This study included all patients treated for infectious keratitis who underwent corneal scrapings in the Cornea and External Disease Department of HSPE-SP between July 1997 and July 2018. HSPE-SP is the referral hospital for 1.2 million public employees from 163 cities of the State of São Paulo and covers mainly an urban area with a humid subtropical climate.

The exclusion criteria were immune-related, viral, Acanthamoeba or fungal ulcers, endophthalmitis, blebitis, dacryocystitis, scleritis, uveitis, conjunctivitis, and endothelitis.

We collected information about age, sex, predisposing systemic and ocular factors, clinical classification of ulcers, empirical treatment, the need for systemic antibiotics or surgery, corticosteroids and anti-collagenase use, duration of treatment, initial and final visual acuity, bacterioscopy and culture results, and antibiotic sensitivity test results. Ocular trauma, contact lens wear, ocular scarring disease, corneal transplant, bullous keratopathy, meibomitis, lid abnormalities, and herpetic keratitis were considered ocular risk factors. Systemic risk factors included systemic immunosuppression, either secondary to diseases or immunosuppressive agents, such as diabetes mellitus or individuals affected by human immunodeficiency viruses, and chronic alcoholism, drug use, rheumatoid arthritis, neoplasia, atopy, encephalopathy, or coma.

The ulcers were classified as severe if they had a central or paracentral location, a diameter of 4 mm or more, involved more than one-third of the corneal thickness, and/or presented with a hypopyon. Moderate ulcers were peripheral, measured between 2 and 4 mm, affected only the anterior third of stromal depth, and had no hypopyon. Mild ulcers were < 2 mm in diameter.

At our institution, in cases of moderate and severe corneal ulcers, we scraped and cultured a sample and administered a topical combination of cephalothin or cefazolin (50 mg/mL) and gentamicin or tobramycin (14 mg/mL) every hour or prescribed a fourth-generation fluoroquinolone (moxifloxacin, gatifloxacin) as monotherapy. Systemic therapy was prescribed in severe cases with intraocular or scleral involvement, perforation, or gonococcal etiology. Fluorometholone 0,1 % ophthalmic suspension was prescribed after corneal epithelialization in selected cases: severe ulcers with intense inflammation, 2 to 3 days after initiation of broad-spectrum antibiotics with a positive clinical response but the absence of fungal or Pseudomonas suspicion. In cases of corneal melting and unfavorable progress, an oral anti collagenase, such as doxycycline 100 mg b.i.d., was prescribed.

Duration of treatment was defined as the time elapsed between the initial therapy implemented at the hospital and the complete healing of the ulcer (cure). Infection was considered as cured if there were no more inflammatory signs or symptoms, pain, stromal infiltrate, or anterior chamber reaction. Visual acuity was evaluated using the Snellen chart at the first visit and after the cure. For statistical analysis, the data were converted to a logarithmic scale.

Corneal samples from infectious keratitis cases at HSPE-SP were collected in a standardized manner as follows. After topical anesthesia, a modified Kimura spatula was used to directly inoculate the scraped material from the bottom and border of the ulcer onto Gram stain and Giemsa colorations and blood agar, chocolate agar, Sabouraud agar, and enriched thioglycolate medium cultures.

The identification and antimicrobial susceptibility testing were initially performed using panels NMIC/ID 121 and PMIC/ID 105 in the system BD-Phoenix (Becton Dickinson). Susceptibility to ciprofloxacin, ceftriaxone, cefepime, teicoplanin, sulfamethoxazole-trimethoprim, amikacin, gentamicin, erythromycin, clindamycin, and tetracycline was evaluated by a disk diffusion method as recommended by the Clinical Laboratory Standards Institute (CLSI M100–23) and Brazilian Health Regulatory Agency (Anvisa). The susceptibility to vancomycin was determined by the broth microdilution method. S. aureus ATCC 25923 and E. faecalis ATCC 29212 were tested as quality‐control strains and the susceptibility testing results were interpreted based on the CLSI M100-S23 criteria and Anvisa.18

Data were analyzed using IBM Statistical Package for the Social Sciences (SPSS) 21.0 version. Numerical data were presented as mean ± Pattern Deviation (PD) or median and interquartile range, wherever applicable. Shapiro-Wilk and Kolmogorov-Smirnov were used to evaluate whether variables were normally distributed. For normal distribution data, t-Student test was applied. When variables were not normally distributed, Mann-Whitney test was chosen. The comparison of the mean of the quantitative variables between the categories of polytomous qualitative variables were not normally distributed and the Kruskal-Wallis H test was applied. The Wilcoxon test was used to compare two paired numerical data and X2 to compare categorical data. The alpha error was set at 5 %.

From July 1997 to July 2018, 160 cases of infectious keratitis were reported. Table 1 describes the epidemiological characteristics of the study population. The mean age was 65.1 ± 1 8.4 years. Risk factors were identified in 83,1 % of the population of the study (Table 1).

Topical eye drops were used as initial empirical treatment in 159 patients (99.4 %): 105 (66 %) received fortified antibiotics (cephalothin or cefazolin 50 mg/mL and gentamicin or tobramycin 14 mg/mL) and 54 (33.9 %) were treated with only a fourth-generation quinolone. Data were unavailable for one patient. Surgical treatment was later performed in 40 patients (25 %) using the modalities shown in Table 1. Three patients (1.8 %) underwent more than one surgical treatment.

