Safety and Efficacy of Epithelial Island Crosslinking in Keratoconus with Thinnest Pachymetry less than 400µ.

PURPOSE: To evaluate the efficacy and safety of epithelial-island crosslinking (EI-CXL) in keratoconus with corneas thinner than 400 µm.
METHODS: Twenty-six patients (30 eyes) underwent EI-CXL (preserving the epithelium over the thinnest area), using standard protocol (3 mW/cm2 for 30 min). Uncorrected and best spectacle-corrected distance visual acuity (UCDVA, BCDVA), manifest refractive spherical equivalent (SEQ), mean simulated keratometry (Kmean), maximum keratometry (Kmax), and thinnest corneal thickness (TCT) were determined preoperatively and at 1, 3, 6, and 12 months following CXL. Endothelial cell count (ECC) was determined preoperatively and at 6 months. Anterior segment optical coherence tomography (AS-OCT) was done at 1 month to determine the depth of the corneal stromal demarcation line (DL).
RESULTS: After 1 year, mean UCDVA improved from 1.29 preoperatively to 1.17 (P = 0.001) and BCDVA from 0.62 to 0.57 (P = 0.011). Mean manifest SEQ decreased from -7.63 to-7.32D (P = 0.001). Mean Kmean decreased from 54.92 to 54.81D (P = 0.045), and Kmax from 67.60 to 67.42D (P = 0.072), and mean TCT changed minimally from 377.17 to 375.30 µ (P = 0.11). The mean ECC decreased from 2329 to 2268 cells/mm2 (2.6% decrease, P < 0.001). AS-OCT showed a DL in 29 out of 30 eyes at an average depth of 215.9 µ under the spared epithelium, and 299.9 µ in the de-epithelialized cornea.
CONCLUSION: EI-CXL halted keratoconus progression over a 1-year period. This was associated with statistically significant endothelial loss, but less than seen with conventional epi-off CXL in thinner corneas. Copyright:

Introduction

Collagen crosslinking (CXL) is the treatment of choice for halting progressive corneal ectasia. The most commonly used technique is known as the “Dresden protocol.”[1] This is recommended in eyes with corneal thickness of at least 400µ after de-epithelialization, to avoid endothelial toxicity.[23] However, most keratoconic corneas requiring CXL may not fulfil this preoperative requirement, where pachymetry may be less than this threshold.[4]

Several protocols have been proposed to achieve safe CXL in thin corneas. One of these involves the use of hypo-osmolar riboflavin. However, Stojanovic et al. found that although this procedure was safe and halted the progression of keratoconus, the effect was less than that seen with CXL in normal corneas.[5] Another idea was to avoid epithelial debridement to preserve corneal thickness, which was termed transepithelial CXL. Since riboflavin does not penetrate the normal corneal epithelium, several chemical agents were used to pretreat the intact epithelium to make it more permeable to riboflavin, including ethylenediaminetetraacetic acid, trometarol, and benzalkonium chloride (BAK). However, the depth of stromal demarcation line (DL) in cases of transepithelial CXL is approximately 200µ deep, indicating that the actual CXL effect might be less as compared to the standard protocol.[6] Another protocol is iontophoresis-assisted cross-linking (I-CXL). However, studies of I-CXL reported a DL at a depth of 216µ in 35% of the patients, and a failure rate of 20%, as compared to a 7.5% of failure rate following conventional CXL.[7]

Sachdev et al. described a technique for CXL of thin corneas, where they used a stromal lenticule removed from patients undergoing small incision lenticule extraction for myopic correction, customized according to corneal thickness. A similar technique is contact lens-assisted CXL, where a riboflavin-soaked contact lens is applied to the debrided cornea. However, long-term studies are needed to establish the safety and efficacy of these protocols.[8910]

In 2009, Kymionis et al. described a technique that involved epithelial debridement of the keratoconic cornea sparing the epithelium over the apex of the cone as determined on topography.[11] Only two patients were reported in this study. Later on, Mazzotta and Ramovecchi performed almost the same technique and reported that the CXL effect was seen at a depth of 150µ in the epithelium-on area, as compared to 250µ in the epithelium-off area, indicating a definite, but lower CXL effect under the intact epithelium.[12] This study involved ten patients, and the CXL effect was determined by measuring the depth of anterior stroma with keratocyte apoptosis, as seen on confocal microscopy.

In this study, we prospectively evaluated 30 eyes with a similar technique of partial epithelial debridement, and employed anterior segment optical coherence tomography (AS-OCT) to determine the depth of DL after surgery.

