Uddaraju and associates recently published a randomized clinical trial of collagen cross-linking (CXL) on nonresolving deep stromal fungal keratitis. The authors used the Dresden protocol settings (riboflavin 0.1%, ultraviolet A [UV-A] 370 nm at 3 mW/cm ² during 30 minutes) and concluded that the use of CXL for treating deep fungal keratitis resulted in an increased risk of perforation.

Considering the importance of this conclusion, we would like to highlight a number of inadequately evidenced assertions and experimental flaws present in this study.

The results presented within this study are based upon a randomized trial that was stopped early owing to 4 perforations in the CXL group. This problem obviously could have occurred by chance, as perforations are common in corneal fungal infections.

Randomization is an essential element of this study and is important to enable statistical analysis. It assumes reasonable homogeneity in the patient groups, which would have been highly unlikely, with wide differences recognized between such corneal infections. Using these forms of statistical analysis on small and highly heterogeneous groups would be at best prone to large confounding effects and could lead to spurious assertions.

In our opinion, applying the Dresden protocol settings is inadequate for this study. The Dresden protocol was developed primarily to ensure that corneas could be safely treated for keratoconus as long as corneal thickness was 400 μm or more. The absorption rate of 0.1% riboflavin shows logarithmic decay, with most of the energy absorbed in the first 100 μm. Therefore, even if the anterior cornea were perfectly transparent, which obviously is not the case in a deep fungal infection, the intensity of the UV-A light would not be sufficient to treat deep fungal keratitis. This will also reduce, in a similar fashion, the beneficial effect of CXL to reduce proteolytic enzymatic digestion of the corneal collagen.

The authors treated nonresponding fungal keratitis only after 2 weeks of medical treatment. This may have enabled the accumulation of fungal organisms throughout the cornea, including the deepest layers. Subsequent CXL treatment will then result in sudden release of foreign fungal antigen, inducing a short but marked host inflammatory response that potentially could transiently enhance corneal melting.

This study was prematurely stopped based upon 4 perforations, and one could not draw conclusions of either efficacy or lack of efficacy of cross-linking based upon the limited numbers assessed and methodology used. This paper does, however, offer the opportunity to highlight the fact that the conventional Dresden CXL protocol was never designed to target microbial keratitis. New and more effective antimicrobial CXL protocols need to be developed for the treatment of deep corneal keratitis.