February 01, 2014
17 min read

Cross-linking round table, part 2: Epi-on vs. epi-off and measures of success

You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact customerservice@slackinc.com.

At the 2013 European Society of Cataract and Refractive Surgeons meeting, Ocular Surgery News convened a round table of international experts to discuss the current state of corneal cross-linking. The second part of that wide-ranging discussion, moderated by Roy S. Rubinfeld, MD, is featured in this issue of OSN.

Roy S. Rubinfeld, MD: Let’s talk about epi-on vs. epi-off in terms of safety. At this conference, there are numerous abstracts and presentations showing some proprietary new epi-on cross-linking techniques to have efficacy similar to older epi-off techniques, but when there is a complication, say after a perforation, after epi-off, it is simply not pretty.

William B. Trattler, MD: I spoke to a surgeon who shared the details of a serious complication with epi-off cross-linking that occurred in Australia, which appears to be a complication from a compounding pharmacy that used the wrong riboflavin formulation. It was not related to the UV light but rather due to an improper formulation of the riboflavin formulation.

Roy S. Rubinfeld, MD

Pulsed light may be an option to facilitate oxygen diffusion into the cornea during cross-linking, according to Roy S. Rubinfeld, MD.

Image: Salas D

Michael Mrochen, PhD: Outside of the U.S., a high majority of surgeons mix the riboflavin themselves because they want to save money. The severe complications we see, as a manufacturer for cross-linking, are frequently caused by riboflavin that is not CE marked or approved by any authority.

Arthur B. Cummings, FRCS: So this is not normal.

Trattler: Definitely not normal. The challenge for epithelialization is you want to treat the patients with steeper corneas, but the steeper the cornea, the longer it takes for the epithelium to heal.

George O. Waring IV, MD: In the U.S. Food and Drug Administration trials for epi-off cross-linking, we naturally used bandage contact lenses postoperatively. When you have a very steep cornea, the bandage contact lens can act like a mechanical barrier at the apex of the cone and prevent that last bit of epithelialization. So one of the tricks was to take off the bandage contact lens earlier to allow complete re-epithelization.

Trattler: Is there any reason why epi-on should not be as good as epi-off? If we could optimize the riboflavin concentration in the cornea and the UV light enters into the eye, why can’t it be the same? Why should there be a difference?

Roundtable Participants

  • Roy S. Rubinfeld, MD 
  • Moderator

  • Roy S. Rubinfeld, MD


  • Jorge L. Alio, MD, PhD
  • Jorge L. Alió
  • Arthur B. Cummings, FRCS
  • Arthur B. Cummings
  • A. John Kanellopoulos, MD
  • A. John Kanellopoulos
  • Michael Mrochen, PhD
  • Michael Mrochen
  • Karolinne Rocha, MD
  • Karolinne Rocha
  • William B. Trattler, MD
  • William B. Trattler
  • George O. Waring IV, MD
  • George O. Waring IV


Mrochen: There are three factors. One is the UV light dose. Second is the riboflavin concentration in the stroma. And third is the oxygen available for a reaction. If you have riboflavin in the epithelium, you block part of the UV light and get less in the stroma. If you use a higher concentration of riboflavin, more is absorbed in the epithelium or in the upper part of the stroma, and you get less cross-linked tissue. Modifying the riboflavin concentration is risky, and slit lamp investigation is needed to ensure consistent riboflavin distribution in the stroma. When you apply the UV light, oxygen is consumed, and the diffusion rate of oxygen is pretty slow. When the epithelium is on, it seems that there is an additional barrier for oxygen to go back into the cornea. So not only do you have less riboflavin in there, but also there is less diffusion of oxygen going into the cornea during an epi-on treatment.

Trattler: Do you believe that if you “beat up” the cornea a little bit, make it more permeable to riboflavin, that would have an effect on allowing oxygen into the eye?

Mrochen: For me, all epi-on treatments are partial epi-off treatments by breaking up the epithelium. If you change the osmolarity so that the epithelial cells swell and break up, you get more diffusion there. If you take an excimer laser and do 40 µm phototherapeutic keratectomy, you remove the upper layers. If you take any mechanical component, a vibrator or iontophoresis, they all destroy, to some degree, the epithelium.

