Transepithelial CXL (“Epi-On”)

Recently, there has been much discussion regarding transepithelial crosslinking, or the epi-on technique. In epi-on crosslinking approach, the treatment is similar to standard CXL; however, we do not remove the surface epithelial cells (which are analogous to the tiles on a floor). Because those cells remain in place and don’t need to replicate and heal, there are a few potential advantages of the epi-on technique. First, leaving the cells in place may lead to faster visual recovery and earlier return to contact lenses. Also, since there is less disruption to the surface cells, the potential for infection and the potential for corneal haze may be somewhat less.

As seen in the pictures below, in the standard epi-off croslinking technique, the surface layer of cells on the cornea are removed.  Afterwards, they grow back from the edge in a centripetal pattern. Epithelial healing generally takes 4-5 days, compared to the epi-on technique where the cells are left intact leading to faster recovery.

However, there are some disadvantages. Most of these relate to the efficacy of the epi-on procedure compared to the standard technique.

  1. With the epithelial intact, it is more difficult to get complete riboflavin saturation into the cornea. Remember, the cornea absorbs riboflavin much like a sponge absorbs water. With the epithelium in place, riboflavin, a very large molecule, finds it more difficult to diffuse into the corneal tissue.
  2. The epithelium itself can act as a mask absorbing some of the incoming ultraviolet light. This may attenuate ultraviolet light energy deeper in the cornea and make the treatment effect more superficial and less deep and, thus, potentially less effective or longlasting.
  3. Oxygen from the air is required in order to perform a complete crosslinking reaction. With the epithelium intact, oxygen diffusion may be more difficult, and thus we may not get as much of a crosslinking effect.
  4. Although there may be less transient corneal haze formation with the epi-on technique, it’s not yet clear whether this haze really is an unwanted side effect of crosslinking or simply indicates an adequate crosslinking effect.

Here at the CLEI Center for Keratoconus, we have been studying trans-epithelial crosslinking in a formal clinical trial for several years (see below). Using the techniques of the study we are able to adequately achieve good riboflavin saturation into the cornea as seen in the picture below.

As in the standard crosslinking technique, we see a generalized improvement of the height of the keratoconic cone one year after epi-on crosslinking. Although It can’t be directly compared to the results that we have with standard crosslinking techniques because of study designs, it does appear that the results may be less robust using the epi-on approach.  Different crosslinking approaches may be suggested based on your particular problem.  A careful assessment of your ocular situation will allow us to make the best recommendation for your care.

It is important to know that only the standard crosslinking technique is FDA-approved. Epi-on is not an FDA approved technique. It can be done in a formal clinical trial or on an off-label basis after discussion of its risks and benefits. There are a number of clinical trials in the U.S. and internationally looking at other methods of epi-on crosslinking. These include new riboflavin formulations to improve uptake, methods which the ultraviolet power or adjust the ultraviolet dose, and ways to facilitate oxygen diffusion, such as intermittently pulsing the ultraviolet light or using supplemental oxygen around the eye during the procedure. We certainly look forward to results of these number of clinical trials and the evolution of crosslinking as time goes on. Certainly, the advantages and disadvantages of various crosslinking techniques will remain unclear until further clinical trial results become available.

From Dr. Hersh’s discussion of the CLEI Center for Keratoconus clinical trial of epi-on crosslinking for keratoconus:

Background of Transepithelial Corneal Collagen Crosslinking
Transepithelial corneal collagen crosslinking (CXL) has been suggested as an alternative to the standard crosslinking procedure (in which the epithelium is initially debrided). Putative advantages include faster healing, improved patient comfort, and less risk of corneal haze or infectious keratitis. 13-16 In addition, this technique may decrease corneal thinning during the CXL procedure and allow treatment of more severe cases in which corneal thickness may otherwise preclude treatment.17
This single center, randomized controlled clinical trial is one of the largest to date designed to analyze safety and efficacy outcomes of transepithelial collagen crosslinking. In addition, 2 dosing regimens (every 1-minute or 2-minute riboflavin administration during UVA treatment) were compared to determine any advantages or disadvantages to procedure outcomes.