Methods

Study population and type of study

This prospective, interventional study was conducted on 30 eyes of 26 patients diagnosed with progressive keratoconus, all of which had a thinnest point of 360–400µ (including epithelium and stroma) as per Oculus Pentacam (OCULUS, Inc., USA). Patients were recruited from the ophthalmology clinic at Kasr Al Ainy Hospital, Cairo University, during the period from September 2017 till March 2019. The study was approved by the Ethical Committee of Kasr Al Ainy School of medicine, Cairo University.

Inclusion criteria

Patients with progressive keratoconus, defined as an increase in maximum keratometry reading (K max) value of >1 D from baseline, increase in mean simulated keratometry (K mean) >0.75 D from baseline, increase in manifest spherical equivalent (SEQ) >0.5 D from baseline, or a decrease in thinnest corneal point of >2% from baseline, throughout a follow-up period of 1 year

Patients diagnosed with keratoconus younger than 20 years of age without the need for documented progression[13]

Clear cornea.

Exclusion criteria

Patients who have undergone previous corneal collagen cross-linking

Pregnant or lactating females

Patients who have undergone previous corneal refractive procedures

Patients with a history of herpetic keratitis

Patients with associated retinal pathologies.

Preoperative visit

History taking

Uncorrected distance visual acuity (UCDVA) and best spectacle-corrected distance visual acuity (BCDVA)

Manifest refractive SEQ

Slit-lamp examination of the anterior and posterior segments with full dilatation and tonometry

Scheimpflug imaging with Oculus Pentacam (OCULUS, Inc. USA)

Specular microscopy using Konan Medical CellChek X8 (KONAN MEDICAL, Inc., USA)

Patients signed an informed consent for the procedure conforming with the declarations of Helsinki.

Procedure

Topical anesthesia is instilled in the patient's eye, and the horizontal and vertical meridians are marked while the patient is in the upright position at the slit lamp. This is to avoid the improper identification of corneal thinnest point location in the lying position that occurs due to cyclotorsion

After the patient lies down on the operating table, additional topical anesthesia is instilled, and the speculum is applied.

Method of epithelial island crosslinking

Defining the area of the epithelial island (where epithelium will not be debrided):

A color-coded thickness map with a millimeter scale is obtained from the patient's tomographic data [Figure 1]

Figure 1

Preoperative tomography with millimeter scale (red circle represents the epithelial Island, blue circle represents the pupillary area)

The circle indicating a thickness of 450µ is identified on the color scale. This circle is then identified in the thickness map [the red circle in Figure 1]. The epithelium within this circle will be left intact

Two horizontal lines are projected from the uppermost and lowermost points of this circle to the vertical millimeter scale to estimate the vertical diameter of this circle [approximately 3 mm in the case depicted in Figure 1]. Two vertical lines are then likewise projected from the rightmost and leftmost points of the circle to the millimeter scale, to estimate the horizontal diameter of the circle (also approximately 3 mm)

Finally, the relationship between this circle and the pupil is noted on the thickness map. In almost all cases, the epithelial island [red circle in Figure 1] was found to be shifted slightly inferiorly and temporally from the pupil [blue circle in Figure 1].

Marking the epithelial island on the patient's cornea:

A caliper was set at the desired diameter of the epithelial island. The caliper was used to mark the circle of the epithelial island using a pair of ink marks in the horizontal, vertical, left oblique, and right oblique meridians, slightly inferior and temporal to the pupil [figure 2a and the blue circle in figure 2b]. A 9mm marker was used to delineate the peripheral boundary of the debridement area [figure 2a and the red line in figure 2b]. The epithelium was then debrided starting from the marked central circle to the 9-mm line. [Figure 2a].

Figure 2

(a) Figure showing the surgeon's view of the left eye. The epithelial island is slightly below and temporal to the pupil. (b) Figure showing the same eye as Figure 2a, with an overlay (blue circle) indicating the boundaries of the pithelial island, and another (red circle), indicating the peripheral boundary of the debrided epithelium. The area of debrided epithelium is soaked with riboflavin

The cornea is soaked with 0.1% riboflavin with hydroxypropyl methylcellulose, by instilling 2 drops every 2 min, for 10 min

The ultraviolet (UV)-A irradiation is performed at 3 mW/cm2 for 30 min of total UV-A exposure time (six sequential steps of 5 min each) and energy dose of 5.4 J/cm2. We used the C. S. O. VEGA CBM-X-LINKER (C. S. O., Italy)

The cornea is then rinsed with balanced salt solution, covered with a therapeutic contact lens for 5 days and medicated with moxifloxacin, dexamethasone 0.1% and sodium hyaluronate 0.2% eye drops four times daily. Following the first 5 days after surgery, these drops were then tapered over a period of 2 weeks.

