Purpose
To determine risk factors and clinical signs that may differentiate between bacterial, fungal, and acanthamoeba keratitis among patients presenting with presumed infectious keratitis.
Design
Hospital-based cross-sectional study.
Methods
We examined the medical records of 115 patients with laboratory-proven bacterial keratitis, 115 patients with laboratory-proven fungal keratitis, and 115 patients with laboratory-proven acanthamoeba keratitis seen at Aravind Eye Hospital, Madurai, India, from 2006-2011. Risk factors and clinical features of the 3 organisms were compared using multinomial logistic regression.
Results
Of 95 patients with bacterial keratitis, 103 patients with fungal keratitis, and 93 patients with acanthamoeba keratitis who had medical records available for review, 287 (99%) did not wear contact lenses. Differentiating features were more common for acanthamoeba keratitis than for bacterial or fungal keratitis. Compared to patients with bacterial or fungal keratitis, patients with acanthamoeba keratitis were more likely to be younger and to have a longer duration of symptoms, and to have a ring infiltrate or disease confined to the epithelium.
Conclusions
Risk factors and clinical examination findings can be useful for differentiating acanthamoeba keratitis from bacterial and fungal keratitis.
Acanthamoeba keratitis is a relatively rare, difficult-to-treat infection of the cornea that can result in severe vision loss. Studies have identified several risk factors for acanthamoeba keratitis, including contact lens wear, orthokeratology, water exposure, and certain contact lens solutions. Although most acanthamoeba keratitis research has been conducted in industrialized countries, acanthamoeba keratitis also occurs in developing countries, often in non–contact lens–wearing individuals.
Acanthamoeba keratitis is frequently misdiagnosed as herpetic or fungal keratitis, and is subsequently treated incorrectly, which can lead to poor outcomes. Case series of acanthamoeba keratitis have identified several important clinical signs, such as pseudodendrites, perineural infiltrates, and ring infiltrates. However, we are unaware of any studies that have compared the clinical findings of acanthamoeba keratitis with those of bacterial and fungal keratitis. Clinical signs can be especially useful for differentiating the cause of infectious keratitis when microbiological testing is not available—which is frequently the case in developing countries. In this study, we compare the risk factors and clinical signs of laboratory-proven bacterial, fungal, and acanthamoeba keratitis cases from a tertiary eye care hospital in south India, in an attempt to improve differentiation of these forms of keratitis.
Methods
We obtained approval for this retrospective cross-sectional study from the Committee for Human Research at the University of California, San Francisco, and from the Institutional Review Board at Aravind Eye Hospital, Madurai. The research adhered to the tenets of the Declaration of Helsinki.
We identified all cases of smear- or culture-proven acanthamoeba keratitis from the microbiology database at Aravind Eye Hospital, Madurai, India, from January 1, 2006 to June 30, 2011. As controls, we identified a random sample of fungal and bacterial keratitis cases, matched to acanthamoeba keratitis cases based on the year of presentation (ie, the number of fungal and bacterial cases chosen for a particular year was the same as the number of acanthamoeba cases detected that year). During this period of time, results of cultures and smears showed a fungal organism in approximately 35% of keratitis cases, a bacterial organism in 20%, and a parasitic organism such as acanthamoeba in 1%. Using a review of the literature as a guide, we prespecified certain risk factors and diagnostic signs to be of interest and extracted information on these variables from the patient’s medical record using a standardized data collection form, masked to the identity of the organism. We were able to mask data extractors by having a separate chart reviewer cover all references to the microbiological diagnosis with adhesive paper. We recorded information on demographics, medical history, visual acuity at presentation, and clinical examination at presentation. We used only clinical information documented before microbiological evaluations were performed (ie, clinical examinations were masked to laboratory results). It should be noted that medical records could not be located for all patients listed in the microbiological database.