Jorge L. Alió, MD, PhD: We have published a real epi-on technique in which you inject riboflavin in the tunnels. You have only one line of femtosecond laser incision, which is in the periphery. When you put the riboflavin in the tunnels, immediately, within seconds, the whole cornea is impregnated with riboflavin. Simply speaking, you are injecting a hydrosoluble substance in a hydrophilic environment. The impregnation is automatic, the epithelium is untouched, and you have cross-linking.

We have been observing in this channel injection that we have a demarcation line that is about 300 µm, which is a little bit less dense but still an evident demarcation line. Using confocal microscopy, you see evidence that you have a good, active healing process. Epi-on works. The problem is making sure that it is not toxic to the epithelium and not risky for the patient. Complications in my hands have been basically infiltrates and infection. I had a lens with one infection with gram-positive cocci.

Karolinne Rocha, MD: What is the average depth of the demarcation line observed with this technique compared with the standard epi-off treatment?

Alió: We have not done too many epi-on cases. We only do this procedure when we implant rings at the same time because then we have a tunnel. We have experience with pockets and with segments. Segments are done with Kerarings (Mediphacos) or Intacs (Addition Technology) and pockets created for the implantation of MyoRings (Dioptex). I do not think that pockets are the best solution because we disrupt the lamella. But in the periphery, to make a 6- or 7-mm femtosecond corneal tunnel seems to be a good technique to get the riboflavin into the corneal stroma without disrupting or removing the corneal epithelium and to achieve cross-linking. My work is published in Journal of Refractive Surgery.

Demarcation line

Waring: There is also probably a spatial component to the epithelium. Not only is there an anatomic and physiologic barrier, but also a variable thickness and therefore further distance to travel. There is a reasonable body of evidence to suggest that demarcation lines are deeper in epi-off. In epi-on, the demarcation lines are more anterior and can be less apparent. We have demonstrated in our lab that the effect of cross-linking is stronger in the anterior stroma than it is in the posterior stroma. Anterior stroma and posterior stroma behave like two different corneas, and this is due to the differences in organizational structure. Is it more important to cross-link the stronger cornea, the anterior cornea, or the posterior cornea where keratoconus starts, the weaker part? Presumably with epi-off, you may be able to get a deeper treatment to the weaker cornea that needs it more.

Trattler: You want to show improvement in corneal shape and improvement in vision, but the demarcation line may be a sign of toxicity, an area where there is not enough oxygen. So what is the demarcation line, and why is it important?

Waring: I think the demarcation line is a treatment effect of some sort, whether it is bad or good. It is probably a collection of dead keratocytes.


Alió: The demarcation line is the only evidence we can see on confocal microscopy that the treatment is working, but we do not know how dense it should be.

Rubinfeld: So, the three key elements required for cross-linking are oxygen, riboflavin, and UV light, and the goal is to produce singlet oxygen, which stimulates cross-linking. If you have low oxygen and high UV, you produce mostly hydrogen peroxide and other reactive oxygen species that kill cells and produce demarcation lines, thus showing that treatment is working. But is this pathway of killing keratocytes necessary?

Mrochen: No one wants to kill keratocytes, but keep in mind that any LASIK or PRK procedure leads to loss of keratocytes within the stroma.

Rubinfeld: Oxygen diffusion into the cornea, while the UV light remains on, is very slow. But if you pulse the light so that molecular oxygen can diffuse from the ambient tear film and the surface of the eye into the cornea, then you preferentially expedite the pathway of producing singlet oxygen, which is safe, effective, controlled cross-linking, with no demarcation line. Isn’t the goal here to produce singlet oxygen to get amides to convert to aldehydes and strengthen corneas?

Mrochen: In terms of the demarcation line, newer technologies that have entered the market that produce some 100 µm of demarcation line, rather than the 300 µm or 250 µm demarcation line of earlier techniques, are associated with less clinical success. So, some people today would say, “If I cannot see a demarcation line, then nothing happened in the cornea, and so I don’t have an effect.”

Rubinfeld: Rather, it seems that they are saying, “If I don’t see a negative side effect of killing cells, then I haven’t had a positive effect, that is, strengthening of the cornea.”

