Incidence and Visual Outcomes of Endophthalmitis After Intravitreal Injection of Dexamethasone Implant vs Ranibizumab

Abstract

Purpose:

To compare the incidence and visual outcomes of endophthalmitis after injection of an intravitreal dexamethasone implant and injection of intravitreal ranibizumab.

Methods:

This retrospective cohort study assessed endophthalmitis in eyes receiving an intravitreal injection of a 0.7 mg dexamethasone implant (DEX group), 0.5 mg ranibizumab (R5 group), or 0.3 mg ranibizumab (R3 group) between January 1, 2016, and May 31, 2018, at 2 large retina practices in the United States.

Results:

Suspected endophthalmitis occurred in 5 eyes after 4973 DEX injections, 43 eyes after 163 974 R5 injections, and 6 eyes after 18 954 R3 injections. Suspected endophthalmitis was significantly more common in the DEX group (1/995) than in the R5 group (1/3813) (P = .008) but not than in the R3 group (1/3159) (P = .10). Visual acuity outcomes were similar in the 3 groups.

Conclusions:

Suspected endophthalmitis might be more common after 0.7 mg dexamethasone injections than after 0.5 mg ranibizumab injections. Culture-positive endophthalmitis rates were similar across all 3 medications.

Introduction

Intravitreal injection (IVI) of pharmacologic agents, including antivascular endothelial growth factor (anti-VEGF) agents and steroid medications, is a first-line treatment for many retinovascular diseases. Based on US Medicare data, the use of IVI has markedly increased during the past 20 years, from less than 3000 injections in 2000 to more than 3 million in 2016. ¹ Several anti-VEGF inhibitors are commercially available for the treatment of diabetic retinopathy, ² neovascular age-related macular (nAMD),^3,4 retinal vein occlusion (RVO), ⁵ and myopic choroidal neovascularization. ⁶ This treatment is not without risks, the most notable being endophthalmitis, which can result in permanent vision loss and blindness. ⁷ Large studies have shown the incidence of endophthalmitis after IVI to be approximately 1 in 2000 injections,^8,9 with smaller studies reporting rates ranging from 1 in 1050 to 1 in 19 000.^10,11

Intravitreal administration of a dexamethasone implant was approved by the US Food and Drug Administration for the treatment of diabetic macular edema (DME), macular edema secondary to RVO, and noninfectious posterior uveitis. ¹² Endophthalmitis after dexamethasone implant insertion has been reported in case studies^13,14; however, few large studies have reported the incidence. The GENEVA study found no endophthalmitis cases in a series of 1830 implant injections. ¹⁵ One case series of more than 3500 injections reported an incidence of approximately 1 in 720 injections, ¹⁶ while the MEAD study reported an incidence of approximately 1 in 3000 injections. ¹⁷

Differences in the administration technique, needle gauge, underlying disease processes, and pharmacologic properties might be a reason for the differences in the endophthalmitis incidence between steroids and anti-VEGF drugs. The larger wound created during dexamethasone implant injection as well as the local immunosuppressive effect of steroid medication could affect infection rates. ¹⁸ This study compared the incidence and visual outcomes of suspected endophthalmitis and culture-positive endophthalmitis after dexamethasone implant injection and after IVI of ranibizumab.

Methods

Inclusion and Exclusion Criteria

Patients were included if they received at least 1 IVI of 0.5 mg ranibizumab (R5 group), 0.3 mg ranibizumab (R3 group), or a 0.7 mg dexamethasone implant (DEX group) between January 1, 2016, and May 31, 2018, at Mid Atlantic Retina/The Retina Service of Wills Eye Hospital, Philadelphia, Pennsylvania, or Associated Retinal Consultants/William Beaumont Hospital Retina Service, Royal Oak, Michigan. The 0.5 mg ranibizumab injections were administered using a prefilled syringe (pR5 group) or conventional syringe preparation (cR5 group). All 0.3 mg ranibizumab syringes were conventionally prepared.

Patient selection was performed using International Classification of Disease and Current Procedural Terminology billing codes. Cases of endophthalmitis were confirmed by review of the patients’ medical records. Data collected from these records included patient age, date of the causative injection, indication for injection therapy, date of vitreous or anterior chamber tap and intravitreal antibiotic injection, culture results, and visual acuity (VA) at the time of the causative injection and at the 3-month and most recent follow-up visits.