Microbiological methods for the Aravind Ocular Microbiological Laboratory have been described previously. In general, all patients with presumed infectious keratitis undergo corneal scraping for smear and culture. Gram staining and potassium hydroxide (KOH) wet mount are routinely performed for all smears. Routine culture media include sheep’s blood agar, chocolate agar, potato dextrose agar, and brain-heart infusion broth without gentamicin. For ulcers in which acanthamoeba is suspected clinically and/or when smears are KOH positive for amoebic cysts, further corneal scrapings are performed for culture on non-nutrient agar overlaid with Escherichia coli .
We created univariate multinomial logistic regression models with causative organism as the response variable (acanthamoeba, bacteria, or fungus), and each of the baseline risk factors or clinical features as the explanatory variable. Stromal infiltrate size was calculated as the geometric mean of the longest diameter and its perpendicular extent, as recorded in the medical record. For the purposes of this study, feathery infiltrate borders indicate that the words “feathery” or “fluffy” were documented in the medical record. Satellite lesions indicate that the word “satellite” was written in the chart, whereas multifocal lesions indicate that multiple discrete lesions were drawn. In general, ophthalmologists at the study site use the term satellite lesion to refer to a smaller infiltrate adjacent to a larger main infiltrate. All satellite lesions were by definition also classified as multifocal. Pseudodendrite indicates that the word “pseudodendrite” or “dendrite” was written in the medical record. We realize that the pseudodendrite is an ill-defined entity but use that term herein since it has been widely used in the acanthamoeba keratitis literature. Visual acuity was converted to logMAR units. We assessed for overall differences between the 3 organisms with a likelihood ratio test, and performed pairwise comparisons for any variables with P < .001. To account for potential confounding, we entered all variables into a multivariate multinomial logistic regression model. We used a backwards stepwise algorithm for model selection, removing variables with the highest likelihood ratio test until all variables had a P value <.01. We kept variables with a P < .01 in the multivariate model to account for important confounders, but only declared as statistically significant those variables with P < .001.
Results
From January 2006 to June 2011, a total of 115 acanthamoeba keratitis cases were listed in the microbiological database, of which 93 (81%) had medical records available for review. We randomly selected 115 bacterial and 115 fungal keratitis cases from the same time period, and were able to identify microbiological and medical records for 95 (83%) of the bacterial cases and 103 (90%) of the fungal cases ( P = .16). Organisms were generally detected on both the smear and culture ( Table 1 ). Bacterial cases were most commonly caused by Streptococcus pneumoniae (36/95, 38%) and Pseudomonas aeruginosa (28/95, 29%); fungal ulcers were most commonly caused by Fusarium species (32/103, 31%) and Aspergillus species (26/103, 25%) ( Table 2 ).
| Acanthamoeba Total = 93 | Fungus Total = 103 | Bacteria Total = 95 | |
|---|---|---|---|
| Culture-positive | 85/92 (92.4%) | 102/103 (99.0%) | 94/94 (100%) |
| Gram-positive | 79/92 (85.9%) | 83/97 (85.6%) | 76/93 (81.7%) |
| KOH-positive | 67/75 (89.3%) | 77/85 (90.6%) | 3/58 a (5.2%) |
| Organism | Number (%) |
|---|---|
| Bacteria | |
| Streptococcus pneumoniae | 36/95 (38%) |
| Pseudomonas aeruginosa | 28/95 (29%) |
| Nocardia species | 6/95 (6%) |
| Staphylococcus aureus | 4/95 (4%) |
| Staphylococcus epidermidis | 4/95 (4%) |
| Diphtheroids | 4/95 (4%) |
| Viridans group streptococci | 3/95 (3%) |
| Streptococcus pyogenes | 3/95 (3%) |
| Klebsiella species | 2/95 (2%) |
| Moraxella catarrhalis | 1/95 (1%) |
| Enterococcus species | 1/95 (1%) |
| Atypical Mycobacterium species | 1/95 (1%) |
| Acinetobacter species | 1/95 (1%) |
| Aeromonas hydrophilia | 1/95 (1%) |
| Culture negative (gram-positive cocci) | 1/95 (1%) |
| Fungi | |
| Fusarium | 32/103 (31%) |
| Aspergillus flavus | 19/103 (18%) |
| Aspergillus fumigatus | 7/103 (7%) |
| Curvularia | 8/103 (8%) |
| Exerohilum species | 4/103 (4%) |
| Bipolaris | 3/103 (3%) |
| Scedosporium species | 3/103 (3%) |
| Candida albicans | 1/103 (1%) |
| Lasiodiplodia species | 1/103 (1%) |
| Rhizopus species | 1/103 (1%) |
| Cladosporium species | 1/103 (1%) |
| Unidentified hyaline | 17/103 (17%) |
| Unidentified dematiaceous | 5/103 (5%) |
| Culture negative (KOH positive) | 1/103 (1%) |
Risk factors and clinical characteristics for each of the 3 classes of organisms are summarized in Table 3 , along with the omnibus P values from the univariate multinomial logistic regression models that assessed for overall differences between the 3 organisms. Pairwise comparisons for those risk factors and clinical features with evidence of an overall difference (defined as P < .001) are shown in Table 4 .