Clinical Barriers to Transepithelial Crosslinking
Though of potential clinical advantage, there are several theoretical and practical hurdles to transepithelial crosslinking being as effective in mitigating keratoconic progression as standard CXL. First, the epithelium itself is a barrier to diffusion of riboflavin, a large molecule, into the corneal stroma. There are evolving methods, however, to enhance diffusion, such as those we have used in this study. These include eliminating dextran from the riboflavin carrier, treating with adjunctive benzalkonium chloride containing solutions or other permeability enhancers, and improving contact of the riboflavin with the cornea by means of a soaked pledget.
To date, dextran has been included in most riboflavin formulations which are used clinically, based on its documented efficacy in the standard CXL procedure with the epithelium removed.10 However, inclusion of dextran in the riboflavin solution substantially diminishes its ability to penetrate the epithelium. In a study of rabbit eyes using riboflavin in a dextran solution,13 transepithelial crosslinking was less effective than standard CXL. In other laboratory research using riboflavin in dextran solution,18,19 stromal concentrations of riboflavin failed to reach levels sufficient for effective crosslinking. In contrast, riboflavin in solutions without dextran seem to facilitate penetration through the epithelium20,21. In further efforts to improve the epithelial permeability to riboflavin diffusion, adjunctive agents such as benzalkonium chloride have been shown to be advantageous.22-25 Similarly, ethylenediaminetetraacetic acid (EDTA) can increase permeability to hydrophilic molecules such as riboflavin.26,27 Novel methods to facilitate riboflavin uptake such as iontophoresis are also being developed.28,29
Notwithstanding adequate stromal absorption of riboflavin, other factors may mitigate the crosslinking effect in the setting of an intact epithelium. First, both the surface riboflavin film30 and the riboflavin-soaked epithelial layer may absorb incident UVA light, attenuating the UV power/stromal depth relationship. Consequently, this may decrease the depth at which the threshold power for a crosslinking effect is met and the actual crosslinking would be less deep and robust than with the epithelium absent. Supporting this thesis is evidence that cytotoxic keratocyte damage is restricted to a more anterior, approximately 200 um, stromal depth in the transepithelial crosslinking approach.13,14 To diminish attenuation of UV power by surface riboflavin and riboflavin-soaked epithelium, it has been suggested that the riboflavin within the epithelial layer be removed before UV exposure by rinsing the corneal surface, and then suspending further riboflavin administration during the UV administration.31 Although this was not specifically studied in our investigation, the potential advantage of this technique is belied by our findings of more robust improvements in maximum K and uncorrected visual acuity in the every 1 minute compared to 2-minute riboflavin administration randomized subgroup; that is, there was a trend to better outcomes with greater, not lesser, administration of riboflavin during UV exposure.
In addition to the UV-barrier effect, the epithelium may diminish oxygen diffusion into the corneal stroma, further attenuating the crosslinking effect by limiting the crosslinking reactions that occur through oxygen-dependent pathways.32 Investigators are studying the effects of pulsed-light protocols to allow greater oxygen diffusion as well as oxygen supplementation as modalities to enhance crosslinking in transepithelial procedures.33
Finally, as noted in our previous work on standard crosslinking, the role of wound healing in the ultimate clinical effect of the procedure is unclear2. After the standard CXL procedure, there is a typical corneal stromal haze and demarcation line that follows a generally consistent time course; haze appears and is maximum at 1 month through 3 months and diminishes to baseline over the course of a year16. Indeed, haze appears to be a normal concomitant of the standard crosslinking procedure, and is first observed as dust-like change in the anterior corneal stroma which evolves into a mid-stromal demarcation line.34 Crosslinking associated corneal haze is most likely a measure of back-scattered and reflected light, causing decreased corneal transparency, and likely demarcates the depth of the actual crosslinking effect.35 This may be an effect of biologic changes such as keratocyte apoptosis, keratocyte repopulation, and changes within the collagen fibrils and surrounding glycosaminoglycans. Such haze is not seen, generally, in the transepithelial procedure, a finding touted as an advantage of the latter. However, whether stromal haze is, indeed, an unwanted side effect of CXL or, rather, is a proxy for beneficial stromal healing which may enhance a clinically effective crosslinking effect is unclear.