Intravitreal Injection Technique

Standard protocol was used for the intravitreal implant injection of 0.7 mg dexamethasone (Ozurdex, Allergan Inc) and intravitreal injection of 0.5 mg and 0.3 mg ranibizumab (Lucentis, Genentech). All injections were performed in an office setting. Topical anesthesia alone was used in the majority of cases, with some cases receiving a subconjunctival injection of 2% lidocaine. Topical 5% or 10% povidone–iodine was used for antisepsis. The dexamethasone implant insertion was performed using a 22-gauge needle inserted 3.5 to 4.0 mm from the limbus. The ranibizumab injection was performed using a 30-gauge needle inserted 3.5 to 4.0 mm from the limbus. The use of a bladed lid speculum or manual lid retraction, conjunctival displacement, and quadrant of injection were based on the individual physician’s preference.

Endophthalmitis Treatment Protocol

In eyes with clinical signs of infectious endophthalmitis (ie, suspected endophthalmitis), a pars plana vitreous aspiration was immediately attempted and an intravitreal antibiotic was injected with a 25- or 27-gauge needle. Anterior chamber paracentesis was performed to obtain aqueous fluid if vitreous fluid could not be aspirated. The intravitreal antibiotic regimen included vancomycin (1 mg/0.1 mL) and ceftazidime (2.25 mg/0.1 mL); intravitreal amikacin (400 mcg/0.1 mL) was used as a substitute for ceftazidime in patients allergic to penicillin. Patients were also prescribed topical antibiotics and prednisolone acetate 1% drops, with some patients receiving atropine sulfate 1% as well. Each patient’s clinical course was monitored daily until improvement was seen, at which point the drops and frequency of follow-up visits were gradually reduced. Pars plana vitrectomy was performed in some cases with severe infection or lack of clinical improvement. Oral steroids were not administered to any patient as part of the endophthalmitis treatment.

Outcome Measures

The primary outcome was the incidence of suspected endophthalmitis and culture-positive endophthalmitis postinjection in the R5, pR5, cR5, R3, and DEX groups. Secondary outcomes included Snellen VA and microbial spectrum analysis. VA outcomes were evaluated 3 months after the endophthalmitis diagnosis and at the final follow-up visit and were measured as the mean change in VA from baseline. The baseline VA was defined as the VA at the time of the causative injection. Cases with fewer than 3 months of follow-up data were not included in the 3-month visual outcome analyses but were included in the final follow-up analyses.

Logarithm of the minimal angle of resolution (logMAR) equivalents of Snellen VA were used for statistical analysis. As established in previous studies, logMAR values of 2.3 and 2.6 were used for acuity levels of counting fingers and hand motions (HM).^19,20 Cases with visual outcomes of light perception (LP) and no light perception (NLP) were excluded from logMAR analyses of visual outcomes because these descriptors of stimulus perception were not considered measures of true VA. ²¹ For comparative analyses, suspected cases of endophthalmitis included both culture-positive cases and culture-negative cases.

Statistical Analysis

All endpoints were evaluated by intravitreal medication group, including R5, R3, and DEX. The R5 group was further divided based on injection preparation, including cR5 and pR5, for analysis of each endpoint. Clinical variables were analyzed using Excel (Microsoft Corp), and statistical analysis was performed using SPSS Statistics (IBM). Descriptive statistics were calculated for continuous variables. Comparisons of the incidence of endophthalmitis between groups were performed using the χ² test or Fisher exact test. Visual outcomes were compared using the t test. The follow-up duration and mean VA were compared using analysis of variance. In all cases, a P value less than 0.05 was considered statistically significant. One case of endophthalmitis occurred after a single eye received ranibizumab and dexamethasone injections the same day; this case was excluded from statistical analysis because confounding would be unavoidable.

Results

Injection Counts and Baseline Characteristics

A total of 4973 DEX injections were included in this study. DME was the leading indication for these injections. Other indications included central RVO (CRVO), branch RVO (BRVO), uveitis, nAMD, and other diagnoses (Table 1).

Table 1.

Baseline Characteristics by Treatment Group.