Postop visits

An initial visit was done the 2nd day following the procedure to confirm the absence of infection. A second visit was done 5 days postoperatively to remove the contact lens and confirm full re-epithelialization. Examinations were then carried out at the 1st , 3rd , 6th , and 12 months postoperatively, where patients were subjected to the following:

Determination of UCDVA and BCDVA using subjective refraction techniques

Manifest refractive SEQ

Slit-lamp examination and intra-ocular pressure measurement by applanation tonometry

Scheimpflug imaging using Oculus Pentacam (OCULUS, Inc., USA).

AS-OCT was done at the 1-month follow-up visit to measure the depth and extent of the DL. It was done using RTVue-100 (Optovue, Inc., Fremont, CA, USA) spectral-domain machine after attaching the cornea-anterior module-L. Raster scans centered around the apex of the cone were acquired, and the DL was identified as a continuous hyperrefective line within the stroma. The depth of this line from the anterior corneal surface was measured using a caliper, both centrally in the area of the epithelial island and in the peripheral cornea outside this zone.

Specular microscopy was done at the 6-month follow-up visit using Konan Medical CellChek X8 (KONAN MEDICAL, Inc., USA). The patient fixates a target inside the eyepiece and the central area of the cornea was scanned automatically and the endothelial cell count (ECC) among other parameters, such as hexagonal cell percentage, the coefficient of variation of endothelial cells and the pachymetry, was analyzed and displayed.

Statistical analysis

Descriptive statistical analysis was done using  IBM Statistical Package for the Social Sciences (SPSS) for Windows version 24 software (Armonk, New York: IBM Corp, USA). Statistically significant differences are those whose P value is 0.05 or less. Friedman analysis of variance test was used to compare all related values across all follow-up periods. Related samples Wilcoxon signed-rank test was used to compare the preoperative and the 1-year postoperative values.

Results

Demographic data

Ten males and twenty female patients were recruited with a mean age of 22.23 ± 3.49 years (range 14–28 years)

Postoperative epithelial healing

All patients had full re-epithelialization by the 4th postoperative day. None of the patients had delayed epithelial healing, nor stromal edema, and there were no cases of infection.

Mild stromal haze, indicating a treatment effect, was seen in all cases. In the area of preserved epithelium, the haze tended to be more superficial.

Topographical data

UCDVA, BCDVA, manifest refractive SEQ

At 1 month, all three parameters decreased below the preoperative level, and increased thereafter, returning to a point close to the preoperative level at 1 year. The difference between preoperative levels and levels at 1 year was statistically significant for all 3 parameters (UCDVA P = 0.001, BCDVA P = 0.011, SEQ P = 0.001) [Table 1].

Table 1

Changes in the different parameters and their statistical significance

	Mean±SD		P
	Preoperative	1 year after surgery
LogMAR UCDVA	1.29±0.36	1.17±0.42	0.001
LogMAR BCDVA	0.62±0.32	0.57±0.34	0.011
SEQ (diopters)	−7.63±2.91	−7.32±2.93	0.001
K mean (diopters)	54.92±5.07	54.81±4.98	0.045
K max (diopters)	67.600±4.34	67.42±4.18	0.072
Thinnest point (µm)	377.17±11.42	375.30±9.97	0.110
ECC	2329±490.68	2268±472.66	<0.001

K mean: Simulated mean keratometry, K max: Maximum keratometry, UCDVA: Uncorrected distance visual acuity, BCDVA: Best-corrected distance visual acuity, SEQ: Spherical equivalent, SD: Standard deviation, LogMAR: Logarithm minimum angle of resolution, ECC: Endothelial cell count

Mean simulated keratometry (Kmean) and maximum keratometry (Kmax).

At 1 month, Kmean and Kmax increased above the preoperative level, and decreased thereafter, returning to a point close to preoperative levels at 1 year. The difference between preoperative values and values at 1 year after surgery was statistically significant for Kmean (P = 0.045), but insignificant for mean Kmax (P = 0.072) [Table 1].

Topographic thinnest point

At 1 month, there was a decrease in the thinnest point thickness as compared to baseline. Thereafter, the value increased gradually reaching a point close to the baseline at 1 year. The difference between preoperative levels and levels at 1 year was not statistically significant (P = 0.110) [Table 1].

