To investigate the utility of corneal biopsy in the evaluation of infectious keratitis; to compare results of culture and histopathologic examination of the same specimens; to investigate potential factors related to positive biopsy results.
Retrospective, observational case series.
We reviewed medical records for all patients who underwent corneal biopsy because of infectious keratitis at the Jules Stein Eye Institute from June 1989 through June 2009. In general, biopsy specimens were both cultured and examined histopathologically. Lesion size, lesion progression, and interval from presentation to biopsy were analyzed as possible predictors of positive biopsies.
Organisms were identified in 20 of 48 (42%) consecutive biopsies (positive cultures in 9 of 47 cases [19%]; positive histopathologic examination in 19 of 47 cases [40%]). Culture and histopathologic results were concordant in 30 of 46 biopsies (65%) for which both techniques were performed; 10 of the 16 discordant cases were culture-negative/histopathology-positive, while the remaining 6 had positive but discordant results for the 2 techniques (cultures all showed bacteria; histopathologic examination showed fungi or cysts consistent with Acanthamoeba sp.). Corneal biopsy revealed microorganisms in 12 of 27 patients (44%) with negative cultures of corneal scrapings obtained at presentation. None of the potential risk factors was statistically associated with positive biopsies.
Corneal biopsy can be useful for identifying the cause of infectious keratitis in selected cases. Culture and histopathologic examination can provide complementary information, but discordant results may occur. Acanthamoebic and fungal infections are more likely to be identified by histopathologic examination.
Identifying the causal organism in cases of infectious keratitis can be difficult. Corneal biopsy has been recommended for cases in which keratitis persists or progresses, despite antimicrobial therapy. Biopsies provide more sample for culture than initial scraping of the ocular surface, and they provide access to deeper stromal infections than possible with superficial scraping alone. They also provide an opportunity for histopathologic analysis, although there has been conflicting information in the literature as to whether culture or histopathologic examination of a corneal biopsy specimen is more useful. Corneal biopsies are not without risk; they may result in corneal perforation, stromal scarring, and irregular astigmatism. It is therefore important to establish the indications for biopsy, understand the limitations of the procedure, and determine the most appropriate processing of biopsy specimens.
It has been routine practice at the Jules Stein Eye Institute to divide corneal biopsy specimens and send portions both to the Clinical Microbiology Laboratory for culture and to the Ophthalmic Pathology Laboratory for histopathologic examination. We have performed a comprehensive review of corneal biopsies performed at UCLA over a 20-year period to assess the utility of this procedure. Specifically, we undertook this project with the following goals: to compare results of culture and histopathologic examination on the same specimen; to determine whether results of these techniques differed on the basis of the type of infectious agent; to identify factors that predicted positive biopsy results; to compare biopsy results to initial cultures of corneal scrapings; and to determine the clinical relevance of a negative corneal biopsy.
We reviewed medical records for all patients with suspected infectious keratitis who underwent corneal biopsies at the Jules Stein Eye Institute from June 1989 through June 2009. Cases were identified from a registry of specimens in the Ophthalmic Pathology Laboratory; included were those specimens identified as being associated with “corneal biopsy” or “diagnostic keratectomy.” For patients who underwent multiple biopsies, data from first biopsies were used for analyses, unless otherwise indicated.
Although this study was not prospectively designed, most cases were managed in a typical manner, per protocols of the UCLA Cornea Service, as described below. At presentation, superficial corneal scrapings are obtained with a spatula or surgical blade and inoculated directly onto blood, chocolate, and Sabouraud dextrose agar, and into Middlebrook or Lowenstein-Jensen and thioglycollate transport media. Material from the scraping is also placed on at least 2 glass slides for staining and microscopic examination. Plates, tubes, and slides are transported immediately to the UCLA Clinical Laboratories. A specimen is also sent for inoculation onto non-nutrient agar with Escherichia coli overlay for culture of Acanthamoeba sp.
Indications for biopsy include: 1) lack of clinical response to initial antimicrobial therapy chosen on the basis of initial culture or smear results (or to broad-spectrum antibiotics in those cases for which cultures and smears are negative); or 2) progression of lesions (eg, enlargement of stromal infiltrates or corneal thinning), despite treatment, even if there are some other signs of response. Prior to biopsy, antimicrobial therapy is generally discontinued for approximately 24 hours.