Treatment Group	Total Injections (n)	Injections Administered by Indication (%)
		nAMD	BRVO	CRVO	DME	Myopic Degeneration	Uveitis	Other
Dexamethasone 0.7 mg implant	4973	4.4	16.2	16.5	39.9	0.00	10.3	12.8
Ranibizumab 0.5 mg (combined)	163 974	81.0	8.5	6.5	0.27	0.42	0.00	3.3
Conventional ranibizumab 0.5 mg	88 556	80.1	7.9	6.0	0.30	0.35	0.00	4.7
Prefilled ranibizumab 0.5 mg	75 418	81.2	9.3	7.0	0.24	0.49	0.00	1.8
Ranibizumab 0.3 mg a	18 954	0.30	0.06	0.06	99.8	0.00	0.00	0.02
Total	187 901	70.8	7.9	6.1	11.4	0.36	0.27	3.3

Abbreviations: BRVO, branch retinal vein occlusion; CRVO, central retinal vein occlusion; DME, diabetic macular edema; nAMD, neovascular age-related macular degeneration.

^aSome R3 injections were administered to treat multiple indications with a single procedure (eg, DME and CRVO treated with one R3 intravitreal injection), causing the total number of R3 intravitreal injection to be less than the sum of the injections given for each indication.

A total of 163 974 R5 injections were included in this study, with 88 556 cR5 injections and 75 418 pR5 injections. The leading indication for treatment was nAMD. Other indications included BRVO, CRVO, DME, myopic degeneration, and other diagnoses (Table 1).

A total of 18 954 R3 injections were included in this study. The leading indication was DME. Additional indications included nAMD, BRVO, CRVO, and other diagnoses. In 0.18% of R3 injections, 1 injection was administered to treat 2 of the aforementioned indications in the same eye, causing the total number of R3 injections to be less than the sum of the injections given for each indication (Table 1).

Incidence of Endophthalmitis

Suspected endophthalmitis

Five cases (0.101%) of suspected endophthalmitis occurred in the DEX group (Table 2). The indication for DEX injection was DME in 2 cases, BRVO in 2 cases, and posterior uveitis in 1 case. Forty-three cases (0.026%) of suspected endophthalmitis occurred in the R5 group, 28 cases (0.032%) in the cR5 group, and 15 cases (0.020%) in the pR5 group. In the cR5 group, the treatment diagnosis was nAMD in 21 cases, BRVO in 4 cases, and CRVO in 3 cases. In the pR5 group, the diagnosis was nAMD in 12 cases and BRVO in 3 cases. Six cases (0.032%) of suspected endophthalmitis occurred in the R3 group; the diagnosis in all cases was DME.

Table 2.

Incidence of Suspected Endophthalmitis by Treatment Group.

Treatment Group	Total Injections (n)	Cases of Suspected Endophthalmitis (n)	Incidence (%)	Odds Ratio (95% CI)	P Value
Dexamethasone 0.7 mg implant	4973	5	0.101	—	—
Ranibizumab 0.5 mg (combined)	163 974	43	0.026	3.84 (1.52-9.69)	.008
Conventional ranibizumab 0.5 mg	88 556	28	0.032	3.18 (1.23-8.24)	.03
Prefilled ranibizumab 0.5 mg	75 418	15	0.020	5.06 (1.84-13.93)	.002
Ranibizumab 0.3 mg	18 954	6	0.032	3.18 (0.97-10.42)	.10

The incidence of suspected endophthalmitis was statistically significantly higher in the DEX group than in the R5 group (P = .008). There was a higher incidence of suspected endophthalmitis in the DEX group than in the cR5 group (P = .03) and pR5 group (P = .002). The rate of suspected endophthalmitis was similar between the pR5 group and cR5 group (P = .19). The difference in the incidence of suspected endophthalmitis between the DEX group and the R3 group did not reach statistical significance (P = .10). Table 2 shows the odds ratios (ORs) and 95% CIs.

Culture-positive endophthalmitis

In the DEX group, 2 of the 5 endophthalmitis cases were culture positive (overall incidence 0.040%) (Table 3). The indication for the DEX injection was DME in 1 case and BRVO in the other. In the R5 group, 18 of the 43 cases (0.011%) were culture positive, with 15 cases (0.017%) occurring after the cR5 injection and 3 cases (0.004%) after the pR5 injection. The indication for the cR5 injection was nAMD in 12 cases, BRVO in 2 cases, and CRVO in 1 case. The indication for the pR5 injection was BRVO in 2 cases and nAMD in 1 case. Three (0.016%) of the 6 suspected endophthalmitis cases in the R3 group were culture positive, and the treatment diagnosis was DME in all these cases.

Table 3.

