1. Introduction
Antimicrobial therapy is one of the most important medical discoveries of the twentieth century, and it has prevented millions of cases of premature morbidity and mortality from bacterial infection. Since antimicrobials have become widely available, the burden of resistance among bacteria has increased in tandem, especially within the past 10 years. In endophthalmitis, where rapid control of infection is important for maintaining sight, reported rates of antimicrobial resistance have risen in northern American states [
1,
2]. This finding mandates an updated review of local trends in antimicrobial resistance and its impact on long-term visual function.
Local patterns of antimicrobial resistance are a major consideration in the choice of empiric antimicrobial therapy for endophthalmitis. Empiric intravitreal administration of antimicrobials, such as Vancomycin plus either Amikacin or Ceftazidime, is widely adopted, as they cover Gram-positive and Gram-negative microorganisms, respectively [
3,
4,
5]. In India, there have been few case reports of emerging antimicrobial resistance against these antibiotics [
6,
7,
8,
9]. Historical reports from Auckland (1983–1991) and Queensland, Australia (1998–2013) have provided assurance that this trend has not spread to the Australasian region [
10,
11].
Microorganisms evolve to develop antimicrobial resistance through two mechanisms: (1) mutations in genes associated with drug target sites and (2) horizontal gene transfer of foreign DNA containing resistance determinants [
12]. When used, antibiotics exert selection pressure for microorganisms possessing resistance mechanisms, which now have a survival advantage and are able to transfer forward resistance genes. This relationship is demonstrated in the hospital environment, where antimicrobial use is associated with the emergence of multiresistant strains [
13,
14]. From a broader perspective, trends in antimicrobial resistance relate to the inappropriate use of antibiotics in agriculture, community infections, healthcare policy, infection control and host migration [
15,
16,
17].
Interpretation of antimicrobial resistance in clinical practice is a relative phenomenon with many layers of complexity. The establishment of clinical susceptibility breakpoints relies on the in vitro activity of an antibiotic against a sizeable bacterial sample, but it does not take into consideration the pharmacodynamics and pharmacokinetics of the drug, which vary according to the route and site of antibiotic administration. In vitro resistance does not necessarily equate to treatment failure in the context of intraocular pathogens, as high intraocular concentrations of antibiotics following an intravitreal bolus dose in the early stages of infection can overwhelm mechanisms of resistance [
18]. Wu et al., found that antimicrobial resistance conferred no additional risk of vision loss provided all patients received empiric intravitreal antibiotics [
5].
This study aims to describe the implicated microorganisms in endophthalmitis, their antimicrobial susceptibility profile, and clinical outcomes in a New Zealand population following a standardized treatment protocol that incorporates empiric treatment with intravitreal Vancomycin and Ceftazidime.
3. Results
Three hundred eighty-nine cases of endophthalmitis were included during the study period. The median age at presentation was 70.0 years [IQR 58.1–80.0], and 203 (52.2%) were female. The underlying cause of endophthalmitis was cataract surgery in 117 (30.1%), intravitreal anti-vascular endothelial growth factor (VEGF) injection in 97 (24.9%), endogenous in 57 (14.7%), post vitrectomy in 32 (8.2%), post glaucoma surgery in 29 (7.5%), following corneal infection or surgery in 5 (1.3%), and following trauma or other procedures in 52 (13.4%).
The median duration of symptoms at presentation was one day [IQR 1–3 days]. The median time following surgery to presentation was 7.5 days [IQR 3–42] for cataract surgery, 4 days [IQR 2–6.5] for intravitreal injection, 1206 days [IQR 38–2746] for glaucoma surgery, 3 days [IQR 2–10] for post vitrectomy, and 3 days [IQR 2–16 days] for corneal surgery. The median time to presentation following trauma was 5 days [IQR 2–8]. The mean presenting BCVA was 1.88 ± 0.80 LogMar. The median BCVA was hand movement [IQR 20/400–hand movement]. Hypopyon was present in 202 (51.9%), and red reflex was present in 147 (37.8%) and absent in 242 (62.2%).