Mrochen: What we lack is a real measure. We need another pathway to evaluate the response clinically.

Waring: Excimer lasers also kill keratocytes, and with 28 million excimer procedures performed worldwide, we know that this is OK.

Mrochen: Yes, 100 µm to 150 µm.


Rocha: Dr. Salomão and Dr. Wilson have shown that very little myofibroblast generation is detected after corneal collagen cross-linking. After killing all the keratocytes, there is fibroblast activation and a transient, mild fibroblast generation for a few months. This study was published in the Journal of Refractive Surgery in June 2011. The corneal haze observed after cross-linking is different from the haze seen after PRK. The stromal haze associated with cross-linking is normally related to transient corneal fibroblast generation, and in most cases it progressively decreases at 3 to 12 months and does not affect the clinical outcomes.

Persistent stromal haze after cross-linking has been correlated to thin corneas, advanced keratoconus, steeper keratometry and an older age.

Mrochen: I understand haze in refractive surgery, but for these keratoconus patients, is the haze really an adverse event?

Alió: That depends on the level. You have heavy healers, and these people have a scar.

Cummings: These patients with heavier haze, though, often end up with much more flattening. There is some correlation there. The heavier the haze is, the more flattening you are going to see.

Waring: At about 1 month after epi-off treatment, I expect to see the opacity of the demarcation line, and that is a good sign. That means I can expect clinical success. That is different from a true haze, however. Then there is a deep haze — the deep opacities that do not go away. I have observed an interesting variation of this with a deep ring-like stromal haze sparing the central cornea and a massive flattening effect, and the patient did great as a result of the Intacs-like flattening. This response has since been published by others in a case report.


Measure of success

Rubinfeld: Can we all agree here that cross-linking can be defined as “effective” as a standalone procedure when it stops the progression of vision loss and the progression of corneal steepening? Some patients also respond with mild improvement in vision. In general, in keratoconus, cross-linking is fantastic for early cone. But we all know this. You cannot really figure out what you have done unless you do a differential map. So how do we know it worked?

Trattler: There is a lot of variability in measurements in a steep cornea. Using a Pentacam (Oculus), variability increases as the cornea gets steeper, even when multiple measurements are obtained over a span of a minute or two.

Mrochen: For those steep corneas, the only way to really get measurement is with Placido.

Trattler: The problem with Placido is that the image breaks up when the cornea gets that steep. I use a Magellan topographer (Nidek), and when the cornea is more than 65 D to 70 D, the image is completely distorted and not helpful or repeatable.

Rubinfeld: I have the Zeiss, and I see similar problems. You get poor data in very steep corneas with any of the usual scanning technologies.

Mrochen: Yes, the moment you get steeper than 68 D to 70 D, you lose it completely on Placido. You see a small island. So how can we compare clinical data outcomes based on corneal topography if we get no readings from any of the devices, not from Scheimpflug or from Placido?

Rocha: At a previous ESCRS meeting, I presented maximum keratometry (Kmax) repeatability data for mild, moderate and severe keratoconus. Significant variability was seen in same-day Kmax measurements, especially in severe cases. Since then, we have suggested a zonal analysis in these eyes, and that minimizes the problem.

The corneal power in a zone centered at the Kmax location was the most reproducible metric for evaluating patients with moderate to severe keratoconus.

Waring: Dr. Rocha has actually shown this in an abstract, that with the zonal analysis you get less variability. And then Cynthia J. Roberts with her C-spot work has also proposed zonal analysis that has excellent reproducibility.

Mrochen: In general, for these measurement principles, one relies on the assumption that the cornea is spherical, which is usually not the case. How can we possibly set validated criteria if the systems we are using to do that are not designed for these patients?

Cummings: We have started assessing a new device called Cassini (i-Optics), a multi-spot, multicolor light-emitting diode (LED) tear film-reflection imaging technology. I have been looking at it mostly in post-refractive cases. Dr. Kanellopoulos has been looking at it in keratoconus cases. I certainly think for steep cones we are getting better data with the Cassini. It has multiple colored LEDs directed from seven different angles with a number of cameras. You do not get the rotation effects of the Scheimpflug, and there is no conversion from elevation. It looks promising.


