Treatment Patterns and Outcomes in Bleb-Related Endophthalmitis: A 9-Year Analysis at a Tertiary Eye Center

Abstract

Purpose:

Clinical presentation, treatment choices, and outcomes in cases of bleb-related endophthalmitis (BRE) at a tertiary care center over a 9-year period are described.

Methods:

A retrospective review was conducted of patients diagnosed with BRE at Duke Eye Center (Durham, North Carolina) from January 1, 2009 to January 1, 2018, with at least 6 months of follow-up, assessing demographic data, initial management, and visual acuity (VA).

Results:

Twenty eyes of 20 patients with BRE were identified. Median time from surgery to presentation was 6.53 years. Presenting VA of light perception only was significantly associated with the decision to pursue pars plana vitrectomy (PPV) as initial treatment (odds ratio 59.4, 95% CI, 2.1-1670.8, P = .016). Twelve eyes (60%) had culture-proven infectious endophthalmitis. Eleven eyes (55%) underwent PPV during treatment; 5 eyes underwent PPV on presentation, and 6 eyes underwent PPV after initial presentation. Compared with pre-endophthalmitis VA, 6 eyes that underwent subsequent PPV had greater VA loss at 6 months than cases not undergoing subsequent PPV (Early Treatment Diabetic Retinopathy Study line loss of 14 vs 4 lines, respectively; P = .044).

Conclusions:

BRE eyes presenting with light-perception VA were more likely to undergo initial PPV; yet many eyes in this study required PPV during treatment. Visual outcomes are often poor in BRE despite intensive management. There was greater VA loss from pre-endophthalmitis VA levels at 6 months in eyes undergoing PPV after initial treatment. Prospective studies are needed to assess the optimal role of PPV in patients with BRE.

Introduction

Endophthalmitis is an uncommon sight-threatening complication of intraocular surgery involving an inflammatory response in the eye that is typically related to the introduction of pathogenic microbes from the environment. Prompt diagnosis and treatment of endophthalmitis are necessary to save the globe and maximize visual recovery and outcome. ^1,2

In certain eyes with glaucoma, trabeculectomy may be necessary to lower intraocular pressure (IOP). During this procedure, a channel is created between the anterior chamber of the eye and the subconjunctival space, providing an alternate outflow pathway for aqueous humor. However, this surgical filtration pathway may also allow exogenous pathogenic bacteria to reach the intraocular space, which can lead to a significant intraocular inflammatory event known as bleb-related endophthalmitis (BRE). BRE is relatively uncommon, with a reported incidence of 0.1% to 2%. ^{3 -7} Eyes that have undergone trabeculectomy may present with endophthalmitis many years after the initial surgery. ³

Initial management of BRE involves the prompt administration of broad-spectrum intravitreal antimicrobials and in some cases, vitrectomy surgery. ⁸ However, clinical guidelines for the initial management of BRE are not well defined, with treatment decisions often varying depending on physician preference.

The Endophthalmitis Vitrectomy Study (EVS) was a landmark prospective clinical trial that assessed the impact of pars plana vitrectomy (PPV) on visual outcomes in patients with endophthalmitis after cataract surgery or secondary intraocular lens placement. ⁹ It is unlikely that extrapolating the EVS’s results for the management of BRE is ideal, because the pathogenesis, clinical presentation, and causative organisms differ between individuals with endophthalmitis related to cataract surgery and those with BRE. ^10,11 There is not a similar prospective trial to inform ophthalmologists regarding the role of PPV in management of BRE and how clinicians can positively influence visual outcomes. Song et al reported that eyes with BRE that underwent vitreous tap with intravitreal antibiotics and those that underwent initial PPV both had poor outcomes. ¹² However, Busbee and colleagues reported that initial PPV led to better visual outcomes. ¹³ Additional studies are necessary to guide clinical decision-making in eyes with BRE. This study aims to evaluate the clinical characteristics, management decisions, and outcomes in cases of BRE at a tertiary academic medical center over a 9-year period.

Methods

We conducted a retrospective review of 133 eyes of 130 consecutive patients that were diagnosed with endophthalmitis at the Duke Eye Center (Durham, North Carolina) from January 1, 2009 to January 1, 2018. All participants were aged 18 years or older and had at least 6 months of follow-up from initial presentation. We recorded demographic information, ocular comorbidities, presumed etiology of endophthalmitis, initial treatment choices, microbiologic yield, and corrected visual acuity (VA) at presentation and at 6 months. Baseline, or pre-endophthalmitis VA, was collected from the most recent clinic visit prior to development of endophthalmitis. Individuals with endophthalmitis presumed to be related to a prior trabeculectomy were identified from this cohort. All cases were diagnosed with BRE by an experienced retina specialist based on clinical examination findings consistent with BRE (ie, blebitis with signs and symptoms of endophthalmitis and vitreous involvement on examination) and no other more likely cause. Elevated IOP was defined as greater than 21 mm Hg and low IOP was defined as less than 9 mm Hg.