A positive culture was obtained in 207 eyes (53.2%). An anterior chamber tap was performed in 167 eyes (42.9%) and was positive in 50 (29.9%). A vitreous tap was performed in 228 (58.6%) and was dry in 38. Of the 190 successful vitreous taps, culture results were positive in 83 (43.7%). Vitrectomy was performed in 247 eyes (63.5%) and was culture positive in 105 (42.5%). If vitrectomy was the primary procedure, it was culture positive in 62 of 122 (50.8%), whereas if vitreous tap and inject was the primary procedure, vitrectomy was positive in 42 of 138 (30.4%).
Organisms were Gram-positive in 162 eyes (79.8%), Gram-negative in 29 eyes (14.2%) and fungal in 12 eyes (5.9%) (
Table 1). All Gram-positive microorganisms were tested for susceptibility to Vancomycin. In the nine Gram-negative microorganisms tested for Ceftazidime susceptibility, only one (11.1%) demonstrated intermediate susceptibility. Resistance to at least one antimicrobial agent was present in 89 culture results (43.0%), and multidrug resistance (resistance to ≥3 drugs) was present in 23 culture results (11.1%). In terms of fluoroquinolone resistance, two samples (3.2%) were resistant to ciprofloxacin, and all five (0%) microorganisms tested against moxifloxacin were susceptible. Resistance to at least one antimicrobial agent and multidrug resistance for
Staphylococcus aureus was 63.2% and 5.3%, respectively. Similarly, for coagulase-negative staphylococci (CoNS), it was 69.1% and 17.6%, respectively.
There were no predictors for any resistance or multidrug resistance observed in patient demographics (age, gender) or in clinical presentation (cause of endophthalmitis, presenting vision, hypopyon, red reflex). No increase in the likelihood of resistant microorganisms was observed over time (OR 0.974
p = 0.345), and no association was observed between year of presentation and multidrug resistance (OR 1.037
p = 0.419). The frequency of resistance and multidrug resistance by year are reported in
Table 2.
The primary procedure was a tap and inject in 251 eyes (64.5%) and a vitrectomy in 138 eyes (35.5%). In those receiving a primary tap and inject, a further 58 underwent a secondary vitrectomy <24 h from presentation. Late vitrectomy (>24 h) occurred in 33 eyes (28.7%) with no antimicrobial resistance and in 19 eyes (21.8%) with at least one antimicrobial resistance (p = 0.270). There was also no association observed between late vitrectomy and multidrug resistance (25.6% with no multidrug resistance vs. 27.3% with multidrug resistance, p = 0.862).
Median follow-up time was 10.1 months [IQR 2.3–34.1]. Retinal detachment occurred in 44 eyes (11.3%), and enucleation/evisceration in 29 eyes (7.5%). Mean visual acuity was 1.13 ± 1.03 LogMar at three months, 1.15 ± 1.07 LogMar at six months, and 1.16 ± 1.10 LogMar at nine months. Median visual acuity was 20/80 [IQR 20/40–hand movement] at three months, 20/100 [IQR 20/30–hand movement] at six months, and 20/100 [IQR 20/30–hand movement] at nine months. At the last recorded follow-up, mean visual acuity was 1.18 ± 1.06 LogMar, and median visual acuity was 20/100 [IQR 20/30–hand movement]. Severe vision loss (≤20/200) occurred in 167 eyes (42.9%).
Risk factors for retinal detachment are reported in
Table 3. On univariate analysis, the following variables were associated with retinal detachment: younger age (OR 0.971
p < 0.001); weekend presentation (OR 2.151
p = 0.025); presenting visual acuity (OR2.977
p = 0.002); hypopyon (OR 2.274
p = 0.028); and absence of red reflex (OR 0.400
p = 0.014). On multivariate analysis, the following variables were associated: younger age (OR 0.977
p = 0.028); hypopyon (OR 3.239
p = 0.048); and antimicrobial resistance (OR 2.455
p = 0.048).
Risk factors for enucleation or evisceration are reported in
Table 4. On univariate analysis, only presenting visual acuity was associated with increased risk of enucleation or evisceration (OR 2.965
p = 0.013). On multivariate analysis, there were no associated risk factors observed. No significant association was observed between antimicrobial resistance and risk of enucleation.