A. John Kanellopoulos, MD: The metrics that we are using to establish keratoconus diagnosis and progression are, to say the least, primitive. In the U.S., for every keratoconic patient an ophthalmologist sees, there are about 50 that optometrists see. We looked at several metrics — visual acuity, steepest keratometry, corneal thickness — and we found that there is poor correlation as far as diagnosing keratoconus and establishing progression. So it is important, both bibliographically and practically, to establish what progression is, in order to say that it has progressed or not. There are anterior curvature symmetry indices available on the Pentacam, and on some other topographies as well, that show very sensitively the level of asymmetry. And we have found that they are far more sensitive than keratometry and pachymetry for picking up changes in the cornea. The Cassini, a multi-color spot reflection topography, may be better, particularly for keratoconus, because in the advanced cases in which you have Bowman’s degradation, the Pentacam may be biased and not very precise. When cornea clarity is compromised, Placido-based topography and/or Cassini may be better. On another note, when the cone is very central, Placido is “blind” in the center. So the Cassini may offer the best of both worlds in such cases.

Alió: I think you have brought up an important issue. Is Pentacam the adequate machine for keratoconus? Probably some newer technologies do a better job. We have experience with the Sirius CSO aberrometer. The corneal aberrometries are a much more mathematically and technically effective method of follow-up. And I think that Pentacam does not offer the quality of the image that I see with a Sirius, which offers more data about the two corneal surfaces.

Kanellopoulos: The OCTs may actually bring higher accuracy to this application and field of study. I am aware that Opto-Vue may be coming out with an anterior and posterior curvature map. This is produced with 25,000 shots a second, which some spectral-domain OCTs can offer nowadays. As far as detecting changes, that will turn out to be the ultimate keratoconic criteria. It is a direct measurement and unobstructed view of the interface fluid with Descemet’s. It will be a Pentacam on steroids.

Figure 1 

Figure 1. Slit lamp image of non-homogeneous, limited stromal riboflavin loading with previous transepithelial riboflavin formulation. Upper (yellow) arrow shows an area of limited stromal loading. Lower (white) arrow shows little or no riboflavin in the stroma.

Images: Rubinfeld RS

Figure 2 

Figure 2. Slit lamp photo of excellent, homo-geneous epithelium-on riboflavin stromal loading with a new proprietary (CXLO LLC) transepithelial riboflavin formulation after 15 minutes of loading.

Figure 3 

Figure 3. Cobalt blue slit lamp photo of homogeneous, transepithelial (epi-on) proprietary CXLO riboflavin formulation. Arrow shows limited riboflavin in the intact epithelium, which allows UVA to penetrate into the well-loaded stroma.


Rubinfeld: Are you suggesting that OCT will be able to give you a biomechanical measurement?

Kanellopoulos: Once you cross-link a cornea, you lose the game of measuring keratic refractive changes because whenever we cross-link we get an effectiv.e refractive change. We change index of refraction as well as the curvature. We cannot achieve equal cross-linking everywhere, so we create cross-linking differentials. We are doing some work with Avedro using specifically topography-guided manipulation of these differentials in cross-linking and getting dramatic 3 D changes, steepening and flattening, and toric changes, while also focusing on specific irregularities. But once you cross-link, it is very difficult to assess the level of cross-linking.

Rubinfeld: So are you saying that the Pentacam may be a good diagnostic device, but in terms of measuring the treatment, it does not work very well?

Kanellopoulos: Once the cornea is cross-linked, it may be off, as the change in cornea clarity may affect its accuracy. But it is a brilliant tool, and the Pentacam is still my No. 1 tool for qualitatively, not quantitatively, evaluating keratoconus.

Rubinfeld: So some devices are very good at diagnosis but not that good at measuring effect of treatment.

