Article Text

Download PDFPDF

Predisposing factors, clinical and microbiological insights of bacterial keratitis: analysis of 354 cases from a leading French academic centre
  1. Clara Bertret1,
  2. Juliette Knoeri1,
  3. Loic Leveziel1,
  4. Tristan Bourcier2,
  5. Françoise Brignole-Baudouin1,
  6. Lilia Merabet1,
  7. Nacim Bouheraoua1,
  8. Vincent Michel Borderie1
  1. 1Hôpital National des 15-20, Paris, France
  2. 2Ophthalmology, UDS, Strasbourg, France
  1. Correspondence to Professor Vincent Michel Borderie, Hôpital National des 15-20, Paris 75571, France; vincent.borderie{at}


Aims To report an epidemiological update of bacterial keratitis (BK) in a tertiary ophthalmology centre over 20 months compared with a previous study on the same timeframe from 1998 to 1999.

Methods 354 patients with BK documented by microbiological corneal scraping or resolutive under antibiotics treatment from January 2020 to September 2021 were analysed retrospectively.

Results One or several risk factors were found in 95.2% of patients: contact lens wear (45.2%), ocular surface disease (25.0%), systemic disease (21.8%), ocular trauma (11.9%) and ocular surgery (8.8%). The positivity rate of corneal scrapings was 82.5%, with 18.2% polybacterial. One hundred seventy-five (59.9%) bacteria were Gram-negative, and 117 (40.1%) were Gram-positive. The most common bacteria were Pseudomonas aeruginosa (32.5%), Moraxella spp (18.1%) and Staphylococcus aureus (8.2%). Final visual acuity (logarithm of the minimum angle of resolution) was associated with age (r=+0.48; p=0.0001), infiltrate size (r=+0.32; p<0.0001), ocular surface disease (r=+0.13; p=0.03), ocular trauma (r=-0.14; p=0.02) and contact lens wear (r=−0.26; p<0.0001). Gram-negative bacteria were responsible for deeper (r=+0.18; p=0.004) and more extensive infiltrates (r=+0.18; p=0.004) in younger patients (r=−0.19; p=0.003). Compared with the previous period, the positivity rate of corneal scrapings and the proportion of Gram-negative bacteria, especially Moraxella spp, increased. All P. aeruginosa and Moraxella spp were sensitive to quinolones, and all S. aureus were sensitive to both quinolones and methicillin.

Conclusion Contact lens wear remained the leading risk factor. The bacteria distribution was reversed, with a predominance of Gram-negative bacteria and increased Moraxella spp.

  • epidemiology
  • contact lens
  • cornea
  • infection
  • treatment medical

Data availability statement

Data are available on reasonable request.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

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.


  • Wearing contact lenses is the leading risk factor for bacterial keratitis.


  • We report an inversion of the Gram-positive-Gram-negative balance with increased Moraxella spp over 20 years and an increase in the positivity rate of corneal scrapings.


  • The rate of antibiotic resistance is very low with protocols using three broad-spectrum antibiotics to treat severe bacterial keratitis.

  • Implementation of a strict protocol for corneal scraping results in a high positivity rate of microbiological tests.


Infectious keratitis represents a significant cause of corneal opacification, potentially leading to vision loss.1 Visual prognosis depends on a quick diagnosis and appropriate antimicrobial therapy. The most common germs are bacteria in many countries such as Brazil (78.9%),2 Canada (86%),3 the UK (92.8%),4 Australia (93.1%),5 except China (52.7%)6 and India (35.7%).7 The main symptoms are acute ocular pain, photophobia, redness and decreased vision. Bacterial keratitis (BK) rarely occurs without predisposing factors such as trauma, contact lens (CL) wear, ocular surface disease (OSD), systemic immunosuppression or ocular surgery.8 The most commonly isolated bacteria are Gram-positive, mainly coagulase-negative Staphylococci belonging to the ocular surface commensal flora.1 However, some studies have identified Pseudomonas aeruginosa as the leading bacterium due to climatic and geographical variations.4 5 9 10 Intensive broad-spectrum topical antibiotics are first used to treat severe BK, followed by an adaptation to the identified pathogen or the clinical response.11 Management of BK is currently challenged by the emergence of antimicrobial resistance (AMR), particularly in the USA and Asia.12–14 Epidemiological studies in Europe report the rate of AMR in BK, but no data are available in France.4 10

An epidemiological BK study was conducted in our tertiary centre between 1998 and 1999.15 The present study aimed to compare two large series of patients suffering from BK with a 20-year gap to identify potential changes in BK risk factors, clinical presentations, prognosis and microbiological epidemiology. The secondary objective was to evaluate the sensitivity of bacteria to antibiotics.