Incidence of Culture-Positive Endophthalmitis by Treatment Group.

Treatment Group	Total Injections (n)	Cases of Culture-Positive Endophthalmitis (n)	Incidence (%)	Odds Ratio (95% CI)	P Value
Dexamethasone 0.7 mg implant	4973	2	0.040	—	—
Ranibizumab 0.5 mg (combined)	163 974	18	0.011	3.66 (0.85-15.80)	.22
Conventional ranibizumab 0.5 mg	88 556	15	0.017	2.37 (0.54-10.39)	.52
Prefilled ranibizumab 0.5 mg	75 418	3	0.004	10.11 (1.69-60.54)	.03
Ranibizumab 0.3 mg	18 954	3	0.016	2.54 (0.42-15.21)	.61

The incidence of culture-positive endophthalmitis was statistically significantly higher in the DEX group than in the pR5 group (P = .03) but not than in the R5 group (P = .22) or cR5 group (P = .52). The rate of culture-positive endophthalmitis was statistically significantly lower in the pR5 group than in the cR5 group (P = .02). The incidence was similar between the DEX group and R3 group (P = .61). Table 3 shows the ORs and 95% CIs.

Follow-up Duration

Suspected endophthalmitis

The mean duration of follow-up in cases of suspected endophthalmitis was 11.0 months (range, 1.7-25.3 months) in the DEX group, 10.7 months (range, 0.5-28.4 months) in the R5 group, 13.2 months (range, 0.9-28.4 months) in the cR5 group, 5.9 months (range, 0.5-15.4 months) in the pR5 group, and 12.7 months (range, 4.2-30.3 months) in the R3 group. The follow-up was 3 months or longer in all 5 DEX cases, 92.9% of cR5 cases (n = 26), 86.7% of pR5 cases (n = 13), and all 6 R3 cases. The follow-up duration was statistically significantly shorter in the pR5 group than in the cR5 group (P = .002); the difference was the result of the more recent availability of pR5 compared with cR5. However, there was no statistically difference in follow-up duration between the DEX group and the R5 group (P = .93), cR5 group (P = .57), pR5 group (P = .11), or R3 group (P = .52).

Culture-positive endophthalmitis

The mean duration of follow-up in cases of culture-positive endophthalmitis was 13.5 months (range, 1.7-25.3 months) in the DEX group, 10.4 months (range, 0.5-27.5 months) in the R5 group, 12.0 months (range, 0.9-27.5 months) in the cR5 group, 2.7 months (range, 0.5-4.9 months) in the pR5 group, and 10.9 months (range, 4.2-19.4 months) in the R3 group. The duration was 3 months or longer in all 2 DEX cases, 86.7% of cR5 cases (n = 13), 66.7% of pR5 cases (n = 2), and all 3 R3 cases. There was no significant difference in follow-up duration between treatment groups (P = .41).

Visual Outcomes

Suspected Endophthalmitis

There was no statistically significant difference in logMAR VA between the DEX group, R5 group, and R3 group at baseline (P = .93), 3 months (P = .17), or the final follow-up (P = .26). Table 4 shows the mean VA outcomes by treatment group.

Table 4.

Visual Acuity in Endophthalmitis Cases by Treatment Group.

	Suspected Endophthalmitis LogMAR VA			Suspected Endophthalmitis LogMAR VA
Treatment Group	Baseline	3 Mo a	Final	Baseline	3 Mo a	Final
Dexamethasone 0.7 mg implant
Mean	0.46	0.45	0.73	0.35	0.48	0.30
Range	1.00, 0.00	NLP, 0.00	NLP, 0.00	0.40, 0.30	NLP, 0.48	NLP, 0.30
Ranibizumab 0.5 mg (combined)
Mean	0.65	1.05	0.89	0.65	1.50	1.11
Range	2.60, 0.00	NLP, 0.00	NLP, 0.00	2.30, 0.10	NLP, 0.18	NLP, 0.00
Conventional ranibizumab 0.5 mg
Mean	0.64	1.26	1.02	0.58	1.52	1.14
Range	2.60, 0.10	NLP, 0.10	NLP, 0.00	2.30, 0.10	NLP, 0.18	NLP, 0.00
Prefilled ranibizumab 0.5 mg
Mean	0.67	0.66	0.64	0.93	1.39	0.96
Range	2.30, 0.00	LP, 0.00	NLP, 0.00	2.30, 0.10	2.60, 0.18	2.30, 0.18
Ranibizumab 0.3 mg
Mean	0.54	1.19	1.6	0.80	2.6	2.45
Range	LP, 0.30	LP, 0.48	NLP, 0.18	LP, 0.30	LP, 2.30	LP, 2.30
P	.93	.17	.26	.08	— b	— b

Abbreviations: LogMAR, logarithm of the minimal angle of resolution; LP, light perception; NLP, no light perception; VA, visual acuity.