Risk factors for severe vision loss (≤20/200) are reported in
Table 5. On univariate analysis, the following were associated with increased risk of vision loss: younger age (OR 0.989
p = 0.042); poor presenting vision (OR 3.451
p < 0.001); and absence of red reflex (OR 0.362
p < 0.001). On multivariate analysis, poor presenting vision was associated with increased risk of vision loss (OR 3.323
p < 0.001), and early vitrectomy (<24 h) was associated with reduced risk (OR 0.565
p = 0.020). There was no association between antimicrobial resistance or multidrug resistance with severe visual acuity loss.
4. Discussion
In light of increasing antimicrobial resistance globally, this study describes trends in antimicrobial resistance in New Zealand and its effect on patient outcomes. Vancomycin and Ceftazidime provided comprehensive coverage over Gram-positive and Gram-negative organisms, respectively. Resistance to at least one organism was present in nearly half of the samples, and multidrug resistance in approximately 10%. There was no trend towards increasing antimicrobial resistance over time. Whilst antimicrobial resistance was not associated with an increased risk of vision loss or enucleation, it was associated with increased risk of retinal detachment.
Empiric treatment plays a crucial role in limiting vision loss by halting microbial proliferation and should not be delayed. In the maxim of “Time is Retina”, Michael et al., retrospectively reviewed 374 eyes with endophthalmitis and found that treatment within 2 h with intravitreal antibiotics was associated with a better final visual outcome [
27]. Historically, dual therapy, with Vancomycin directed against Gram-positive organisms and either Amikacin or Ceftazidime against Gram-negative organisms, has been established as a rational choice to cover a heterogenous group of causative organisms [
3,
28]. Whilst this combination of antibiotics is validated by low resistance rates in our study population, the same does not apply to other centres that report emerging cases of antimicrobial resistance [
6,
7,
8,
9,
29]. In comparison with the ARMOR surveillance study, there was at least a two-fold rate increase in multidrug resistance in CoNS (46.3%) and
Staphylococcus aureus (41%) isolates [
30]. Such resistant organisms are associated with poor visual outcomes and warrant regular reviews of local empiric antibiotic selection.
Antimicrobial stewardship constitutes coordinated interventions designed to improve and measure the appropriate use of antimicrobials by promoting the selection of the optimal antimicrobial drug regimen, dose, duration of therapy, and route of administration [
31]. New Zealand fares better than other nations, with comparatively lower rates of antimicrobial resistance and conservative antibiotic usage, and this acts as a strong motivator for collaborative efforts to stop the emergence of further resistance. Nationwide initiatives include: (1) antimicrobial prescribing guidelines enabling prescribers to select an effective agent at the correct dose with the narrowest spectrum, fewest adverse effects and lowest cost [
32], (2) continuous surveillance of all clinical laboratories’ susceptibility testing for emerging antibiotic resistance, (3) infection control policy that controls the spread of resistant strains and decreases overall use of antimicrobials [
17,
33], (4) all antibiotics supplied by prescription only, (5) infectious disease approval is required for last-line antibiotics, and (6) antibiotics in farming animals are regulated for therapeutic and prophylactic purposes, as opposed to large-scale use as growth promoters in other parts of the world [
16,
31,
34].
Local protocols instituted in the studied eye department include: (1) avoiding the use of topical antimicrobials before or after intravitreal injections, (2) utilising intracameral cefuroxime for infection prophylaxis following routine cataract surgery instead of topical antimicrobials, and (3) judicious use of topical or oral antimicrobials for ocular infections. These local measures may decrease the presence of resistant organisms within a single individual’s ocular flora but do not make much difference in levels of resistance within the community. There is good evidence that topical antimicrobials do not minimise the risk of endophthalmitis following intravitreal injection [
35]. On the contrary, prolonged use of antimicrobials eliminates the natural flora and favours proliferation and potential infection with antimicrobial-resistant organisms [
36,
37]. Intracameral cefuroxime at the conclusion of routine cataract surgery provides the best protection against post-operative endophthalmitis, with topical antimicrobials conferring little added benefit [
38,
39,
40]. Finally, the use of fluoroquinolones, such as topical ciprofloxacin and oral moxifloxacin, is strictly regulated. Prescribing rights are exclusive to tertiary ophthalmology service providers, and oral moxifloxacin is reserved for use in penetrating eye trauma.
