Article Text

Download PDFPDF

Incidence and clinical features of post-injection endophthalmitis according to diagnosis
  1. Nadim Rayess1,
  2. Ehsan Rahimy1,
  3. Chirag P Shah2,
  4. Jeremy D Wolfe3,
  5. Eric Chen4,
  6. Francis C DeCroos5,
  7. Philip Storey1,
  8. Sunir J Garg1,
  9. Jason Hsu1
  1. 1The Retina Service of Wills Eye Hospital, Philadelphia, Pennsylvania, USA
  2. 2Ophthalmic Consultants of Boston, Boston, Massachusetts, USA
  3. 3William Beaumont Hospital, Royal Oak, Michigan, USA
  4. 4Retina Consultants of Houston, Houston, Texas, USA
  5. 5Southeastern Retina Associates, Chattanooga, Tennessee, USA
  1. Correspondence to Dr Jason Hsu, Department of Ophthalmology, Mid Atlantic Retina, The Retina Service of Wills Eye Hospital, Thomas Jefferson University, Philadelphia, PA 19107, USA; jhsu{at}


Purpose To compare the incidence and clinical features of endophthalmitis after intravitreal antivascular endothelial growth factor (VEGF) therapy for diabetic eye disease, neovascular age-related macular degeneration (AMD) and retinal vein occlusion (RVO).

Methods Multicentre, retrospective, consecutive case–control study. All patients treated with intravitreal bevacizumab, ranibizumab or aflibercept for diabetic eye disease, neovascular AMD or RVO between 1 January 2009 and 30 September 2013 at three retina practices were included in this study. The total number of anti-VEGF injections administered for the three indications was calculated using billing records. Endophthalmitis cases were identified using both endophthalmitis log sheets and billing records. Patient charts were reviewed to confirm that endophthalmitis was directly related to anti-VEGF injection and to record clinical features and culture results.

Results During the study period, a total of 353 978 intravitreal anti-VEGF injections were performed. Presumed infectious endophthalmitis occurred in 119 of 296 017 injections performed for neovascular AMD (1/2487, 0.040%), 12 of 24 541 for diabetic eye disease (1/2045, 0.049%) and 4 of 32 418 for RVO (1/8104, 0.012%). χ2 analysis found endophthalmitis rates to be higher in diabetic eye disease compared with RVO (p=0.010) and higher in neovascular AMD compared with RVO (p=0.014), while diabetic eye disease and neovascular AMD (p=0.517) had similar rates. The average age of the overall neovascular AMD patient population (81.9 years) was significantly older than the diabetic eye disease (64.7 years, p<0.001) and RVO (73.4 years, p<0.001) populations.

Conclusions Endophthalmitis rates appear to be lower in eyes with RVO compared with diabetic eye disease and neovascular AMD, possibly due to impaired immunity in diabetics and waning immunity in the generally older AMD population.

  • Infection
  • Microbiology
  • Retina

Statistics from


Intravitreal antivascular endothelial growth factor (VEGF) injections are the primary treatment of neovascular age-related macular degeneration (AMD), diabetic macular oedema (DME) and macular oedema secondary to retinal vein occlusion (RVO). The most significant complication of intravitreal injections is post-injection endophthalmitis, which occurs in approximately 0.01–0.08% of cases and can be a cause of significant vision loss.1–6

Systemic risk factors for postoperative endophthalmitis have been assessed in a number of other studies,7 ,8 and one series demonstrated that the majority of post-vitrectomy endophthalmitis cases occurred in patients with diabetes mellitus.7 This increased incidence observed in patients with diabetes was attributed to their overall impaired cellular and humoral immune responses, predisposing them to systemic infections, potentially including endophthalmitis.7

Although patients with diabetes may be at an increased risk of developing endophthalmitis following intraocular surgery, to our knowledge, no study has assessed whether the risk of developing endophthalmitis after intravitreal injection differs depending upon the underlying diagnosis. This multicentre case–control study stratified endophthalmitis rates based on the following treatment indications: diabetic eye disease, neovascular AMD and RVO, and sought to determine whether indication for anti-VEGF injection affected the incidence of post-injection endophthalmitis.


