Corneal culture and antibiotic susceptibility results for microbial keratitis in the Mid-Atlantic region of the United States, 2016–2020

Abstract

Objective:

To examine the microbial distribution and antimicrobial susceptibility of culture-positive microbial keratitis at a large tertiary referral center in the mid-Atlantic region of the United States.

Methods:

Retrospective review of culture-positive microbial keratitis cases at the Wilmer Eye Institute from 2016 through 2020.

Results:

Of the 474 culture-positive microbial keratitis cases, the majority were bacterial (N=450, 94.9%), followed by fungal (N=48, 10.1%) and Acanthamoeba (N=15, 3.1%). Of the 450 bacterial isolates, 284 (69.5%) were Gram-positive organisms while 157 (28.4%) were Gram-negative. The most common bacterial species isolated was coagulase negative Staphylococcus spp. (N=154, 24.8%), and the most common Gram-negative isolate was Pseudomonas aeruginosa (N=76, 12.3%). Among fungi, the most common isolates were Candida (N=25, 45.4%) while Fusarium (N=6, 10.9%) and Aspergillus (N=3, 5.5%) were less common. Of the 217 bacterial isolates tested for erythromycin susceptibility, 121 (55.7%; about 60% of coagulase negative staphylocci and corynebacteria tested) showed resistance to erythromycin.

Conclusions:

Microbial keratitis in the Baltimore Mid-Atlantic region of the United States is most commonly caused by bacteria, with fungi and acanthamoeba being less common. Gram-positive bacterial infections predominate. Among fungal keratitis cases, Candida species are more commonly encountered than are filamentous species. Use of erythromycin as infection prophylaxis should be reexamined. Findings from our study may guide empiric treatment in this geographic region.

INTRODUCTION

Microbial keratitis, or infectious corneal ulceration, causes unilateral blindness for 1.5 to 2 million people per year globally.¹ This figure is likely an underestimate and does not account for many more cases of moderate or severe visual impairment. Timely identification of causative organisms, in vitro assessment of antimicrobial susceptibility, and initiation of appropriate treatment are necessary to prevent vision-threatening complications such as corneal scarring, perforation, and endophthalmitis.

The distribution of causative organisms varies based on geography, climate, and patient sociodemographic features,^2–8 and seasonality.^9–12 Risk factors for microbial keratitis include contact lens wear, trauma, ocular surface disease,¹³ and recent eye surgery. Since many cases of microbial keratitis are not cultured, knowledge of a region’s microbial distribution may help guide empiric treatment.^14–17 To improve clinical practice, we sought to examine the microbial distribution and antimicrobial susceptibility of culture-positive microbial keratitis at a large, tertiary referral center in the Mid-Atlantic region of the United States.

MATERIALS and METHODS

This retrospective review of culture results of cases of microbial keratitis at the Wilmer Eye Institute, Johns Hopkins University School of Medicine, from June 14, 2016, through June 14, 2020, was undertaken as a quality improvement project following expedited approval by the Johns Hopkins institutional review board (IRB00220717). The start date marked the adoption of electronic medical records (EMR) by the emergency department where after-hour ophthalmic emergencies are seen.

Identification of cases

A clinical informatics specialist queried the EMR to produce a list of ICD-10 codes consistent with microbial keratitis of bacterial, mycotic, and acanthamoebal etiologies. Diagnosis codes inconsistent with culture-confirmed microbial keratitis for the purposes of this study were removed. A list was finalized of patients diagnosed with any of the approved ICD-10 codes (H16.001, H16.002, H16.003, B60.19, B64, H16.009, H16.011, H16.012, H16. 013, H16.019, H16.02, H16.021, H16. 022, H16.023, H16.029, H16.03, H16.031, H16.032, H16.033, H16.039, H16.04, H16.041, H16.042, H16.043, H16.049, H16.06, H16.061, H16.062, H16.063, H16.069, H16.071, H16.072, H16.079, H16.1, H16.101, H16.102, H16.109, H16.2, H16.201, H16.203, H16.209, H16.31, H16.311, H16.312, H16.313, H16.319, H16.8, B89, H16.9) had corneal microbiologic culture orders placed by ophthalmologists, and had positive corneal culture results.