Material and methods

Study design

The study period ranged from January 2020 to September 2021. All patients hospitalised in the French National Vision Hospital (Hôpital National des 15-20, Paris, France) infectious ocular disease unit for BK, defined as a corneal stromal infiltrate with an epithelial defect documented by microbiological corneal scraping or resolutive under antibiotics treatment, were included. Only the most affected eye was included for bilateral BK. A total of 354 BK in 354 eyes were included. Nine BK were bilateral. Hospitalisation for infectious keratitis was decided if the patient presented severity criteria on examination in the emergency or consultation units. The severity criteria were the same as those already described in 1998–199911: location of the infiltrate within the 3 mm of the visual axis, infiltrate size >2 mm, anterior chamber inflammation, worsening under 24-hour antibiotic treatment, corneal transplantation, single eye patient, immunosuppressed patient, bilateral keratitis. Diagnosis and management of BK were performed according to written standard operative procedures.

Microbiological analysis

An ophthalmologist performed corneal scrapings at the earliest stage using a sampling kit including different tools, that is, a sterile scalpel blade 15T and an e-Swab (Copan Diagnostics, Murrieta, California, USA). Samples from e-Swab were inoculated under sterile conditions on various solid and liquid media that facilitate the growth of bacteria, including chocolate agar (PVX), 5% sheep blood agar, Schaedler broth and a Sabouraud medium. Specimens from the scalpel blade were smeared on two slides and stained with May Grünwald Giemsa (MGG). The bacterial cultures were incubated in 5% CO2 at 35±2°C and examined daily for growth for at least 7 and 15 days. The culture was considered positive when a bacterium was identified on at least one of three media. The isolated bacteria were identified by biochemical galleries, and the susceptibility to antibiotics was evaluated by the diffusion disc method and interpreted according to the guidelines established by the French Society of Microbiology and the European Committee on Antimicrobial Susceptibility Testing. Bacterial multiplex PCR was performed in cases of suspected BK with negative direct examination and culture. Fungal, Acanthamoeba and viral co-infections were systematically and simultaneously investigated.

Corneal imaging

In some cases with atypical clinical features or history, in vivo confocal microscopy was performed to rule out Acanthamoeba and filamentous fungal keratitis.

Data collection

The following data were collected: age, sex, duration of the symptoms before consultation, antibiotics or other treatments before consultation, initial and final visual acuity (logarithm of the minimum angle of resolution (logMAR)), causative bacteria and its susceptibility to antibiotics, ocular complications and surgical management. Infiltrate size (<5 mm2, 5–15 mm2, >15 mm2), infiltrate location (peripheral, central), infiltrate depth (<1/3, 1/–-2/3, >2/3 of the total corneal thickness), anterior chamber inflammation (Tyndall scored from 1+ to 4+, hypopyon) were assessed by slit-lamp examination. The following risk factors were evaluated: corneal trauma, CL wear (soft, rigid, scleral), OSD (bullous keratopathy, recurrent keratitis, peripheral ulcerative keratitis, endothelial decompensation, exposure keratitis, palpebral malposition, neurotrophic keratitis, dry eye syndrome), systemic diseases (diabetes, chronic alcoholism, systemic immunosuppression, psychiatric illness) and prior ocular surgery.