The participating centres involved in this study included The Retina Service of Wills Eye Hospital, Mid Atlantic Retina, Philadelphia, Pennsylvania, USA; Associated Retinal Consultants at William Beaumont Hospital, Royal Oak, Michigan, USA; and Retina Consultants of Houston, Bellaire, Texas, USA. Each site used billing records and endophthalmitis logs to identify cases of endophthalmitis secondary to intravitreal anti-VEGF injections that occurred between 1 January 2009 and 30 September 2013. Patient clinical records were reviewed to confirm that the endophthalmitis was directly related to an antecedent intravitreal anti-VEGF injection. The date of the causative injection, vitreous/anterior chamber tap with intravitreal antibiotic injection, indication for anti-VEGF injection, visual acuity (VA), patients’ age and culture results were recorded from patient charts.

Billing records were used to determine the total number of intravitreal injections of bevacizumab (Avastin; Genentech, South San Francisco, California, USA), ranibizumab (Lucentis; Genentech) and aflibercept (Eylea; Regeneron, Tarrytown, New York, USA) administered for each of the following diagnoses: neovascular AMD, DME, proliferative diabetic retinopathy (PDR), branch retinal vein occlusion (BRVO) and central retinal vein occlusion (CRVO). The age of all patients that received intravitreal anti-VEGF injections for each of the three diagnoses was obtained using billing records.

Inclusion and exclusion criteria

All patients diagnosed with presumed infectious endophthalmitis were included in this study. Endophthalmitis was defined as any case in which clinical suspicion was high enough to warrant either a vitreous/anterior chamber tap with injection of intravitreal antibiotics or a primary pars plana vitrectomy (PPV) as manifest by pain, significantly decreased VA within 7 days of the intravitreal injection or demonstrated signs of intraocular inflammation on examination (generally ≥2+ anterior segment cellular reaction and/or posterior segment vitritis). Only patients who had received an intravitreal injection of bevacizumab, ranibizumab or aflibercept for neovascular AMD, DME/PDR or BRVO/CRVO were included. Furthermore, patients with presumed non-infectious endophthalmitis who were treated with topical steroids instead of a tap and inject were excluded.

Injection technique

The injection procedure occurred in an office-based setting. All eyes were prepped using two cycles of topical anaesthesia and 5% povidone–iodine (Betadine 5%; Alcon Labs, Fort Worth, Texas, USA) drops applied to the ocular surface. Use of a sterile lid speculum, conjunctival displacement and quadrant of intravitreal injection were at the discretion of the treating physician. Some patients also received subconjunctival lidocaine 2% injection followed by additional 5% povidone–iodine. The injection was performed using a 30-gauge or 31-gauge needle at approximately 3.5–4.0 mm from the limbus. During the study period, patients initially received post-injection topical antibiotic prophylaxis for up to 4 days following the intravitreal injection. This transitioned to a period where the use of post-injection antibiotics was determined by each physician. By the end of the study period, there was a trend towards not prescribing them.

Statistical analysis

The primary outcome of this study was to compare endophthalmitis incidence following anti-VEGF injections according to the underlying ocular diagnosis. Pearson's χ2 analysis was performed to compare the incidence of endophthalmitis among the three diagnoses. Furthermore, unpaired two-tailed t-test analysis was used to detect any difference in the age of patients receiving anti-VEGF therapy according to their underlying diagnosis. Statistical analysis was performed using GraphPad Software (GraphPad, La Jolla, California, USA).


A total of 135 cases of endophthalmitis from a total of 353 978 intravitreal anti-VEGF injections (1/2622 injections, 0.038%) were included in this study. Clinical characteristics leading up to the diagnosis of endophthalmitis and the following treatment strategies implemented are summarised in table 1 for the three diseases. Patients with neovascular AMD received 296 017 anti-VEGF injections, and there were 119 cases of presumed infectious endophthalmitis (1/2487 injections, 0.040%). Patients in this group who developed endophthalmitis had a mean age of 80.3±9.3 years, whereas the average age of the total neovascular AMD population receiving anti-VEGF injections was marginally older, 81.9±9 years old (p=0.053). Patients with DME/PDR received 24 541 anti-VEGF injections, from which 12 cases of presumed infectious endophthalmitis developed (1/2045 injections, 0.049%). This subset of DME/PDR endophthalmitis patients had a mean age of 63.7±10.6 years. Similarly, the average age of the DME/PDR patient population receiving anti-VEGF injections was 64.7±11.8 years old (p=0.769). Finally, patients with BRVO/CRVO received 32 418 anti-VEGF injections, from which four cases of endophthalmitis were identified (1/8104 injections, 0.012%). The mean age of RVO patients who developed endophthalmitis was 80.3±8.1 years, which was older than the average age of the RVO patient population receiving anti-VEGF injections (73.4±12.2 years), although this was not statistically significant (p=0.258). Table 2 summarises the visual outcomes, and table 3 describes the microorganisms isolated from the culture-positive cases for the various diagnoses.