This review was performed in two phases—first, a manual chart review in the EMR to select eligible patients and enter clinical information that could not be automatically extracted, and second, automated extraction of microbiology data into the final dataset.

Manual entry dataset

From a list of all patients with positive bacterial or fungal corneal culture results during the observational period (N=638), 5 ophthalmologists (IK, NS, FW, MS, LH) performed a detailed chart review to ascertain which cases entailed culture-positive microbial keratitis in one or both eyes. We excluded eyes with positive culture results that were not thought to be caused by true cases of microbial keratitis, such as conjunctival cultures for purulent conjunctivitis, corneal cultures performed for persistent epithelial defects caused by neurotrophic keratopathy but without any stromal infiltrate, or endogenous endophthalmitis cases that did not initially present with any corneal involvement. Two ophthalmologists reviewed each chart; a third ophthalmologist adjudicated where there was disagreement. Cultures with any growth were deemed positive. Clinical judgment was exercised to annotate the “start” and “end” dates for each infection, i.e., when the patient presented for care and when antimicrobial therapy was stopped. These dates delineated the time interval of active infection. Multiple positive culture results from the same eye within a continuous time interval were counted as a single infection. For example, if a patient had a culture-positive case of microbial keratitis, achieved re-epithelialization, discontinued antibiotics (signifying resolution of this infection), and then returned at least 4 weeks later with findings consistent with a new infection and was re-cultured, then these 2 culture results were considered 2 different infections. During chart review we abstracted data on presence of risk factors for microbial keratitis, including contact lens wear and recent eye trauma. We also evaluated occurrence of key adverse clinical outcomes such as receipt of corneal gluing and keratoplasty (therapeutic and optical).

Automated extraction dataset

Corneal culture results between the start and end dates of infection as determined by chart review were extracted from the EMR. If an eye underwent multiple rounds of corneal culture for a single infection, the earliest positive culture result was recorded. Separate corneal cultures performed during separate time intervals were counted separately. The type of culture performed, specific organisms isolated from each culture category, and antimicrobial susceptibility if performed were merged with the manual entry dataset (StataCorp, College Station, TX).

Statistical analysis

Patient demographics such as sex, race, and age were examined. The unit of analysis was the case of microbial keratitis (i.e., a patient with simultaneous bilateral infections or multiple infections in one eye over non-overlapping time periods could contribute multiple data entries). We calculated the proportion of all culture-positive microbial keratitis cases that grew bacteria or fungi or acanthamoeba; the prevalence of specific organism isolates within each culture category; the number of specific organism isolates tested for susceptibility to specific antimicrobial agents; and the proportion of isolates that demonstrated resistance to specific agents. Continuous variables are presented as mean ± standard deviation or median (range); categorical variables are presented as frequencies. All analyses were performed using SAS and Stata version 16.

Microbiological evaluation

Bacterial aerobic cultures were performed using blood and chocolate agar plates that were incubated at 35°C in 10% CO₂ and examined daily for a week. Identification of isolates and antibacterial susceptibility testing (categorized as susceptible, intermediate, or resistant) were performed using an automated method (BD Phoenix PMIC-108, Becton Dickinson, Franklin Lakes, NJ) with testing guidelines and serum standards set forth by the Clinical and Laboratory Standards Institute (CLSI).¹⁸ If susceptibility was not determined, disk susceptibility or E-test was used. Antibiotic groups were chosen by the CLSI and the antibiotic stewardship committee. Intermediate susceptibility was categorized as “resistant” for the purposes of this project.

Fungal susceptibility to polyenes and triazoles was assessed using microbroth dilution according to guidelines and serum standards set by CLSI and the antibiotic stewardship committee.