Kanellopoulos: We do not have good criteria of what keratoconus is. Let’s start with even a more basic fact. Kmax is irrelevant for assessing progression. By definition, the steepest keratometry measurement in keratoconus is not going to be in the center. So what happens as it progresses is the cone steepens but the central part of the cornea flattens. It is very deceiving, measuring the central “steepest” part of the cornea when the progression itself will distort it. Furthermore, most ophthalmologists may not be very familiar with how to read each different kind of corneal topography. Having topography-guided technology released in the U.S. with FDA approval may become challenging because most people do not fully understand the maps, in my opinion. Most clinicians will not be able to tell the difference between Placido-derived topography and Pentacam-derived topography. They are both color maps. They see a cone. They do not make the clinical correlation whether this cornea is opaque, is dry, has scar, has degradation. Nevertheless, ideally the newcomers in this field may learn from our mistakes and innovations to use cornea maps and “zap” them on the cornea. I nevertheless think there is a learning curve there to be tackled.

Alió: We prefer to work on data derived by corneal aberrometry to quantify a corneal topography in keratoconus cases. There are too many elements involved. The evolution of the technology has come faster than the medical education, so we need to educate in the subjective interpretation of corneal topography. We have the technology, but many doctors do not have the adequate understanding about the interpretation of corneal topographies in special cases.

Waring: We sent out 15 patients’ topographic maps in a masked fashion to a group of international experts. Half the cases were natural-occurring ectasia, and half were iatrogenic after laser vision correction. When asked to identify which was which, few cases were able to be distinguished.


Mrochen: A surgeon recently wrote me that he wanted topography-guided treatment for a patient. I forwarded the data to several friends, and I got back four completely different opinions from four experienced corneal refractive surgeons.

Invasive pharmacology

Rubinfeld: If we could determine that you could definitely get sufficient riboflavin, UV light and oxygen into the cornea without weakening the cornea with a mechanical approach, would epi-on be more popular if data showed equal efficacy? Epi-on is clearly safer with less discomfort and much faster visual recovery, about a day. Would it become the standard of care?

Kanellopoulos: The way we are cross-linking is primitive, in my opinion. Nowhere in medicine are we trying to increase permanence in biomaterials by making them “titanium” strong. The cornea has to move. Seeing all this sophisticated movement, we have to try, in my opinion, cross-linking patterns and look not only at the new strength achieved for the cornea, but also its ability to move and oscillate under the physiologic changes. Because if the cornea is not moving, then that dynamic activity may be transferred elsewhere and affect other aspects of the eye physiology.

Alió: This is a nondiscriminating treatment now, but I am sure that we can selectively activate drugs and finally create a much more selective treatment. This is one of a new series of treatments that could be included into the term of “invasive pharmacology of the cornea.”

See part 1 of this round table at Healio.com/Ophthalmology.

Alió JL, et al. J Refract Surg. 2011;doi:10.3928/1081597X-20110705-01.
Kanellopoulos AJ, et al. Case Rep Ophthalmol. 2013;doi:10.1159/000356123.
Roberts C. Biomechanical and topographic analysis in crosslinking. escrs.org/milan2012/programme/Handouts/IC-8/IC-08_Roberts_Biomech%20CXL%20Handout.pdf. Presented at: European Society of Cataract and Refractive Surgeons meeting; 2012; Milan.
Rocha KM, et al. J Refract Surg. 2013;doi:10.3928/1081597X-20130129-08.
Salomão MQ, et al. J Refract Surg. 2011;doi:10.3928/1081597X-20101201-02.
Santhiago MR, et al. Cornea. 2012;doi:10.1097/ICO.0b013e3182068c42.
Söğütlü E, et al. Cornea. 2012;doi:10.1097/ICO.0b013e31822481df.
For more information:
Jorge L. Alió, MD, PhD, can be reached at Vissum, Instituto Oftalmologico de Alicante, Avda. de Denia, s/n, 03016 Alicante, Spain; email: jlalio@vissum.com.
Arthur B. Cummings, FRCS, can be reached at Wellington Eye Clinic, Suite 36 Beacon Hall, Beacon Court, Sandyford, Dublin, Ireland; email: abc@wellingtoneyeclinic.com.
A. John Kanellopoulos, MD, can be reached 115 E. 61st St., Suite 1B, New York, NY 10021; email: ajk@brilliantvision.com.
Michael Mrochen, PhD, can be reached at IROC Science, Technoparkstrasse 1, 8005 Zurich, Switzerland; email: michael.mrochen@irocscience.com.
Karolinne Rocha, MD, PhD, can be reached at Cole Eye Institute, Cleveland Clinic Foundation, 2022 East 105th St., Cleveland, OH 44195; email: karolinnemaia@hotmail.com.
Roy S. Rubinfeld, MD, can be reached at Re:Vision, Roy Rubinfeld, MD, PO Box 30845, Bethesda, MD 20824; email: rsrubinfeld@gmail.com.
William B. Trattler, MD, can be reached at 8940 N. Kendall Drive, #400, Miami, FL 33176; email: wtrattler@gmail.com.
George O. Waring IV, MD, can be reached MUSC Storm Eye Institute 167 Ashley Ave., Charleston, SC 29425; email: georgewaring@me.com.
Disclosures: Alió and Cummings have no relevant financial disclosures. Kanellopoulos has financial interests in Avedro, WaveLight/Alcon and Optovue. Mrochen is an IROC Innocross shareholder. Rocha is a consultant for CXL Ophthalmics LLC. Rubinfeld has financial interests in CXLUSA LLC, CXLO LLC and CurveRight LLC. Trattler has a financial interest in CXLO. Waring is on the medical advisory board of CXLUSA.