| Characteristic | Acanthamoeba (N = 93) | Fungus (N = 103) | Bacteria (N = 95) | P a |
|---|---|---|---|---|
| Risk Factors | ||||
| Age, y; mean ± SD | 38 ± 16 | 43 ± 16 | 50 ± 18 | <.001 |
| Female sex, n (%) | 41 (44%) | 41 (40%) | 29 (30%) | .17 |
| Symptom duration, days; mean ± SD | 33 ± 62 | 10 ± 13 | 13 ± 39 | <.001 |
| Trauma, n (%) | 55 (59%) | 62 (60%) | 60 (63%) | .84 |
| Vegetative trauma, n (%) | 27 (29%) | 32 (31%) | 33 (35%) | .70 |
| Past ocular surgery, n (%) | 5 (5%) | 12 (12%) | 22 (23%) | .001 |
| Cataract extraction, n (%) | 4 (4%) | 9 (9%) | 15 (17%) | .03 |
| Keratoplasty, n (%) | 0 (0%) | 1 (1%) | 2 (2%) | .10 |
| Other b , n (%) | 1 (1%) | 2 (2%) | 5 (5%) | .19 |
| Topical antibiotic use, n (%) | 59 (63%) | 48 (47%) | 28 (29%) | <.001 |
| Topical antifungal use, n (%) | 37 (40%) | 34 (33%) | 19 (20%) | .01 |
| Topical steroid use, n (%) | 8 (9%) | 6 (6%) | 6 (6%) | .89 |
| Native medicine use, n (%) | 10 (11%) | 11 (11%) | 10 (11%) | 1.00 |
| Topical breast milk use, n (%) | 3 (3%) | 10 (10%) | 7 (7%) | .17 |
| Topical castor oil use, n (%) | 1 (1%) | 4 (4%) | 4 (4%) | .33 |
| Eye contact with tongue, n (%) | 3 (3%) | 1 (1%) | 1 (1%) | .43 |
| Contact lens wear, n (%) | 3 (3%) | 0 (0%) | 2 (2%) | .10 |
| Clinical characteristics | ||||
| Visual acuity (logMAR); mean ± SD | 1.46 ± 0.61 | 1.18 ± 0.71 | 1.36 ± 0.66 | .009 |
| Infiltrate size; mean ± SD | 5.6 ± 3.0 | 4.6 ± 3.1 | 4.6 ± 3.3 | .06 |
| Stromal involvement in posterior one-third, n (%) | 32 (34%) | 36 (35%) | 42 (44%) | .29 |
| Hypopyon, n (%) | 36 (39%) | 43 (42%) | 52 (55%) | .06 |
| Pseudodendrites, n (%) | 7 (8%) | 2 (2%) | 1 (1%) | .04 |
| Epitheliopathy without stromal disease n (%) | 12 (13%) | 2 (2%) | 2 (2%) | .001 |
| Feathery edges, n (%) | 5 (5%) | 20 (19%) | 7 (7%) | .004 |
| Satellite lesions, n (%) | 4 (4%) | 3 (3%) | 0 (0%) | .05 |
| Multifocal lesions, n (%) | 17 (18%) | 10 (10%) | 5 (5%) | .02 |
| Ring infiltrate, n (%) | 28 (30%) | 5 (5%) | 4 (4%) | <.001 |
| Perineuritis, n (%) | 3 (3%) | 0 (%) | 0 (0%) | .03 |
a Overall comparison of the 3 groups in univariate multinomial regression.