RESULTS

A total of 474 cases of culture-positive microbial keratitis among 429 patients accessed from 638 patient records were observed from June 14, 2016, to June 14, 2020. The mean age of the 429 patients was 55.1+/−20.2 years. There was relative gender parity (53.2% male) and racial diversity (58.1% Caucasian, 27.7% Black, 5.3% Hispanic patients, 4.9% Asian, and 4.0% other).

All cases in this study received initial treatment with antimicrobial agents; a smaller proportion of eyes proceeded to therapeutic keratoplasty when medical treatment failed (N=54, 11%). Bacterial isolates grew from a majority of culture-positive cases of microbial keratitis (N=450, 94.9%), fungi grew from a smaller number of culture-positive cases (N=48, 10.1%), and Acanthamoeba species from fewer cases (N=15, 3.1%). Cumulative percentages exceed 100% because of polymicrobial infections (Table 1). Supplementary Table 1 describes polymicrobial keratitis which included bacteria-bacteria (N=85), bacteria-fungi (N=12), and fungi-fungi (N=2) with no polymicrobial infections involving acanthamoebal coinfection.

TABLE 1.

Corneal Culture Categories and Most Common Organisms Isolated

Culture category	N (%)a	Organisms growing	N (%)b
Bacterial culturec	450 (94.9%)	Gram-Positive Organisms	431 (69.5)
		Coagulase-negative Staphylococcus species	154 (24.8)
		Cutibacterium (Propionibacterium) species	100 (16.1)
		Corynebacterium species	39 (6.3)
		Viridans group Streptococci	35 (5.6)
		Methicillin-susceptible Staphylococcus aureus	29 (4.7)
		Methicillin-resistant Staphylococcus aureus	18 (2.9)
		Streptococcus pneumoniae	16 (2.6)
		Other Gram-Positive Organisms	40 (6.5)
		Gram-Negative Organisms	157 (25.3)
		Pseudomonas aeruginosa	76 (12.3)
		Serratia marcescens	11 (1.8)
		Enterobacter species	10 (1.6)
Fungal culture	48 (10.1%)	Yeast Organisms	30 (54.5)
		Candida species	25 (45.4)
		Filamentous Organisms	19 (34.5)
		Fusarium species	6 (10.9)
		Aspergillus species	3 (5.5)
		Alternaria species	3 (5.5)
		Paecilomyces lilacinus	3 (5.5)
		Other fungal organisms	15 (27.3)
Acanthamoebal culture	15 (3.1%)	Acanthamoeba species	15 (100)

^aUnit of analysis is case of keratitis. N (%) refers to the number of cases of microbial keratitis with positive cultures of each category (bacterial, fungal, acanthamoebal) of all 474 culture-positive keratitis cases. Denominator is 620 organisms. Cumulative percentages exceed 100% because of polymicrobial infections.

^bUnit of analysis is culture category. N (%) refers to the number of positive organism isolates found within each culture category, 620 bacterial isolates from 450 culture-positive keratitis cases, 55 fungal isolates from 48 culture-positive keratitis cases, and 15 acanthamoeba isolates from 15 culture-positive keratitis cases.

^cN=13 of 620 bacterial isolates (2.1% of all bacterial isolates) could not be speciated, meaning the microbiology laboratory could classify the isolates as either Gram positive or Gram negative.

Among the 620 organisms that grew from the 450 cases of culture-positive bacterial keratitis, the most common organisms isolated were coagulase-negative Staphylococcus species (154 (24.8%)), Cutibacterium (Propionibacterium) species (100 (16.2%)), and Pseudomonas aeruginosa (76 (12.3%)) (Table 1). The majority of the 620 organisms were Gram-positive (N=431, 69.5%), Gram-negative bacteria accounted for the minority of organisms (N=157, 28.4%). A smaller number of bacteria could not be speciated (N=13, 2.1%), meaning the microbiology laboratory could not classify them as Gram positive or Gram negative. Among 48 cases of culture-positive fungal keratitis, the most common organisms were Candida species, followed by non-speciated filamentous fungi.