Table 1

Post-injection endophthalmitis according to diagnosis: timeline and treatment strategies

Table 2

Post-injection endophthalmitis according to diagnosis: visual outcomes

Table 3

Post-injection endophthalmitis according to diagnosis: microbiologic spectrum

Patients receiving injections for neovascular AMD had a significantly higher risk of developing endophthalmitis compared with those with BRVO/CRVO (p=0.014). Similarly, individuals receiving intravitreal injections for DME/PDR had a significantly higher risk of developing endophthalmitis compared with patients with BRVO/CRVO (p=0.010). Conversely, when compared with each other, there was no difference in endophthalmitis rates between the neovascular AMD and DME/PDR subgroups (p=0.517).

The neovascular AMD patient population in this study had a mean age of 81.9±9 years, which was significantly older than both patients with DME/PDR (64.7±11.8 years; p<0.001) and patients with RVO (73.4±12.2 years; p<0.001). Patients receiving anti-VEGF therapy for RVO were found to be significantly older than patients with DME/PDR (p<0.001).


This multicentre, retrospective consecutive case–control study evaluated the rates of endophthalmitis due to intravitreal anti-VEGF injections stratified by underlying treatment indication. Specifically, we compared endophthalmitis rates for the three most common diagnostic indications for injections: neovascular AMD, DME/PDR and BRVO/CRVO. Our study demonstrated that patients receiving intravitreal injections for RVO had the lowest overall incidence of developing endophthalmitis (0.012%). Conversely, patients receiving injections for DME/PDR and neovascular AMD had similar overall incidences of endophthalmitis (0.049% and 0.040%, respectively), and fall within range of the previously reported post-injection endophthalmitis rates of 0.01–0.08%.5 ,9 However, patients with neovascular AMD and DME/PDR had a significantly higher endophthalmitis rate than patients with RVO. With the paradigm shift towards using intravitreal anti-VEGF injections as the treatment of choice for several common retinal diseases, it is helpful to determine whether certain populations with a particular diagnosis are at greater risk of developing endophthalmitis.

We hypothesise that the difference in rates observed may be a consequence of differences in the immune response between the groups. Since DME and PDR are a complication of the cumulative damage from diabetes mellitus, these patients are more likely to have had poor diabetic control for a number of years. Moreover, diabetes is a systemic disease that negatively impacts the immune system and decreases wound healing. From a systemic standpoint, diabetics have an increased prevalence of infections.10 ,11 This observation could be logically explained by defects in the innate immune system, an increased adherence of microorganisms to diabetic cells, microvascular abnormalities and a greater number of invasive procedures that diabetics require.10

Some may argue that due to the immune privilege of the eye any amount of bacteria would lead to endophthalmitis irrespective of systemic immune functioning. However, Maxwell et al12 demonstrated that a certain threshold concentration of bacteria had to be inoculated into the vitreous cavity in order to induce endophthalmitis, suggesting that immune response in fact plays an important role. Patients with diabetic eye disease may have a poorer immune response, and therefore, require a lower threshold of bacterial inoculation to develop endophthalmitis.

Another explanation may be an alteration of the conjunctival flora in diabetics. A study by Martins et al13 demonstrated that patients with diabetes had a higher positive conjunctival culture rate than patients without diabetes. In addition, patients with diabetes with retinopathy had an even higher conjunctival culture positivity rate compared with those without retinopathy.13 Therefore, an altered conjunctival flora and/or higher bacterial load could also explain the increased rate of endophthalmitis observed in our study.

Other studies have suggested that endophthalmitis following invasive ophthalmic procedures such as PPV or cataract surgery may occur more frequently in patients with diabetes.7 ,14 Cohen et al7 evaluated 18 cases of endophthalmitis following PPV. Their study demonstrated that 11 (61%) of these patients had diabetes mellitus, suggesting that diabetics are at an increased risk of developing endophthalmitis following intraocular surgery.7 Our study further supports the idea that endophthalmitis may be more prevalent in diabetics in the setting of office-based intravitreal injections.

