To assess the efficacy of corneal cross-linking (CXL) as an adjuvant to appropriate antifungal therapy in nonresolving deep stromal fungal keratitis.
Randomized clinical trial.
Eyes with culture-positive deep stromal fungal keratitis not responding to appropriate medical therapy for a period of 2 weeks were randomized to receive either adjuvant CXL or no additional treatment. Antifungal medical therapy was continued in both groups. The prespecified primary outcome was treatment failure at 6 weeks after enrollment, defined as perforation and/or increase in ulcer size by ≥2 mm.
The trial was stopped before full enrollment because of a marked difference in the rate of perforation between the 2 groups. Of the 13 cases enrolled in the study, 6 were randomized to the CXL group and 7 to the non-CXL group. Five eyes in the CXL group and 3 eyes in the non-CXL group experienced treatment failure by 6 weeks ( P = .56). In a secondary analysis, the CXL group experienced more perforations than the non-CXL group (4 vs 0, respectively; P = .02).
CXL used as adjuvant therapy for recalcitrant deep stromal fungal keratitis did not improve outcomes.
Fungal keratitis is an important cause of ocular morbidity, especially in developing countries. Fungal corneal infections are often worse than those attributable to bacteria. Perforation is not uncommon; a recent clinical trial comparing natamycin and voriconazole drops for fungal keratitis found that 16% of eyes perforated or required therapeutic keratoplasty. Risk factors for poor outcomes include larger and deeper ulcers, so adjuvant therapies for advanced fungal ulcers may be warranted in order to prevent perforation. Corneal cross-linking (CXL) is one promising adjuvant therapy, since the corneal strengthening caused by this technique may inhibit corneal melting and prevent perforation. Recent reports have supported the use of CXL in infectious keratitis: several observational studies have reported resolution of bacterial and fungal keratitis after treatment with CXL, and a nonrandomized interventional study showed that eyes with bacterial, fungal, or amoebic keratitis treated with CXL had less subsequent perforation or reinfection than eyes not treated with CXL.
We are unaware of any studies that have specifically assessed the efficacy of CXL for recalcitrant deep stromal fungal keratitis. This is a substantial gap in knowledge, since these deep fungal ulcers may benefit the most from additional interventions to prevent perforation. Therefore, we undertook a therapeutic exploratory randomized clinical trial to assess the efficacy of CXL as an adjuvant therapy to topical antifungal drugs in patients with culture-proven deep stromal fungal keratitis.
Subjects and Methods
Ethical approval was obtained for our randomized controlled trial from the Aravind Eye Care System Institutional Review Board and the trial conformed to the tenets of Declaration of Helsinki. The trial was registered with clinicaltrials.gov ( NCT02328053 ).
Study participants were enrolled in the trial at Aravind Eye Hospital in Madurai, India from February 2, 2014 to October 30, 2014. All patients with a corneal ulcer presenting to our institution undergo comprehensive clinical and microbiological smear examinations and treatment is initiated based on the results of these tests. As part of our treatment protocol, all patients with superficial ulcers who are smear positive for fungal filaments are initially treated with 5% natamycin (Myconat Ophtho Remedies, Allahabad, India) eye drops hourly during waking hours. For larger ulcers (more than 5 mm in the longest diameter) and/or those deeper than one third of corneal stroma, 1% voriconazole (Vozole; Aurolab, Madurai, India) eye drops are added to the topical natamycin therapy. For the present trial, eyes with culture-positive fungal keratitis that had not improved after 2 weeks of medical therapy and that had an infiltrate measuring 5 mm or more in diameter and involving the posterior two thirds of the cornea at the 2-week visit were enrolled ( Supplemental Figure 1 , available at AJO.com ).
After obtaining informed consent but before randomization, slit-lamp examination was performed to document the longest diameter and widest perpendicular width of the infiltrate, as well as the most posterior depth of the stromal infiltrate. The infiltrate size was summarized as the geometric mean of the length and width. Anterior segment photographs were taken. In vivo confocal microscopy (HRT3-RCM; Heidelberg Engineering, GmbH, Heidelberg, Germany) directed at the center of the stromal infiltrate was used to document the corneal depth of the most posterior fungal hyphae. Uncorrected visual acuity was measured by certified refractionists using a Snellen chart.