Comparing Gram positive versus Gram negative bacterial keratitis cases, we noted that Gram positive bacterial keratitis cases had a higher proportion of patients with a history of contact lens wear (37.2% vs. 27.7%) and a higher proportion of patients undergoing optical keratoplasty (8.1% vs. 3.0%, p=0.008) and therapeutic keratoplasty (26.9% vs. 9.6%, p<0.001). Comparing fungal versus non-fungal keratitis cases, we noted that fungal keratitis cases were significantly more likely to undergo corneal gluing (9.3% vs. 2.8%, p=0.044) and therapeutic keratoplasty (37.2% vs. 14.6%, p<0.001). There were no differences in the proportion of patients aged <50 or ≥50 comparing bacterial versus fungal keratitis cases.

Although antimicrobial susceptibility testing was inconsistently performed during the observation period, susceptibility to commonly used topical antibiotics was performed in nearly all isolates of coagulase-negative Staphylococcus species (153 of 155 isolates tested), Pseudomonas aeruginosa (76 of 76 isolates tested), methicillin-sensitive Staphylococcus aureus (30 of 30 isolates tested), and Streptococcus pneumoniae (16 of 16 isolates tested). The remaining isolates underwent inconsistent antimicrobial susceptibility testing. Results of antimicrobial susceptibility testing are shown in Table 2. Fungal isolates and the proportion that demonstrated resistance are shown in Table 3.

TABLE 2.

Antibiotic Resistance Profiles for Most Common Bacterial Isolates

Organism	Number of isolates	Antibiotic	Number resistant/Number tested	Percentage resistance (%)
Coagulase-negative Staphylococcus species	154	Erythromycin	91/154	59.1
		Oxacillin	65/154	42.2
		Tetracycline	22/153	14.4
		Vancomycin	0/153	0
		Clindamycin	48/151	31.8
		Linezolid	0/26	0
		Moxifloxacin	1/4	25
		Levofloxacin	1/1	100
		Trimethoprim-Sulfamethoxazole	1/1	100
		Gentamicin	0/1	0
Cutibacterium (Propionibacterium) species	100	Ciprofloxacin	0/8	0
		Penicillin	1/8	12.5
		Rifampin	1/8	12.5
		Ceftriaxone	0/5	0
		Tetracycline	0/5	0
		Vancomycin	0/5	0
		Moxifloxacin	0/1	0
Pseudomonas aeruginosa	76	Aztreonam	9/75	12
		Ciprofloxacin	5/75	6.7
		Meropenem	2/75	2.7
		Piperacillin-Tazobactam	2/75	2.7
		Cefepime	0/75	0
		Ceftazidime	0/75	0
		Tobramycin	0/75	0
		Gentamicin	0/75	0
		Imipenem	1/2	50
		Levofloxacin	1/1	100
Corynebacterium species	39	Erythromycin	11/18	61.1
		Gentamicin	1/18	5.56
		Penicillin	4/17	23.5
		Ciprofloxacin	2/3	66.7
		Moxifloxacin	0/1	0
Viridans group Streptococci	35	Penicillin	2/5	40
		Levofloxacin	0/1	0
		Linezolid	0/1	0
		Daptomycin	0/1	0
Methicillin-susceptible Staphylococcus aureus	30	Erythromycin	13/29	44.8
		Clindamycin	12/29	41.3
		Tetracycline	2/29	6.9
		Trimethoprim-Sulfamethoxazole	1/28	3.6
		Daptomycin	0/2	0
		Levofloxacin	0/1	0
Methicillin-resistant Staphylococcus aureus	18	Clindamycin	12/18	66.7
		Trimethoprim-Sulfamethoxazole	1/18	5.6
		Linezolid	0/12	0
		Daptomycin	0/2	0
Streptococcus pneumoniae	16	Moxifloxacin	0/16	0
		Penicillin	0/16	0
		Amoxicillin	0/16	0
		Ceftriaxone	0/16	0
		Vancomycin	0/16	0
		Erythromycin	6/16	37.5
		Tetracycline	2/16	12.5
		Trimethoprim-Sulfamethoxazole	2/16	12.5
		Clindamycin	1/15	67

TABLE 3.