^aCases with less than 3 months of follow-up data were not included in 3-month visual outcomes analysis but were included in the final visual outcomes analysis.

^bStatistical comparison of visual outcomes at 3 months and the final follow-up could not be performed because only 1 case of post-dexamethasone culture-positive endophthalmitis had a visual outcome other than LP or NLP. Therefore, the sample size in the culture-positive dexamethasone group was insufficient for statistical comparison.

Table 5 and Figure 1 show the change in logMAR VA in cases of suspected endophthalmitis in all treatment groups. The change in VA was not significantly different between the DEX group and the other treatment groups at the 3-month follow-up or the final follow-up (Table 5).

Table 5.

Visual Outcomes of Suspected Endophthalmitis by Treatment Group.

Treatment Group	3-Mo Change in LogMAR VA				Final Change in LogMAR VA
	Mean	Range	Cases, n (%) a	P Value vs DEX	Mean	Range	Cases, n (%) a	P Value vs DEX
Dexamethasone 0.7 mg	+0.06	NLP, +0.30	5 (100)	—	−0.23	NLP, +0.30	5 (100)	—
Ranibizumab 0.5 mg (combined	−0.49	NLP, +0.22	39 (90.7)	.15	−0.33	NLP, +0.30	43 (100)	.77
Conventional ranibizumab 0.5 mg	−0.67	NLP, +0.22	26 (92.9)	.10	−0.43	NLP, +0.30	28 (100)	.62
Prefilled ranibizumab 0.5 mg	−0.13	LP, +0.08	13 (86.7)	.04	−0.12	NLP, 0.00	15 (100)	.60
Ranibizumab 0.3 mg	−0.39	LP, 0.00	6 (100)	.21	−0.82	NLP, +0.22	6 (100)	.36

Abbreviations: DEX, dexamethasone; LogMAR, logarithm of the minimal angle of resolution; LP, light perception; NLP, no light perception; VA, visual acuity.

^aCases with less than 3 months of follow-up data were not included in 3-month visual outcomes analysis but were included in final visual outcomes analysis.

Figure 1.

Change in VA from baseline after suspected endophthalmitis. Baseline VA is defined as the VA at the time of the causative intravitreal injection.

Abbreviations: LogMAR, logarithm of the minimal angle of resolution; VA, visual acuity.

In the DEX group, 1 eye had a VA of NLP at the 3-month and final follow-ups. One case of endophthalmitis in the cR5 group had a VA of NLP at 3 months, and 3 cases had a VA of NLP at the final follow-up. Of the cases of endophthalmitis in the pR5 group, 1 had LP VA at 3 months and at the final follow-up; 1 case had a VA of NLP vision at 6 weeks and was subsequently lost to follow-up. One case of endophthalmitis in the R3 group resulted in a VA of LP at 3 months and the final follow-up; however, the VA was LP at the time of the causative injection. For statistical analysis, this case was considered to have a 0.00 change in logMAR VA over time. Two other cases in the R3 group had LP acuity at 3 months, 1 of which worsened to NLP by the final follow-up; the other improved to HM VA.

Culture-Positive Endophthalmitis

In the R5 group, the mean change in VA was −0.97 (range, NLP to 0.00) at 3 months and −0.59 (range, NLP to +0.12) at the final follow-up. In the cR5 group, the mean change in VA was −1.10 (range, NLP to 0.00) and −0.72 (range, NLP to +0.12), respectively. One case of postinjection endophthalmitis in the cR5 group had a visual outcome of NLP at 3 months, which increased to 2 cases by the final follow-up. In the pR5 group, the mean change in VA was −0.19 (range, LP to −0.08) at 3 months and −0.03 (range, −0.08 to 0.00) at the final follow-up. No patient in the pR5 group with culture-positive endophthalmitis had a visual outcome of NLP at 3 months or at the final follow-up.