Literature on Transepithelial Crosslinking
Published results of transepithelial CXL efficacy have shown both positive and negative results of the procedure. They are difficult to compare since they are not uniform in design; different riboflavin formulations and adjunctive agents have been reported, different dosing regimens have been used, and varied techniques have been promulgated. For instance, Caporossi and associates26, using riboflavin/dextan with EDTA and trometamol as permeability enhancers, reported initial improvement in vision after transepithelial CXL, but a subsequent worsening of the maximum K value at 24 months. Gatzioufas and colleagues36 used riboflavin in a 0.01% BAC-containing solution and a 10 minute exposure to 9 mW/cm2, and reported progression in 46% of eyes. Koppen and associates37, using riboflavin containing dextran and BAC, found improvement in CDVA, but worsening of the maximum K. Demonstrating more positive results, Filippello and coworkers27 found general topography improvement after tranepithelial crosslinking using riboflavin/dextran, EDTA and trometol. In another study, Stojanovic’s group21 reported improvement in the UDVA, CDVA, and maximum K value after transepithelial crosslinking aided by superficial mechanical disruption of the epitheliuim. In a pediatric population, Salman38 found improvement in UDVA and maximum K value; they used riboflavin/dextran with EDTA, BAC, and trometol. Leccisotti and Islam39, using riboflavin/dextran and gentamicin, BAC, and EDTA, reported an improvement in the CDVA and mean K value. Rechichi and colleagues.40 used a corneal disruptor device to create and found a statistically significant improvement in UDVA and CDVA, and improvement in the mean simulated K value and steepest simulated K value over 12 months.

Study Results
In our study, we used riboflavin 0.1% solution without dextran and adjunctive administration of proparacaine with BAC to increase the permeability of the epithelium. In addition, we found that the application of an 8 mm pledget soaked in riboflavin for 15 minutes improved riboflavin uptake, likely a result both of enhanced contact as well as mild trauma to the epithelium. Although most eyes achieved saturation in one hour, some required up to 80 minutes for complete saturation.
We used the maximum K value on corneal topography as our primary efficacy outcome. Maximum K was chosen since it measures a salient feature of keratoconus; ie the steepness of keratoconic topographic distortion. Moreover, topographic maximum K affords an objective, quantitative endpoint. Overall, maximum K decreased by an average of 0.45 D over 1 year, a clinically modest, but statistically significant finding. Looking at individual eyes, most eyes remained stable whereas 13% improved by 2 or more diopters and 5% increased by 2 or more diopters; 15% increased by 1 or more diopters. Although, these latter eyes might be considered treatment failures because cone progression was not stabilized, it is unclear what the natural evolution of the disease might have otherwise revealed; it is possible that disease progression was slowed but not completely, or that progression, indeed, proceeded apace.
In addition to the primary efficacy measurement of maximum K, changes in CDVA may point to additional efficacy or, conversely, to safety concerns after crosslinking. In this study, there was no change in average CDVA over one year. Given the general improvement in maximum K and improvement in UDVA, we might have expected improvement in CDVA as well. However, the topography improvement, though significant, was modest and, thus, may not have manifest a clinically significant effect on corrected vision. Among patients receiving CXL, 24% gained 2 or more lines of CDVA, whereas 13% lost 2 lines or more. Curiously, of the 11 eyes with CDVA loss, maximum K was stable in 7, improved by 2D or more in 2 eyes and steepened by 2 diopters or more in 2 eyes. Thus, there was no obvious association of loss in CDVA with continued topographic progression. If we use the combined outcomes of continued worsening of maximum K associated with loss of CDVA, 2 eyes (2.4%) in the study can be considered to have suffered advancing disease even after the CXL procedure. Retreatment was suggested for these two patients, but they deferred and were lost to subsequent followup.
Of clinical note, this technique was typically associated with substantial punctate keratopathy of the epithelium after the procedure, and with frank epithelial erosion in 10 eyes. Because the native corneal epithelium acts as a barrier to diffusion of a large molecule such as riboflavin, the essential challenge of transepithelial crosslinking is to chemically or mechanically effect a microtrauma to the epithelium to enhance its permeability. Thus, it is likely that the lengthy time of riboflavin administration, chemical manipulation with BAC, and the minor trauma secondary to pledget application leads to transient epitheliopathy or even frank epithelial erosion in a number of cases. Although patient subjective pain was not prospectively assessed, 29 patients (35%) reported discomfort at the one day postoperative visit; the therapeutic contact lens was retained in these patients. Thus, a number of patients in this study were continued with a therapeutic lens for a total time similar to that of standard CXL with epithelial removal.