Antifungal Resistance Profiles for Fungal Organisms

Organism	Number of isolates	Antifungal agents	Number resistant/ Number tested	Percentage resistance (%)
Candida species	23	Voriconazole	1/6	16.7
		Amphotericin B	0/6	0
		Fluconazole	0/5	0
		Nystatin	0/1	0
		Micafungin	0/1	0
		Natamycin	0/1	0
Fusarium species	3	Amphotericin B	0/1	0
		Voriconazole	0/1	0
Trichosporon species	2	Amphotericin B	0/1	0
		Fluconazole	0/1	0
		Micafungin	0/1	0
		Posaconazole	0/1	0
		Voriconazole	0/1	0

No susceptibility testing was performed for the following microbes: Aspergillus species, Alternaria species, Pacilomyces lilacinus, Bipolaris species, Epicoccum species, Trichosporon species.

DISCUSSION

This study describes microbial distribution of culture-positive microbial keratitis at our institution, with a focus on bacterial and fungal infections. Our findings may inform choice of empiric treatment or prophylactic therapy for microbial keratitis among this large, diverse, and densely populated region comprising nearly 60 million people.

Microbial burden in the Mid-Atlantic region

Geographic differences in pathogen distribution can vary widely, and this knowledge can guide empiric therapy. Such differences have largely been attributed to variations in climate, with warm and humid climates associated with greater proportion of fungal keratitis and more temperate climates with a greater proportion of bacterial keratitis.^3,19 A study in south India found that fungal keratitis (63%) outnumbered bacterial keratitis (35.7%), which may arise from trauma related to land clearing and agricultural work.³ At institutions in the United States,^2,20,21 when less frequent infections like Acanthamoeba are excluded, bacteria comprise 85–90% of culture-positive cases while fungi account for about 10–15%. These results were noted at our institution as well as at locations outside the US.^22–24

Bacterial keratitis

At most US and non-US institutions including ours,^20–29 Gram-positive bacterial isolates are more common than Gram-negative isolates (Table 4). For our calculations of bacterial keratitis in our study as well as studies listed for comparison (Table 4), only cases of bacterial keratitis, instead of all microbial keratitis, were used in the denominator. Empiric therapy at these institutions should therefore include appropriate Gram-positive coverage. In contrast, Gram-negative bacteria are slightly more common than Gram-positive bacteria in more humid climates such as Miami.³⁰ Gram-negative bacteria might be more common in areas with a higher prevalence of extended wear contact lens use although there is no epidemiologically determined data of the prevalence of extended wear lenses for our city/state/region. Among cases of culture-positive bacterial keratitis at our institution, the most common isolates were coagulase-negative Staphylococcus species, followed by Cutibacterium (Propionibacterium) species. Coagulase-negative staphylococci predominated in Boston (45.5%)²⁷ and Los Angeles (35–45%).²¹ Although Cutibacterium species were found commonly found in Durham, NC (9%),³¹ we postulate that Cutibacterium (formerly Propionibacterium) species, less frequently mentioned in studies of microbial keratitis at other institutions, could have been contaminant from the skin and eyelid flora.

TABLE 4.