BK medical management

All hospitalised BK-suspect patients received three fortified topical antibiotics (ie, piperacillin (20 mg/mL), gentamicin (15 mg/mL) and vancomycin (50 mg/mL)), administrated hourly for 48 hours, and followed by one drop every 2 hours until improvement. This association of three fortified antibiotics is commonly used in France or Europe.16 Piperacillin or ticarcillin is used for its action against most Gram-positive and Gram-negative bacteria. Vancomycin is used as a second antibiotic to act against methicillin-resistant Staphylococcus and Streptococcus pneumoniae with reduced sensitivity to penicillin. Gentamicin is used as a third antibiotic for its rapid bactericidal action in synergy with piperacillin and vancomycin. Topical dual antibiotic therapy was adapted according to the clinical response, germ and antibiogram as soon as the microbiological results were available. The therapeutic protocol was not modified except for replacing ticarcillin with piperacillin compared with the period of 1998–1999.15

Statistical analysis

Statistical analysis was performed using MedCalc software Ltd V.19 (8400 Ostend, Belgium). Descriptive data were reported in mean and SD. LogMAR visual acuity was used to calculate mean and SD. Analysis of variance, χ2 test, Wilcoxon rank-sum test and Mann-Whitney U test assessed relationships between initial BK characteristics, care protocol and outcomes. The relationships between final visual acuity, risk factors and BK initial clinical characteristics were evaluated with multiple linear regression.


Table 1 summarises population characteristics at baseline and clinical outcomes. CL wear was the most common risk factor found in 160 BK (45.2%) (table 2). A previous BK history was observed in 28 BK (7.9%) and associated with the presence of systemic disease (p=0.002). Females were predominant in the CL wear group (62.5%; p<0.0001) as opposed to the trauma group with 76.2% males (p=0.0005). The rate of microbiological identification was 82.5% (292/354). Two hundred nineteen germs (61.9%) were identified by direct examination, 258 (72.9%) by culture and 9 (2.5%) by PCR multiplex. 9.6% (34/354) of the corneal samples were identified on direct examination with no further growth in culture, 15.8% (56/354) were grown on only one of the three media (33 coagulase-negative staphylococci (CoNS), 20 Propionibacterium acnes, 3 Corynebacterium spp) and 57.0% (202/354) on at least two of the culture media. Prior antibiotics did not modify the microbiological positivity rate (84.0% vs 77.1%; p=0.12). Gram-negative bacteria (n=175; 59.9%) were predominant compared with Gram-positive bacteria (n=117; 40.1%) (table 3). The most frequent bacteria were P. aeruginosa (95; 32.5%). No Staphylococcus aureus resistant to methicillin was noted, even in recurrent BK (table 4). Gram-negative bacteria featured a higher Tyndall effect than other germs (p<0.0001). Moraxella BK featured larger and deeper infiltrates than other bacteria (p<0.0001) and was more frequently associated with chronic alcoholism (p=0.002). Among the most frequent germs, mean visual acuity (logMAR) was in ascending order: S. pneumoniae (1.59±1.06), Moraxella spp (1.30±1.02), P. aeruginosa (0.70±0.81), S. aureus (0.60±0.62), CoNS (0.48±0.74), P. acnes (0.27±0.59). Corneal scrapings were polybacterial in 53/292 (18.2%) cases. The predominant bacteria in polybacterial culture were Moraxella spp (19/53; 35.8%) only associated with Gram-positive bacteria (7 Staphylococcus coagulase-negative, 5 S. aureus, 5 P. acnes, 2 Corynebacterium spp). These polyBK were more frequently associated with systemic disease (p=0.048), CL wear (p=0.038), older patients (59.2±22.6 years vs 50.0±21.0; p=0.004), and worse final visual acuity (1.18±1.01 vs 0.75±0.92; p=0.006) compared with monobacterial keratitis.

Table 1

Population characteristics at baseline and clinical outcomes (n=354) compared with the previous study 1998–1999 (n=291)

Table 2

Population characteristics by main risk factors

Table 3

Bacteria identified in patients with bacterial keratitis

Table 4

Antibiotic susceptibility rates of the main bacteria isolated

Amniotic membrane transplantation (AMT) was performed for delayed epithelial healing in 78 BK (22.0%). The most common BK requiring AMT was S. pneumoniae keratitis (40.0%), followed by Moraxella (34.0%), P. aeruginosa (27.0%) and S. aureus (16.7%) keratitis. BK requiring AMT featured poorer initial and final visual acuity (p<0.001). Corneal perforation occurred in 18 BK (5.1%). The average age of patients with perforated BK was higher (67.8 vs 50.5 years; p=0.0001). Thirteen therapeutic penetrating keratoplasties (3.7%) and three eviscerations (0.8%) were performed (table 5).