One hundred and six patients (66.2 %) were cured with initial therapy. Forty-five patients (28.1 %) did not improve clinically and required a change of therapy: 26 (57.7 %) were using the fortified combination and 19 (42.2 %) were on quinolone. Data were unavailable for the nine remaining patients. No statistically significant difference was found between treatment with fortified combination versus fourth generation quinolone and clinical response (p = 0.16). Ten cases (20.8 %) out of the 45 patients with clinical deterioration had negative bacterioscopy and culture results and were subsequently treated with vancomycin (25 mg/mL) and amikacin (25 mg/mL). Three of those ten patients (30 %) treated with vancomycin and amikacin had infection resolved with topical treatment only. Seven of them (70 %) had to be submitted to corneal transplantation, and posteriorly three patients had to be eviscerated. In 19 cases (42.2 %), the etiological agent was isolated in the culture and guided the choice of medication, those patients achieved infection resolution with topical treatment. In 16 cases (35.5 %), other treatments were used: patients initially using quinolone started cephalothin and gentamicin, those already using a fortified combination initiated individual different treatments: moxifloxacin, gatifloxacin, acyclovir, ceftazidime, azithromycin, amphotericin B 0.15 %. Clinical evolution varied widely, but four eyes were eviscerated.

Intravitreal injection of vancomycin with ceftazidime was administered to six patients (3.7 %). Oral antimicrobial therapy was used as adjuvant treatment in five patients (3.1 %): amoxicillin in two patients (1.2 %), and ciprofloxacin in three patients (1.8 %). Three patients (1.8 %) were treated with an intravenous antimicrobial agent, two (1.2 %) patients used ciprofloxacin and one (0.6 %) patient was advised vancomycin with ceftriaxone. Subconjunctival vancomycin injection was administered to three patients (1.8 %). Nine patients (5.6 %) evolved to endophthalmitis. A total of six eyes (3.7 %) were eviscerated, five of which (3.1 %) were previously amaurotic.

Seventy-five patients (46.8 %) were treated with topical corticosteroid and/or an oral anticollagenolytic: 8 patients (5 %) used both these medications, 41 patients (25.6 %) used only topical steroids, and 26 patients (16.2 %) used oral doxycycline.

The median time between symptom onset and treatment initiation was 5 (3‒9.5) days, varying from 1 to 60 days for 55 patients. Nineteen patients (34.5 %) waited for 1 to 3 days, 20 patients (36.4 %) waited for 4 to 7 days, and 16 patients (29.1 %) waited for more than 7 days before initiation of treatment. The mean duration of treatment was 22.5 ± 9 days. We compared duration of treatment of patients using quinolone versus fortified antibiotic (Table 2) and no statistically significant difference was found (p = 0.531). We also compared duration of treatment for Gram-positive versus Gram-negative bacteria (Table 2) and no statistically significant difference was found (p = 0.767).

A comparison between pre- and post-treatment visual acuity was made for 76 patients. Information about visual acuity was unavailable for the remaining patients. The mean vision improvement was of two lines and two letters, or a mean variation of −0.25 logMAR (z = 3.36; p < 0.001). Thirty-eight patients (50 %) had the same vision, 28 (36.8 %) had improved visual acuity, and 10 (13.1 %) had worsened vision despite treatment. Eighteen eyes (23.7 %) were previously amaurotic.

Pre- and post-treatment visual acuity for S. aureus, S. pneumoniae and P. aeruginosa were compared to search for a correlation between the etiology and visual prognosis, but no statistically significant difference was found (Table 3).

The culture was positive in 69 case (43.1 %). Gram-positive bacteria were the most common isolate (60.3 %): Staphylococcus spp. (37 %) was the first one in frequency followed by Streptococcus spp. (19.1 %). All isolated microorganisms are discriminated in Table 4. More than one specimen was identified in five samples (7.1 %).

Gram-positive bacteria were less susceptible to the tested antibiotics than Gram-negative bacteria. Only six Gram-positive specimens (13.6 %) were sensitive to all the tested antibiotics. All microorganisms were sensitive to cephalothin and vancomycin, which are frequently used as empirical therapies. Resistance to oxacillin and penicillin was observed in seven cases (15.9 %). All Gram-positive bacteria were sensitive to all tested quinolones, except for a Staphylococcus intermedius strain resistant to moxifloxacin.

Gram-negative bacteria were sensitive to all the tested antibiotics, except for the following cases of resistance: a Serratia marcescens resistant to ceftriaxone and nitrofurantoin; a Citrobacter spp. resistant to cephalothin, and cefoxitin; Enterobacter resistant to amoxicillin-clavulanate, and cefuroxime; a Morganella morganii resistant to amoxicillin-clavulanate, and sulfamethoxazole-trimethoprim.

Data were analyzed in order to establish whether there was an increase in antibiotic resistance for S. aureus, S. pneumoniae and P. aeruginosa over time. No statistically significant difference was found. Resistance over time of tested antibiotics for S. aureus, S. pneumoniae e P. aeruginosa is shown in Table 5.