Each study participant was subsequently randomized in a 1:1 ratio to receive either adjuvant CXL or no other treatment in addition to his or her topical antifungal therapy as per the existing protocol of Aravind Eye Care System. Randomization was performed using the rand and sort functions in Excel (Microsoft, Redmond, Washington, USA). Allocation was concealed in sequentially numbered opaque sealed envelopes by the study coordinator. Patients randomized to the CXL arm were admitted to the hospital and CXL was performed according to the Dresden protocol under strict aseptic precautions and using topical anesthesia with 0.5% proparacaine (Aurocaine; Aurolab, Madurai, India). For the first 30 minutes the cornea was soaked with 0.1% riboflavin drops (riboflavin dextran solution; Intacs XL, Dorset, UK) every 2 minutes. Over the next 30 minutes the cornea was exposed to ultraviolet-A radiation of 370 nm with 3 mW/cm 2 (IROC-UVX, Zurich, Switzerland) along with continued instillation of riboflavin drops. Topical antifungal therapy was restarted 1 hour after the procedure. Given its uncertain effectiveness as well as its potential to confound the results of the trial, intrastromal voriconazole injections were not included in the treatment algorithm and were not performed for any of the ulcers in the trial. Both groups were followed every 2–3 days. If the ulcers worsened or perforated, therapeutic keratoplasty was performed ( Supplemental Table , available at AJO.com ).
The prespecified primary outcome of the trial was treatment failure at 6 weeks following enrollment, defined as a composite outcome of perforation and/or increase in infiltrate size by ≥2 mm from the time of enrollment. A secondary outcome was uncorrected visual acuity measured at 6 weeks following enrollment, using the Snellen chart, by masked refractionists certified for the study. Ophthalmologists assessing the clinical outcomes were masked to treatment allocation.
The primary intention-to-treat analysis compared treatment failure at 6 weeks using a Fisher exact test at a significance level of .05. Other outcomes were compared with the Wilcoxon rank sum test for continuous variables and Fisher exact test for categorical variables. We estimated that enrolling 31 study participants in each treatment group would provide 80% power to detect a 33% absolute difference between the treatment arms, assuming a 5% level of significance and that half the ulcers in the control group would experience treatment failure. No stopping rules were prespecified at the outset of the trial. However, given the difference in the perforation rate between the 2 groups, the investigators opted to halt further enrollment and stopped the trial before all study participants were enrolled. No statistical tests were performed until after the investigators decided to end the trial ( Supplemental Figure 2: CONSORT flow diagram , available at AJO.com ).
The enrollment characteristics of the 13 study participants are shown in Table 1 . Overall, patients were middle-aged (median age 41 years, interquartile range [IQR] 39–56 years) with a slight male preponderance. The causative organism was fairly evenly distributed between Aspergillus species, Fusarium species, and unidentified hyaline and dematiaceous fungi. We found no statistically significant differences between the 2 treatment groups at enrollment, although the CXL group tended to have slightly larger infiltrates than did the non-CXL group (median infiltrate size 6.79 mm, IQR 6.71–7.48 in the CXL group vs 5.48 mm, IQR 5.00–5.70 in the non-CXL group; P = .12).
|Enrollment Characteristic||Treatment Group||P Value a|
|CXL N = 6||No CXL N = 7|
|Age, median (IQR)||39.5 (35–41)||56 (40–62)||.10|
|Male sex, n (%)||3||5||.59|
|Infiltrate diameter, mm, median(IQR) b||6.79 (6.71–7.48)||5.48 (5.00–5.70)||.12|
|Deepest filament, μm, median(IQR) c||353 (320–411)||360 (290–401)||.83|
|Hypopyon size, mm, median (IQR)||1.3 (1–3)||1.0 (1–3)||.82|
|Visual acuity, median (IQR)||HM (HM-HM)||HM (5/60-HM)||.31|
The prespecified primary outcome was treatment failure at 6 weeks, which was experienced by 5 eyes (95% confidence interval [CI] 35.9%–99.6%) in the CXL group and 4 eyes (95% CI 18.4%–90.1%) in the non-CXL group ( P = .56; Table 2 ). Therapeutic keratoplasty was used both for tectonic repair of perforation and for removal of infected tissue of worsening ulcers, and was performed for all 9 treatment failures and none of the 4 healed ulcers. When the individual components of the primary outcome were analyzed separately, perforations occurred in 4 eyes of the CXL group compared with none in the non-CXL group ( P = .02) and increase in infiltrate size >2 mm without perforation occurred in 1 eye in the CXL group vs 4 eyes of the non-CXL group ( P = .27). Of note, the eyes that perforated in the CXL group did not have an increase in infiltrate size before perforation. Visual acuity at 6 weeks was worse in the CXL group (median, hand motions) compared with the non-CXL group (median, 2/60), although this difference was not statistically significant ( P = .08). Compared with visual acuity at enrollment, visual acuity at 6 weeks had improved in 2 eyes of the CXL group and in 6 eyes of the non-CXL group ( P = .10).
|Outcome at 6 Weeks||Treatment Group||P Value a|
|CXL N = 6||No CXL N = 7|
|Treatment failure, n (%) b||5||4||.56|
|Perforation, n (%)||4||0||.02|
|Increased infiltrate size, n (%) c||1||4||.27|
|Visual acuity, median (IQR)||HM (6/60-LP)||2/60 (6/12-HM)||.08|
|Improvement in visual acuity d||2||6||.10|
|Lines visual acuity improvement, median (IQR)||0 (0–7.8)||3.2 (0.2–7.8)||.25|