Table 5

Demographic, risk factors, clinical and microbiological characteristics of 16 bacterial keratitis requiring emergency surgery

In multivariate analysis, poorer final logMAR visual acuity was associated with age (r=0.48; p=0.0001), infiltrate size (r=0.32; p<0.0001), OSD (r=0.13; p=0.03). CL wear (r=−0.26; p<0.0001), ocular trauma (r=-0.14; p=0.02) and absence of risk factor (r=−0.14; p=0.02) were associated with better final visual acuity. Gram-negative bacteria were responsible for deeper (r=0.18; p=0.004) and more extensive infiltrates (r=0.18; p=0.004) in younger patients (r=−0.19; p=0.003) compared with Gram-positive bacteria.

A comparison with the previous study is presented in table 1. Gram-negative bacteria were predominant (59.9% vs 17.0%) as opposed to Gram-positive bacteria (40.1% vs 83.0%), with an increase in Moraxella spp (18.1% vs 0.3%). P. aeruginosa was more frequent among CL wearers (50.0% vs 12.0%). The main risk factor remained CL wear, followed by OSD. All patients received fortified antibiotic eye drops vs 213 (71%) in the previous study. Compared with the 1998–1999 BK series, the 2020–2021 BK series featured deeper infiltrates with more anterior segment inflammation, more frequent corneal vascularisation, poorer initial visual acuity, a higher microbiological identification rate, shorter hospitalisation time and required more therapeutic keratoplasties.


Changes in BK risk factors, microbiological agents and clinical presentation could be shown over 20 years. An increase in the prevalence of Moraxella spp was noted in our series as in the UK.17 18 Moraxella spp are associated with systemic diseases, particularly diabetes, chronic alcoholism and OSDs.17 19 20 Chronic alcoholism was also a risk factor for Moraxella keratitis in our study. As in the Zafar study, 42.3% of the Moraxella keratitis patients had a hypopyon.19 The infiltrates were deeper and needed more AMT for delayed epithelial healing. AMT improves the healing process and reduces corneal haze and neovascularisation. It is associated with better visual recovery and less corneal vascularisation at 6 months than topical treatment alone.21 A high number of AMTs (22.0%) were performed in our population to accelerate healing and prevent corneal perforation.

CL wear remained the leading risk factor for BK, with 160 cases (45.2%).8 22 The incidence of BK secondary to CL wear ranges from 0.33% in the West Bengal area23 to 50.3% in France.15 In the CL group, patients were younger, consulted quickly, had a better visual prognosis and had more Gram-negative BK, especially P. aeruginosa, compared with the other groups. The study’s high rate of CL wearers may explain the high rate of P. aeruginosa. CL wear induces hypoxia and hypercapnia of the cornea, making these eyes more susceptible to infection. Mechanisms of biofilm formation, epithelial disruption and alteration of the antimicrobial quality of the tear film also favour the development of BK.24

Numerous studies report polymicrobial keratitis with widely varying rates, from 2.4%25 to 23.9%.5 In the study by Lim et al,26 the rate of identified polybacterial keratitis was 3.0% with two Gram-negative bacteria as the most frequent combination, and in the study by Ting et al, 10.7% with Streptococcus spp and coagulase-negative Staphylococci combination. As in Jones’s study, polybacterial keratitis, particularly Moraxella spp, was associated with organisms commonly isolated from normal conjunctival flora.27 The real pathogenic impact of these associated commensal germs can be questioned.

The 82% microbiological positivity rate was significantly higher compared with the previous period (68%)15 and recent series (49.3% in China9 and 64.3% in Saudi Arabia28). The inoculation protocol was modified between 1998 and 1999 and 2020 and 2021. Previously, all culture media were inoculated by the ophthalmologist with a dry swab in the examination room, whereas in the present study period, they were inoculated under sterile conditions by the microbiologist using e-Swab in the laboratory. Thanks to its enhanced capillary action and hydraulic liquid absorption, e-Swab enables several culture media to be inoculated with a single sample.29 There is no significant difference in microbiological results between direct and indirect inoculation by e-Swab of culture media, except for direct examination with Gram stain. To correct this, in our protocol, specimens from the scalpel blade were smeared directly on two slides after the corneal scraping with e-Swab.30 A second solid culture medium of blood agar type was added to interpret contamination. This change in the inoculation procedure and replacing the dry swab with an e-Swab allowed greater control of culture media storage conditions and limited external contamination of culture media.