Biopsies are performed in a minor procedure room using topical anesthesia. In most cases, a partial-thickness incision is made with a 2-mm or 3-mm trephine at a leading edge of the corneal infiltrate, incorporating both infected and adjacent, clear cornea. In selected cases, an alternative, free-hand method is used, in which a sharp blade creates a keratotomy at the edge of the infiltrate, after which a surgical blade is used to create a lamellar incision into the infiltrate and the overlying tissue is removed with scissors. Biopsy specimens are bisected; half is placed in unpreserved saline for transport to the UCLA Clinical Laboratories for culture, while the other half is placed in 10% buffered formalin for transport to the Ophthalmic Pathology Laboratory. Material for culture is ground with a mortar and pestle and mixed with sterile trypticase soy broth; this suspension is placed onto appropriate media for isolation of bacteria, fungi, mycobacteria, and Acanthamoeba sp. Material for histopathologic examination is processed routinely for paraffin embedding, and 5-μm-thick sections are cut at 250-μm-step intervals. Sections are prepared with hematoxylin-eosin, Gomori methenamine silver, periodic acid–Schiff, Ziehl-Neelsen, and Gram stains for identification of organisms.
In general, patients are treated topically at presentation with 2 fortified antibiotics for broad-spectrum coverage until results from initial scrapings are obtained, at which time treatment is individualized, based on those results. Subsequent treatment decisions are based on best medical judgment. Therapy is changed on the basis of biopsy results when initial antimicrobial agents are deemed not appropriate treatment for the organisms identified by the biopsy; if culture and histopathologic examination provide conflicting results (see definition of discordant biopsy results below), we have generally chosen antimicrobial agents on the basis of histopathologic examination results.
Age and sex were identified for each patient; no other demographic data were collected. The following data about corneal infections were collected: greatest diameter of corneal stromal infiltrates; location of the infiltrates (central, peripheral); and change in clinical signs during follow-up (enlargement of infiltrates; new foci of infiltration). Culture results for corneal scrapings obtained at presentation (initial evaluation at UCLA) were recorded. We did not consider the results of microscopic examination of smears made of the material from corneal scrapings at presentation.
We recorded the following treatment information: antimicrobial therapy prior to initial evaluation at UCLA and initial antimicrobial therapy at UCLA. We identified the interval from initial evaluation at UCLA to biopsy and the indication for each biopsy (progression [enlargement of infiltrate, new focus of infiltration], lack of response, other).
We evaluated each case for the following outcomes after primary biopsy: perforation at the time of biopsy; change in medical therapy; repeat scraping for culture; repeat biopsies; other surgical intervention (evisceration, penetrating keratoplasty [PK]); and healing of ulceration. The intervals from biopsy to each of these outcomes were determined. For patients who underwent PK, we determined whether organisms had been identified in the host corneal tissue removed at the time of surgery.
A biopsy was considered to be positive if organisms were identified by either culture or histopathologic examination. Concordant biopsy results were those in which there was agreement between culture results and histopathologic examination results (both negative or both positive and consistent with the same organism). Discordant biopsy results were those in which either culture or histopathologic examination results (but not both) were positive or those with positive culture and positive histopathologic examination results that suggested different organisms. Healing was defined as resolution of infiltrates. A therapeutic PK was one required before healing because of perforation, impending perforation, or progression with risk of scleral infection.
Data Analysis and Statistical Techniques
To evaluate whether the 2 techniques (culture vs histopathologic examination) were different in their abilities to identify each type of organism, we compared the distribution of positive results, categorized by organism type, using the McNemar test, which assesses the symmetry of the distribution for each technique, as a measure of the relative ability of each to identify organisms. It does not address the accuracy of either technique. Agreement between culture and histopathologic examination results was assessed by kappa statistics. The Kruskal-Wallis test was used to evaluate 2 potential risk factors for positive biopsies (time from presentation to biopsy; lesion size [greatest diameter of the stromal infiltrate]) and to compare times from biopsy to resolution based on biopsy results. The Fisher exact test was used to evaluate indications for biopsy as a risk factor. For patients who underwent multiple biopsies, results of positive second biopsies were used to calculate intervals from biopsy to outcomes, if first biopsies were negative.
