Figure 1). In these 6 eyes, the most recent OCT before endophthalmitis was performed at an average interval of 75 days before developing endophthalmitis. On these scans, 1 eye had SRF, whereas 5 eyes had no fluid and were compact. Two eyes exhibited pigment epithelial detachment. In 4 eyes with compact OCT before endophthalmitis, the mean time since IRF or SRF was observed was 347 days. One eye remained quiescent for 9.3 months before CNV activity recurred and treatment with intravitreal anti-VEGF was resumed, but this time with intravitreal ranibizumab and less frequently (now every 12 weeks instead of every 4 weeks).

Table 1.

Individual Characteristics of Eyes With Neovascular Age-Related Macular Degeneration That Had Quiescent Choroidal Neovascularization After Acute Endophthalmitis.^a

	Patient 1	Patient 2	Patient 3	Patient 4	Patient 5	Patient 6
Injection type	Aflibercept	Aflibercept	Aflibercept	Aflibercept	Aflibercept	Aflibercept
Injection interval	Every 4 wks	Every 4 wks	Every 4 wks	Every 4 wks	Every 4 wks	Every 8 wks
SRF on most recent OCT (days prior to endophthalmitis)	No (94)	Yes (68)	No (158)	No (7)	No (29)	No (92)
Etiology of endophthalmitis	Injection	Injection	Injection	Injection	Injection	Baerveldtb
Endophthalmitis treatment	Vancomycin, ceftazidime	Vancomycin, ceftazidime	Vancomycin, ceftazidime, voriconazole	Vancomycin, ceftazidime	Vancomycin, ceftazidime	Vancomycin, ceftazidime, Baerveldt removal
Sterile endophthalmitis?	Yes	Yes	No (coagulase-negative Staphylococcus)	No (Streptococcus pneumoniae)	Yes	Yes
BCVA at 6 mo after endophthalmitis	20/126	20/200	20/60	20/320	20/64	20/32
Need for anti-VEGF injections after endophthalmitis	No	No	No	No	Yes (ranibizumab every 12 wks)	No
Quiescent period, mo	21	54	25	37	9	71

Abbreviations: anti-VEGF, antivascular endothelial growth factor; BCVA, best-corrected visual acuity; OCT, optical coherence tomography; SRF, subretinal fluid.

^aDoses: Aflibercept 2 mg in 0.05 mL (40 mg/ml), vancomycin 1 mg (10 mg/mL), ceftazidime 2.25 mg (22.5 mg/mL), voriconazole 0.1 mg (1 mg/mL), ranibizumab 0.5 mg in 0.05 mL (10 mg/mL).

^bExposed Baerveldt-related case: last aflibercept injection 40 days prior to diagnosis with endophthalmitis.

Figure 1.

Spectral domain-optical coherence tomography image of the macula in the right eye of a patient who did not receive any intravitreal antivascular endothelial growth factor injections after a single episode of acute endophthalmitis that developed following an intravitreal aflibercept injection. There is no evidence of choroidal neovascularization activity on optical coherence tomography.

Four of 6 eyes (66%) had culture-negative endophthalmitis, and 2 of 6 (33%) were culture-proven. Isolated species included coagulase-negative Staphylococcus in 1 case and Streptococcus pneumoniae in the other case. In these 6 eyes, the mean age was 82 years (range, 76-86 years) and 50% were female. All 6 eyes had been treated with regularly scheduled intravitreal aflibercept injections for at least 5 years before the development of endophthalmitis. Five of 6 eyes (83%) were receiving monthly injections, and 1 of 6 (17%) was receiving injections every 2 months. The etiology of endophthalmitis was related to an intravitreal injection in 5 of 6 eyes (83%) and to an exposed Baerveldt glaucoma drainage device in 1 of 6 eyes (17%). In the injection-related cases, the mean time from injection to presentation with endophthalmitis was 4 days. In the Baerveldt-related case, the last aflibercept injection was 5 weeks before developing endophthalmitis. Endophthalmitis was presumed to be Baerveldt related in this case because there was evidence of tube exposure with fibrin in the exposed section of tube as well as a significant anterior chamber inflammatory response. For treatment of presumed infectious endophthalmitis, this patient underwent administration of intravitreal antibiotics and Baerveldt removal. A vitrectomy was not performed.

In these 6 eyes, mean BCVA at the time of presentation with endophthalmitis was 20/1229 (logMAR = 1.788) and mean BCVA at 6 months was 20/100 (logMAR = 0.698), with a mean gain of approximately 11 lines (change in logMAR = –1.09).