b Amniotic membrane (acanthamoeba group, n = 1); retinal detachment repair (fungus group, n = 1); dacryocystectomy (bacteria group, n = 1); the remainder were unspecified.
| Explanatory Factor | Odds Ratio (95% Confidence Interval) a | ||
|---|---|---|---|
| Acanthamoeba vs Bacteria | Acanthamoeba vs Fungus | Fungus vs Bacteria | |
| Risk factors | |||
| Age, per decade | 0.64 (0.53-0.77) | 0.82 (0.69-0.97) | 0.78 (0.66-0.92) |
| Symptom duration, per week | 1.13 (1.02-1.25) | 1.23 (1.08-1.40) | 0.92 (0.80-1.05) |
| Topical antibiotic use | 3.97 (2.16-7.29) | 1.90 (1.07-3.36) | 2.09 (1.16-3.76) |
| Clinical characteristics | |||
| Ring infiltrate | 9.80 (3.28-29.3) | 8.44 (3.10-23.0) | 1.16 (0.30-4.46) |
a Univariate multinomial logistic regression with causative organism as the outcome; odds ratios are reported for acanthamoeba keratitis relative to a bacterial keratitis reference group and to a fungal keratitis reference group, and for fungal keratitis relative to a bacterial keratitis reference group; results with P < .05 in bold.
In pairwise comparisons, there appeared to be more differentiating features of acanthamoeba keratitis than for either bacterial or fungal keratitis. Risk features of acanthamoeba keratitis that were significantly different from both fungal keratitis and bacterial keratitis included younger age, longer symptom duration, prior use of topical antibiotics, and presence of a ring infiltrate ( Table 4 ). Risk factors associated with bacterial keratitis relative to fungal or acanthamoeba keratitis included older age and lack of prior topical antibiotic use.
In the multivariate model, several features of acanthamoeba keratitis were significantly different from both fungal keratitis and bacterial keratitis ( Table 5 ). Patients with acanthamoeba keratitis were younger than patients with bacterial keratitis or fungal keratitis, and had a longer duration of symptoms before being treated. In terms of clinical signs, acanthamoeba keratitis was more likely to have disease confined to the epithelium and a ring infiltrate. The multivariate model revealed fewer discriminating features for either bacterial or fungal keratitis; only age was significantly different among all 3 organisms, with older age a risk factor for fungal keratitis relative to acanthamoeba keratitis, and for bacterial keratitis relative to both fungal and acanthamoeba keratitis ( Table 5 ).
| Explanatory Factor | Odds Ratio (95% Confidence Interval) a | |||
|---|---|---|---|---|
| Acanthamoeba vs Bacteria | Acanthamoeba vs Fungus | Fungus vs Bacteria | Omnibus P Value | |
| Risk factors | ||||
| Age, per decade | 0.62 (0.50-0.78) | 0.78 (0.63-0.95) | 0.80 (0.67-0.96) | <.001 |
| Symptom duration, per week | 1.10 (1.00-1.21) | 1.17 (1.04-1.32) | 0.94 (0.83-1.07) | <.001 |
| Clinical characteristics | ||||
| Visual acuity, per unit logMAR | 1.96 (1.10-3.49) | 2.37 (1.37-4.08) | 0.83 (0.52-1.32) | .005 |
| Epitheliopathy without stromal disease | 12.9 (2.45-67.6) | 17.1 (3.24-89.9) | 0.75 (0.10-5.64) | <.001 |
| Multifocal lesions | 5.90 (1.87-18.6) | 3.22 (1.24-8.39) | 1.83 (0.59-5.69) | .004 |
| Ring infiltrate | 11.0 (3.42-35.3) | 9.26 (3.23-26.6) | 1.19 (0.30-4.65) | <.001 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