Separately, age may be an additional factor that impacts a patient's overall immune system, with older patients having a weaker immune response.15 Immunological studies have demonstrated an impaired T-cell response in older, yet healthy patients. This is related to a decrease in naive T-cell number, with a relative increase in hypofunctioning memory T cells.15 Additionally, elderly patients have impaired stimulation of T cells that can be in part attributed to an increased production of anti-inflammatory cytokines such as interleukin-10 by antigen-presenting cells.15 Furthermore, elderly patients are more likely to have chronic disorders or receive medications that impair their immune response, increasing their susceptibility to infections.16

In our study, patients receiving injections for neovascular AMD were significantly older (mean, 81.9 years) than those receiving injections for DME/PDR (mean, 64.7 years) and RVO (mean, 73.4 years). We hypothesise that the more advanced age of individuals with neovascular AMD may lead to an increased risk of endophthalmitis due to a combination of poorer immune response and increased concentration of bacterial conjunctival flora in older patients17 similar to the patients with DME/PDR. This could account for the comparable endophthalmitis rates that were noted in patients with neovascular AMD (1/2487 injections, 0.040%) and DME/PDR (1/2045 injections, 0.049%).15 ,16 It is important to note that the average age of neovascular AMD patients (80.3±9.3 years) who developed endophthalmitis was clinically similar to the average age of the neovascular AMD population (81.9±9 years). However, statistically it was the closest of the three groups to achieving statistical significance (p=0.053). This is because the neovascular AMD population was the most powered group in the study.

When comparing patients with DME/PDR and patients with RVO, patients with DME/PDR in our study may be on average somewhat younger, yet their apparent risk of endophthalmitis was higher. This suggests diabetes mellitus is likely a major contributing factor for decreased immunity. Overall, our study suggests that patients receiving intravitreal anti-VEGF therapy for DME may be at a higher risk of developing endophthalmitis than patients with RVO. Similarly, the higher endophthalmitis rates detected in patients with neovascular AMD compared with patients with RVO could be attributed to a decrease in their immune response as they represent a generally older population age group than patients with RVO.

In this series, endophthalmitis was diagnosed on average around 3.2–4.0 days following the causative injection. Furthermore, the overall mean VA was counting fingers at presentation. Interestingly, patients with RVO were more likely to regain their lost VA than those with neovascular AMD or diabetic eye disease. This difference in response to therapy could be explained by a more preserved immune system in patients with RVO compared with patients with neovascular AMD and diabetic eye disease. However, this difference could also be attributed to the limited number of cases of endophthalmitis that occurred in patients with RVO.

There are several limitations to this multicentre case–control study on post-injection endophthalmitis. First, as this involved multiple treating physicians across three institutions, we do not know whether differences in institution or even physician injection protocols may have contributed to differences in endophthalmitis rates. The retrospective nature of this study prevented us from controlling for outside factors that a standardised protocol would have allowed. Notably, the use of post-injection prophylaxis was not controlled and was dependent on physician preference. During the early part of the study period, post-injection antibiotic prophylaxis was a standard practice. By the end of the study period, there was a trend towards not prescribing post-injection antibiotics. These changes in physician preference towards prescribing postprocedural prophylaxis reflect the evolving literature on this topic, which demonstrated no risk reduction in endophthalmitis with their use.18–21 Additionally, we did not exclude patients who had multiple diagnoses. Therefore, it is likely that some patients receiving intravitreal injections for one diagnosis (eg, RVO) may have also carried other diagnoses (eg, diabetes mellitus). However, we would expect the presence of diabetes in patients with RVO to be a confounder that would tend to overestimate the endophthalmitis rates in patients receiving injections for RVO. Despite this, our results still demonstrate a lower endophthalmitis rate in patients with RVO.

Another limitation of the study is the difference in total number of injections according to each diagnosis. There were fewer injections for DME and RVO in comparison to neovascular AMD. Despite these limitations, this is the largest study to date investigating post-injection endophthalmitis, having characterised endophthalmitis rates from a sample of >350 000 injections. Additionally, this is the first study to look into differences in endophthalmitis rates according to indication for intravitreal injection.

The results of this multicentre study on post-injection endophthalmitis rates demonstrate that patients receiving injections for diabetic eye disease and neovascular AMD appear to be at a higher risk of developing endophthalmitis than patients with BRVO/CRVO. We hypothesise that the effects of diabetes mellitus and ageing on weakening the immune system and altering the conjunctival flora may play additional roles in determining a patient's risk of developing endophthalmitis following intravitreal injections.


View Abstract


  • Funding J. Arch McNamara Memorial Fund for Retina Research through Wills Eye Hospital was used to fund the central IRB. The sponsor or funding organisation had no role in the design or conduct of this research.

  • Competing interests None declared.

  • Ethics approval Institutional Review Board (IRB) approval for this multicentre, retrospective consecutive case series was obtained through the Wills Eye Hospital and Western IRBs.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • At a glance
    Keith Barton James Chodosh Jost Jonas