In the R3 group, the mean change in VA was −0.65 (range, LP to 0.00) at 3 months and −1.10 (range, LP to 0.00) at the final follow-up. As mentioned, 1 case with postoperative culture-positive endophthalmitis in the R3 group had a VA of LP at the time of the causative injection, at the 3-month follow-up, and at the final follow-up. For statistical analysis, this case was considered to have a change of 0.00 at each time period. An additional case with culture-positive endophthalmitis had a visual outcome of LP at 3 months that subsequently improved to HM by the final follow-up.

The case of endophthalmitis after DEX injection and R3 injection in the same eye on the same day was culture positive with a treatment indication of DME. The eye lost 0.60 logMAR VA at the 3-month follow-up and 0.20 logMAR VA at the final follow-up (follow-up 392 days). This eye was not included in the statistical analyses (Table 4).

Conclusions

In our study, the incidence of suspected endophthalmitis was 1 in 995 injections in the DEX group, 1 in 3813 injections in the R5 group, and 1 in 3159 injections in the R3 group. The incidence of suspected endophthalmitis was significantly higher in the DEX group than in the R5 group, including the conventional and prefilled preparations of the latter medication. The incidence of suspected endophthalmitis was similar in the R3 group and the DEX group.

The incidence of culture-positive endophthalmitis was 1 in 2487 injections in the DEX group, 1 in 9110 injections in the R5 group, and 1 in 6318 injections in the R3 group. Overall, the incidence of culture-positive endophthalmitis was not significantly different between the DEX group and the R3 and R5 groups. However, the pR5 group had a significantly lower risk for culture-positive endophthalmitis than the DEX group. The final visual outcomes in suspected endophthalmitis cases were similar between groups.

In part as a result of the more recent approval of the 0.7 mg dexamethasone implant as a treatment for retinal diseases and fewer treatment indications, it is used less frequently than anti-VEGF medications, resulting in small sample sizes and fewer studies that report endophthalmitis rates after this procedure. Although our sample of approximately 5000 procedures did not provide sufficient statistical power to characterize some aspects of the procedure, ours is among the largest studies to date to assess the incidence of endophthalmitis after dexamethasone injection. Based on the current literature, the incidence of suspected endophthalmitis after dexamethasone injection ranges from 0% to 1.0%.^15,17,22 Our incidence of 0.101% is comparable to previously reported values.

Our study also aimed to quantify the rate of culture-positive endophthalmitis after dexamethasone injection, which few previous reports have characterized. One study reported 2 cases of culture-positive endophthalmitis out of approximately 3400 dexamethasone injections, with an incidence of 1 in 1733 injections (0.06%). ¹⁶ In our study, the incidence was 0.04%. Our comparative analysis found that the rate of culture-positive endophthalmitis after dexamethasone implant injection was not significantly different from that after conventional 0.5 mg and 0.3 mg ranibizumab IVI. However, culture-positive postinjection endophthalmitis was significantly more frequent after dexamethasone injection than after prefilled 0.5 mg ranibizumab injection.

Results in recent studies indicate that prefilled ranibizumab might also have lower rates of endophthalmitis than conventional injections.^{23 –25} Indeed, our rate of culture-positive endophthalmitis was significantly higher after conventional 0.5 mg ranibizumab injection than after prefilled 0.5 mg ranibizumab injection, although the suspected endophthalmitis rates were similar. This difference in culture-positive endophthalmitis might be related to decreased exposure of prefilled syringes to bacterial flora, in particular human oral flora. However, the dexamethasone implant is prefilled in the syringe; thus, the difference in endophthalmitis rates between the dexamethasone injection and prefilled 0.5 mg ranibizumab injection might be the result of other differences, as discussed next.

Our rate of suspected endophthalmitis after ranibizumab IVI (1/3733 injections) is comparable to that reported in other studies. The CATT trial reported an endophthalmitis rate of 1 in 2703 ranibizumab IVI, ² and the MARINA Study Group found a postranibizumab IVI endophthalmitis rate of 1 in 2083 IVI. ²⁶ Other large studies report similar rates of 1 in 1206 to 1 in 2857 IVI.^{27 –29} The rate of culture-positive endophthalmitis after ranibizumab IVI was 1 in 8711 IVI in our study. This is consistent with results reported in the literature, with incidences ranging from 1 in 3962 to 1 in 8250 IVI.^{28 –30} Of note, the incidence of suspected endophthalmitis was 0.032% in both the cR5 group and the R3 group in our study. However, a statistically significant difference was found between the DEX group and cR5 group but not between the DEX group and R3 group. The latter statistical comparison is likely limited by the sample size in the DEX group and R3 group.