Endothelial cell count

The mean ECC decreased from 2329.93 ± 490.68 cells/mm2 preoperatively, to 2268.60 ± 472.66 cells/mm2 6 months after surgery, representing a mean decrease of 2.6%. This was statistically significant (P < 0.001) [Table 1].

Table 1 shows the summary of the changes occurring in the afore-mentioned parameters between the preoperative and 1-year visits.

Demarcation line

One month after surgery, AS-OCT was used to analyze the presence and depth of DLs under the spared epithelium, as well as in the peripheral de-epithelialized cornea. It was possible to detect a DL in 29 out of the 30 eyes (96.7%).

A DL was detected under the spared epithelium at an average depth of 215.93 ± 85.48 microns. Under the peripheral cornea, the average depth was 299.90 ± 114.30µ [Figure 3]. The difference was statistically significant (P < 0.001)

Figure 3

Bar chart showing depth of demarcation line under epithelialized and non-epithelialized areas

Discussion

Corneal collagen cross linking with the use of photosensitizing riboflavin and UV-A has demonstrated satisfactory results in stabilizing keratoconic and keratectatic corneas. A major limitation of the technique is that its application is contraindicated in corneas with stromal pachymetry <400 µm. This is due to the fact that an irradiance of 0.37 mW/cm2 has been found to be cytotoxic for the endothelial cell layer in these thin corneas. As a result, the 400µ limit is considered the safe limit to protect the endothelium and intraocular structures from the adverse effects of UV irradiation, and it has been established as a clinical standard.[214] Unfortunately, the very patients who are in need of CXL are the same ones who have thin corneas often below the threshold of that considered to be safe for CXL treatment.[11]

Kymionis et al. were the first to describe a customized epithelial debridement CXL technique on two patients (one with keratoconus and the other with post-LASIK ectasia). Both patients had a pachymetry of <400µ (380 and 375µ). The patients were then followed-up for 9 months and the topography showed stability [Figure 4].[11]

Figure 4

Anterior segment optical coherence tomography showing demarcation line 288 µm under spared epithelium and 380 µm under peripheral cornea

Mazzotta and Ramovecchi[12] performed a similar technique on ten patients with progressive keratoconus with average thinnest point of 384 µm (range 368–391 µm). After 1 year of follow-up, there was a slight decrease of average K (49.7D to 49.1D), a slight change in thinnest point, and an overall improvement of visual acuity by an average of 1 decimal equivalent. All changes were not statistically significant.

These results agree with our findings of a statistically significant decrease of our average K (P = 0.045) from baseline to the 1-year follow-up, as well as a statistically insignificant decrease in K max values. Moreover, their study agrees with the improvement in visual acuity observed in our study. In addition, our study has demonstrated a statistically insignificant change in thinnest point thickness (P = 0.110).

Mazzotta et al. evaluated the efficacy of the technique using in vivo confocal microscopy, whereas in our study, we used AS-OCT. Mazzotta et al. detected keratocyte apoptosis 1 month after treatment. The mean depth of this event was 160µ under the epithelial island (range 130–180µ), and 250µ under the peripheral cornea (range 230–270µ).[12] In our study, the mean depth of DL was 215.93 ± 85.48µ under the epithelial island, and 299.90 ± 114.30µ under the peripheral de-epithelialized cornea. Regarding safety to the corneal endothelium, Mazzotta and Ramovecchi[12] did not detect any endothelial cell loss. In our study, the mean ECC decreased from 2329.93 ± 490.68 cells/mm2 preoperatively to 2268.60 ± 472.66 cells/mm2 6 months after surgery (mean decrease of 2.6%) (P < 0.001). Regarding endothelial cell loss, our results are superior to those obtained with conventional CXL with complete epithelial debridement, in corneas thinner than 400µ.[1516]

We hypothesize that this contrast in variation of CXL effect and endothelial cell loss might be due to the different method of identifying the apical thin area of the cornea intraoperatively. In their study, Mazzotta and Ramovecchi[12] used a Maloney keratometer to localize the apical region, and then applied a Bores 3.25 mm zone cone marker as a standard for all patients undergoing treatment. Therefore, they spared a larger diameter of epithelium over the apex before debridement than in our study. Moreover, Mazzotta and Ramovecchi[12] used an intra-operative pachymeter, and whenever a thickness of <350µ was encountered, the application of hypotonic riboflavin for 10 min was done. Their technique might serve better in a more precise identification of the thin area, sparing more epithelium, as well as more correction of intra-operative corneal dehydration, but with a lessened depth of the CXL effect. However, our study proved the efficacy of the technique used as the depth of the DL detected under de-epithelialized areas approached that detected in conventional CXL. The only patient in our study who did not show a DL might be explained by the fact that the haze resulting from CXL effect might have backscattered the light of the AS-OCT, preventing the detection of a continuous DL. Moreover, the sample size of the study conducted by Mazzotta et al. was smaller (10 patients) when compared to that in our study (30 eyes).