Comparison of Transepithelial Crosslinking to Standard CXL
One of the major debates in keratoconus management is the relative efficacy of transepithelial crosslinking to the standard procedure. Therefore, although not directly comparable because of study design and entry criteria, it is of interest to compare our transepithelial CXL results with our clinical results of standard CXL with removal of the epithelium. In our previous study of CXL outcomes,2 maximum K flattened by 2.0 D after one year compared to 0.45 D in the current study, suggesting that the standard procedure is more efficacious. However, the study cohorts for these two studies have important differences. First, the previous study was restricted to patients with documented progression of keratoconus over the prior 2 years compared to the current study in which all keratoconus patients could be entered. Second, the average maximum K value of the entry population in the current trial was 57.0 D, compared with 60.4 diopters in the previous study, a difference which may impact expected outcomes. In previous work41, we have shown that preoperative maximum K is independently associated with greater 1 year postoperative improvement in maximum K; that is, eyes with worse KC tend to have a more robust topographic flattening response to CXL. These population discrepancies, therefore, may account for some of the outcomes differences between the two studies. However, maximum keratometry did worsen by 2 or more diopters in only 1 of 49 eyes (2%) in the previous study compared to 4 of 82 eyes (5%) in the current transepithelial study, and it improved by 2 or more diopters in 17 of 49 eyes (35%) in the previous study compared to 11 of 82 eyes (13%) reported herein, suggesting that there is, indeed, less efficacy of this transepithelial approach compared to standard CXL. Similarly, there was no significant change in mean keratometry in our study population, compared with a decrease of approximately 1 D found in our previous trial of standard crosslinking.2 This, similarly, may be attributable to a somewhat less robust effect from a transepithelial procedure. Also, the mean K, which is measured from the central cornea, is generally a flatter area since maximum K is typically eccentric in keratoconus42 – reiterating our previous findings41, flatter corneas tend to have less of a flattening effect from CXL, so this could also contribute to the finding of no change in mean K in our study.
Clinical trials have directly compared transepithelial to standard CXL, typically finding more robust results in standard treatment groups. Soeters et al performed a randomized trial in 61 patients comparing a standard to a transepithelial protocol.43 They found one year topography stability in the transepithelial group compared to significant flattening of 1.2 to 1.5 D in the standard group, but also noted that 23% in the transepithelial group showed progression of maximum K after one year, compared to no longer term progression in the standard group. In a randomized study comparing standard CXL with iontophoresis-assisted crosslinking, Bikbova and Bikbov found better stabilization and regression of keratometry values in the standard group.44 Of note, the average depth of the demarcation line was 292 um in the standard group compared to only 172 microns in the transepithelial group. Rush and Rush45, in a randomized trial of standard versus transepithelial crosslinking using an enhanced riboflavin solution found improvement of 1.52 diopters in the steep keratometry reading in the standard group, compared to only 0.54 diopters in the transepithelial group. Cerman and colleagues46, similarly, found more improvement in maximum, flat, and steep keratometry with standard as compared to transepithelial treatments.
In addition to ultimate outcomes differences, there is a difference in the timecourse of wound healing comparing the transepithelial technique to standard crosslinking with epithelial removal. In standard CXL, all 3 outcomes indicators – maximum K, CDVA, UDVA – tend to worsen at one month, resolve to baseline at 3 months, and improve thereafter.2 This tempo reflects healing responses to standard crosslinking as evidenced by the natural history of crosslinking associated stromal haze16. Both epithelial and stroma wound healing mechanisms may account for this postoperative timecourse. The epithelial thickness profile in native keratoconus typically shows a doughnut pattern; the epithelium is attenuated over the cone apex and thickened paracentrally.47,48 Thus, the epithelium tends to mask and mitigate the stromal cone of KC. With corneal de-epithlialization in the standard CXL procedure, the more profound stromal irregularity is revealed, with an increase in measured maximum K. As the epithelium heals and remodels, topography improvement is seen. Stromal healing, too, may play a part in the timecourse of outcomes in the standard crosslinking procedure. In vitro and ex vivo studies have shown that standard collagen crosslinking leads to an almost immediate loss of keratocytes in the corneal stroma.49 Confocal microscopy in patients with keratoconus shows that activated keratocytes repopulate the corneal stroma starting at two months, and stromal repopulation is almost complete at six months.50
As seen in our study of transepithelial crosslinking, the standard post-CXL timecourse showing worsening of outcomes at one month is not seen. This is a clinically advantageous finding. Likely, it is related to the retention of the epithelial layer, with a much diminished epithelial healing response; it may also reflect altered stromal responses since epithelial removal may also militate stromal healing reactions. So, while one of the purported advantages of transepithelial crosslinking is the general lack of a stromal haze response, it remains unclear if the biologic and wound healing reactions signaled by the formation of stromal haze are actually unwanted or are, indeed, beneficial to the clinical topographic and disease stabilization effect of the crosslinking procedure.