Nine of 15 (60%) eyes did not exhibit extended remission of NVAMD following acute endophthalmitis and required continued treatment of active CNV (Table 2). Mean follow-up period for these 9 eyes was 31.7 months. In these 9 eyes, 7 (78%) continued to receive intravitreal anti-VEGF injections, beginning after a mean of 3.33 months following the diagnosis of acute endophthalmitis. Two eyes (22%) did not continue intravitreal injections for NVAMD; one underwent photodynamic therapy with verteporfin, and the other did not receive further treatment because of poor visual potential after endophthalmitis (light perception only at 6 months). Of the 7 eyes that continued intravitreal anti-VEGF for active NVAMD, 5 (71%) received monthly injections, 1 eye was extended to injections every 2 months, and another was extended to injections every 3 months. Four of 7 eyes (57%) were switched to a different anti-VEGF agent following acute endophthalmitis, whereas 3 of 7 (43%) continued receiving the same medication (2 ranibizumab, 1 aflibercept). Of those 4 that changed medications, 2 were switched from aflibercept to ranibizumab, and the other 2 from ranibizumab to bevacizumab.

Table 2.

Characteristics of Eyes With Choroidal Neovascularization Activity After Acute Endophthalmitis (n = 9).^a

Injection used before endophthalmitis	77% aflibercept
	23% ranibizumab
Injection interval before endophthalmitis	100% every 4 wks
SRF on most recent OCT (mean days prior to endophthalmitis)	33% (414)
Mean time to NVAMD treatment after endophthalmitis	3.33 mo
Injection interval after endophthalmitis (n = 7)b	72% every 4 wks
	14% every 8 wks
	14% every 12 wks
Anti-VEGF used after endophthalmitis (n = 7)b	57% ranibizumab
	29% bevacizumab
	14% aflibercept

Abbreviations: anti-VEGF, antivascular endothelial growth factor; NVAMD, neovascular age-related macular degeneration; OCT, optical coherence tomography; SRF, subretinal fluid.

^aDoses: Aflibercept 2 mg in 0.05 mL (40 mg/mL), ranibizumab 0.5 mg in 0.05 mL (10 mg/mL), bevacizumab 1.25 mg in 0.05 mL (25 mg/mL).

^bSeven of 9 eyes continued anti-VEGF therapy after endophthalmitis. One eye discontinued injections in favor of photodynamic therapy, and 1 eye discontinued injections because of poor visual outcome after endophthalmitis (light perception only).

Overall, these 9 eyes had a mean BCVA of 20/1862 (logMAR = 1.969) at the time of presentation with endophthalmitis, and a mean BCVA of 20/100 (logMAR = 0.699) at 6-month follow-up, with a mean gain of approximately 13 lines (change in logMAR = –1.27). Of these 9 eyes, the most recent OCT evaluation was an average of 52 days before developing endophthalmitis. Three of 9 eyes had evidence of SRF, whereas 6 eyes were compact. Of these 6 eyes, 1 eye had no prior OCT demonstrating fluid (initial OCT found subretinal hyperreflective material and pigment epithelial detachment that warranted treatment), and in the other 5 eyes, the mean time since fluid was observed on OCT was 414 days.

Mean time since IRF or SRF was observed on OCT in eyes without evidence of active CNV was not significantly different between eyes with recurrent or quiescent NVAMD after endophthalmitis (414 days vs 347 days, P = .611).

Conclusions

Following a single episode of acute endophthalmitis, 40% of eyes with active CNV due to NVAMD became quiescent and did not require further treatment over an average follow-up period of about 36 months. There are limited reports of this phenomenon in the literature, and the exact mechanism for this extended remission remains unclear. ^{13 -15} In 2017, Kally et al hypothesized that complement factor H may play a role in NVAMD quiescence following endophthalmitis. ¹⁴ In 2019, Kokame and colleagues suggested that upregulation of guanylate-binding-protein after endophthalmitis may induce an antiangiogenic state that favors extended remission of NVAMD. ¹⁵ Another study reported 2 eyes that developed quiescent NVAMD after vitrectomy for endophthalmitis and postulated that removal of the vitreous gel may have contributed to decreased CNV activity; ¹³ however, subsequent studies reported similar findings in nonvitrectomized eyes. ^14,15

Angiogenic factors governing CNV activity in eyes with NVAMD are complex and likely involve interactions among a multitude of inflammatory molecules. VEGF may be the most prominent mediator of neovascularization in eyes with NVAMD, as evidenced by the efficacy of anti-VEGF therapy in the treatment of CNV. However, many other intraocular cytokines may also play a contributory or homeostatic role in CNV activity, including angiopoietin, angiogenin, interferon γ-induced protein-10 (IP-10), and monocyte chemoattractant protein-1 (MCP-1), among others, all of which are significantly upregulated in eyes with NVAMD compared with controls. ^16,17 Additionally, after initiation of anti-VEGF therapy for NVAMD, interleukin-6 (IL-6), IP-10, and angiopoietin levels in the eye have been shown to increase. ^17,18