Bacterial and fungal distribution in microbial keratitis at other academic ophthalmology departments in the United States^*

	Wilmer Eye Institute, Baltimore, MD	Massachusetts Eye & Ear Infirmary Boston, MA27,37	Wills Eye Hospital, Philadelphia, PA2	Bascom Palmer Eye Institute, Miami, FL30,36	St. Louis University, St. Louis, MO20	University of California, San Francisco25	University of Pittsburgh 26	Doheny Eye Institute, Los Angeles (faculty practice)21	Los Angeles County Hospital-University of Southern California (public hospital)21
Most common bacterial isolate	Coagulase-negative Staphylococcus spp. (24.8%)	Coagulase-negative Staphylococcus spp. (45.5%)	Pseudomonas aeruginosa (24%)	Pseudomonas aeruginosa (26%)	Pseudomonas aeruginosa (26%)	Methicillin-susceptible Staphylococcus aureus (20%)	Pseudomonas aeruginosa (18.6%)	Coagulase-negative Staphylococcus spp. (41%)	Coagulase-negative Staphylococcus spp. (30%)
Most common fungal isolate in fungal keratitis	Candida spp. (45%)	Fusarium spp.10 (41%)	Candida spp. (% unavail)	Fusarium spp. (62%)	Fusarium spp. (35%)	---	---	Candida spp. (14%)	Candida spp. (29%)
Bacteria vs. fungi	90% vs. 10%	---	85% vs. 15%	---	84% vs. 16%	---	---	91% vs. 9%	89% vs. 11%
Gram-positive vs. Gram-negative in bacterial keratitis	69.5% vs 28.4%	87% vs 13%	---	48% vs 50%	59% vs. 41%	65% vs 35%	56% vs 44%	70% vs 21%	68% vs 21%
Pseudomonas aeruginosa prevalence in bacterial keratitis	12%	3%	24%	26%	21%	11%	17%	13%	10%

^*For proportions of bacterial keratitis in our study as well as other studies included for comparison, the number of cases of bacterial keratitis, rather than all cases of microbial keratitis, was used as the denominator.

Pseudomonas aeruginosa was the most common Gram-negative bacteria at our institution (12%) as was the case elsewhere, but the proportion of bacterial keratitis cases caused by Pseudomonas in our population was half that in Miami (26%),³⁰ St. Louis (21%),²⁰ and Philadelphia (24%).² Possible explanations for the large difference in Pseudomonas prevalence between Baltimore and Philadelphia, which are separated by only 90 miles, might be differences in patient demographics or in prevalence of contact lens wear. The smaller proportion of Pseudomonas aeruginosa at our institution compared with others is unlikely to be a result of any institutional recommendations against extended wear contact lenses,³² especially as our catchment as a tertiary care center is large and non-cornea specialty trained doctors may be treating patients with contact-lens related keratitis successfully with 4^th generation fluoroquinolones such that such patients do not require care at our tertiary center. In addition, daily disposable lens wear has increased significantly worldwide and is associated with a significant decrease in risk of infection or inflammation compared with two-week or monthly replacement lenses.^33,34 Without more detailed data, it is difficult to draw more detailed conclusions regarding the proportion of Pseudomonas keratitis and contact lens wear from our data set. At the University of Southern California,²¹ the resident clinic at Los Angeles County Hospital and the faculty practice are separated by less than 0.5 mile, yet the pathogen distributions varied widely between these socioeconomically divergent populations, perhaps reflecting differences in exposures or access to care (Table 4). A large department in Boston had a much higher prevalence of coagulase-negative Staphylococcus (45.5%) and a much lower prevalence of Pseudomonas aeruginosa (2.7%) than did other US institutions.²⁷ This finding is true in the United Kingdom as well.^{24, 35}