Several mechanisms could contribute to an increased rate of endophthalmitis after dexamethasone injection compared with ranibizumab injection. The injection needle used for dexamethasone implant insertion is 22 gauge compared with 30 or 31 gauge for most ranibizumab IVIs. The larger gauge might result in wound gape, leakage, and hypotony, potentially increasing the risk for infection. It is also possible that the corticosteroid in dexamethasone leads to local immunosuppression, facilitating bacterial growth. Furthermore, DME was the indication for 39.9% of dexamethasone injections and 99.7% of 0.3 mg ranibizumab injections in our study. Less than 1% of 0.5 mg ranibizumab injections were administered to treat DME.

Results in previous the studies indicate that patients with diabetes might have a higher risk for endophthalmitis after anti-VEGF IVI than patients with RVO, ⁹ which could account for the greater similarity in the suspected endophthalmitis rates in our R3 group and DEX group. The difference in endophthalmitis rates between the DEX group and R5 group might be partially the result of the underlying diagnosis, not only the result of differences in the inherent risks of each injection procedure. Similarly, the difference in endophthalmitis rates between the R5 group and R3 group might have been the result of the relative immunocompromised state of diabetic patients in the R3 group.

We found that the visual outcomes of patients diagnosed with endophthalmitis were not significantly different after injection of dexamethasone or injection of ranibizumab in suspected cases. The mean decrease in VA after suspected endophthalmitis at the final follow-up was approximately 2 lines in the DEX group, 3 lines in the R5 group (4 lines in cR5 group; 1 line in pR5 group), and 8 lines in the R3 group. In culture-positive cases, the mean decrease in VA was approximately 1 line in the DEX group, 6 lines in the R5 group (7 lines in cR5 group; 0 lines in pR5 group), and 11 lines in the R3 group. This is consistent with the results in other studies of visual outcomes after IVI-associated endophthalmitis, which show that the majority of patients do not regain baseline VA.^{9,31 –33}

A limitation of this retrospective cohort study is that not all known potential confounders were measured and reported. One of the primary outcome measures (VA) was assessed using Snellen notation, which is not as precise or accurate as the standardized Early Treatment Diabetic Retinopathy Study vision score. Because 2 institutions participated in this study, slight variations in IVI protocol might have been used at each site. Individual physicians’ practices regarding a no-talking policy and injection technique, including the quadrant of injection, conjunctival displacement, and the use of a bladed lid speculum or manual lid retraction, might have varied. In addition, because of the differences in needle gauge, eyes in the DEX group might have received viscous anesthetic agents or subconjunctival anesthetics more frequently than eyes receiving ranibizumab.

Furthermore, loss of patients to follow-up might have led to a lower measured incidence of endophthalmitis than the true value. Loss to follow-up might also have altered our recorded visual outcomes and the statistical power of our visual outcome analyses. However, given the tertiary-care nature of both institutions in this study, it is unlikely that many patients sought follow-up elsewhere. Our analyses of VA outcomes might have been affected by the wide range of follow-up durations among the cases in our study. By including patients with both short-term follow-up (<3 months) and long-term follow-up (> 1 year), confounding factors affecting visual decline or improvement (eg, progression of cataract or retinal disease, including epiretinal membrane, retinal detachment, or advancement of underlying macular pathology) could be introduced. In addition, there might have been an imbalance in known and unknown prognostic factors between the DEX group and anti-VEGF groups. Furthermore, although the sample size of our study was considerable, endophthalmitis is a rare event that might require an even larger dataset to provide adequate power for analysis.

Just as IVIs of anti-VEGF agents have increased in use for the treatment of retinal disease, the use of dexamethasone implants is likely to continue to rise in coming years because of their utility in treating chronic ocular conditions such as DME, RVO, and posterior uveitis. Our study found a higher incidence of suspected endophthalmitis after dexamethasone implant insertion than after 0.5 mg ranibizumab IVI. In addition, dexamethasone implant injection was associated with higher rates of culture-positive endophthalmitis than prefilled 0.5 mg ranibizumab injection. However, visual outcomes after suspected endophthalmitis treatment were comparable between eyes receiving a dexamethasone implant and those receiving a ranibizumab injection.