Following both sterile and infectious endophthalmitis, a number of cytokines are also upregulated acutely, including IL-6, IP-10, IL-8, and MCP-1. ^19,20 There are currently no studies that have examined the long-term effects of endophthalmitis on the intraocular inflammatory milieu. However, IL-8 and MCP-1 are distinctly proangiogenic mediators. MCP-1 induces angiogenesis by stimulating migration of smooth muscle cells and mesenchymal cells toward endothelial cells. ²¹ IL-8 has been implicated in ocular angiogenesis and increased vascular permeability, and its polymorphisms have been shown to alter anti-VEGF treatment responses in eyes with NVAMD. ^22,23 IL-6, which is upregulated both in endophthalmitis and after anti-VEGF therapy, has also been implicated as a proangiogenic cytokine that stimulates vascular proliferation with defective pericyte coverage. ²⁴ On the other hand, IP-10 has antiangiogenic activity in vitro and is upregulated in patients with NVAMD at baseline, after anti-VEGF injection, and also after endophthalmitis. ^16,18,19,25

Although the interactions among these immune mediators likely play a role in determining NVAMD disease activity in response to anti-VEGF therapy and also after acute endophthalmitis, it remains unclear why a subset of eyes with NVAMD exhibit long-term quiescence of CNV after endophthalmitis. Perhaps the severe acute inflammatory response associated with endophthalmitis affects the stability of the CNV in a way that is not yet well understood. Our findings suggest that inflammation associated both with culture-proven infectious and culture-negative acute endophthalmitis may contribute to an extended remission of CNV in some eyes with NVAMD.

Because the pathogenesis of NVAMD is closely linked to impaired angiogenesis and vascular leakage in response to hypoxic cellular signaling, it is possible that endophthalmitis may affect CNV activity by altering local oxygen demand or retinal perfusion. ²⁶ Prior studies suggest that reduced choriocapillaris flow may be associated with the development of type 3 CNV. ²⁷ Similarly, complement factor H polymorphisms have been associated with retinal thrombosis and an increased risk of developing age-related macular degeneration. ²⁸ As such, it is possible that the sharp increase in proangiogenic cytokines due to endophthalmitis induces a state of vascular remodeling that contributes to the stabilization of existing CNV. ²⁹ It is also plausible that tissue damage as a result of severe inflammation may decrease oxygen demand and angiogenic drive. Finally, disruption of the blood-ocular barrier owing to endophthalmitis may lead to long-term changes in intraocular cytokine profiles that may modulate CNV activity. ³⁰ These hypotheses, although theoretical, could be potential avenues for further inquiry to better elucidate the effects of acute endophthalmitis on intraocular cytokine profiles, vascular perfusion, and neovascular activity.

It may be that 6 eyes that exhibited CNV quiescence after NVAMD and were treated for at least 5 years with intravitreal anti-VEGF therapy would not have required continued treatment of NVAMD even in the absence of acute endophthalmitis. However, we observed that 2 of 6 eyes in the quiescent group had evidence of active CNV on the most recent OCT prior to endophthalmitis (average 75 days prior), whereas 3 of 9 eyes in the recurrent group had active CNV (average 52 days prior). In addition, eyes that were compact on the most recent OCT did not differ significantly in the time since fluid was last observed on OCT (414 days in the recurrent group vs 347 days in the quiescent group, P = .611). This suggests that prior NVAMD activity on anti-VEGF therapy may not influence the outcome of disease quiescence after acute endophthalmitis. Larger studies will have greater statistical power to support this hypothesis.

Limitations included the retrospective nature of this study and small sample size. We were unable to determine definitive causation between endophthalmitis and CNV quiescence. In addition, CNV activity in this study was determined by examination findings, OCT imaging, and clinical judgment. We were unable to confirm CNV quiescence by fluorescein angiography because it was not performed in the cases presented here. Finally, our study inclusion dates include the dates of the national outbreak of sterile inflammation after aflibercept, ⁸ so some of our culture-negative (presumed sterile) cases may have been associated with aflibercept-induced sterile inflammation. It is unclear whether patients experiencing aflibercept-associated inflammation would be more or less likely to exhibit NVAMD quiescence, because 13 of 15 eyes both with quiescent and recurrent NVAMD were receiving aflibercept before developing endophthalmitis in our data set. Despite these limitations, our findings describe a relatively underreported phenomenon that warrants further investigation and could potentially elucidate new mechanisms or targets for the treatment of NVAMD.

In summary, we describe 6 of 15 eyes with active NVAMD that became quiescent following an episode of acute endophthalmitis. Intraocular inflammation in the setting of endophthalmitis may be associated with clinically reduced disease activity in a subset of eyes with NVAMD. In those with persistent CNV activity after endophthalmitis, less-frequent intravitreal injections may be required.