Predisposing Factors, Microbial Characteristics, and Clinical Outcome of Microbial Keratitis in a Tertiary Centre in Hong Kong: A 10-Year Experience

Purpose. To study the risk factors, microbial profile, antibiotic susceptibility pattern, and outcome for microbial keratitis over the past 10 years in a tertiary center in Hong Kong. Methods. All cases with corneal scraping performed in Queen Mary Hospital, Hong Kong from January 2004 to December 2013 were included. Clinical outcome was defined as poor if the final visual acuity (VA) was abnormal or worse than presenting VA, a major complication occurred, or therapeutic keratoplasty was required. Results. 347 scrapes were performed in the 10-year period growing 130 microorganisms (32.3% culture positive rate). Contact lens use was the commonest risk factor. The commonest isolates were coagulase-negative Staphylococcus and Pseudomonas aeruginosa. Fluoroquinolone susceptibility was tested in 47 Gram-negative bacteria with 93.6% susceptibility (100% for Pseudomonas). 90.7% of cases had good visual outcome. Multivariate logistic regression showed age (p = 0.03), trauma (p = 0.006), and ulcer size >3 mm (p = 0.039) to be independently associated with poor outcome. Conclusion. There was no shifting trend in the isolate distribution or emergence of resistant strains in our study. Contact lens wear was the commonest risk factor, with Pseudomonas being the most frequent isolate in this group. It remained 100% susceptible to fluoroquinolones and 97% cases had good visual outcome.


Background
Microbial keratitis is one of the commonest causes of corneal blindness worldwide. The mainstay of diagnosis is corneal scraping. However, the clinician has to decide on the antibiotic regime before the culture and antibiotic susceptibility results are available. This decision is often based on patient's demographics, risk factor profile, and the local microbial distribution pattern. As geographical and climatic factors result in regional differences in the pattern of microbial isolates, local epidemiologic studies are necessary [1]. The antibiotic susceptibility profile is also important in view of reports on the emergence of methicillin or fluoroquinolone resistance in North America and Asia in the past decade [2][3][4][5]. The in vivo resistance correlates with a worse clinical outcome as well [6].
In Hong Kong, the last large-scale epidemiological study on microbial keratitis was over 15 years ago [7]. The aim of this study was to investigate for any changes in the microbiological and antibiotic susceptibility pattern, risk factors profile, and clinical outcome in a Hong Kong tertiary referral center over the past 10 years. susceptibility testing. Any positive growths from the specimens would be reported by the microbiologist. The antibiotic susceptibility was determined according to the CLSI (Clinical and Laboratory Stand Institute) criteria.

Data Collection.
All culture and antibiotic susceptibility results obtained from the corneal scrapings were retrieved from the Department of Microbiology database. Cases notes were reviewed. Data collection included patient demographics, risk factors profile, and the clinical progress. Cases with incomplete clinical data were excluded from the analysis. These included cases that defaulted follow-up after initial presentation or with incomplete documentation of risk factors or progress in the clinical notes. Cases with absence of corneal infiltrates were also excluded from analysis. Risk factors considered, including contact lens use, trauma, history of corneal and ocular surface disorders, and presence of systemic illness (which either resulted in immunocompromised state or poor self-care), were shown in Table 1. For clinical features, size and location of corneal ulcer and presence of hypopyon in the anterior chamber were recorded. Size of lesion was categorized as small if less than 3 mm or large if more than 3 mm. Presence of good initial response to antibiotic was defined as improvement in pain or reduction in size of ulcer or infiltrate in the first 48-72 hours. Clinical outcome was defined as good if the visual acuity at final visit was better than 0.7 (Snellen), or, in cases where there was poor premorbid visual acuity from preexisting ocular pathology, there should be no drop in final visual acuity as compared with presenting visual acuity. Poor outcome was defined if there was a drop in visual acuity or if a major complication occurred, which included corneal perforation, endophthalmitis, or had therapeutic keratoplasty performed.

Analysis.
Statistical analysis was performed with SPSS version 21 (IBM). The number of each cultured isolate and antibiotic resistance pattern were tested with Spearman correlation for any changing trends over the 10-year period. The microbial pattern in different risk factor groups was compared with chi-square test. Univariate and multivariate analyses by logistic regression were used to examine for any risk factors affecting the outcome. The study received approval by the Institutional Research Ethics Board of the University of Hong Kong (reference number: UW14-043) and complied with the Declaration of Helsinki.