Comparison of Dosing Regimens
We also looked at riboflavin dosing regimens during the UV portion of the treatment – either every one minute or every two minute riboflavin administration during the 30 minute UV administration. There are potentially conflicting advantages and disadvantages to greater or lesser riboflavin dosing during UV administration. More frequent dosing may retain a higher riboflavin concentration in the cornea and thus facilitate more robust crosslinking. Conversely, the surface riboflavin film and the riboflavin within the epithelium may partially mask the incoming UV and dissipate the power transmitted to the stroma, mitigating the crosslinking reaction.30
When directly comparing the 2 treatments, statistical analysis showed no significant differences in the changes in maximum or mean K values, CDVA, or UDVA between eyes randomized to riboflavin administration either every 1 or 2 minutes. However, in within-groups analyses, only the 1-minute cohort showed statistically significant improvement in maximum K. The one minute group improved by -0.73 diopter, compared to the two minute cohort which improved by only -0.14D. Regarding this finding, it should be noted that there was a significant difference in preoperative mean maximum K between groups; the one minute group, preoperatively, was significantly less steep (55.2 D) than the two minute group (59.2 D). Again, in previous work we have shown that steeper corneas tend to flatten more robustly with crosslinking.41 Thus, based on this, the two minute subgroup would have been expected to have a more substantial topography change, simply based on the steeper preoperative maximum K. That this was not found in the results supports our finding that the more frequent dosing regimen, indeed, leads to more robust results. While this certainly is not a dispositive finding with regard to riboflavin dosing in the transepithelial technique, it does belie the suggestion that rinsing the surface of riboflavin and adding no more riboflavin during the UV phase will improve outcomes by diminishing epithelial screening of the incoming UV power.31 Further investigation on this front is warranted before any clinical conclusions can be drawn.
A limitation of this study is the lack of a control group, or treatment group with epithelial removal. Further controlled clinical trials evaluating different transepithelial protocols are essential to optimize the procedure. Variables such as riboflavin formulation, dosing regimen (including the option of adding no additional riboflavin during UV administration), UV power and time of administration, and oxygen availability may impact the outcomes of the transepithelial procedure. Moreover, when interpreting our results, it is important to note that inclusion criteria for this study admitted all keratoconic patients; although patients were chosen in whom crosslinking was felt to be indicated (that is, those patients with documented progression or at risk of progression), actual documentation of progression was not within the inclusion criteria. Because we had found, in our previous work, that patients with steeper corneas have a greater likelihood of topography improvement, study criteria did not exclude patients without documented progression in whom the likelihood of clinically significant topography improvement suggested a treatment benefit. Also, given the longterm nature of keratoconus progression, followup studies of greater than one year are essential in assessing the ultimate efficacy of this procedure, particularly since some studies of transepithelial crosslinking have demonstrated loss of effect over time.43 And, importantly, further controlled trials comparing the transepithelial to the standard approach with epithelial removal are necessary to determine the relative risks and benefits of each.


Picture of man with Doctor

Thank you so much Doc! When I first came to you I was 20/50 now I'm 20/10. I'm ready to have a great year!

Randy Foye
Pro Athlete, Utah Jazz #8
Female Boxer

My experience at The Cornea and Laser Eye Institute has been amazing. All of the staff have been helpful and courteous...

Phd Student, MMA fighter
Man with Doctor

Dr. Hersh and his staff were excellent. I was comfortable from the time I walked in until the time I left after the procedure.

Luke Petitgout
Pro Athlete, New York Giants Football