Kasetsuwan and associates recently published a randomized clinical trial on the use of PACK (Photoactivated chromophore for keratitis) cross-linking (CXL) for moderate to severe infectious keratitis. The authors compared 2 groups of patients, receiving either standard antimicrobial treatment, or standard treatment and PACK-CXL. Main outcome measures were the size of stromal infiltrates at 30 days after treatment and the size of the epithelial defect. The authors conclude that PACK-CXL does not provide an adjuvant effect to standard antimicrobial therapy.
Considering the importance of this new field, we would like to highlight a number of aspects in the current study that merit further attention.
First, it seems that the statistical analysis is seriously flawed: The authors used a sample size of 15 eyes per group, leading to an expected power of 80%. However, it is surprising that the reported “median difference” does not reflect the actual difference between the “medians” of both groups. The authors should have reported the median deviation in order to verify if the initially estimated power was actually reached. The presented data give the impression of being underpowered. Also, one third (n = 11) of all cases yielded negative laboratory results, and should have been excluded from analysis, leaving the study with a mere n = 19 for both arms. For our own ongoing noninferiority multicenter trial ( Clinicaltrials.gov NCT02717871 ), we compare the effect of PACK-CXL alone vs the effect of antimicrobial treatment alone. This study setup should allow detection of differences more easily than the setup chosen by Kasetsuwan. Yet, the Department of Biostatistics of the University of Geneva has calculated an n of 252 eyes to be significant.
Second, it does not make sense to additionally separate and analyze subgroups of bacterial and fungal keratitis—which further decreases the statistical power (sample size of 5 or 10 eyes, respectively). At 7 days the standard deviation was even higher than at 30 days, and the actually required sample size would probably be in the range of 130 eyes. These results cannot be considered meaningful.
Third, the authors mention application of fluorescein stain prior to PACK-CXL to measure the size of the ulcer. We have shown unambiguously that fluorescein competes with riboflavin for energy at 365 nm, and the presence of fluorescein will reduce the effectiveness of PACK-CXL.
Fourth, the cross-linking settings used are effective only to a depth of 300 μm. Depth of the ulcers was assessed at the slit lamp, instead of proper assessment by anterior segment optical coherence tomography.
Fifth, the primary outcome should not be the size of the stromal infiltrate at 30 days, but rather the speed of epithelial closure.
Considering these major inconsistencies, we believe that the conclusions taken in this study are seriously flawed. PACK-CXL is a promising new field, and it is of utmost importance to collect and publish data using a solid study design. Multicenter trials with a large sample size are required to address this topic. This will also allow separate analysis of the efficiency of PACK-CXL in exclusively laboratory-confirmed bacterial and fungal keratitis.