Results
347 corneal scrapings were performed in the 10-year period. 43.4% were males and 53.5% were females. Mean age was 46.2 ± 21.1 (range 10-96; median 46, interquartile range 27-63). The peak incidence occurred in the 21-30 subgroup (68 cases (19.6%), mean 25.1 ± 2.9) and the 51-60 subgroup (56 cases (16.1%), mean 54.6 ± 2.7). After exclusion, 260 cases were available for analysis of risk factors and 241 cases for clinical outcome (19 cases defaulted follow-up before healing of ulcer). 3.1. Culture Results. The distribution of microbial isolates was shown in Figure 1 and Table 2. 32.3% (112) scraping sample was culture positive and grew 130 microorganisms. 16 cases had polymicrobial growth. The most common bacterial growths were coagulase-negative Staphylococcus, Pseudomonas aeruginosa, and Staphylococcus aureus. Neither the Gram-positive nor the Gram-negative group demonstrated a shifting trend over the 10-year period ( = 0.634 and 0.722; = 0.172 and −0.129, resp.).   treatment. 38% of these cases were combined with aminoglycosides (gentamicin). 6.5% (17 cases) started with combination of fortified antibiotics (ceftazidime plus tobramycin or vancomycin). 3.5% (9 cases) received antifungal agents on top of initial antibacterial treatment (as these were referred as suspected fungal cases). Topical antifungal agents employed included topical natamycin and/or 0.1% amphotericin B. Systemic antifungal agents with voriconazole, ketoconazole, or fluconazole were used in 3 cases. 90% (239) of cases showed good initial response with improvement in either pain, infiltrate size, epithelial defect size, or amount of hypopyon. 12% (32) of cases needed to step up treatment, either due to lack of treatment response after 48-72 hours or as guided by the antibiotic susceptibility result of the corneal scrape.

Antibiotic
90.7% had good outcome and 9.3% (24 cases) had poor outcome. Two cases developed endophthalmitis resulting in blind eye. One culture-negative case had corneal biopsy performed at a later stage which found Acanthamoeba. One initially culture-negative case required enucleation and histology revealed fungal infection. Therapeutic penetrating keratoplasty was performed in one case.
Logistic regression was performed to analyze where the clinical outcome was affected by the factors shown in Table 3. Older age group (average 62.7), history of trauma, and presenting lesion size >3 mm were found to be associated with poor outcome with both univariate ( = 0.05, 0.009, and 0.044, resp.) and multivariate ( = 0.03, 0.006, and 0.039, resp.) analyses.