Fungal keratitis

A significantly higher proportion of fungal cases required gluing or therapeutic keratoplasty than did non-fungal cases possibly from delayed presentation or diagnosis and suboptimal treatment. Candida spp. were the most common isolates of all culture-positive fungal keratitis cases seen in Baltimore, Philadelphia,² and Los Angeles.¹⁹ Fusarium spp. were the predominant fungi in Miami (62%),³⁷ Taiwan (52%),²² South India (42%),³ St. Louis (41%),²⁰ and Boston (35%).³⁷ These differences in geographical distribution of filamentous versus yeast organisms have important therapeutic implications. The Mycotic Ulcer Treatment Trial I (MUTT I) conducted in south India found that topical natamycin was more effective than topical voriconazole, a finding largely derived from improved outcomes in keratitis caused by Fusarium, which comprised a large proportion of cases in South India.³⁸ MUTT II found no benefit to adding oral voriconazole to topical antifungal agents in 240 patients with filamentous fungal ulcers.³⁹ However, the therapeutic efficacy of topical natamycin versus voriconazole or amphotericin B in non-filamentous fungal keratitis remains unclear.^{2, 21} At institutions in temperate climates like ours where Candida species comprise nearly half of all cases of fungal keratitis, topical voriconazole remains widely used for empiric treatment of suspected fungal keratitis and as guided therapy for non-filamentous keratitis.⁴⁰ The high proportion of Candida keratitis at our institution may support the use of adjuvant oral and/or intrastromal voriconazole⁴¹ (which are off-label use) for severe or refractory cases, but the latter has not been universally adopted. Adequately powered randomized studies evaluating the efficacy of these interventions for severe Candida keratitis have not been conducted due to high variation in clinical severity, difficulty with precise measurement of outcomes, and low patient recruitment. More studies on optimal treatment regimens for non-filamentous fungal keratitis are needed.

Antimicrobial resistance

Antibiotic susceptibility testing for commonly used topical antibiotics was performed in nearly all isolates of coagulase-negative Staphylococcus species, Pseudomonas aeruginosa, methicillin-sensitive Staphylococcus aureus, and Streptococcus pneumoniae. Antibiotic susceptibility was not consistently tested for remaining isolates, which limits scope of inference regarding antimicrobial resistance within these other groups. In particular, fluoroquinolone resistance was incompletely tested across most isolates even though the ARMOR study showed increased fluoroquinolone resistance across time.⁴² Despite these data limitations, two findings deserve mention.

Erythromycin is commonly used to prevent secondary bacterial infection in corneal abrasions, large dendritic herpetic lesions, viral stromal keratitis with overlying epithelial defects, persistent epithelial defects in neurotrophic keratopathy, and severe dry eye or exposure keratopathy. It also is prescribed in other settings such as primary care clinics and intensive care units. Of the 217 bacterial isolates tested for erythromycin susceptibility, 121 (55.7%) were resistant to erythromycin. Resistance to erythromycin was noted across several bacterial species including coagulase-negative Staphylococci (59.1%), Corynebacterium species (61.1%), methicillin-susceptible Staphylococcus aureus (44.8%), Streptococcus pneumoniae (37.5%), and Enterobacter species (100%). Even though macrolides such as erythromycin are not used for bacterial keratitis treatment, our findings indicate that the widespread use of erythromycin for prophylaxis should be reconsidered.

A significant proportion (65 of 154 tested, 42.2%) of the coagulase negative staphylococci (CoNS) were resistant to oxacillin, which is often used as an indicator for methicillin resistance. Previous studies have postulated and demonstrated that resistance genes, namely the mecA gene, can be shared by CoNs with methicillin-sensitive Staphylococcus aureus (MSSA), converting MSSA into methicillin resistant Staphylococcus aureus (MRSA).⁴³ Given this finding, empiric antimicrobial coverage of MRSA should be considered.

Strengths and limitations

Strengths of this study include its large sample size collected over five years at a single large tertiary referral centre serving the Mid-Atlantic region of the US and examination of the complete spectrum of microorganisms isolated from corneal culture including bacteria, fungi, and parasites. This study has limitations because of its retrospective design and inherent limitations in culture-based approaches to diagnosing microbial keratitis. Contaminants, namely Cutibacterium as well as other eyelid flora, may have altered assessments of the most common bacterial isolates in this project. Since the microbiologic yield of corneal cultures has consistently been shown to be only 40–60% across multiple studies,^3,22,30,,44 our focus on only culture-positive microbial keratitis may not reflect the etiologies of culture-negative microbial keratitis at our institution. Future research should examine differences among culture-positive, culture-negative, and non-cultured microbial keratitis. We expect a higher proportion of ulcers without prior antibiotic treatment to be culture-positive for bacteria than ulcers that are pre-treated before presentation to our institution. Purulent cases also might be more likely to be culture-positive than culture-negative. Beyond such differences, we cannot hypothesize what differences we would expect to see; the differences in clinical course among culture-positive, culture-negative, and non-cultured microbial keratitis are a topic for future study.

