Corneal crosslinking without epithelial removal

Abstract

Stulting et al.1 recently reported 2-year data from a proprietary epithelium-on (epi-on) corneal crosslinking (CXL) protocol clinical trial. They stated, “Epithelium-on CXL using this new protocol halted the progression of keratoconus and ectasia after LASIK. It was safer and provided more rapid visual recovery than CXL with epithelial removal…” We have serious concerns about the trial design and execution. Despite clear editorial guidance from this journalA and international standards bodies,2,B the trial was never registered with a clinical trials database. Nine other trials from the authors’ group (CXLUSA) have been registered on clinicaltrials.gov, some (unlike this trial) with an epi-off control arm. It is standard clinical practice to perform CXL on adults with keratoconus only when progression occurs. Instead of using Scheimpflug images to determine progression, the current study relied on medical records and a highly unusual definition of progression; that is, “progression was documented by…a progressive increase in myopic spherical equivalent (>1 D), cylinder (>1 D), or Kmax [maximum keratometry] (>1 D) in the previous 2 years…” Furthermore, only 56% of all patients were within this definition of progression and 44% were not (stable topography). Treating patients with nonprogressive keratoconus and then claiming treatment success seems pointless. Five hundred ninety-two eyes started the study. However, the rates of loss to follow-up and dropout were enormous, on the level of 39% to 41% year on year. After 24 months, only 22.4% of eyes present at the start of the study received the principal outcome assessments of uncorrected distance visual acuity (UDVA), corrected distance visual acuity (CDVA), and Kmax. Furthermore, 4 in every 10 patients failed to receive aberrometry assessments. Some analyses presented used a “consistent cohort” (115 eyes) comprising “eyes with preoperative and 3-, 6-, 12-, and 24-month examinations to avoid artifacts created by analyzing different eyes at different timepoints.” Although this is a more robust dataset, only 19.4% of the patients were enrolled. Progression was defined as “an increase of more than 1 diopter (D) Kmax and a loss of more than 1 line of CDVA in the same eye,” a measure that excludes patients who have progressed but can still read the charts. (We all have had such patients.) O’Brart et al.3 reported that patients can “gain” a line of vision over repeated assessments merely by learning the charts. This puts into perspective the statement by Stulting et al. that “the mean UDVA and CDVA improved by 1 to 1.5 Snellen lines at 1 and 2 years postoperatively (P < .0001).” The article states that the “mean Kmax decreased by 0.48 D at 2 years postoperatively (P = .0002)”; however, this value is within the inter-measurement variability of the instrument used (Pentacam, Oculus Optikgeräte GmbH).4 Their statement that “no eyes progressed” does not tally with the reported finding of “at 24 months postoperatively, Kmax increased by more than 1.0 D in 11 (8.3%) of 133 eyes…” The authors surely realize that their definition of progression after treatment is nonsensical because they used a different standard to determine progression before surgery. Despite that, if only 56% of the cohort had keratoconus progression at enrollment and 8.3% of eyes progressed by the common definition, this yields a failure rate of 14.8%, similar to every other published epi-off trial to date. Observations such as “no eyes progressed” (clearly, they did), “no vision-threatening events were observed” (1 case of hydrops that required keratoplasty), and “the new technique avoided the potential complications of epithelial removal for CXL” (30 eyes had epithelial defects) are, we believe, inconsistent with the data the authors present. Despite all of this, it is possible that the CXLUSA protocol could work. Per the basic tenets of the scientific method, if others were allowed to replicate their work (perhaps under a nondisclosure agreement to protect commercial sensitivities) to validate their claims, this would help alleviate many of our concerns. Unfortunately, every request to date (by us and others) has been refused. Ultimately, we believe that the claims made by Stulting et al.1 cannot be substantiated by this study because of serious flaws in the methodology.