We identified 48 consecutive patients who underwent at least 1 corneal biopsy. Table 1 shows demographic factors, corneal lesion characteristics, and corneal biopsy–related factors for this group of patients. Table 2 shows biopsy results. Organisms were identified by culture or histopathologic examination in 20 of 48 primary biopsies (42%). In 1 case, a specimen was not sent for culture; in another, a specimen was not sent for histopathologic examination. Cultures were positive in 9 of 47 cases (19%), while histopathologic examinations were positive in 19 of 47 biopsies (40%). Histopathologic examination was more likely to identify fungi and Acanthamoeba sp. than culture ( Table 2 ). No cultures were positive for Acanthamoeba sp.; among the 8 cases with presumed Acanthamoeba sp. on histopathologic examination, some had only cystic spaces with apparent double walls between lamellae of corneal stroma (but without nuclei) that were suggestive, but not diagnostic, of Acanthamoeba sp.
|Male sex, n (%)||21 (44%)|
|Mean ± SD||55.5 ± 15.8 years|
|Median (range)||57 (29 to 89) years|
|Corneal lesion characteristics|
|Size (largest dimension; n=36) a|
|Mean ± SD (n = 36) a||5.1 ± 2.3 mm|
|Median (range) (n = 36) a||5.0 (1.1 to 10) mm|
|Location (n = 38) a|
|Median interval from onset of keratitis to corneal scraping at initial UCLA examination (range) (n = 37) a||15 (1 to 240) days|
|Median interval from scraping to biopsy (range) (n = 40) a||19 (0 to 275) days|
|Positive culture of initial corneal scrapings (n = 42) a||15 (36%)|
|Indication for biopsy (n = 45) a|
|Lack of improvement||23 (55%)|
|Positive results c||20 (42%)|
|Number of second biopsies d||5|
|Perforation during procedure (n = 53) e||3 (6%)|
a Number of cases for which information was available, if less than 48.
b Two cases of amputation of LASIK flap for confirmation of diagnosis; 1 case to confirm report of positive fungal isolate from outside facility.
c Either positive culture or organisms seen on histopathologic examination.
d Three of 5 patients had negative first biopsies; 2 of the second biopsies were positive by culture or by histopathologic examination; details are provided in the text.
|Organisms Identified by Biopsy||Culture of Biopsy Specimen (n = 47)||Histopathologic Examination of Biopsy Specimen (n = 47)||P Value a|
|Bacteria||6 (12.8%)||4 (8.5%)||.53|
|Mycobacteria b||2 (4.3%)||1 (2.1%)||NE|
|Fungus||1 (2.1%)||6 (12.8%)||.025|
|Acanthamoeba sp . c||0||8 (17.0%)||NE|
|Total||9 (19.1%)||19 (40.4%)||.002|
a McNemar test. This test evaluates whether two techniques perform differently on the same sample; only those cases for which both techniques had been used (n = 46) were assessed. Without knowing the true distribution, one cannot assess which of the reported distributions is more accurate.
b One of the 2 culture-positive cases was the same case for which mycobacteria were identified on histopathologic examination. The other culture-positive case was not examined histologically; the lack of other histopathologic examination–positive cases prevented statistical comparison using the McNemar test. If one assumes a binomial distribution, with the proportion of cultures positive for mycobacteria among 47 specimens being the same as the results for histopathologic examination (2.1%), then the probability of obtaining 2 or fewer positive cultures for mycobacteria would be .92.
c Results cannot be compared statistically using the McNemar test because there were no cultures positive for Acanthamoeba sp. If one assumes a binomial distribution, with the proportion of cultures positive for Acanthamoeba sp. among 47 specimens being the same as the results for histopathologic examination (17.0%), then the probability of obtaining no positive cultures for Acanthamoeba sp. would be .0002.