Would you treat keratoconus with corneal collagen cross-linking — epi-on or epi-off — in the absence of documented progression?


Consider patient profile when deciding

Peter S. Hersh, MD

Peter S. Hersh

In some cases, yes. Of course, the main reason to use collagen cross-linking is to stabilize the cornea and prevent progression, but cross-linking may have other beneficial clinical effects in selected patients. We did a multifactorial analysis to determine which patients may do better or who may do worse with cross-linking, and we found that there are some predictors of topography improvement as well as some predictors of correctable visual acuity.

First, a higher preoperative maximum keratometry reading was independently associated with more improvement in the cone height after cross-linking. Specifically, eyes with a maximum K of 55 D or more were 5.2 times more likely to flatten by at least 2 D at 1 year. So there is a subset of patients who may achieve clinically significant flattening of the cone from cross-linking. Such flattening could help, for instance, in contact lens fitting.

Second, we looked at factors that might lead to improvement in vision, and we found that a worse preoperative spectacle-corrected vision was independently associated with two or more lines of improvement of spectacle-corrected vision, such that eyes that had spectacle-corrected vision of 20/40 or worse were 5.9 times more likely to improve their correctable vision by two lines or more at 1 year. So we are able to identify patients who in addition to being stabilized may improve as well.

We therefore may want to take this into account preoperatively. Even in some patients with no documented progression, we may still consider cross-linking if they fit the profile of somebody in whom we may improve corneal topography or in whom we may improve spectacle-corrected visual acuity. With this in mind, we are currently working on a patient treatment algorithm to help guide patient selection for cross-linking.

Peter S. Hersh, MD, is an OSN Refractive Surgery Board Member. Disclosure: Hersh is a consultant of Avedro Inc.


Slow progressors may still benefit

John A. Hovanesian, MD, FACS

John A. Hovanesian

Generally yes. When critiquing research, my former professors would say, “The enemy of truth is short-term follow-up.” The advent of topography 20 years ago has now shown us that many patients with keratoconus have meaningful but very slow topographic progression of keratoconus over many years. Furthermore, it was thought that keratoconus progressed only early in life (into a patient’s late 20s) and became stable around age 30. Longer-term studies have shown that this, too, is not always true; some patients in their 60s have demonstrable, progressive corneal steepening that stabilizes with cross-linking. So denying a patient the visual and stability benefits of cross-linking on the basis that his or her keratoconus is “stable” may well be doing the patient a disservice.

Furthermore, the risks associated with cross-linking are very minimal with recent techniques. “Epi-on” or transepithelial cross-linking creates only minimal corneal trauma and all the same benefits as older, epi-off techniques. Delayed healing and infection are almost non-issues. Besides some discomfort the night of the procedure, most patients tolerate this procedure extremely well.

When you stack up the risks against the well-documented visual benefits of collagen cross-linking and realize that many “stable” keratoconus patients are indeed progressing, it is clear why European surgeons uniformly apply this procedure to nearly all patients with visually significant keratoconus.

John A. Hovanesian, MD, FACS, is OSN Cataract Surgery Section Editor. Disclosure: Hovanesian is an investigator in the CXLUSA collagen cross-linking study.