Discussion
The overall positive culture rate in this series was 32.3%. When compared with other keratitis studies published in the past decade (Table 4), their culture growth rate of corneal scrapings ranged from 25.6% to 78% [1, 8, 10-12, 14-17, 21, 22]. Ours was among the lower yield range. One main reason was that our hospital was a regional tertiary center. Most cases that presented to us have already been started on topical antimicrobial therapy. Our center also received referral from private ophthalmologists, and this group of partially treated patients often has already received broad-spectrum antibiotics therapy, commonly topical levofloxacin or moxifloxacin. Another reason was the indication of performing corneal scraping was not limited to infectious keratitis. In some cases, the clinical diagnosis was noninfective, such as peripheral ulcerative keratitis or herpetic keratitis. Clinically there was absence of suppurative infiltrate. Nevertheless, corneal scraping was performed in order to exclude presence of secondary bacterial infection as this had important implication in the management plan. These cases, as expected, yielded negative culture result and contributed to the lower overall rate of positive growth.
We had a slightly higher proportion of Gram-positive than Gram-negative bacteria, with coagulase-negative Staphylococcus (CNS) and Pseudomonas being the most common growths, respectively. Most other studies in developed countries also had similar findings where there were a higher proportion of Gram-positive bacteria, except two studies from the United Kingdom [14,16]. But, in the studies performed in Asian developing countries, they had more fungal keratitis as majority were traumatic keratitis [18][19][20]. On the other hand, in developed Asian regions, such as Shanghai, Beijing, Taiwan, Japan, and Hong Kong (current study), the findings resembled those of developed European and American regions, where the most common risk factors were contact lens use and fungal keratitis only consisted of a small portion of cases [3,10,21,23]. This highlighted the fact that wide geographical variation existed even within the same continent, and the clinician should pay attention to the background and travel history of the patient especially in international cities like Hong Kong.
The Hong Kong Keratitis Study Group studied 223 cases in 1997 to 1998 [7]. The positive culture rate was 35%, which was very similar to our series. The main risk factors were history of ocular surface disease followed by trauma, and the most prevalent causative bacteria were Pseudomonas. Compared to them, our series had a higher portion of contact lens-related keratitis. The shift in the risk factor for keratitis could be due to an increasing number of teenagers and young patients using contact lens and with poor contact lens habit. This suggestion could be supported by another study that looked into 18 cases of pediatric keratitis (aged under 18) from 2001 to 2010 in Hong Kong [24]. In their series, contact lens use was the most common risk factor. In our study, 15% (17 patients) of the contact lens-keratitis subgroup were aged 18 or younger. In both their study and our contact lens-keratitis subgroup, Pseudomonas followed by coagulasenegative Staphylococcus was the most common pathogen. This showed consistency in the microbial pattern in contact lens-related keratitis in Hong Kong. Pseudomonas also remained fully susceptible to third-generation cephalosporin, aminoglycosides, and quinolone antibiotics in both studies, which spanned over 15 years. This was very reassuring as these were the most commonly used agents by ophthalmologists in Hong Kong, and no increase in resistant strains was reported.   Journal of Ophthalmology In our study, fluoroquinolone susceptibility was tested in 47 Gram-negative bacteria with 93.6% susceptibility. When compared with a recent 10-year study in UK, they reported a 94.4% susceptibility to ciprofloxacin for their Gram-negative bacteria [16]. In another recent 11-year study in Toronto, they found a 97.4% susceptibility to ciprofloxacin for their Gramnegative isolates, with 97.8% for Pseudomonas aeruginosa [1]. Therefore, our fluoroquinolone susceptibility rate was actually quite similar to those reported by overseas studies. As for Pseudomonas aeruginosa, we found 100% susceptibility to fluoroquinolone in our series. The low resistant rate of Gram-negative bacteria in our series, including Pseudomonas aeruginosa, may be related to the relatively low usage of quinolone in general in Hong Kong. There were two main reasons for the low usage of quinolone. First of all, the most commonly prescribed antibiotic eye drops in our primary care setting was chloramphenicol instead of fluoroquinolones. Fluoroquinolone eye drops were mainly only prescribed by ophthalmologists and therefore abuse of these antibiotic agents was less of an issue. Secondly, in patients with systemic infections such as pneumonia, our physicians always try to avoid the use of quinolone as a first line agent. This is because of the high incidence of tuberculosis in Hong Kong, and the widespread use of fluoroquinolone is discouraged to prevent the emergence of quinoloneresistant tuberculosis. The zero resistance to fluoroquinolone in our Pseudomonas cases could also explain the good initial treatment response (90%) in our series. We would continue fluoroquinolone monotherapy as the initial first line therapy in most of our infectious keratitis cases. In fact, Shalchi et al. gave same recommendation in their 10-year review in United Kingdom [16]. A recent meta-analysis also supported the use of fluoroquinolone as the first choice for empirical treatment in most cases of suspected bacterial keratitis too [25].
Older age, larger ulcer size, and traumatic cases were associated with poor outcome with a loss of final visual acuity. The older age group patients often had coexisting keratopathies, ocular surface diseases, or systemic diseases. Other study also reported worse outcome in patients over 60 years old [9]. We would recommend antibiotic with broad Gram-positive coverage in older patients since we found a significantly higher proportion of Gram-positive isolates in both presence of keratopathy and presence of systemic conditions. Larger lesion size (>3 mm) was also associated with poor outcome, as this group of patients often had delayed presentation or factitious pathogens, although we found no correlation of microorganisms with the ulcer size. We would recommend admission for fortified antibiotic for these patients to ensure a better compliance to therapy and closer monitoring.
Limitations. This study was limited by its retrospective design. The decision of performing corneal scraping in clinically bacterial keratitis cases was not unified. Some early cases with small lesions (<1 mm) were treated empirically without obtaining corneal scraping. As a result, there was very likely an underreporting of the number of bacterial keratitis cases and cultured isolates. The study was also carried out in a tertiary referral centre and therefore was treating a higher portion of more difficult or treatment-refractory cases.
The number of underlying risk factors might have been overlooked by the attending clinician as well. For example, the presence of keratopathies or systemic illness might not have been performed consistently in a retrospective study. For the antibiotic susceptibility testing, different groups of bacteria were tested with different sets of antibiotics. For example, fluoroquinolones were not routinely tested in Gram-positive isolates. Therefore, we could not directly compare the percentage of fluoroquinolone resistance with other studies. Regarding the clinical outcome, we could only rely on parameters such as final visual acuity and cases with complications. The exact number of days for complete resolution of infiltrate and reepithelialization could not be determined as the follow-up intervals varied among different clinicians and in different cases.

Conclusions
This study was the largest case series of consecutive corneal scrapings in Hong Kong. It identified contact lens wear as the most common risk factor with Pseudomonas being the most prevalent pathogen in this group. Its susceptibility to most first line antibiotic agents remained 100% throughout the decade, and monotherapy with fluoroquinolones was an effective first line treatment in most of our cases. More aggressive treatment approach should be employed in keratitis belonging to the older age group, larger presenting ulcer size, and traumatic cases in view of their association with a poorer outcome.

Disclosure
This study was presented as an e-Poster at the World Cornea Congress VII 2015 in San Diego, USA.