Statistics were performed using XLSTAT (Addinsoft). Categorical variables were compared using χ² analysis, and continuous variables were compared using either a 2-tailed t test (for 2 sample groups) or analysis of variance testing (for 3 or more sample groups). Logistic regression analysis was used to identify associations between binary outcomes and continuous predictor variables. A P value of less than .05 was used in all analyses to define statistical significance. Snellen VA values were converted to logMAR for statistical analysis. Changes in logMAR VA and Early Treatment Diabetic Retinopathy Study (ETDRS) lines were calculated.

Results

A total of 133 eyes of 130 patients with endophthalmitis were identified. Of these, 20 eyes (15.0%) of 20 patients had presumed infectious endophthalmitis related to prior trabeculectomy. All eyes met the definition of BRE and had evidence of vitreous involvement. These eyes had no recent history of trauma or ocular surgery within 3 months prior to presentation with endophthalmitis and were afebrile. Median follow-up for these patients was 3.83 years. All eyes received intravitreal antimicrobials following initial diagnostic testing with aqueous tap, needle vitreous tap, or mechanical vitreous biopsy with PPV.

Demographic data, presenting examination findings, and initial treatment choices are detailed in Table 1. Twelve cases (60%) were culture positive; Streptococcus species bacteria were most commonly isolated (Figure 1).

Table 1.

Baseline Characteristics in Eyes That Developed Endophthalmitis After Trabeculectomy (n = 20).

Characteristic	Value
Right eye, %	50
Female sex, %	55
Average age (range), y	73 (48-85)
Median time to presentation	2384 d (6.53 y)
MMC used in trabeculectomy	100% (20/20)
Elevated IOP (> 21 mm Hg)	30% (6/20)
Low IOP (< 9 mm Hg)	15% (3/20)
Hypopyon	65% (13/20)
Aqueous tap	30% (6/20)
Needle vitreous tap	60% (12/20)
Initial PPV	20% (5/20)
Subsequent PPV	30% (6/20)
Average time to subsequent PPV, d	25
Intravitreal vancomycin	100% (20/20)
Intravitreal ceftazidimea	95% (19/20)
Culture positive	60% (12/20)

Abbreviations: IOP, intraocular pressure; MMC, mitomycin C; PPV, pars plana vitrectomy.

Doses: vancomycin 1 mg (10 mg/mL), ceftazidime 2.25 mg (22.5 mg/mL).

^a One patient received intravitreal amikacin for gram-negative coverage because of a penicillin allergy and concern for potential cross-reactivity with ceftazidime.

Figure 1.

Microbes associated with culture-positive endophthalmitis after trabeculectomy (n = 12). Six of 12 bacteria are oral flora (all Streptococcus species).

Pre-endophthalmitis mean VA was 20/66. Mean VA at presentation was 20/4220 and mean VA at 6 months was 20/378 with a mean gain of 10 ETDRS lines across all groups from presentation to 6 months. From pre-endophthalmitis VA to 6 months, patients lost an average of 8 ETDRS lines of vision. Three eyes were no light perception vision and 1 eye had light perception (LP) only vision at 6 months. Presenting logMAR VA was not significantly associated with logMAR VA at 6 months (r ² = 0.13, P = .11). Presenting VA was significantly different among eyes undergoing initial aqueous tap (20/4856), needle vitreous tap (20/2973), or PPV (20/16000, P = .0002). At 6 months, the difference in mean VA between initial diagnostic procedure groups approached statistical significance (P = .07), with the most improvement in the needle vitreous tap group (13 ETDRS lines gained).

Five eyes underwent PPV on presentation and 6 eyes underwent PPV after initial presentation. Compared with eyes that did not undergo PPV on presentation, the 5 eyes that underwent PPV on presentation had significantly worse VA on presentation (20/16000 for PPV on presentation vs 20/2706 for no PPV on presentation, P = .037) and trended toward worse vision at 6 months (20/1252 for PPV on presentation vs 20/253 for no PPV on presentation, P = .27) (Table 2). There was no difference in VA change from pre-endophthalmitis VA to 6 months between those who underwent PPV and those who did not (7 ETDRS line loss vs 9 ETDRS line loss, respectively; P = .71).