Older age and larger infiltrate size were known risk factors for poor visual prognosis, as in the Portsmouth corneal ulcer study,31 Nottingham’s study32 and the study by Miedziak et al.33 OSD was also associated in our series with a poor visual prognosis, particularly the delay between the first symptoms and the emergency department assessment. Conversely, CL wear, ocular trauma and the absence of risk factors were protective prognostic factors.

AMR to common broad-spectrum antibiotics was low compared with studies in the USA13 or Asia.6 14 This finding may not apply to BK after keratoplasty. We recently reported that bacterial resistance is common in those patients, and it may be associated with prolonged use of preventive topical antibiotics that result in the selection of resistant bacteria.30

The most commonly identified resistant species in the literature were P. aeruginosa and Staphylococcus spp for moxifloxacin and S. aureus for methicillin (MRSA).14 34 35 We found only three resistant P. aeruginosa (one to piperacillin, one to ceftazidime and one to gentamicin), and no MRSA over the period 2020–2021, similar to the very low MRSA rate of 0.07%–1% in the UK.4 17 All Moraxella spp were sensitive, as in the study by Termote et al in Vancouver.36 The systematic use of three fortified topical antibiotics (ie, piperacillin (20 mg/mL), gentamicin (15 mg/mL) and vancomycin (50 mg/mL)) for severe keratitis and the frequent use of topical quinolones in combination with another antibiotic could explain this low rate. Patients with recurrent BK (7.9%) had no further AMR in our study.

There is no international consensus on the treatment of BK. Despite the risk of increasing bacterial resistance, fourth-generation quinolones such as moxifloxacin are now widely used as monotherapy in South America, Africa and Oceania.16 A randomised comparative study between fourth-generation fluoroquinolones and fortified antibiotics in the treatment of BK showed no differences.37 However, the resistance rate to quinolones is high in India38 and increases in the USA.35 In our study, all severe BK were treated initially with three topical fortified antibiotics, followed by a dual antibiotic therapy adapted to the antibiogram according to the clinical course.

Our present study features some limitations. This monocentric retrospective series aimed to describe BK in patients requiring hospitalisation. The inclusion of BK referred to a tertiary care centre led to a selection bias towards the most severe cases. In contrast to the 1998–1999 study15 and the literature,2 9 17 18 31 Gram-negative bacteria were the most frequent contaminants. The included BK cases were more severe than 20 years ago. The predominance of Gram-negative bacteria may explain this severity.24 Besides, Ung et al reminds us that without standardised criteria for laboratory identification, the worldwide microbiological results should be interpreted with caution because most commensal organisms of the eyelids and ocular surface are Gram-positive and more likely to contaminate media.1 24 36


P. aeruginosa, favoured by CL wear, was the principal causative agent of severe BK. An increase in Moraxella spp was noted over 20 years. Teaching patients about CL hygiene should remain a priority for practitioners.

Data availability statement

Data are available on reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

The Ethics Committee of the French Society of Ophthalmology approved this retrospective study (IRB 00008855), and informed consent was obtained from all subjects. The described research adhered to the tenets of the Declaration of Helsinki.



  • Contributors Conceptualisation: JK, LL and VMB. Data curation: CB, JK, LL and LM. Formal analysis: JK, CB, LL and VMB. Funding acquisition: VMB. Investigation: JK, CB and LL. Methodology: JK, LM, VMB, NB and TB. Project administration: JK and VMB. Guarantor: VMB Supervision: VMB, JK and LL. Validation: JK, LL, VMB, LM, NB, FB-B and TB. Writing—original draft: CB, JK, LL and LM. Writing—review and editing: CB, JK, LL, LM, VMB, NB, FB-B and TB.

  • Funding This work was supported by Sorbonne Université, Paris, France.

  • Competing interests None declared.

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