Our cohort of culture-positive endophthalmitis (n = 207) is the largest to date that addresses conflicting views about the impact of antimicrobial resistance on visual outcomes. Whilst Wu et al., (n = 99) found that antimicrobial resistance was not associated with a risk of vision loss, Choi et al., (n = 82) found that resistance to Vancomycin or third-generation cephalosporins was associated with a 75% lower chance of achieving VA better than counting fingers [
5,
41]. Further complicating matters is the lack of concordance between in vitro antimicrobial sensitivities and clinical response, which is reflective of greater concentrations of antimicrobials following intravitreal administration. Following an intraocular injection of Vancomycin at the same treatment dose for endophthalmitis, concentrations exceeded the minimum inhibitory concentration (MIC) of endophthalmitis-causing Gram-positive bacteria four-fold for up to 26 h. [
18] Alternative measurements, such as the minimal bactericidal concentration or time-kill curve of antibiotics, may provide greater value in infections in immune-privileged sites such as the inner chamber of the eye [
42].
The association between antimicrobial resistance and higher rates of retinal detachment would suggest underlying involvement of other virulence factors. In a rabbit model of
Staphylococcus epidermidis-induced endophthalmitis, Kaspar et al., found that antimicrobial-resistant strains caused more inflammation and destruction of the infected retina than antimicrobial-susceptible strains. One plausible explanation is that virulence factors are genetically transferred together with resistance vectors, accumulating in the longer-surviving bacteria [
43]. These bacteria are better capable of ocular tissue invasion, surviving in the intraocular compartment, breaking down the blood–retinal barrier and triggering a destructive immune response in the retina [
44]. The fact that weekend presentations are also associated with retinal detachments could be reflective of either greater symptom severity or longer delay to initial presentation. Previous work by our research group has not observed any delay in treatment with weekend presentations nor a decreased likelihood of receiving a vitrectomy over the weekend [
27].
Visual outcome based upon the proportion of patients without severe vision loss is better than previously reported, which could be attributable to differences in study population, research protocols and treatment regimens. Wu et al., conducted a study utilising a similar protocol and found a higher rate of severe vision loss compared to our cohort (42.9% vs. 67.7%) and a worse mean final BCVA (2.19 LogMar vs. 1.18 LogMar) [
5]. As presenting vision and culture positivity rates were equivocal, other possible explanations for this difference could be the higher degree of reported antimicrobial resistance and lower rates of early vitrectomy (50.4% vs. 4.0%). Early vitrectomy, in particular, has been shown to be predictive of better long-term visual outcome [
45].
There was a steep rise in both incidence rates and percentage of culture-positive cases from 2019 to 2021, coinciding with the start of the COVID-19 pandemic. Whilst the number of elective cataract and vitreoretinal surgeries decreased, intravitreal injections continued largely unabated during the COVID pandemic. Mandatory mask wearing, social distancing and lockdowns may also have played a role in the disease spectrum. Finally, the year-to-year variation in the percentage of culture-positive cases may be associated with the likelihood of diagnosing endophthalmitis versus sterile inflammation, as all those managed with intravitreal antibiotics and sampling were included in our study cohort.
The limitations of this study are inherent to retrospective analysis, which include incomplete data, selection bias and lack of randomisation. However, it is the largest cohort to date with comprehensive long-term data allowing for visual prognostication based on baseline factors and treatment. Although our protocol for antimicrobial susceptibility follows global standards, subtle variations in antimicrobial selection represent a potential source of sampling bias. Lower quantities of antimicrobials selected for testing result in underdetection of antimicrobial resistance, and vice versa. In line with local antimicrobial stewardship initiatives, first-line antibiotics assume priority in testing and reporting to promote their clinical use.