Table 2.

Mean Corrected Visual Acuity in Patients Presenting With Bleb-Related Endophthalmitis Who Underwent No PPV During Treatment (n = 9), PPV on Initial Presentation (n = 5), and PPV After Initial Presentation (n = 6).

Vitrectomy status	VA at presentation	VA at 6 mos	Mean logMAR changea	ETDRS lines gained
No PPV	20/1603	20/115	–1.144	11 lines
PPV on presentation	20/16000	20/1252	–1.107	11 lines
PPV after initial presentation	20/5935	20/830	–0.854	8.5 lines
P valueb	.032	.21	.89

Abbreviations: ETDRS, Early Treatment Diabetic Retinopathy Study; PPV, pars plana vitrectomy; VA, corrected visual acuity.

^a Change in logMAR over 6 months.

^b Calculated by analysis of variance.

Of the 6 eyes that underwent subsequent PPV, the mean time from initial diagnostic intervention to subsequent PPV was 25 days (median, 4 days). Of the 6 eyes that underwent subsequent PPV, 4 underwent PPV within 1 week after initial treatment, and 2 underwent PPV following a significant delay after initial treatment (66 and 69 days, respectively). The 4 cases that underwent subsequent PPV within the first week had no clinical improvement with tap-and-inject with intravitreal antibiotics alone, so PPV was performed as a therapeutic option; Streptococcus species were isolated in all 4 cases. For the 2 cases that underwent delayed PPV, there was initial improvement with tap-and-inject alone, but clinical worsening with decreased vision occurred about 2 months after initial presentation. In these 2 delayed PPV cases, 1 grew Pseudomonas aeruginosa from PPV cultures and 1 was culture negative.

Intraocular cultures were drawn on presentation in these patients, but repeat diagnostic procedures were not performed prior to subsequent PPV in these 6 cases. In 3 cases with PPV cultures drawn, 1 was positive for P aeruginosa (initially culture negative), 1 was positive for Streptococcus viridans (initial aqueous culture positive for S viridans), and 1 PPV culture was negative (initial vitreous tap culture negative). There was no difference in presenting VA between eyes that underwent subsequent PPV compared with those that did not (20/5935 vs 20/3646, respectively; P = .57), and there was also no difference in VA between these groups at 6 months (20/830 vs 20/270, P = .41). Eyes that underwent subsequent PPV had a significantly greater change in VA from pre-endophthalmitis VA levels at 6 months, likely reflecting a more severe clinical course in these cases (14 ETDRS line loss in cases undergoing PPV after initial presentation vs 4 ETDRS line loss in cases undergoing no PPV or PPV on presentation, P = .044).

Eyes that presented with VA of LP were significantly more likely to undergo initial PPV (odds ratio 59.4; 95% CI, 2.1-1670.8; P = .016); however, presenting VA was not associated with subsequent PPV after initial management (P = .31). At presentation, presence of elevated IOP, low IOP, hypopyon, culture-proven etiology, or age older than 75 were not associated with undergoing an initial PPV (P > .05). In addition, elevated IOP on presentation, hypotony on presentation, hypopyon on presentation, or age older than 75 were not associated with undergoing a subsequent PPV (P > .05). There was a trend between culture-proven cases and an increased likelihood of undergoing subsequent PPV, but this finding did not reach statistical significance (P = .09). Cases that grew oral flora (ie, Streptococcus species other than S agalactiae) were not more likely to undergo initial PPV (P = .99) or subsequent PPV (P = .25) than other culture-positive cases.

Elevated IOP on presentation, hypotony on presentation, and presence of hypopyon were not associated with worse presenting VA (P > .05). Cases that were culture proven did not have significantly worse VA at presentation (P = .31) or at 6 months (P = .85) compared with culture-negative cases. Culture-positive cases that grew oral flora trended toward worse presenting VA (P = .07) but had similar VA at 6 months compared with culture-positive cases that grew other organisms (P = .93). Age older than 75 was not associated with worse VA at 6 months (P = .22).