As shown in Table 3 , agreement between culture and histopathologic examination results was poor on a statistical basis (kappa = 0.011). Both culture and histopathologic examination results from the same biopsy were available in 46 cases. Results were concordant in 30 cases (65%) and discordant in 16 cases. Among the 30 concordant cases, both culture and histopathologic examination were negative in 28 cases, while both were positive in 2 cases (fungi in 1 case, mycobacteria in the other). Among the 16 discordant cases, only histopathologic examination was positive (ie, culture negative) in 10 cases; bacteria were seen in 4 cases, cysts consistent with Acanthamoeba sp. in 4 cases, and fungi in 2 cases. In the 6 discordant cases for which both cultures and histopathologic examinations were positive, all cultures grew bacteria, while histopathologic examinations revealed cysts consistent with Acanthamoeba sp. (n = 4) or fungi (n = 2). The bacteria in 4 of these 6 cases are commonly considered to be contaminants (coagulase-negative staphylococci in 3 cases, Propionibacterium sp. in 1 case); the other 2 positive cultures grew Pseudomonas stutzeri , which is less virulent than Ps. aeruginosa but has been reported to cause keratitis, and Acinetobacter lwoffi , an opportunistic pathogen, widely distributed in nature, that can colonize healthy tissues. Among the 4 cases with cysts on histopathologic examination and bacteria on culture, 3 were confirmed subsequently to be infected with Acanthamoeba sp. at therapeutic PK; the other was lost to follow-up after the biopsy. Of the 2 cases with fungus on histopathologic examination and bacteria on culture, 1 healed and the other had no organisms at therapeutic PK. For both, treatment had been switched to antifungal agents in response to biopsy results. For each of the culture-negative discordant cases with Acanthamoeba sp. (n = 4) or fungus (n = 2) on histopathologic examination, there were outcome data (organisms identified on host tissue at therapeutic PK or healing in response to a change in therapy) to support the histopathologic diagnoses. There were no discordant cases in which both culture and histopathologic examination identified bacteria, but in which the cultured bacteria were not consistent with the morphology and staining characteristics of the bacteria seen on histologic examination ( Table 3 ).
|Culture Results||Histopathologic Examination|
|Concordant a||Discordant b|
a Both culture and histopathologic examination identified organisms; morphologic and staining characteristics on histopathologic examination were consistent with culture results.
b Both culture and histopathologic examination identified organisms; however, morphologic and staining characteristics on histopathologic examination were not consistent with culture results.
Table 4 shows the relationship between culture results of corneal scraping at initial UCLA examination and subsequent biopsy results. Results for culture of initial scrapings were known for 42 cases; results were not available or scrapings were not performed in 6 cases. Among the 15 cases with positive cultures on initial scrapings, subsequent biopsies were positive in 3 cases. In 2 of these cases, biopsy results confirmed the results of the initial cultures (1 case of mycobacterium and 1 case of fungus); in the third case, culture of initial scrapings grew Staph. epidermidis, whereas histopathologic examination of the biopsy specimen identified a fungus (biopsy culture was discordant, also growing Staph. epidermidis , which was believed to be a contaminant). Of the 27 cases with negative cultures of initial scrapings, 12 biopsies were positive. Of the 26 cases with negative cultures of initial scrapings that also had culture of biopsy specimens, only 4 biopsy cultures were positive (3 bacteria and 1 mycobacterium). Of the 26 cases with negative cultures of initial scrapings that also had histopathologic examination of biopsy specimens, 12 were positive, 9 of which were nonbacteria. Of the 3 cases with bacteria on histopathologic examination only, 1 showed gram-positive bacteria; 1 showed gram-negative bacteria; and 1 had substantial crush artifact and findings of both gram-positive and gram-negative bacteria, none of which were intracellular. Of these 3 cases, 1 had been using topical antibacterial and antifungal drugs before the time of initial scraping, 1 had been using only an antifungal drug, and 1 had not been treated. Viewed another way, among the 9 cases with positive cultures of biopsy material, 7 had undergone cultures of initial scrapings. In 3 cases, biopsy cultures revealed the same organism as initial scrapings (1 each of Fusarium sp., M. chelonae , and Staph. epidemidis ). The initial cultures had been negative in the other 4 cases (biopsy positive for bacteria in 3 cases, for M. chelonae in 1 case).
|Initial Scraping a||Biopsy Results|
|Culture (n = 41)||Histopathologic Examination (n = 41)|
|Bacteria||Fungus||Acanthamoeba sp.||Mycobacteria||Negative||Bacteria||Fungus||Acanthamoeba sp.||Mycobacteria||Negative|
|Bacteria (n = 13)||1||0||0||0||12||0||1||0||0||12|
|Fungus (n = 1)||0||1||0||0||0||0||1||0||0||0|
|Mycobacteria b (n = 1)||0||0||0||1||0||—||—||—||—||—|
|Negative c (n = 27)||3 d||0||0||1||22||3 e||3||5||1||15|