Kymionis et al.,[3] performed conventional epi-off CXL on thin corneas (<400µ after complete epithelial removal) and reported a drop in endothelial cell density averaging 10% postoperative. In contrast, sparing the epithelium over the thin area in our study resulted in a lessened endothelial cell loss averaging 2.63%. This variation in endothelial cell loss highlights the importance of custom epithelial debridement in CXL of thin corneas.

The finding of a proper DL under the spared epithelium in our study agrees with its detection in transepithelial CXL studies with enhanced riboflavin penetration. Bonnel et al.,[17] evaluated the DL of transepithelial iontophoresis-assisted CXL and detected the stromal line at a mean depth of 246.67 ± 50.72µ (range 183–339µ) by an examiner and 241.89 ± 62.52µ (range 163–358µ) by a second examiner. Another study evaluating transepithelial accelerated CXL, using BAK as a riboflavin penetration enhancer, has detected an average stromal line at depth of 210.63 ± 27.50µ.[18] We believe these results reinforce the idea that in epithelial island CXL, the peripheral debrided cornea potentiates and enhances the diffusion and penetration of riboflavin to the thin area under the spared epithelium, explaining our finding of a proper DL under the spared area at an average depth of 215.93 ± 85.48µ.

Other CXL techniques have been described for thin corneas. These methods include contact lens-assisted CXL which was described by Jacob et al. for keratoconic corneas having pachymetry of 350–400µ after epithelial removal. They reported no significant changes to visual acuity and a stable topography with no statistically significant change in K max after a mean follow-up of 6.1 ± 0.3 months. The present study also reported no significant endothelial cell loss (P = 0.063).[9] While this technique may prove to be effective in longer-term studies, we suspect its efficacy may be hindered by the fact that only 50% of the irradiation reaches the anterior stroma resulting in a shallow DL.[9]

The armamentarium for thin corneas also encompasses transepithelial CXL with riboflavin penetration enhancers, which have so far gathered mixed results, and iontophoresis assisted CXL. However, most of these studies agreed upon a relatively shallow DL when compared to the conventional cross linking protocol, thus questioning the efficacy of these procedures.[19202122] Moreover, Hafezi reported a case of failure of corneal CXL for progressive keratoconus after preoperative stromal swelling using hypoosmolar riboflavin solution in extremely thin cornea.[23]

Recently, Hafezi et al. described a technique for thin corneas, where the total energy delivered is tailored to the thinnest corneal location. This is achieved by varying the duration of irradiation with a constant fluence of 3 mW/cm², which is the standard fluence used in the Dresden Protocol. Although, this would be probably safer to the endothelium, the main concern would be a reduction of CXL effect, due to reduction of the total energy delivered. Long-term studies should determine the long-term efficacy of this technique.[24]

Our study showed an initial worsening was observed in visual acuity, SEQ, mean, and maximum K values 1 month following the procedure. Other studies have shown the similar patterns in conventional and epithelial island CXL.[122526] We believe these changes might be due to the recovery and remodeling processes that take place in the cornea following cross-linking.

Many clinical trials examining corneal collagen cross linking have suggested criteria to establish keratoconus progression. These criteria include an increase in K max value of >1 D from baseline, increase in K mean >0.75 D from baseline, increase in manifest SEQ >0.5 D from baseline, and a decrease in thinnest point of >2% from baseline.[27] Applying those criteria on this study's results, we believe the patients enrolled in our study show stabilization of keratoconus progression in the 1-year follow-up.

However, our study is limited by a relatively small sample size, a relatively short 1-year follow-up period, the lack of wave-front analysis, and the absence of a control group. Moreover, an intra-operative pachymeter could have helped in a more precise identification of the cone, which may further lessen endothelial cell loss.

Our study demonstrates the safety and efficacy of epithelial island CXL in corneas thinner than 400µ. In terms of safety, endothelial cell loss is less than conventional epithelium-off CXL in these corneas. In terms of efficacy, the depth of DL is comparable to conventional CXL, unlike most other modalities where a shallow DL is seen. Moreover, our study suggests that some CXL effect is still achievable in the area of spared epithelium. However, longer-term studies of this technique are needed to assess the longevity of the cross-linking effect.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.