Conclusions

Current practice patterns for the management of BRE are not well defined in the literature, and there are limited prospective studies to guide management of endophthalmitis following intraocular surgery aside from the EVS, which is limited to cases following cataract surgery or secondary intraocular lens implantation. Our data show that EVS guidelines are largely followed in individuals presenting with BRE, with LP vision significantly predicting the decision to perform initial vitrectomy. However, 6 patients who did not undergo vitrectomy on presentation underwent subsequent vitrectomy. Notably, 4 patients with endophthalmitis who underwent vitrectomy within 1 week of initial presentation were culture positive for Streptococcus species bacteria, which often cause poor visual outcomes despite intensive management. ¹⁴ Streptococcus-related BRE, although often sensitive to commonly used intravitreal antibiotics, may be less likely to improve with antibiotics alone, which may partially explain the clinical decision to pursue vitrectomy within a week after initial treatment. ¹⁴

These findings suggest that LP vision alone may not predict whether a patient with BRE will require a vitrectomy during the course of presentation. This finding is consistent with our hypothesis that initial management of BRE should be prospectively studied to develop clinical guidelines specific to BRE and promote the best possible visual outcomes for affected patients.

Owing to the retrospective nature of this study, we are unable to conclude whether eyes with LP vision on presentation would have had a different visual outcome if they did not undergo initial PPV. However, our results did not reveal a significant difference in visual outcomes at 6 months among those that underwent initial PPV and those that did not in the entire cohort with BRE. Although we observed significantly greater VA loss among eyes undergoing subsequent PPV, this difference likely reflects the greater clinical severity of these cases. However, this difference could also be partly explained by a greater delay in time to PPV. Further research may clarify whether early PPV leads to improved visual outcomes compared with subsequent PPV performed after initial treatment.

Previous studies have demonstrated IOP elevation on presentation in BRE as well as difficulty controlling either excessively high or low IOP in eyes after endophthalmitis related to glaucoma-filtering surgery. ^15,16 However, in our cohort, we observed similar numbers of cases with IOP elevation and low IOP on presentation with BRE. Neither IOP elevation on presentation nor low IOP on presentation was associated with worse VA at 6 months.

Among the 12 culture-positive cases, 6 eyes (50%) grew bacteria associated with common oral flora, ¹⁷ and all of these were of the Streptococcus species. Prior studies have demonstrated that oral Streptococcus species bacteria are commonly isolated in cases of endophthalmtis ^18,19 and may be associated with worse visual outcomes in cases of intravitreal injection-related endophthalmitis. ²⁰ In our cohort, culture-positive cases associated with oral flora trended toward worse presenting VA; however, these cases had similar VA at 6 months compared with other culture-positive cases. We did not observe a significant difference in initial VA or VA at 6 months between culture-positive and culture-negative cases.

This study is limited by its retrospective nature. We were unable to directly compare outcomes between 2 treatment choices (ie, vitrectomy vs tap-and-inject) in similar groups of patients without conducting a prospective clinical trial similar to the EVS; however, given that our institution has treated only 20 cases (with at least 6 months of follow-up) of BRE over 9 years, it would be exceedingly difficult to conduct a prospective study at a single tertiary center. Our study is also limited by its relatively smaller sample size, which reflects the low incidence of BRE. Finally, given the retrospective nature of this study, we are unable to directly compare outcomes between individuals who presented with LP vision and who did or did not undergo PPV. It may be that PPV leads to greater visual improvement than tap-and-inject alone, but because eyes with worse vision often underwent PPV, there is selection bias. Worse vision on presentation is a confounder for our final VA results. Studies designed prospectively with intervention and control arms (similar to the EVS) may be necessary to define best practices in BRE.

We found that management of BRE at our institution largely follows the guidelines of the EVS, which recommends initial PPV if presenting VA is LP or no LP. However, it is unclear whether initial PPV in individuals with BRE who present with LP vision would substanstially improve visual outcomes after endophthalmitis. Given that the EVS was conducted in a patient population with endophthalmitis after cataract surgery or secondary intraocular lens placement, we cannot conclude that its findings can be directly extrapolated to individuals with BRE. Additionally, we observed that a significant proportion of individuals required vitrectomy during the course of treatment (either on presentation or subsequently after initial tap-and-inject). This suggests that BRE may not resolve with intravitreal antibiotics alone in some cases, regardless of presenting VA. Future research may examine the impact of initial vs late vs no PPV in eyes with BRE.

Our study underscores the benefit of timely consultation with a retina specialist when BRE is suspected in an eye with signs of blebitis and vitreous involvement. Multicenter prospective clinical trials comparing vitrectomy vs tap-and-inject in individuals with BRE may yield optimal practice guidelines to improve outcomes in this complication, which often results in poor visual outcomes despite optimal management.