We focused on bacterial and fungal pathogens in this report; analysis of acanthamoebal cases will be done separately. At our institutions (and likely most American institutions) we do not examine for protozoa as part of routine practice because our laboratories are not adequately equipped. Also, we did not examine for patterns such as microbial etiologies or susceptibilities in polymicrobial cases, which would be of interest studies from various geographical locations reporting their experience with microbial keratitis. The proportions of polymicrobial infections had a large range (which we calculated if not reported in the publication): 4.2% for mixed bacterial/fungal in Philadelphia; 8.2% in our Baltimore series; 9.7% at Bascom Palmer (130/1338) for bacterial cases; 10% at St. Louis University for bacterial, fungal, and acanthamoeba; 19.2% in Boston for bacterial cases; 22% at Doheny and 29% at University of Southern California in Los Angeles for bacterial, fungal, and Acanthamoeba cases. The publications from San Francisco and Pittsburgh did not allow such a calculation as they analyzed organisms from all cultures without mention of number of cases of keratitis. Additional future investigation into polymicrobial cases is warranted at all sites.

A key limitation of this study’s retrospective design is the inconsistent data collected regarding antimicrobial resistance testing. Although 4 bacterial isolates were almost universally tested for resistance to commonly prescribed antibiotic classes, several other isolates underwent rare or inconsistent antimicrobial resistance testing. Prevalence of antimicrobial resistance may also be over-reported in this report, as pathogens isolated from pre-treated cases of microbial keratitis may be more likely to demonstrate resistance possibly from reduced culture yield or altered microbial load prior to obtaining culture. In addition, the proportion of resistant isolates may be inflated because isolates with intermediate susceptibility to an antimicrobial agent were classified as “resistant” prior to statistical analysis. Because clinical response does not always correlate with in vitro susceptibility testing, clinical response to certain widely used antibiotics such as erythromycin and 4^th generation fluoroquinolones should be examined in future studies. Finally, assessment of antibiotic resistance using minimum inhibitory concentrations (MIC) used for systemic antibiotics may over-estimate resistance as antibiotic concentrations achieved in corneal tissue using eye drops are typically far greater than systemic concentrations achieved via oral or parenteral routes.⁴⁵ The correlation between clinical ophthalmic response and MIC for coagulase-negative staphylococci and streptococci has been lower than for Pseudomonas and Staphylococcus aureus.⁴⁶

As this study’s primary purpose was to examine the distribution of microbial etiologies of culture positive microbial keratitis, this analysis did not delve into details of cases where medical treatment failed and surgery was performed, which will be investigated in the future. Contact lens wear was not systematically recorded by physicians and not in this retrospective chart review, which limits assessment of which organisms were most associated with contact lens wear.

Taken together with findings from other regions in the US, our results demonstrate regional differences in bacterial and fungal pathogens causing microbial keratitis. Gram-positive bacterial keratitis predominated in the Mid-Atlantic region, while prevalence of Pseudomonas isolates was half that of other US institutions. Candida, and not filamentous fungi, was the most common cause of fungal keratitis in this region. Because of the high prevalence of in vitro resistance to erythromycin, its prophylactic use should be re-examined in our patient population. These results may be used to guide empiric treatment in the Mid-Atlantic region. As in other reports, temporal changes in regional microbial distribution and antimicrobial susceptibility may require repeat analysis, clinical correlation, and modification of clinical practices.